robert harley who has consummate ability to elucidate just about any description of a sound compares jitter to image stabilizing binoculars. its in the newest issue-#194 with the meridian speaker on the front, august.
...regards...tom
New issue of what? Not everyone lives where you live, and I'd like to read that.
New issue of what? Not everyone lives where you live, and I'd like to read that.
"TAS" The Absolute Sound...
But image stabilization? That's a stretch.
But image stabilization? That's a stretch.
One could call it the opposite. Image stabilisation addresses
spatial but digital audio jitter is
temporal distortion. The audio equivalent of image stabilisation would instead be fixing the spatial image between two stereo channels.
A better "analog world" example for jitter correction would be the stabilization of a spiral spring's circular motion with a pendulum in clocks.
robert harley who has consummate ability to elucidate just about any description of a sound
including imaginary ones.
Audiophile gibberish belongs in general audio, not scientific discussion. Moved.
"including imaginary ones" WOW, i've found another genius here at h2audio forum!
sorry hlloyge, i wasnt thinking that everyone might not know that abbreviation. i usually post at audioasylum.com where most of us crazies know about it. tas has been doing a great job of evaluating audio equipment since '73 and coined the term 'high end audio'.
robert harley is a digital recording engineer who used to work for reference recordings i believe or a similarly fine company. he is also an analog lover of the first magnitude. when one review equipment at the level he does, MANY more details and nuances become quite obvious.
as for the stretch, perhaps so to illustrate the point, not make a direct equivalency statement. i for one could only imagine what exactly jitter is and how it might affect the sound. i think RH came pretty close with that verbiage.
It's easy to bandy about sarcasm with the word genius just as it is easy for someone to call a subjective audiophile delusional. So far I've not seen ample evidence to conclude either is right as of now.
However, arguing from an appeal to authority won't buy you much in these parts. It's more likely to elicit challenges to any and all people who claim to distinguish one thing from another to prove it through a double-blind test.
If you can't handle such challenges, this is probably not the forum for you.
I'd appreciate an in-depth description of what Harley actually wrote before the inevitable throwdown that is about to occur.
I'd appreciate an in-depth description of what Harley actually wrote before the inevitable throwdown that is about to occur.
Hear! Hear!
robert harley who has consummate...
You had me at "robert harley".
robert harley is a digital recording engineer who used to work for reference recordings i believe or a similarly fine company. he is also an analog lover of the first magnitude. when one review equipment at the level he does, MANY more details and nuances become quite obvious.
Is he? Well, I haven't heard of him anyways, but I know of jitter, so it would be nice if I could find that article to read it and possibly expand knowledge.
Harley was the one that concluded that the glass CD sounded more "analog-like" and other such things, so he's an expert in analog and an analog lover. He's got some articles at our favorite analog-lovers publication Stereophile.
Oh, and this (http://www.avguide.com/forums/blind-listening-tests-are-flawed-editorial?page=1) piece of groundbreaking argument against DBT.
I'd appreciate an in-depth description of what Harley actually wrote before the inevitable throwdown that is about to occur.
Yup yup yup
robert harley who has consummate ability to elucidate
There comes a point when you're just communicating so badly you're going to get mocked for it, regardless of substance. This would be it.
this place is inviting as a toilet with a razor blade seat. i intend to invoke my will upon you, i just wont sit down. it seems this must be where audio annex is coming to rest. heheheh. cant fool me.
this place is inviting as a toilet with a razor blade seat
Show us some science, and we'll love you forever. Show us more of the same subjectivist nonsense and your welcome will be warm as a Canadian winter. That's just how it works. It's like walking into a Mac forum talking about how great Windows 7 is.
This is not exactly a great situation to try to be an apologist for the HA party line by explaining why our derison is so justified. Especially when some people really are being rude. (I love ya, krab, I really do, but was that really necessary?) But there are about 5 billion reasons why we should be contemptful of anything Harley says without engineering or psychoacoustic justification.
tommy, many of us (and I am specifically talking at least half of the posters on this thread) have a far more technical grasp of the mechanics of jitter than we are anticipating reading about in TAS. A good number of us have read Hawksford's papers on jitter simulation. A few of us know how to actually implement a jitter simulator. And most of us know how the audibility of jitter can be evaluated in a by-the-books psychoacoustic (or if you will, flat earth) fashion and grasp why such a large gap exists between what is considered audible in the engineering literature and what is considered audible in the high-end literature.
So, you make a post here - on a forum on scientific discussion - mentioning what we're all pretty sure is going to be a fluff piece that is going to be technically wrong... and you expect us to take it seriously? Of course we're going to chuckle (and mock).
You liken HA to a Audio Annex? I mean, really? HA is like the adult room compared to the kiddie rooms of, say, Audio Asylum. If you post, back your sh*t up, or be educated. Don't make personal insults and don't get emotional or personal. Don't make claims of audibility unless you can prove only aural senses are involved. These are not hard rules for educated adults to follow.
But hey.. you're an adult! And you want to learn! Pull up a chair! Let's get down to bidness. In terms of relating jitter to image stabilizing binoculars... image destabilization can be treated mathematically as an image convolution (http://en.wikipedia.org/wiki/Convolution) with a profoundly asymeetric kernel. Just to make this brutally clear:
(http://files.audiamorous.net/images/cfl-orig.jpg) x (http://files.audiamorous.net/images/conv-kernel.jpg)= (http://files.audiamorous.net/images/cfl-conf.jpg)
(CC-BY-NC-SA 2.0: attribution (http://www.flickr.com/photos/vlastula/))
Convolution is a filtering operation, like a lowpass filter, or an eq, or an antialiasing filter. It doesn't really map very well to the domain of temporal distortions, at least not from the little bits of text you are quoting. The notion that it smears or occludes detail in a recording is only true in the most absolute general sense - that all distortions can do things like that - but if I were to try to formulate an accurate analogy between digital audio jitter and digital photography, I would think a better analogy would be taking a photograph with an image sensor whose pixel cells were radically misaligned with respect to one another:
(http://files.audiamorous.net/images/cfl-random.jpg)
But Harley didn't mention any of that in his article, did he? Of course not. I guess nitpicking on analogies is still nitpicking, but without seeing the article, it does not give me all that much confidence in it.
this place is inviting as a toilet with a razor blade seat
Show us some science, and we'll love you forever. Show us more of the same subjectivist nonsense and your welcome will be warm as a Canadian winter. That's just how it works. It's like walking into a Mac forum talking about how great Windows 7 is.
Um... comparing HA to a Mac forum is rather insulting. Of us.
Yeah, bad analogy, I know... I was trying to capture the flavour, but apparently unsuccessfully. Quitting caffeine is rough.
The most comprehensive, insightful, and freely available text about jitter and clock recovery is this (page 11) (http://www.theaudiocritic.com/back_issues/The_Audio_Critic_21_r.pdf) by Robert W. Adams. Pretty much said it all in 1994.
Edit:
Haha, what a funny coincidence! I just found that the above article even has a one page prolog explicitly showcasing this Robert Harley and how he was making a fool of himself by totally misunderstanding jitter and misleading others about it.
tas isnt scientific american nor does it claim to be anywhere near that pub. maybe RH didnt hit the nail on the head but tried an analogy for the non technical crowd. and perhaps this particular board at HA was a poor choice for an initial hobbyist post.
i can at least be respectful of the fact that none of you stooped to using emoticons. thank you axon for lightening the load on my neurons.
...regards...tom
A "good explanation of jitter in tas" eh? I've yet to see much other than obfustication in TAS. Well, I suppose seeing is believing, but if you don't mind I won't be holding my breath in the meantime.
It's easy to bandy about sarcasm with the word genius just as it is easy for someone to call a subjective audiophile delusional. So far I've not seen ample evidence to conclude either is right as of now.
You didn't read his Stereophile manifesto anti blind testing?
If you want some more giggles, pick up a used copy of his 'Complete Guide to High End Audio'; I perso nally was deeply, er, impressed by his views on the effect of copper crystal directionality on cable sound (and no, I'm not going to dig the book out an quote it again). But since I've been called rude for noting Harley's woo tendencies, here's a reviewer who gave it a '3' out of 5....trying to put the best face on it.
http://www.amazon.com/Complete-Guide-High-...By=addThreeStar (http://www.amazon.com/Complete-Guide-High-End-Audio/product-reviews/0964084961/ref=cm_cr_pr_hist_3?ie=UTF8&showViewpoints=0&filterBy=addThreeStar)
This book is:
Almost what the High-End Audio industry needs
It's important for me that a book be written on the advantages of high-end audio. It's the business I'm in and it is dear to me on a personal level. There are many excellent portions and handy information pieces scattered about the book and that's the good news.
I will refrain from personalizing my complaints and stick to the issues as I see them. To write this book as the author, you should know Ohm's Law. Harley does not. This is made evident in several examples. Amperes, voltage and wattage are all part of a greater equation that appears to mystify the author. The basic laws of physics and simple electrical concepts need be firmly grasped prior to making an endeavor such as this. There are many elements of "Dark Science" in the high-end audio realm and a mystique that is largely relevant. This book does a strong job of handling that delicate balance between science and myth, that is so important to this industry. Along the way however it forgets to "check the science"
That's too bad, but not a total loss...
A serious explanation of negative feedback as used in power amplifiers would have been pretty easy to put down for the record. Most power amp manufactures have fascinating solutions to the problems associated with negative feedback. A breakdown of a few of the key developments in this area would have been excellent. An opportunity missed. Instead he uses an example of a negative feedback amplifier and calls it just the opposite! At that point in the book I admit I was a bit frustrated.
The good parts are many!
It's an enjoyable read when the author sticks to what he actually knows, acoustics and auditioning gear. I learned much and felt the key points were illustrated clearly and in the contexts of meaningful application. I am not saying "don't buy"
I guess I'm saying this book missed a huge opportunity simply by not getting some important parts right. That's all.
Btw, Axon...that bottom pic, c'est tres Seurat!
(http://www.poster.net/seurat-georges/seurat-georges-seine-grande-jatte-2602499.jpg)
You didn't read his Stereophile manifesto anti blind testing?
That should be a rhetorical question, otherwise one might consider me a liar. I should have known better based on the looks of the responses and the little that I did read since I gave mine.
It would seem that we need not have another anti-audiophool fest in this discussion since it appears that hifitommy has realized that he's amongst a group of people who understand digital audio to a depth that is beyond what is written in magazines designed to part people from their money.
robert harley who has consummate ability to elucidate just about any description of a sound compares jitter to image stabilizing binoculars. its in the newest issue-#194 with the meridian speaker on the front, august.
...regards...tom
Almost nobody here is able to read a copy of this article. If Robert Harley has indeed given a clear explanation of the sound of jitter then it should be straightforward for you to put in your own words the basis for his reasoning. If you attempt to do this you will almost certainly find that the foundations on which he bases his reasoning is not established knowledge but uncontrolled observations at best or unsupported wishful thinking at worst.
robert harley who has consummate ability to elucidate just about any description of a sound compares jitter to image stabilizing binoculars. its in the newest issue-#194 with the meridian speaker on the front, august.
I just read it and it is a laughable comparison. The two phenomena have entirely different causes and anyone who has ever looked throught a decently mounted telescope should understand why. He might have gotten more mileage from a comparison with the modern image correction optics on the large ground based telescopes which is now common, but it is no more apt a comparison.
I don't think of openly misleading the reader, intentionally or unintentionally, as "good writing" myself.
That he can even make such a comparison suggests to me that he really has no idea of what digital jitter is.
That he can even make such a comparison suggests to me that he really has no idea of what digital jitter is.
Why would you assume that? Robert Harley earns his living as a writer/journalist/editor and his target audience is audiophiles like hifitommy. In this case the article would seem to be a good one assuming hifitommy's post is genuine which I would judge likely.
Nobody writing for an audiophile audience can deal with the audibility of jitter in a straightforward manner because it has become an important belief in the audiophile industry/community. To expect audiophile articles to be written for the audience here is daft but a significant number of posters seem to have this expectation.
To expect audiophile articles to be written for the audience here is daft but a significant number of posters seem to have this expectation.
I don't see why one, who had understood the concept, would be forced to use
false analogies just to educate a dumb audience. They just need to be simple, not false.
If he knows better and just sells lies in the believe that's what his customers want, when they buy magazines to inform themselves, then so be it. But when his worshippers show up here and wave about his insight, whack'em! That's a natural side effect of believing in a quack. They have a choice.
I think he's saying if Harley understood it is irrelevant cause he needs to come to the conclusion that jitter is a problem cause it's a deeply held belief in the audiophile world. If he understands it or not is not gonna be a deterrent.
Clearly what Harley is doing is pandering to the belief among audiophiles that jitter is a problem. To do that he has presented a phony analogy to something they can relate to.
If it were a real analogy then the argument would fail, because jitter is not a real problem and the analogy would show that.
From a marketing standpoint it makes sense.
There's nothing much sexy about a grandfather clock. Image stabilization is useful, and I guess, a cool upper end teckie consumer thing to have. If your only interest is selling something, or selling people on something, why not link it to that?
From a marketing standpoint it makes sense.
The sale and marketing of real world phenomena for profit (e. g. Coca Cola) makes sense.
But he sale and marketing of the
explanation of real world phenomena for profit is an assault against science.
Science itself is not perfect but quite successful at self-correcting its progress by adhering to a strict contract of scientific standard. Its results often too complex for the average public, so "explanation layers" between science and consumer make sense; and why should media supplying this service not make profit? That's fine. This tips as soon as you allow this "explanation layer" to cut loose from its scientific input and produce arbitrary output still under the scientific flag. This diverts consumer money away from companies actually bringing forward science to quacks and makes science look like some arbitrary, exchangable mind-set just like creationism.
The sale and marketing of real world phenomena for profit (e. g. Coca Cola) makes sense.
But he sale and marketing of the explanation of real world phenomena for profit is an assault against science.
Science itself is not perfect but quite successful at self-correcting its progress by adhering to a strict contract of scientific standard. Its results often too complex for the average public, so "explanation layers" between science and consumer make sense; and why should media supplying this service not make profit? That's fine. This tips as soon as you allow this "explanation layer" to cut loose from its scientific input and produce arbitrary output still under the scientific flag. This diverts consumer money away from companies actually bringing forward science to quacks and makes science look like some arbitrary, exchangable mind-set just like creationism.
I agree.
Seems to me he's selling Coke.
The purpose of marketing "information" is to provoke a feeling (a warm and positive non-rational response) toward a product or service. It has nothing whatsoever to do with rational thought and as such surely has no place on this forum. That goes for all marketing and PR.
Elections are analogous to the audio world. The more money that is spent on marketing and PR the more the focus is on (a) personal feelings toward the individual candidate (do you like him, is he a stand-up guy, would you have a beer with him) as against (b) the party's manifesto (i.e. the substantive rational expression of policy). You can't enjoy a beer with healthcare policy.
The same goes with audio. Very often what is called audiophile seems to me nothing more than the regurgitation of marketing / PR (a) whereas the focus at HA is (b) - rational and scientific information. The two are completely at odds with oneanother.
All marketing is merely emotional manipulation, and often it's hard to fight against precisely because it bypasses the rational. Which is why many of the recent audiophile threads on HA are like yelling at someone that part of the car's spec is not a gorgeous woman and no, she won't fall in love with you even if you do buy it. If someone has fallen in love with the fantasy (e.g. the audiophile and his cables), rationality is not going to cure them, because that's the very thing they've
chosen to abandon to gain the warm fuzziness of their delusions.
The sale and marketing of real world phenomena for profit (e. g. Coca Cola) makes sense.
Well it makes sense in the same way as "a criminal is someone with insufficient capital to form a corporation" makes sense. It "makes sense" when conscience is removed from the equation, and consciences don't get incorporated. But in fact, it has the opposite effect - it makes
nonsense, in that it profoundly distorts market mechanisms, which rely on
rational choices based on as close as possible to perfect information (i.e. hard spec style data) about products and services.
I think it would be helpful to make a stronger connection between the way businesses sell their products (marketing) and all this irrational BS that keeps popping up on HA - they are profoundly connected IMO.
C.
[EDIT: grammar]
That he can even make such a comparison suggests to me that he really has no idea of what digital jitter is.
Why would you assume that?
Well if one equates jitter in digital audio to the shaking of a hand held binocular then one either profoundly misunderstands one of these two phenomina or is fibbing. Since the fact that optical magnification will also magnify the shaking of one's hand is so obvious and easy to understand, and wishing to be charitable on the truthfulness question, I was left with the likelyhood that it is digital jitter that he misunderstands, or so it seemed to me at the time.
That he can even make such a comparison suggests to me that he really has no idea of what digital jitter is.
Why would you assume that?
Well if one equates jitter in digital audio to the shaking of a hand held binocular then one either profoundly misunderstands one of these two phenomina or is fibbing. Since the fact that optical magnification will also magnify the shaking of one's hand is so obvious and easy to understand, and wishing to be charitable on the truthfulness question, I was left with the likelyhood that it is digital jitter that he misunderstands, or so it seemed to me at the time.
That is a less charitable elaboration of exactly what I was thinking of in my earlier post.
But in fact, it has the opposite effect - it makes nonsense, in that it profoundly distorts market mechanisms, which rely on rational choices based on as close as possible to perfect information (i.e. hard spec style data) about products and services.
Both
rational choice and
near perfect information are necessary properties of a classic economic model, not necessarily properties of the market itself. This classic model has raised serious doubt for at least two decades, because it doesn't work very well. The follwing was, for example, true before the credit crunch:
Both winners and losers had perfect information that they were dealing in hot air. The also knew they couldn't stop because it was such a cash cow. For many employes this was rational: you can't bring home a safe 6% p.a. when your peers are making that a month (at least) and all you would have to do was increasing exposure to very abstract risks as anybody else was doing anyway.
More contemporary theories as behavioral finance try to integrate that. But it is not said that a market with perfect? information for everybody would be the best? market.
I have recently witnessed, over on the Stereophile forum (an EPA Superfund candidate site if there ever was one), an admission from an aggrieved 'subjectivist' that insistence on 'science' stuff irks audiophiles because it tends to undermine beliefs in which they've invested a lot of emotional energy. I have always thought of this as the 'you're harshing our buzz' argument, and it was interesting to see it propounded by someone on the 'other' side.
It slightly reminds me of an ongoing ruckus in anthropology over the possession of some ancient human remains found out west that are of great scientific interest. Scientific metrics clearly establish them as not closely related genetically to today's Native Americans -- in fact they appear to be more Caucasian than NA - yet some tribes have traditionally revered the remains as 'their ancestors'. The government came down on the side of the tribes, not the scientists. It's a case where science was made to 'stand down' because its findings and goals are just too orthogonal to an emotionally, historically, and politically charged belief.
Apparently some audiophiles dream of receiving such special exception too, but they haven't 'earned' any such respect for their beliefs.
And if Harley's jitter article turns out to employ highly misleading analogy, it's of a piece with the MP3 article by Mr. Atkinson that was mastheaded with a pixellated version of the Sgt. Pepper's cover art -- as if to say to its audiophile baby boomer readership, 'see, what this looks like is what audio sounds like after you lossy compress it -- it destroys what we LOVE.'
I don't see why one, who had understood the concept, would be forced to use false analogies just to educate a dumb audience. They just need to be simple, not false.
I think you may be missing the point by considering the article from your point view and not that of Robert Harley or hifitommy.
Robert Harley is earning his living by writing articles like this. It is not a hobby but a job. The job is editor/journalist/writer for a magazine with the objective of making a profit by being an attractive vehicle for adverts for luxury goods. The topic is expensive home audiophile hardware but this is almost certain to be a lot less important than the job itself. To do the job well the article needs to be attractive to the target audiophile audience the advertisers want to reach and to be content that the advertisers are happy to have associated with their products. A straightforward article on the audibility is jitter is neither of these things. Companies advertising expensive audiophile CD players do want the content to state that jitter is inaudible in consumer grade CD players for very obvious reasons.
Hifitommy also does not want to read articles stating that jitter is inaudible in consumer CD players because jitter is one of the magical properties associated with the luxury goods that interest him. He wants to see articles that reinforce the importance and difficulties of jitter and generally add to the richness of his hobby.
Robert Harley has written an article that Hifitommy enjoyed reading. It is not an article I want to read and would probably consider it silly and factually wrong if I did like the previous one or two articles of his I have seen. I have no problem with this because I am not involved with either the magazine or consider myself part of the audiophile world. Nor do I think Robert Harley or hifitommy should see things from my viewpoint or the scientific viewpoint or some other viewpoint if they are not interested.
If he knows better and just sells lies in the believe that's what his customers want, when they buy magazines to inform themselves, then so be it. But when his worshippers show up here and wave about his insight, whack'em! That's a natural side effect of believing in a quack. They have a choice.
Now here we have agreement.
Both rational choice and near perfect information are necessary properties of a classic economic model, not necessarily properties of the market itself. This classic model has raised serious doubt for at least two decades, because it doesn't work very well. The follwing was, for example, true before the credit crunch:
Both winners and losers had perfect information that they were dealing in hot air. The also knew they couldn't stop because it was such a cash cow. For many employes this was rational: you can't bring home a safe 6% p.a. when your peers are making that a month (at least) and all you would have to do was increasing exposure to very abstract risks as anybody else was doing anyway.
At risk of getting too OT (perhaps if you want to discuss this further we can via PM), but just to say, there's a profound difference between buying a stereo or a car (i.e. the world of consumer goods - where marketing is rife) and the world of speculation / market manipulation, interest rate arbritrage, dumping toxic debt into pension funds and CDOs etc. The former can function with good information and decent competition (and is relevant to the audio consumer and marketing), the latter is all about operating behind the scenes and off-balance sheet (and has no relation to the audio consumer or marketing). That said, good luck with your per millisecond algorithmic trading (http://www.youtube.com/watch?v=g0U1vMUa2sc&eurl=http%3A%2F%2Fmaxkeiser.com%2Fpage%2F2%2F&feature=player_embedded).
C.
I'm surprised that no one has posted the link yet. May I?
TAS article (http://www.enjoythemusic.com/news/)
Scroll down a little ways until you get to the TAS article.
Edit: Here's a more direct link: TAS article (http://www.enjoythemusic.com/tas/)
Seems more like an abstract than an article; maybe the actual article is in the printed mag?
-bandpass
Moderation: Removed unnecessary quote of the previous post.
I'm surprised that no one has posted the link yet. May I?
That's even worse than what I had expected.
Let me give you the executive summary and spare the reading:
- Canon Image Stabilizer binoculars are awesome! Scientifically same amount of information within image, but much better visible without movement. Brain can focus on details without having to waste effort for its own image stabilization.
- Esoteric G-0Rb, a $16,000 rubidium-based external clock, is awesome! Scientifically same amount of information within signal, but much better audible without jitter. Brain can focus on details without having to waste effort for its own jitter correction.
So the main point of the article is "awesomeness". The visual analogy is far-fetched and doesn't educate non tech people one bit. Besides that according to all known data, exchanging the clock in a consumer DAC with a rubidium based one should not be distinguishable! So a better (while still flawed) comparison would be already perfectly stabilized binoculars vs. even more stabilization. Technically maybe a difference, subjectively indistinguishable.
The just advertises that jitter elimination is awesome, because some unrelated, but cool sounding, image stabilization technology is also awesome.
Seems more like an abstract than an article; maybe the actual article is in the printed mag?
The table of comments indicates the article is a maximum of 2 pages. If 1 page is an advertisement then this is probably the whole article. Also, the TOC says it's an editorial, not a proper article.
rpp3po points out that the article doesn't say anything much and the analogy is flawed. But this is only known by the expert. As I'm a layman myself, the article makes sense to me. If I had not been aware of all the nonsense in audioland I probably would have said "That's a easy to understand explanation of jitter." But I've learned to pretty much ignore anything I read in audio magazines.
Speaking of atomic clocks to synchronize digital audio signals, why does he need a $16K device? I've actually done this with a Casio Atomic Watch. The watch receives the atomic time via radio signal. You wouldn't believe the difference in sound versus not using the Casio atomic watch. The tick-tock of the music is more stable. The radioactive atomic fuzziness is less pronounced, and the music has a beautiful glow around it. Without the watch the sound is also very glassy, but I've found this to be minimized by removing my glasses. The atomic microdynamics of the sound are much better.
It's just a one page, 7 or 8 paragraph article. The four title lines are in very large font.
- Canon Image Stabilizer binoculars are awesome! Scientifically same amount of information within image, but much better visible without movement. Brain can focus on details without having to waste effort for its own image stabilization.
- Esoteric G-0Rb, a $16,000 rubidium-based external clock, is awesome! Scientifically same amount of information within signal, but much better audible without jitter. Brain can focus on details without having to waste effort for its own jitter correction.
So the main point of the article is "awesomeness". The visual analogy is far-fetched and doesn't educate non tech people one bit. Besides that according to all known data, exchanging the clock in a consumer DAC with a rubidium based one should not be distinguishable! So a better (while still flawed) comparison would be already perfectly stabilized binoculars vs. even more stabilization. Technically maybe a difference, subjectively indistinguishable.
The just advertises that jitter elimination is awesome, because some unrelated, but cool sounding, image stabilization technology is also awesome.
I've done digital photography with Canon cameras for about 5 years now, and I'll tell you, if audiophiles were "photographiles", they would HATE IS. Actually some birders rant against it, preferring a very sturdy tripod instead, and with some you can't even get into their thick skull that IS is about
practicality more than about ultimate quality. A non-IS lens will tend to be sharper just because the IS lens requires a few more glass elements. IS, strictly speaking, can only introduce distortions to the image. At best, the distortions won't be greater than the film/sensor resolution, and at worst, you'd lose some sharpness and/or gain some aberrations.
IS is just a practical aid (I love it -- the pros greatly outweigh the cons), but it can also limit the design of the lens. Just like with zoom lenses (which are also a practicality/image quality trade-off due to extra elements) they can restrict what you can realistically do with it, like maximum aperture and possibly focus range. For instance, there are no Canon (or any other brand I know) zoom lenses for 35mm film with max aperture larger than f/2.8, and the only IS lens with aperture larger than f/2.8 is the über-expensive and one of the newest Canons, the 200mm f/2.0 IS (http://www.usa.canon.com/consumer/controller?act=ModelInfoAct&fcategoryid=153&modelid=16357) *drool*.
I've done digital photography with Canon cameras for about 5 years now, and I'll tell you, if audiophiles were "photographiles", they would HATE IS. Actually some birders rant against it, preferring a very sturdy tripod instead, and with some you can't even get into their thick skull that IS is about practicality more than about ultimate quality.
It is a "solution" to a problem that doesn't really exist, IMO (much like fancy wires in audio). If you need detail you mount the optic, and if you need portability you use a low enough power so that the shaking isn't too obvious. Can you really see more detail from a 35mm objective image stabilized at 10 or 15 power than you could with the same objective size at 5 or 7 times not stabilized? I rather doubt it, and for most binocular observation high power doesn't help all that much anyway.
Amature astronomers (who do really want extreme detail) simply mount their optics sturdily and usually get rid of the extra lenses needed to "erect" an image since it doesn't matter all that much if you are looking at Mars, and each optical surface, as you point out above, necessarily degrades detail.
Now a real problem for astronomers is the shaking of images that comes from atmospheric turbulence, as this causes real loss of detail. Modern adaptive optics are really helpful there, especially for the professionals who are imaging and rarely use their own eyes directly. This technology really does work and has given new life to many ground based telescopes that would otherwise have been junked, but which can now do cutting edge research.
Before this gets too OT on the practicalities of carrying movie-grade tripods around; surely the key difference is that image motion caused by hand holding powerful lenses is manifest on non-critical observation, causing obvious defects in normal use of the images.
surely the key difference is that image motion caused by hand holding powerful lenses is manifest on non-critical observation, causing obvious defects in normal use of the images.
I would certainly agree.
surely the key difference is that image motion caused by hand holding powerful lenses is manifest on non-critical observation, causing obvious defects in normal use of the images.
I would certainly agree.
Yeah, I didn't say that there was gonna be more detail, I said that there
wouldn't. It is a matter of not having to lug a tripod around. With photography, it even helps with a tripod and somewhat fast shutter speeds, since it can minimize the quick snappy shake of the mirror.
@andy o
I wasn't arguing with you--I've been there with camera shake, and wobbly binoculars. Just pointing out a gross difference between audio jitter and image wobbles. And I use image stabilization, because a good tripod will not fit in a pocket.
I've heard this most dramatically when I reviewed Esoteric's G-0Rb, a $16,000 rubidium-based external clock. That's right: The G-0Rb is an atomic clock in your equipment rack whose sole purpose is to provide a precise clock for the digital-to-analog conversion process. With the push of a button, I was able to compare the conventional clock in the Esoteric P-03/D-03 combination with the rubidium-generated clock. Engaging the G-0Rb brought the soundstage into sharp focus, revealed the size and character of the hall through better resolution of low-level spatial cues, made instrumental timbres sound more natural and “organic,” and resulted in a wholesale increase in involvement in the musical performance.
Phew it's a good thing that blind testing has been so thoroughly discredited by the author . Otherwise he might have been struggling a bit there.
@andy o
I wasn't arguing with you--I've been there with camera shake, and wobbly binoculars. Just pointing out a gross difference between audio jitter and image wobbles. And I use image stabilization, because a good tripod will not fit in a pocket.
Nah, I was agreeing with you. I thought Ed might have misunderstood my previous comment.
...because jitter is not a real problem...
Is this a common belief here at HA?
ATB,
-Elias
Why do you ask?
Can you demonstrate through a double-blind test that it is?
If so, please share.
Why do you ask?
Can you demonstrate through a double-blind test that it is?
If so, please share.
Sort of... We have an ABX box here (we are big believers in DBT testing, as well as the 'no B.S.' approach to audio technology that is embraced here at HA). The real problem with jitter audibility tests is that it is hard to impose jitter on a digital
music signal. In other words, one can easily see the effects of jitter using an AP generator and an FFT. However, setting up a test to compare Beethoven's 9th w/ and w/o jitter is much more difficult (logistically).
Would you care to see FFT's of a converter w/ and w/o jitter?
ATB,
e
I don't understand why it would be so difficult to perform an ABX test for jitter. Why not just upconvert the material to a higher sampling rate, apply jitter mathematically, and compare the before and after files?
Would you care to see FFT's of a converter w/ and w/o jitter?
If you mean "see" in a literal sense, then my answer is no.
8. All members that put forth a statement concerning subjective sound quality, must -- to the best of their ability -- provide objective support for their claims. Acceptable means of support are double blind listening tests (ABX or ABC/HR) demonstrating that the member can discern a difference perceptually, together with a test sample to allow others to reproduce their findings. Graphs, non-blind listening tests, waveform difference comparisons, and so on, are not acceptable means of providing support.
Furthermore, tests for the audibility of jitter should be consistent with differences found in real-world hardware.
I doubt that you'll find anyone here who believes that there is no amount of jitter that can break the threshold of audibility in any given system.
I don't understand why it would be so difficult to perform an ABX test for jitter. Why not just upconvert the material to a higher sampling rate, apply jitter mathematically, and compare the before and after files?
Heh. And if Hawksford's simulator is insufficient,
do tell.
(Personally I've had my doubt as to how accurately a bandlimited interpolation is capable of simulating a digital upsample + sigma delta modulator + zero order hold + antialias, but I can't really back that up.)
Would you care to see FFT's of a converter w/ and w/o jitter?
ATB,
e
Only if they're accompanied by controlled listening tests data to correlate them to audibility.
'Cos as you know, observable does not necessarily mean audible.
There's already too much audiophile hysteria and hand-waving about jitter; uncorrelated difference graphics wouldn't help the situation.
Furthermore, tests for the audibility of jitter should be consistent with differences found in real-world hardware.
Well, this is sort of a moving target, is it not? Real-world hardware can be affected by several external factors... how can one quantify what is 'real-world' and what is not?
All the best,
Elias
Fine, then feel free to provide samples of audible jitter even if the effect has been exaggerated. They will no doubt be more useful than the garbage article referenced in this discussion.
Fine, then feel free to provide samples of audible jitter even if the effect has been exaggerated. They will no doubt be more useful than the garbage article referenced in this discussion.
Ok, I'll try to pull something together. As I mentioned, its hard to simulate jitter in a music file, and I'm sure no one wants to do a DBT with sine-waves. But I'll see what I can do.
Would you all except the following premise: if the digital cable is the only variable, any differences can be attributed to jitter?
ATB,
-e
I don't understand why it would be so difficult to perform an ABX test for jitter. Why not just upconvert the material to a higher sampling rate, apply jitter mathematically, and compare the before and after files?
You would need
very high sample rates to apply small amounts of jitter this way and the result would not necessarily sound comparable to what a specific DAC would output from a dirty clock. Good asynchronous DACs are immune to vast amounts of jitter, anyway.
Good DACs are immune to jitter anyway.
True, but by this definition, there are very few 'good' DAC's.
An FFT may not be able to determine audibility, but it can determine susceptibility. If we put a DAC on an AP machine and increase jitter, the FFT will show whether the DAC is immune to the change. Most are not.
ATB,
-E
Most are not.
I own one whose circuit layout once filled your day, it sure is...
But I'm wondering why not more designs employ an ASRC by default. You don't need $1000 DACs just for jitter immunity. An AD1895, for example, is just $5.36 per piece.
I own one whose circuit layout once filled your day, it sure is...
?? What is it?
But I'm wondering why not more designs employ an ASRC by default. You don't need $1000 DACs just for jitter immunity. An AD1895, for example, is just $5.36 per piece.
Well, you don't need $1000 to buy that chip, but you need $1000 to pay an engineer to design the infrastructure to execute it properly! Not to mention all the other custom circuitry, components and board design. EDIT: (and quality control, 5-yr warranty, etc)
Just because we share a disinterest in hokus pokus electronics doesn't mean we're talking about cook-book electronics either!
All the best,
Elias
Furthermore, tests for the audibility of jitter should be consistent with differences found in real-world hardware.
Well, this is sort of a moving target, is it not? Real-world hardware can be affected by several external factors... how can one quantify what is 'real-world' and what is not?
Measurements of the amount of jitter on a production audio source would be easy enough to do, no? I think that is clearly a reasonable definition of "real-world".
Is this not a case where If X amount is audible there is no point in testing X+1? Or do different
types of jitter exhibit different audio effects.
I don't understand why it would be so difficult to perform an ABX test for jitter. Why not just upconvert the material to a higher sampling rate, apply jitter mathematically, and compare the before and after files?
You would need very high sample rates to apply small amounts of jitter this way and the result would not necessarily sound comparable to what a specific DAC would output from a dirty clock.
Couldn't you just upsample a 1000x, apply the jitter, and then downsample? I don't see why thats so difficult, though it may use a couple seconds of your CPU time to compute.
Of course this is only as good as your model of real jitter, but I doubt a good model of clock noise is so difficult to come up with. You pick any ok one and then just increase the std deviation until you hear something . . .
Is this not a case where If X amount is audible there is no point in testing X+1? Or do different types of jitter exhibit different audio effects.
Different types of jitter DO sound different. More importantly, different converters are susceptible to jitter at different frequencies. Specifically, most are NOT immune at low frequencies (<1 kHz). Well-designed PLL's can settle high-frequency jitter, but doesn't do anything to low-frequency jitter. And, arguably, low-freq is the most detrimental jitter because it modulates the fundamentals to frequencies near-by (+/- the jitter frequency), which can cause things to sound out-of-tune (like one of a trio of piano strings is out of tune).
ATB,
e
The real problem with jitter audibility tests is that it is hard to impose jitter on a digital music signal. In other words, one can easily see the effects of jitter using an AP generator and an FFT. However, setting up a test to compare Beethoven's 9th w/ and w/o jitter is much more difficult (logistically).
Actually, its pretty easy to add jitter to a SP/DIF signal. I've done it with maybe $10 worth of parts on a protoboard plus an analog signal source for the signal that is to be the jitter.
Take a SP/DIF signal and use a high speed comparator or schmidt trigger chip to turn it into a square wave (Real world sp/dif signals are often band-limited and look a lot more like a sine wave than a square wave). Feed that through a low pass filter to create a wave with a significant rise time. Then mix in variable amounts of the analog jitter signal. Finally, use another comparator or schmidt trigger input gate to create the signal with jitter. The analog signal will slide the trigger point of the digital signal back and forth along its rise time slope, and this will change the timiing of the SP/DIF signal. The first comparator to square the input signal is optional, particularly if the SP/DIF signal coming in is very robust.
I made this work one morning and proved that it was working several ways. One of the DACs I had on hand (mid-90s era Denon DA-500) had no jitter resistance at all, and I could add clearly audible vibrato to any audio signal. I could also measure the creation of the sidebands that are chraracteristic of FM distoriton (jitter). The other DAC, a early Y2K Technics SHAC 500 surround processor (very similar to the circutis now found in just about every surround receiver) had utterly maximal jitter resistance.
With no added jitter applied, all spurious responses from the SHAC-500 were 110 dB down or more, and they stayed that way until I jittered the signal so much that the both DACs would lose lock. In the extreme case, the Deneon DA 500 would make music sound like it was being sung between buzzing lips, and the Technics SHAC 500 would sound exactly the same as if nothing unusual was happening at all.
I don't understand why it would be so difficult to perform an ABX test for jitter. Why not just upconvert the material to a higher sampling rate, apply jitter mathematically, and compare the before and after files?
You would need very high sample rates to apply small amounts of jitter this way and the result would not necessarily sound comparable to what a specific DAC would output from a dirty clock.
Couldn't you just upsample a 1000x, apply the jitter, and then downsample? I don't see why thats so difficult, though it may use a couple seconds of your CPU time to compute.
This is in essence how I prepared the variable jitter samples for the now-defunct PCABX web site. I used CEP 2.1 to do the upsampling and downsampling (in steps because extreme ratios can break CEPs resampling algorithm), and also used the Flanger (if memory serves) to actually apply the desired FM modulation. It was one of the time-sensitive EFX.
Very interesting comments.
The problem I generally see with externally generated jitter, either through up-/downsampling or electronically, is the applicability of those results. You can take it so far, that you can hear at least something, i. e. what jitter of model m at gain g does or does not sound like. After that you will have several pairs m, g that seem to be relevant. But those then have to be translated and tested against real world implementations.
Wouldn't it make more sense to route test signals into a common DAC known to be sensible against jitter, one from a low end onboard S/PDIF source and another of very high quality, make more sense? That's about the worst it can get in real life. If that already wasn't audible, testing could end. If it was, other common DACs could be fed with the same signal.
Very interesting comments.
The problem I generally see with externally generated jitter, either through up-/downsampling or electronically, is the applicability of those results. You can take it so far, that you can hear at least something, i. e. what jitter of model m at gain g does or does not sound like. After that you will have several pairs m, g that seem to be relevant. But those then have to be translated and tested against real world implementations.
There is quite a bit of extant testing of real-world equipment that shows what kind of signals produced the jittering. IOW, the spectra of common jitter signals are either known or knowable. It makes sense to test with jitter signals that resemble real-world jitter signals.
Wouldn't it make more sense to route test signals into a common DAC known to be sensible against jitter, one from a low end onboard S/PDIF source and another of very high quality, make more sense? That's about the worst it can get in real life. If that already wasn't audible, testing could end. If it was, other common DACs could be fed with the same signal.
The most sensible thing might be to take advantage of the tests and sensitivity data that is already are the scientific/professional literature. The bottom line is that there's no excuse for a modern audio component to have audible jitter.
Would it be too much to ask for a thread split? The original thread pales in discussion to the substantially cooler discussion that is taking place right now.
Actually, its pretty easy to add jitter to a SP/DIF signal. I've done it with maybe $10 worth of parts on a protoboard plus an analog signal source for the signal that is to be the jitter.
Take a SP/DIF signal and use a high speed comparator or schmidt trigger chip to turn it into a square wave (Real world sp/dif signals are often band-limited and look a lot more like a sine wave than a square wave). Feed that through a low pass filter to create a wave with a significant rise time. Then mix in variable amounts of the analog jitter signal. Finally, use another comparator or schmidt trigger input gate to create the signal with jitter. The analog signal will slide the trigger point of the digital signal back and forth along its rise time slope, and this will change the timiing of the SP/DIF signal. The first comparator to square the input signal is optional, particularly if the SP/DIF signal coming in is very robust.
Strictly speaking, that is a simulation of only a very specific kind of jitter (data dependent/cable dependent jitter), and does not really answer the general problem.
An FFT may not be able to determine audibility, but it can determine susceptibility. If we put a DAC on an AP machine and increase jitter, the FFT will show whether the DAC is immune to the change. Most are not.
And, arguably, low-freq is the most detrimental jitter because it modulates the fundamentals to frequencies near-by (+/- the jitter frequency), which can cause things to sound out-of-tune (like one of a trio of piano strings is out of tune).
To some (admittedly slight) degree, aren't these statements inconsistent? A FFT is enough to establish masking thresholds for a steady state signal, and a STFT can probably get you most of the way there with real music, right?
If we really are going to break this down to instantaneous frequencies, I think the effect is going to be extremely difficult to reconcile with masking issues. 30ns amplitude jitter with a 500hz jitter signal yields a sideband amplitude of.... -210db? I mean, I'm aware Benjamin/Gannon actually did find thresholds in this vicinity, but the sheer magnitudes involved make me really suspicious about claims at any lower levels.
Would it be too much to ask for a thread split? The original thread pales in discussion to the substantially cooler discussion that is taking place right now.
All the more reason to keep it right here, IMO. It's one thing to condemn an article as being nothing more than irrelevant fluff and question the motivations behind its creation; another still to give people further insight into why it's irrelevant fluff. I'm happy to see that this thread can actually be useful.
EDIT: Now if another mod or admin deems the split appropriate, that's cool with me too. I can definitely see splitting as useful if it improves the chances of getting good information from a title-based search.
Actually, its pretty easy to add jitter to a SP/DIF signal. I've done it with maybe $10 worth of parts on a protoboard plus an analog signal source for the signal that is to be the jitter.
Take a SP/DIF signal and use a high speed comparator or schmidt trigger chip to turn it into a square wave (Real world sp/dif signals are often band-limited and look a lot more like a sine wave than a square wave). Feed that through a low pass filter to create a wave with a significant rise time. Then mix in variable amounts of the analog jitter signal. Finally, use another comparator or schmidt trigger input gate to create the signal with jitter. The analog signal will slide the trigger point of the digital signal back and forth along its rise time slope, and this will change the timiing of the SP/DIF signal. The first comparator to square the input signal is optional, particularly if the SP/DIF signal coming in is very robust.
Strictly speaking, that is a simulation of only a very specific kind of jitter (data dependent/cable dependent jitter), and does not really answer the general problem.
I really don't think so. With the procedures I have described, you have a free choice of the signal to use to add jitter.
Data-dependent jitter would involve using the data being transmitted to add jitter. I call that "self jitter". While that is possible with the technique I used, it isn't what I actually did.
In the experiments I did either hardware or software, I used an independent tone generator to add jitter to a variety of seperately generated tones and musical signals. I chose this means because it more closely represents the most common forms of jitter that I have seen in actual equipment.
Note that if you try to repeat my hardware approaches to adding jitter in controlled kinds and amounts, it might not seem to be working at all with most modern DACs. They will buffer the signal and make the jitter go away!
Different types of jitter DO sound different.
Agreed, and with a bullet! Low frequency jitter sounds like vibratro.
Those of us who were forced to listen to vinyl in the days when it ruled had to spend years listening to music with a number of different sources of audible, low frequency jitter.
The literature of audiblity (e.g. Zwicker and Fastl) describes FM distortion as "roughness".
I would describe the jitter due to an off-center punched LP as having cyclic roughness. The FM distortion due to warps is roughness with a faster cycle. FM distortion due to the interaction of offset arms and bass signals is a different kind of roughness.
More importantly, different converters are susceptible to jitter at different frequencies.
Agreed, and that includes jitter in their own clocks.
Specifically, most are NOT immune at low frequencies (<1 kHz).
That would be IME untrue. The most common kind of converter that consumers listen to are in surround decoders. Surround decoders must have relatively large buffers, and they also seem to have clock regeneration circuits that are effective at producing a stable clock regardless of the kind of jitter. That is what my experiments showed. I used jitter frequencies down into the 20 Hz range.
Well-designed PLL's can settle high-frequency jitter, but doesn't do anything to low-frequency jitter.
A PLL can be designed to address low frequency jitter. I know that DAC chip clock jitter circuits generally address only high frequency jitter, but I believe that is because they are more likely to encounter that problem, even in systems where the clock was stabilzed well in previous circuits.
And, arguably, low-freq is the most detrimental jitter because it modulates the fundamentals to frequencies near-by (+/- the jitter frequency), which can cause things to sound out-of-tune (like one of a trio of piano strings is out of tune).
The actual perception of music being out-of-tune involves very low jitter frequencies, generally subsonic frequencies. At audible jitter frequencies, it sounds like roughness or vibrato. At high frequencies, the vibrato effect is less but the sense of roughness remains. The most common jitter frequencies that are observed in real world equipment are IME related to the power line.
One of the ironies of high end audio illogic is that after the high end ragazines went on a tear wailing incessantly about digital jitter for years, some impressionable audiophiles returned to listening to vinyl to avoid having their enjoyment of music harmed by digital jitter. Of course, vinyl generally has orders of magnitudes more jitter than good digital, and most of it is at low frequencies where it is really audible and nasty.
In short, what the high end audio ragazines did starting in the 80s was to train people to be hysterical about parts-per-million poison, and then directed them to start a daily regimen of parts-per-thousand of the same or worse poison.
Note that if you try to repeat my hardware approaches to adding jitter in controlled kinds and amounts, it might not seem to be working at all with most modern DACs. They will buffer the signal and make the jitter go away!
Are you sure about re-clocking (considerable buffering without some form of re-clocking doesn't make sense)? S/PDIF does not transmit a bit rate field, so the receiving clock would have to continuously monitor the incoming rate and adjust to its mid term average or run out of sync. Do you know any DACs with considerable amounts of latency? That would also be a side effect of buffering. All DACs I know are pretty close to realtime, thus cannot be building up long queues internally.
Apart from better PLL implementations I don't think that re-clocking is the standard for modern DACs. Or do you have some reference?
ASRC re-clocks the input to a target sample rate and works very well. But that's a feature generally only available on discrete chips, not integrated into common DACs.
Apart from better PLL implementations I don't think that re-clocking is the standard for modern DACs. Or do you have some reference?
ASRC re-clocks the input to a target sample rate and works very well. But that's a feature generally only available on discrete chips, not integrated into common DACs.
Exactly. Some modern converters (even 'hi-end') don't even employ a local crystal, simply using the clock from the AES receiver.
Atb,
-e
Note that if you try to repeat my hardware approaches to adding jitter in controlled kinds and amounts, it might not seem to be working at all with most modern DACs. They will buffer the signal and make the jitter go away!
Are you sure about re-clocking (considerable buffering without some form of re-clocking doesn't make sense)?
Re-clocking and buffering go hand-and-hand.
If you are going to change the timing of clock pulses, the corresponding data has to held someplace until its clock pulse comes up for output.
S/PDIF does not transmit a bit rate field, so the receiving clock would have to continuously monitor the incoming rate and adjust to its mid term average or run out of sync.
That is in fact what usually happens. Unless you resample the data, the final conversion clock pulses have to have the same average frequency as the origional data. This is usually the case.
In SP/DIF the clock pulses for final conversion to analog are obtained from the input data. If the input data is jittered (distorted in the time domain) by one means or another, the analog data that is output will have jitter unless you purify the clock. The usual approach is to put the data into a buffer and filter the dervied clock pulses in the time domain, usually using a PLL. The purified clock is used to run the DAC.
My old DA500 probably did not do a lot to purify the clock it derived from its SP/DIF input. A surround decoder is DSP-driven. Pure stereo operation is actually a degenerate compatibility mode that is obtained by bypassing a most of the usual decoding process. To decode a perceptually coded signal, there is a necessary implication of buffering and expansion of the data. The buffering of the data for stereo is done by default.
Do you know any DACs with considerable amounts of latency?
I've actually never checked the latency of a surround decoder, but I'm sure that there is lots of it.
That would also be a side effect of buffering.
That I agree with, and for audio production purposes, latency is usually minimized. However right now, I'm not talking about audio production, I'm talking about consumers listening to music and watching video.
All DACs I know are pretty close to realtime, thus cannot be building up long queues internally.
You're talking about DACs that are being used for audio production. You're also thinking about just the hardware DAC, not the process of playing music.
How much latency is there in playing music with say, Foobar? Well the answer is always huge because there are always seconds, minutes, hours, days and even years between producing music for distribution, and people actually listening to it. When you're playing music on a PC, the buffering process started when the music file was created. There's additional buffering in the music player.
Apart from better PLL implementations I don't think that re-clocking is the standard for modern DACs. Or do you have some reference?
I'm speaking about DACs in the sense of people listening to music on a PC or with a dedicated digital player.
ASRC re-clocks the input to a target sample rate and works very well. But that's a feature generally only available on discrete chips, not integrated into common DACs.
If you look back in my recent posts, you'll see a disclaimer - I said that in this day and age, most people who listen to music through DACs do so in the context of something like a surround receiver. That was too narrow - the list of devices needs to be expanded to PCs and dedicated digital players. The buffering that reduces or eliminates jitter is generally not in a single-purpose DAC chip, but in the music player taken as a whole.
The initial problems with CD jitter came about when audio's high end foolishly split the CD player up into a transport and a stand-alone DAC. The first generation CD players like the CDP 101 had no audible problems with jitter - I've measured several of them. CD players by definition buffer and reclock the audio data.
Splitting the CD player up into 2 chassis introduced many problems that were not initially addressed. This is a case where the high end created a problem where none existed, ranted and raved about it for over 20 years, and then ignored the fact that the basic problems caused by putting the DAC and the transport into 2 boxes was cured by surround processors and receivers about 8 years ago.
Even in computer audio production, there is *some* buffering. Many device drivers and/or DAW programs have adjustments for latency. There always has to be some buffering that is involved with latency. However, 1 millisecond of latency inplies storing 44 samples, and that is more than enough to handle a wide range of problems caused by data that is jittered.
If you think about the actual processing of grabbing data off a disk and sending it out to a DAC, there is massive amounts of jitter (stop-and-go flow of data) that is implied by getting blocks of data data and putting it into output buffers. As long as the buffer has data in it and it is converted to analog with a steady clock, there's no reason why the data flow shouldn't be as smooth as is desired, and the jitter will be vanishing.
Apart from better PLL implementations I don't think that re-clocking is the standard for modern DACs. Or do you have some reference?
ASRC re-clocks the input to a target sample rate and works very well. But that's a feature generally only available on discrete chips, not integrated into common DACs.
Exactly. Some modern converters (even 'hi-end') don't even employ a local crystal, simply using the clock from the AES receiver.
There's no reason for a standard DAC to have a crystal. The device sending the digital data either provides a separate but parallel clock signal, or the clock is derived from the input digital data stream. The latter is far and away the most common situation.
If you see a crystal in a DAC, either the DAC resamples asynchronously, or the crystal is actually there for the benefit of the ADC that is in the same box. Most so-called DACs that are used in audio production also have an ADC in the same box.
Asynchronous resampling has to violate the principle of bit-perfect data transmission.
Great post! Thanks for the overview.
Arnold - when you say that "Re-clocking and buffering go hand-and-hand" I just want to clarify that any time the data arrive serially, as in S/PDIF, there must necessarily be buffering of one data point, so that the bits that make up that data point can be reassembled before they are clocked into the DAC. There could also be further delay of up to one sample period to allow the PLL to filter the clock to counteract jitter.
However, when most people talk about "buffering", I don't think this is what comes to mind. To me buffering implies storage of multiple data points between the incoming data and the DAC. This kind of buffereing is totally unnecessary if all you are doing is correcting jitter in data transmission, which will always be a fraction of a bit clock. Correction of gross jitter in the device which is sending the data may require multi-point buffereing, but now you are talking about trying to fix broken hardware. This would also be required if you are sharing bandwidth and have variable latency in transmission.
There's no reason for a standard DAC to have a crystal. The device sending the digital data either provides a separate but parallel clock signal, or the clock is derived from the input digital data stream. The latter is far and away the most common situation.
You agree that clock recovery from the input stream (S/PDIF, AES/EBU) without a separate word clock is the most common situation. You also agree that jitter correction by buffering requires a crystal (you fill a buffer from a possibly jittered stream and pull from it with a clean clock running at the same average rate) and that most DACs don't have a separate crystal.
...it might not seem to be working at all with most modern DACs. They will buffer the signal and make the jitter go away!
So how can "most modern DACs" recover a jittered incoming signal by "buffering" when they don't have a crystal besides employing a PLL?
I agree that jitter is a non issue for common integrated systems such as CD players. But I'm interested how jitter elimination for S/PDIF is supposed to work without re-clocking approaches as ASRC.
You also agree that jitter correction by buffering requires a crystal (you fill a buffer from a possibly jittered stream and pull from it with a clean clock running at the same average rate) and that most DACs don't have a separate crystal.
Jitter correction by buffering most definitely does NOT require a separate clock crystal. The control to the PLL in this case would be the amount of data in the buffer. This would allow correcting jitter of much lower frequency than without buffering.
There's no reason for a standard DAC to have a crystal. The device sending the digital data either provides a separate but parallel clock signal, or the clock is derived from the input digital data stream. The latter is far and away the most common situation.
You agree that clock recovery from the input stream (S/PDIF, AES/EBU) without a separate word clock is the most common situation.
Yes.
You also agree that jitter correction by buffering requires a crystal
No.
(you fill a buffer from a possibly jittered stream and pull from it with a clean clock running at the same average rate) and that most DACs don't have a separate crystal.
Yes. One obtains a clean clock from the dirty incoming clock by means something like a PLL.
...it might not seem to be working at all with most modern DACs. They will buffer the signal and make the jitter go away!
So how can "most modern DACs" recover a jittered incoming signal by "buffering" when they don't have a crystal besides employing a PLL?
A PLL is all that is needed.
I agree that jitter is a non issue for common integrated systems such as CD players. But I'm interested how jitter elimination for S/PDIF is supposed to work without re-clocking approaches as ASRC.
Just clean up the dirty incoming clock. Store the data in a buffer. Build in some latency. Clock the data out with a cleaned up version of the incoming clock. Cleaning up dirty clocks has been done effectively for decades.
Jitter correction by buffering most definitely does NOT require a separate clock crystal. The control to the PLL in this case would be the amount of data in the buffer. This would allow correcting jitter of much lower frequency than without buffering.
Agreed.
How does one use a PLL to correct for a noisy clock? My limited use of PLL has been for clocking chips, never noise reduction. I naively assume that a clock that is feed by an unstable oscillator into a PLL would cause the DAC's clock to very closely follow the unstable clock, but not actually stabilize the frequency. Do you need to add some filtering or does the PLL handle this directly?
All PLLs utilize a loop filter to some degree, to lowpass filter the phase comparator output.
Exactly. Long-term, the filtered clock's frequency will be the average of the noisy clock's, but with much less variance from the mean.
Ah that makes perfect sense. Thanks.
Thank's all for the feedback! I found this very interesting and read up a little on PLL design. It works exactly as you have described.
Several sources claim, though, that this works only for jitter >10kHz while the most audible jitter would be <10kHz. Wolfson, for example, provides PLL circuits with separate crystals (http://www.wolfsonmicro.com/news/258) to eliminate jitter down to 100Hz. Are there really crystal-less implementations with comparable performance?
Thank's all for the feedback! I found this very interesting and read up a little on PLL design. It works exactly as you have described.
Several sources claim, though, that this works only for jitter >10kHz while the most audible jitter would be <10kHz.
Source? Link?
My interpretation of such statements is that they are talking about stabilizing jitter without a large buffer to hold data while variations at lower frequencies are being handled.
Wolfson, for example, provides PLL circuits with separate crystals (http://www.wolfsonmicro.com/news/258) to eliminate jitter down to 100Hz. Are there really crystal-less implementations with comparable performance?
If you are referring to the following:
"A further benefit of the high performance PLL is the reduction in component count for the overall system. The PLL can be used to synthesise crystal derived clock signals and, without the requirement for any external filter, can operate as a high quality master timing source for the audio system. "
This is a different application than recovering a stable clock from a SP/DIF stream. The application here involves generating stable frequencies from a crystal that are at a wide range of frequencies other than the resonant frequency of the crystal. PLL's are widely used for frequency synthesizers.
If you are referring to the following:
"A further benefit of the high performance PLL is the reduction in component count for the overall system. The PLL can be used to synthesise crystal derived clock signals and, without the requirement for any external filter, can operate as a high quality master timing source for the audio system. "
This is a different application than recovering a stable clock from a SP/DIF stream. The application here involves generating stable frequencies from a crystal that are at a wide range of frequencies other than the resonant frequency of the crystal. PLL's are widely used for frequency synthesizers.
No, I'm not referring to that. It clearly states the intended application on the same page, which is not mainly frequency synthesis, but suppression of jitter on S/PDIF links with better performance than "traditionally" known from transceivers:
The transceivers are structured around an integrated high performance PLL with an intrinsic period jitter of 50ps and jitter rejection frequency of 100Hz. Unlike many other transceivers of this type, the WM8804 and WM8805 contribute a negligible level of jitter to the audio system. Traditionally, S/PDIF transceivers can only suppress jitter above 10kHz and have no effect on the low and medium frequencies, which have the greatest impact on audio quality. Crucially, the WM8804 and WM8805 are capable of suppressing these pre-existing timing issues on the signal at frequencies above 100Hz, removing any unwanted audio distortion or problems in this critical audio range.
As an interconnection standard, users and designers have no control over the quality of the S/PDIF inputs into their systems from third parties. The Wolfson PLL technology enables the WM8804 and WM8805 receivers to lock onto and recover the data and timing from poor quality input signals, thereby allowing the transceivers to accept S/PDIF signals from any source, even if the input signal is severely degraded.
My interpretation of such statements is that they are talking about stabilizing jitter without a large buffer to hold data while variations at lower frequencies are being handled.
Wolfson doesn't make any claims about improved realtime (very small buffer) behavior, but improved jitter rejection quality. So I can't comprehend that interpretation.
The data sheet makes it clear that the chip can be both a transmitter or receiver. As a transmitter, it needs an internal clock, hence the crystal. As a receiver, it synchs to the external data stream and needs no clock of its own.
http://www.wolfsonmicro.com/uploads/docume...8804_Rev4.0.pdf (http://www.wolfsonmicro.com/uploads/documents/en/WM8804_Rev4.0.pdf)
Nope, first page, second paragraph:
A crystal derived, or externally provided high quality master clock is used to allow low jitter recovery of S/PDIF supplied master clocks.
The WM8804's separate clock is used to improve input signal recovery! Besides, a S/PDIF receiver usually needs some form of transmitter, else it would be a dead-end without anything to output into.
In contrast a Cirrus Logic CS8427 is a typical clock-less S/PDIF transceiver. The data sheet (http://www.cirrus.com/en/pubs/proDatasheet/CS8427_F3.pdf) contains diagrams on page 58, which show, that its PLL is only able to eliminate jitter above 10kHz without a separate clean clock.
Nope, first page, second paragraph:
A crystal derived, or externally provided high quality master clock is used to allow low jitter recovery of S/PDIF supplied master clocks.
Crystal oscillators are free-running, and operate at a frequency that is set by the crystal. Thei phase of a crystal oscillator is basically random. It is free-running. It does what comes naturally.
A SP/DIF receiver clock must have the *identical* (e.g. +/- nothing at all) frequency as the input signal. Its phase must also match that of the input signal.
How does a crystal oscillator come to have the identical same frequency and phase as the input signal?
A frequency-synthesized variable frequency oscillator that is based on a crystal generally will produce only frequencies that are related to the crystal's frequency by quotients of integers. Therefore, not *all* frequencies can be generated this way. If the integers are large, then a very large number of dfferent frequencies can be generated, but still not every possible frequency can be generated.
I'm not sure the metaphor is problematic. Metaphors are, after all, literary, not scientific. What is faulty is the assumption that lead to the metaphor, and upon which the metaphor stands: that jitter is as audible at anything close to significant levels. It's questionable whether it is audible at all once it gets past a modern, well-implemented DAC. Audible enough to be heard above the noise floor of an analog lover's reference system? Highly unlikely. But we will never see blind listening tests from that crowd. Only subjective claims.
Tim
30ns amplitude jitter with a 500hz jitter signal yields a sideband amplitude of.... -210db?
The sideband amplitude is a function of the frequency of the audio signal being subjected to jitter. Sideband amplitude is not a function of jitter frequency (as suggested above). The math is also incorrect.
If a 1 kHz tone is subjected to 30ns RMS sinusoidal jitter having a frequency of 500 Hz, the jitter-induced sidebands will have an amplitude of -77.5 dB relative to the amplitude of the 1 kHz tone. This means we will have a sideband at 500 Hz (1kHz - 500 Hz) and at 1500 Hz (1 kHz +500 Hz). Each sideband can easily reach levels that are above the threshold of hearing if the 1 kHz tone is played at reasonably loud levels. The THD+N due to the sidebands is -74.5 dB (relative to the amplitude of the 1 kHz tone). Audibility will be a function of masking curves. Also please note that the distortion caused by jitter is not harmonically related to the audio an should be more audible than harmonic distortion.
If a 10 kHz tone is subjected to 30 ns RMS sinusoidal jitter having a frequency of 500 Hz, the jitter-induced sidebands will have an amplitude of -57.5 dB relative to the amplitude of the 10 kHz tone. The sidebands will occur at 9500 Hz and 10500 Hz.
The distortion due to jitter is given by:
20*log(2*PI()*AudioFrequency*JitterMagnitude)
where AudioFrequency is expressed in Hz
and JitterMagnitude is expressed in Sec
Sideband amplitudes will be 3 dB lower than the combined distortion number calculated with the above formula.
I'm a little bit confused by your math so I tried a simpler case.
Assume that the jitter is a 500 Hz sawtooth with peak-to-peak amplitude of 30 nSec.
Assume that you have a 1 kHz sine wave as your signal and that the jitter is synchronized in such a way that one cycle of the sine wave corresponds with the rising part of the sawtooth, and the next cycle corresponds to the falling part.
Now we have a slope in sample timing offset of +- 30 nSec / 1 mSec (amplitude divided by rise and fal times). This will have the effect of increasing or decreasing the frequency of each sine wave by 0.003%. We now have alternating 999.97 Hz and 1000.03 Hz sine waves.
So where do the signals at 500Hz and 1500Hz come from, and how would it be humanly possible to hear so small a frequency variation in the sine wave?
Sorry if I am way off base here, I am just trying to understand.
a Cirrus Logic CS8427 is a typical clock-less S/PDIF transceiver. The data sheet (http://www.cirrus.com/en/pubs/proDatasheet/CS8427_F3.pdf) contains diagrams on page 58, which show, that its PLL is only able to eliminate jitter above 10kHz without a separate clean clock.
Most AES/EBU or S/PDIF receivers have little or no jitter attenuation below 5 or 10 kHz. The reason for this is that they must meet the AES jitter tolerance test that requires perfect data recovery in the presence of high-amplitude low-frequency jitter. This data-recovery task demands a PLL with a corner frequency of at least 5 kHz. Below 5 kHz the PLL tracks the jitter of the incoming signal. Above 5 kHz the PLL begins to reject the jitter on the incoming signal.
The AES jitter tolerance specifications were based upon measurements of jitter levels on commercially available equipment. Devices that meet the AES jitter tolerance specifications should be able to reliably recover data from nearly all source devices. However, this does not imply that the clock that is recovered by the PLL is suitable for D/A or A/D conversion.
Direct use of the MCLK recovered from a S/PDIF (or AES) receiver will yield jitter-induced distortion products that are only 50 to 70 dB below the amplitude of the audio signal.
A second PLL is required to reduce the jitter on the clock that is recovered by the receiver chip. This second PLL must control the frequency of a low-jitter oscillator. It is common to use a VCXO (voltage controlled crystal oscillator) as the oscillator in this second PLL. Other low-jitter oscillator designs are possible, but the VCXO makes the task somewhat easier. The corner frequency of the second PLL should be much lower than the 5 kHz corner frequency of the first PLL. 1 Hz to 10 Hz corner frequencies are often used in high-quality professional audio equipment. Remember there is no jitter attenuation below the corner frequency of the PLL.
30ns amplitude jitter with a 500hz jitter signal yields a sideband amplitude of.... -210db?
The sideband amplitude is a function of the frequency of the audio signal being subjected to jitter. Sideband amplitude is not a function of jitter frequency (as suggested above). The math is also incorrect.
You are correct; I was using FM math for the sidebands instead of PM. Thank you for the correction. (Methinks I need to read Dunn more often.)
Your math, unlike mine, also makes sense in light of existing listening tests.
So where do the signals at 500Hz and 1500Hz come from, and how would it be humanly possible to hear so small a frequency variation in the sine wave? Sorry if I am way off base here, I am just trying to understand.
You're confusing instantaneous frequency with physical frequency. Like I said, the best source for the basics of the math is probably:
http://www.nanophon.com/audio/jitter92.pdf (http://www.nanophon.com/audio/jitter92.pdf)
So where do the signals at 500Hz and 1500Hz come from, and how would it be humanly possible to hear so small a frequency variation in the sine wave?
Jitter phase-modulates the audio signal and produces upper and lower sidebands. The jitter frequency determines how far the sidebands will be spaced away from the audio signal. Low frequency jitter will produce closely spaced sidebands that tend to be well masked by the original audio signal. High-frequency jitter produces sidebands that are widely spaced above and below the original signal and consequently are not as well masked.
Side band frequency is given by CarierFrequency +/- JitterFrequency when both are sinusoidal
Your example using the sawtooth is actually much more complicated mathematically. The sawtooth jitter will generate many side bands not just the ones you identified.
The sideband amplitude is a function of the frequency of the audio signal being subjected to jitter. Sideband amplitude is not a function of jitter frequency (as suggested above).
For small amounts of modulation, sine wave FM or PM modulation of a sine wave will produce a pair of sidebands that are displaced from the carrier by the modulating frequency. Their amplitude is proportional to the strength of the modulation.
For large amounts of modulation, the sideband structure becomes complex with many sidebands. Their frequencies and the proportioning of energy among them is predicted by the Bessel functions.
Phase modulation and frequency modulation can be accomplished and analyzed by identical means. They are linked by differentiation or in the reverse, by integration of the modulating signal. The modulating signal is differentiated to produce phase modulation by means of a frequency modulator, and vice-versa.
If a 1 kHz tone is subjected to 30ns RMS sinusoidal jitter having a frequency of 500 Hz, the jitter-induced sidebands will have an amplitude of -77.5 dB relative to the amplitude of the 1 kHz tone. This means we will have a sideband at 500 Hz (1kHz - 500 Hz) and at 1500 Hz (1 kHz +500 Hz). Each sideband can easily reach levels that are above the threshold of hearing if the 1 kHz tone is played at reasonably loud levels. The THD+N due to the sidebands is -74.5 dB (relative to the amplitude of the 1 kHz tone). Audibility will be a function of masking curves. Also please note that the distortion caused by jitter is not harmonically related to the audio an should be more audible than harmonic distortion.
If a 10 kHz tone is subjected to 30 ns RMS sinusoidal jitter having a frequency of 500 Hz, the jitter-induced sidebands will have an amplitude of -57.5 dB relative to the amplitude of the 10 kHz tone. The sidebands will occur at 9500 Hz and 10500 Hz.
The sidebands produced by a given amount of jitter are 20 dB larger for the 10x higher frequency carrier, because 30 ns is a 10x (+20dB) larger fraction of the period of the 10 KHz frequency carrier.
The distortion due to jitter is given by:
20*log(2*PI()*AudioFrequency*JitterMagnitude)
where AudioFrequency is expressed in Hz
and JitterMagnitude is expressed in Sec
Sideband amplitudes will be 3 dB lower than the combined distortion number calculated with the above formula.
When you're looking at the magnitude of the spectrum produced by various distortion sources, you really can't tell from a magnitude-only plot whether you're seeing AM or FM distortion. Often real-world distortion is a mixture. Both AM and FM produce sidebands that differ from the carrier by the modulating freqency. They both produce sidebands with equal magnitudes. However, the phase of the sidebands differs, which will cause you to observe unequal sidebands for a mixture of AM and FM of the same carrier by the same modulation.
When you're looking at the magnitude of the spectrum produced by various distortion sources, you really can't tell from a magnitude-only plot whether you're seeing AM or FM distortion. Often real-world distortion is a mixture.
Very true! Our own measurements of the D/A converters built into CD players show that all of the units tested suffered from both phase modulation (FM) and amplitude modulation (AM). We traced these problems to ripple on the DC power supply rails, and to ripple in the ground system. Some of the ripple was AC line related, but much of the ripple was caused by the servos that drive the read head and control the rotational speed of the disk. This ripple was causing phase modulation of the oscillator (jitter), and amplitude modulation of the D/A reference voltage. These problems were common to all of the CD and DVD players that we looked at in our lab. The players ranged in price from $50 to $1200. Our sampling of 4 or 5 players was very small, but it did indicate that AM and FM modulation problems may be very common in CD and DVD players. The worst player produced sideband amplitudes that were only 50 to 60 dB below our recorded test tones.
There's no reason for a standard DAC to have a crystal. The device sending the digital data either provides a separate but parallel clock signal, or the clock is derived from the input digital data stream. The latter is far and away the most common situation.
The reason for adding a crystal is to achieve low-jitter performance at frequencies that exceed the cut-off frequency of the clock-recovery PLL.
The crystal is usually a VCXO (voltage controlled crystal oscillator) that is being controlled by a PLL. The PLL includes a phase comparator and a low pass filter. Above the cut off frequency of the low-pass filter, the jitter performance is determined by the stability of the oscillator. Below the cut-off frequency of the low-pass filter, the jitter performance is determined by the quality of the clock embedded in the digital input signal (AES or S/PDIF). The crystal oscillator makes it much easier to achieve low-jitter performance above the PLL cut-off frequency. The stability of the oscillator is especially important when the PLL cut-off frequency is very low (less than 100 Hz). Low cut-off frequencies are required to eliminate low-frequency jitter, so this makes the use of a VCXO a great solution. But the VCXO solution is not cheap. To cut costs, the VCXO is often omitted, and the clock recovered by the digital audio receiver is simply wired directly to the D/A converter. This is not good practice, but it is very common (and very inexpensive).
If you see a crystal in a DAC, either the DAC resamples asynchronously, or the crystal is actually there for the benefit of the ADC that is in the same box.
VCXOs and crystal oscillators look identical. It is impossible to tell the difference without looking up data sheets on the oscillators.
If the crystal is fixed-frequency, this may be an indication that the DAC resamples asynchronously. Some ASRC (asynchronous sample rate converter) ICs attenuate jitter, but most do not. Most of the ASRC ICs we have tested have very little jitter attenuation below 5 kHz. There are 3 or 4 ASRC ICs that have outstanding jitter attenuation that extends down to a few Hz and these few can outperform two-stage PLL solutions that employ VCXOs.
Asynchronous resampling has to violate the principle of bit-perfect data transmission.
Very true, but the quality of the ASRC is a function of how much DSP horsepower we are willing to expend. The distortion artifacts of the better ASRC devices are below -140dB. These distortion artifacts are as much as 100 dB lower than the jitter-induced sidebands produced by DACs that do not use a VCXO or a fixed-frequency crystal.
When you're looking at the magnitude of the spectrum produced by various distortion sources, you really can't tell from a magnitude-only plot whether you're seeing AM or FM distortion. Often real-world distortion is a mixture.
Very true! Our own measurements of the D/A converters built into CD players show that all of the units tested suffered from both phase modulation (FM) and amplitude modulation (AM). We traced these problems to ripple on the DC power supply rails, and to ripple in the ground system. Some of the ripple was AC line related, but much of the ripple was caused by the servos that drive the read head and control the rotational speed of the disk. This ripple was causing phase modulation of the oscillator (jitter), and amplitude modulation of the D/A reference voltage. These problems were common to all of the CD and DVD players that we looked at in our lab. The players ranged in price from $50 to $1200. Our sampling of 4 or 5 players was very small, but it did indicate that AM and FM modulation problems may be very common in CD and DVD players. The worst player produced sideband amplitudes that were only 50 to 60 dB below our recorded test tones.
Interesting. Have you seen Ian Dennis and Julian Dunn's white paper on possible causes of nonidentical CDP output from bit-identical CDs, from circa 1995?
www.prismsound.com/m_r_downloads/cdinvest.pdf
They posited servo-based effects as well.
(Found no solid evidence of audible effect though)
Interesting. Have you seen Ian Dennis and Julian Dunn's white paper on possible causes of nonidentical CDP output from bit-identical CDs, from circa 1995?
www.prismsound.com/m_r_downloads/cdinvest.pdf
They posited servo-based effects as well.
(Found no solid evidence of audible effect though)
Yes, I am familiar with their work and I spent some time discussing it with Julian at an AES convention. We had conducted our own testing several years later (in 2003) and saw similar problems with CD players.
We have a QSC ABX tester that we used to compare consumer CD players to prototypes of our DAC1 converter. It was fairly easy to score perfectly on the ABX tests. The CD players with modulation problems sounded like they had more midrange when played through their internal D/A converters than when played through the external DAC. This was a rather surprising result given that both devices had nearly identical frequency responses. The CD players tested and the DAC1 prototype both had very flat frequency response between 20 Hz and 20 kHz. The frequency response of the DAC1 extended down to 0.1 Hz, but our playback system was limited to -3dB at 30 Hz.
Subsequent FFT analysis confirmed that the modulation-induced sidebands tended to fill the midrange of the audio spectrum. Some of the added content in the midrange may have been due to high levels of IMD produced by the internal D/A converters and the output stages in the CD players. Based upon our listening tests, we suspect that jitter is often perceived as a difference in frequency response. Specifically, we suspect that jitter-induced sidebands fill in the midrange of the audio spectrum. Much more investigation would be needed to determine audibility thresholds. Nevertheless, we felt we had enough evidence to warrant developing a jitter-attenuation system.
The external prototype DAC was equipped with an Analog Devices AD1896 ASRC which rejects jitter above 1 Hz. At 1 kHz the jitter attenuation of the AD1896 exceeds 100 dB. Jitter tolerance tests and FFT analysis confirmed that the prototype DAC was essentially free from jitter-induced (FM) sidebands as well as AM sidebands.
I am a strong advocate of ABX testing, but I take a very conservative approach to product design. I am not in the business of building perceptual encoders where the goal is just to reduce artifacts to a certain level of audibility (or just below audibility without wasting bandwidth by going too far). Instead I have the luxury of building products that can reduce artifacts to a level that is low enough that I have no doubt that these artifacts are inaudible. This is analogous to a "safety factor" built into a highway bridge. A bridge that is designed to carry 100 Tons may be built with a 5:1 safety factor and may be able to carry about 500 Tons before failing. This safety factor insures that the bridge will not fail under normal use. Safety factors can be built into audio circuits at fairly low cost to insure that certain audio artifacts never reach audible levels. It is often much easier and cheaper to over design than it is to create a design that is just good enough. It is very hard to determine exactly how much jitter attenuation is needed to prevent audibility. It is much easier to design in a generous safety factor.
Manufacturers of inexpensive consumer-grade audio equipment have a different goal: How cheap can we make the product before most people will think it sounds bad? No safety factor is needed because the average consumer will tolerate some audible artifacts.
I highly recommend Julian Dunn's papers on jitter. They are required reading on this subject!
Much more investigation would be needed to determine audibility thresholds. Nevertheless, we felt we had enough evidence to warrant developing a jitter-attenuation system.
But this is exactly evidence that is in pressing need of being made public, IMO. It would not be too surprising if artifacts in the midrange have a lower threshold , as it's where H. sapiens is most sensitive. But I'd like to see such evidence.
I'm not meaning to berate you for focusing on the engineering/design end, but given the eternal subjectivist/objectivist debates, it frustrates me when people in the industry report private DBT results that show difference within 'hot button' categories -- e.g. digital formats, CDPs, cables, amps -- much less ones where 'it was fairly easy to score perfectly' --- but don't actually publish them or provide detail (I'm thinking too of Robert Stuart's brief anecdotal references to such blinds tests in his papers advocating hi-rez formats). Your company makes fine DACs; one would think you'd be eager to publish solid proof that they differ audibly from mass-market stuff...that that overdesign actually pays off in more than just excellent specs.
The extant literature on audibility of jitter is sparse and the results appear highly method-contingent. The extant literature on CDP listening comparison is as tiny or tinier. Both really could use some more systematic scientific investigation of audibility thresholds of measured CDP difference.
But if they published ABX tests, they would be marked in the HiFi community as people who doesn't hold some sort of "true values". I've never seen ABX test results from HiFi companies that make DACs, amplifiers, CD/DVD players, and so on. Maybe I am wrong about this, but I have the feeling that publishing ABX results would actually undermine their efforts in making (and eventually selling) good DAC.
But if they published ABX tests, they would be marked in the HiFi community as people who doesn't hold some sort of "true values". I've never seen ABX test results from HiFi companies that make DACs, amplifiers, CD/DVD players, and so on.
Other than some really bad cheap stuff, and some relatively rare pathological cases, there's not a lot of differences to hear.
Maybe I am wrong about this, but I have the feeling that publishing ABX results would actually undermine their efforts in making (and eventually selling) good DAC.
The days of the DAC as a separate mainstream consumer audio component have been gone for some time. DACs are now routinely encapsulated into power amplfiiers and receivers, which is where they IMO belong.
But if they published ABX tests, they would be marked in the HiFi community as people who doesn't hold some sort of "true values". I've never seen ABX test results from HiFi companies that make DACs, amplifiers, CD/DVD players, and so on. Maybe I am wrong about this, but I have the feeling that publishing ABX results would actually undermine their efforts in making (and eventually selling) good DAC.
Tag McLaren (amps) actually published one on the web some years ago, with the results supporting no audible difference between their gear and much more expensive gear.
I think they changed ownership after that, and the site has vanished.
I think there is a valid market for those in the luxury market who perceive the threshold of audibility as what is the "bare minimum". Enjoying the highest quality engineering available, regardless of minimum specifications of equivalence, is not wrong. I do not own a Benchmark product, but I have heard a DAC-1 once or twice, and am aware of the high reputation Benchmark holds, so I have no reason to doubt they belong in such a tier. I wouldn't mind having such a product, just to avoid even the barest inkling of misapprehension about my DAC - but probably also no small amount of conspicuous consumption.
However, it is a completely different thing to justify such overengineering on the basis of audibility below known thresholds, and there are of course shades of gray there with respect to if thresholds can be exceeded for test signals vs musical content etc. Like, there's nothing wrong about SACD in and of itself, but if you can't prove its superiority with a blind test (or nowdays perhaps several), there is something distinctly wrong about maintaining that claim in light of testing and considerable psychoacoustic justification.
So I find it something of a shame that many high-end audio firms cannot sell their products on their engineering alone, without needing to make claims about what will and won't be an audible improvement, which naturally introduces a conflict with academia. (I'm not talking about Benchmark specifically here - I know nothing about their literature.) Why can't luxury be for its own sake?
It was fairly easy to score perfectly on the ABX tests. The CD players with modulation problems sounded like they had more midrange when played through their internal D/A converters than when played through the external DAC.
Interesting.
How did you match the playback levels of the DACs ?
It was fairly easy to score perfectly on the ABX tests. The CD players with modulation problems sounded like they had more midrange when played through their internal D/A converters than when played through the external DAC.
Interesting.
How did you match the playback levels of the DACs ?
Time-synching is IME far more difficult than levelmatching.
Yes, for instant switching, time synchronisation can be a big problem.
But if the delay between the two dacs (internal and external) is small enough, the problem can be circumvented by muting the amplifier before any switch. In this case, a DC offset between the mass of the devices may introduce an audible click. So the muting must be done with the volume control. It must also be checked that the volume control completely mutes the audio. Some let pass a small signal. In this case it is always possible to mute the amplifier after having turned down the volume.
A switch is then done this way : Amplifier volume to zero / amplifier speakers off / ABX switch switched / amplifier speakers on / amplifier volume restored.
If two different CD players are compared, the synchronisation is more problematic. In this case, a switch consists in
Amplifier volume to zero / amplifier speakers off / both CD players stopped / ABX switch switched / both CD players launched and paused / both CD players manually unpaused at the same time / amplifier speakers on / amplifier volume restored.
Yes, for instant switching, time synchronisation can be a big problem.
To tell the whole story, if nearly instant switching is not available, then the potential sensitivity of the comparison is seriously, perhaps even debilitatingly compromised.
But if the delay between the two dacs (internal and external) is small enough, the problem can be circumvented by muting the amplifier before any switch.
Of course. The problem of transient-free switching was addressed in detail in Clark's 1978 JAES paper. I actually worked out a general procedure for transient-free switching of both inputs and outputs in 1976-7, and provided it to Clark to include in his paper. The benchmark test was to do a switchover of both inputs and outputs of a integrated amplifier from RIAA input to speaker output. I know of no other switching problem that is not either like that problem or a subset of it.
If two different CD players are compared, the synchronisation is more problematic. In this case, a switch consists in
Amplifier volume to zero / amplifier speakers off / both CD players stopped / ABX switch switched / both CD players launched and paused / both CD players manually unpaused at the same time / amplifier speakers on / amplifier volume restored.
You've missed the point. The problem of synchronization is to have 2 CD players that are playing the identical same tracks at the same time offsets within a few dozen milliseconds or less over a useful amount of music for the purpose of comparison.
To tell the whole story, if nearly instant switching is not available, then the potential sensitivity of the comparison is seriously, perhaps even debilitatingly compromised.
A mute/off/switch/on/unmute-procedure without instant-switching capability either resets or randomizes start positions (relative to transients). The former case puts harder burden on the subject's capabilities, the latter adds random noise, which must be accommodated by increasing the number of rounds. So ABX testing without the hassle of time-synching doesn't seem fundamentally flawed, but only shifts effort from technical to procedural expenditure. When properly conducted the probability of false positives can be equal with both approaches. Or am I overlooking something?
To tell the whole story, if nearly instant switching is not available, then the potential sensitivity of the comparison is seriously, perhaps even debilitatingly compromised.
A mute/off/switch/on/unmute-procedure without instant-switching capability either resets or randomizes start positions (relative to transients). The former case puts harder burden on the subject's capabilities, the latter adds random noise, which must be accommodated by increasing the number of rounds.
In both cases the listener is put at a disadvantage. IOW, if nearly instant, transient-free switching is not available, then the potential sensitivity of the comparison is seriously, perhaps even debilitatingly compromised.
To me there is only one solution - get the switching as near-instant, and transient free as you can.
So ABX testing without the hassle of time-synching doesn't seem fundamentally flawed,
That's a decision you get to make. I see you falling into the golden-ear's hands by feeding their prejudices that we bias our tests against possible positive outcomes for the listener.
I see you falling into the golden-ear's hands by feeding their prejudices that we bias our tests against possible positive outcomes for the listener.
That depends on the perspective. For significant positive results you don't necessarily need time-synched switching. Just accomodate the procedure: either reset positions or randomize (+/- several frames) and increase the number of trials. Both can lead to solid positive, but weaker negative results.
Time-synching increases the probability of true positive results (per round), thus strengthens the significance of negative results somewhat. But if you're just after possible positive results between DACs - what this thread was about recently - demanding time-synching* is not
necessary and can only improve chances to find something, when proper non time-synched methods have failed to reveal a difference.
(* instead of simple mute/off/switch/on/unmute-switching)
I'm wondering - we have had two prominent engineers of a well-known, 'non-esoteric' high end DAC manufacture join this thread. Both have shown considerable insight into the matter and claimed hands-on experience with ABX switching equipment. Still they could not yet present anything verifiable regarding audibility.
How about some test samples or protocols?
But if they published ABX tests, they would be marked in the HiFi community as people who doesn't hold some sort of "true values". I've never seen ABX test results from HiFi companies that make DACs, amplifiers, CD/DVD players, and so on. Maybe I am wrong about this, but I have the feeling that publishing ABX results would actually undermine their efforts in making (and eventually selling) good DAC.
We are not afraid to publish specs and/or ABX test results. Our manuals typically contain about 15 pages of graphs and 2 pages of specifications. This is unique in the HiFi community (and often frowned upon), but HiFi is not where our roots are. We began 26 year ago as a manufacturer of Audio for broadcast TV and radio. In the broadcast market buyers live by specifications and test results, and they conduct extensive tests of their own before buying. We sell the same exact products to broadcast, recording studios, and home HiFi. As you can imagine this is an interesting mix! Your forum is refreshing relief from the pseudo-science that is pervasive in the HiFi industry. We attempt to bring this same sort of relief to the more down-to-earth customers in the HiFi industry. We don't play the pseudo-science games.
As for publishing, time is always the issue. Any time spent on publishing takes away from our product development resources. Our research is conducted for the purpose of developing better products. We are always willing to have others pick up where we left off and conduct well-controlled tests that are suitable for publishing.
The goal of our tests is to confirm that a problem may exist and then confirm that a proposed solution is more than adequate to remove audible artifacts. We usually do not attempt to determine audibility thresholds as this is much more difficult, requires many more trials, and more human subjects.
We are considering the creation of an ABX jitter test that could be posted on this forum. The test would allow comparisons between a jitter-contaminated track and an original track. We will keep this forum posted on our progress.
We are considering the creation of an ABX jitter test that could be posted on this forum. The test would allow comparisons between a jitter-contaminated track and an original track. We will keep this forum posted on our progress.
That seems like a coincidence, since I don't expect you have written those four paragraphs within 3 minutes. I'm looking forward to your proposal.
It was fairly easy to score perfectly on the ABX tests. The CD players with modulation problems sounded like they had more midrange when played through their internal D/A converters than when played through the external DAC.
Interesting.
How did you match the playback levels of the DACs ?
The output of our DAC prototype was equipped with precision 10-turn trimmers that allowed us to match gains to better than 0.05 dB. Left was matched to left on the CDP, right to right. We used an audio precision System 2 to normalize the gains at 1 kHz prior to testing. We also used only one input channel on the AP to insure that all channels were fed to the same analyzer channel on the AP. We also shut off the auto ranging on the AP to minimize the gain errors that can occur when the unit auto ranges. We used a CD with a TPDF dithered 1kHz -20 dB FS test tone for calibration.
We are considering the creation of an ABX jitter test that could be posted on this forum. The test would allow comparisons between a jitter-contaminated track and an original track. We will keep this forum posted on our progress.
That seems like a coincidence, since I don't expect you have written those four paragraphs within 3 minutes. I'm looking forward to your proposal.
Yes indeed, I read your post after I posted.
Here is what we could do with the equipment we have (comments and suggestions welcome):
Jitter ABX:
We have low-noise 96 kHz 24-bit recordings that are known to have been made with very low-jitter A/D converters. We can use these as test material. We will also apply the same procedure to a 10 kHz sine wave test tone (to show jitter sideband amplitudes on an FFT analysis).
We have D/A converters in-house that will reproduce these tracks without any significant jitter artifacts (sidebands at least 135 dB below the music signal or test tone).
We have an A/D converter in house that allows us to switch jitter attenuation on and off. When jitter attenuation is on, jitter-induced sidebands will be at least 135 dB below the music signal.
We will use the A/D to apply jitter modulation to the clean analog audio reproduced by the "jitter free" D/A converter. Two versions will be created; one "jitter free" and one with added jitter. The result will be two 96 kHz 24-bit digital files (one with encoded jitter artifacts and one without).
We will use one of our Audio Precision 2722 Test Systems to generate the jitter signal. We can generate sinusoidal, square and/or random jitter functions of varying amplitude (but must limit the choices if we want to get anything done).
I suggest the use of the foobar ABX test plug-in, but careful attention must be paid to the following:
1) Jitter performance of the playback DAC
2) Distortion performance of the playback DAC
3) Data path from foobar to the DAC
4) Playback levels should be documented.
5) Playback equipment should be documented
CD player ABX:
Capture the output of a CD player using a "jitter free" A/D with precisely normalized gains. Capture the output of the same CD player through a "jitter free" D/A feeding the same A/D. Publish original track and captured CD player track, and the captured "clean track" for download and ABX comparison. All three tracks would be 44.1 kHz 16-bit.
Play a 10 kHz TPDF -20 dB FS tone through both chains and capture the results for FFT analysis (to confirm presence of sidebands).
Measure frequency response of both chains.
The CD player test is much simpler and we may want to start with this.
Comments and suggestions welcome please!
We will use one of our Audio Precision 2722 Test Systems to generate the jitter signal.
I think I saw one of these at work
Jitter ABX:
We have low-noise 96 kHz 24-bit recordings that are known to have been made with very low-jitter A/D converters. We can use these as test material.
We will also apply the same procedure to a 10 kHz sine wave test tone (to show jitter sideband amplitudes on an FFT analysis).
We have D/A converters in-house that will reproduce these tracks without any significant jitter artifacts (sidebands at least 135 dB below the music signal or test tone).
We have an A/D converter in house that allows us to switch jitter attenuation on and off. When jitter attenuation is on, jitter-induced sidebands will be at least 135 dB below the music signal.
We will use the A/D to apply jitter modulation to the clean analog audio reproduced by the "jitter free" D/A converter. Two versions will be created; one "jitter free" and one with added jitter. The result will be two 96 kHz 24-bit digital files (one with encoded jitter artifacts and one without).
We will use one of our Audio Precision 2722 Test Systems to generate the jitter signal. We can generate sinusoidal, square and/or random jitter functions of varying amplitude (but must limit the choices if we want to get anything done).
I suggest the use of the foobar ABX test plug-in, but careful attention must be paid to the following:
1) Jitter performance of the playback DAC
2) Distortion performance of the playback DAC
3) Data path from foobar to the DAC
4) Playback levels should be documented.
5) Playback equipment should be documented
CD player ABX:
Capture the output of a CD player using a "jitter free" A/D with precisely normalized gains. Capture the output of the same CD player through a "jitter free" D/A feeding the same A/D. Publish original track and captured CD player track, and the captured "clean track" for download and ABX comparison. All three tracks would be 44.1 kHz 16-bit.
Play a 10 kHz TPDF -20 dB FS tone through both chains and capture the results for FFT analysis (to confirm presence of sidebands).
Measure frequency response of both chains.
The CD player test is much simpler and we may want to start with this.
Comments and suggestions welcome please!
These are IMO the current "rules" for blind testing:
(1) Program material must include critical passages that enable audible differences to be most easily heard.
(2) Listeners must be sensitized to a audible differences, so that if an audible difference is generated by the equipment, the listener will notice it and have a useful reaction to it.
(3) Listeners must be trained to listen systematically so that audible problems are heard.
(4) Procedures should be "open" to detecting problems that aren't necessarily technically well-understood or even expected, at this time. A classic problem with measurements and some listening tests is that each one focuses on one or only a few problems, allowing others to escape notice.
(5) We must have confidence that the Unit Under Test (UUT) is representative of the kind of equipment it represents. In other words the UUT must not be broken, it must not be appreciably modified in some secret way, and must not be the wrong make or model, among other things.
(6) A suitable listening environment must be provided. It can't be too dull, too bright, too noisy, too reverberant, or too harsh. The speakers and other components have to be sufficiently free from distortion, the room must be noise-free, etc..
(7) Listeners need to be in a good mood for listening, in good physical condition (no blocked-up ears!), and be well-trained for hearing deficiencies in the reproduced sound.
(8) Sample volume levels need to be matched to each other or else the listeners will perceive differences that are simply due to volume differences.
(9) Non-audible influences need to be controlled so that the listener reaches his conclusions due to "Just listening".
(10) Listeners should control as many of the aspects of the listening test as possible. Self-controlled tests usually facilitate this. Most importantly, they should be able to switch among the alternatives at times of their choosing. The switchover should be as instantaneous and non-disruptive as possible.
Many of these requirements (1) Program material must include critical passages that enable audible differences to be most easily heard.
(2) Listeners must be sensitized to a audible differences, so that if an audible difference is generated by the equipment, the listener will notice it and have a useful reaction to it.
(3) Listeners must be trained to listen systematically so that audible problems are heard.
(4) Procedures should be "open" to detecting problems that aren't necessarily technically well-understood or even expected, at this time. A classic problem with measurements and some listening tests is that each one focuses on one or only a few problems, allowing others to escape notice.
(5) We must have confidence that the Unit Under Test (UUT) is representative of the kind of equipment it represents. In other words the UUT must not be broken, it must not be appreciably modified in some secret way, and must not be the wrong make or model, among other things.
(6) A suitable listening environment must be provided. It can't be too dull, too bright, too noisy, too reverberant, or too harsh. The speakers and other components have to be sufficiently free from distortion, the room must be noise-free, etc..
(7) Listeners need to be in a good mood for listening, in good physical condition (no blocked-up ears!), and be well-trained for hearing deficiencies in the reproduced sound.
(8) Sample volume levels need to be matched to each other or else the listeners will perceive differences that are simply due to volume differences.
(9) Non-audible influences need to be controlled so that the listener reaches his conclusions due to "Just listening".
(10) Listeners should control as many of the aspects of the listening test as possible. Self-controlled tests usually facilitate this. Most importantly, they should be able to switch among the alternatives at times of their choosing. The switchover should be as instantaneous and non-disruptive as possible.
Many of these requirements relate to ensuring the listener's sensitivity, so leaving them up to the listener makes sense.
There are two requirements that require actions by the people who prepare the data - requirements 3 and 5:
(3) Listeners must be trained to listen systematically so that audible problems are heard.
This means that files for listener training should be provided. The best form of listener training is the same music with the most audible relevant distortion generated by the UUT augmented in such a way that there are samples with unmistakably audible distortion, working in logical steps of less distortion, to the point where the actual distortion of the UUT is the object of the test.
(5) We must have confidence that the Unit Under Test (UUT) is representative of the kind of equipment it represents. In other words the UUT must not be broken, it must not be appreciably modified in some secret way, and must not be the wrong make or model, among other things.
This means that the supplier of the .wav files for testing also needs to provide the results of technical tests of the test environment and the UUT. The standard set of tests for audio gear that is generally availble are the Audio Rightmark tests. The Audio Rightmark tests may be criticized as being limited when it comes to evaluating jitter, so whatever additional tests that the developer of the test materials wishes to apply should be OK as long as it is applied uniformly to the test environment and the UUT.
Oh boy, jitter audibility test! I am looking forward to it. Does all this, Arnold, means that ordinary people can't do ABX test?
And other thing, if the samples aren't 48 kHz or 96 kHz, I wouldn't be able to do the test, since I have Audigy 2 ZX Platinum card, and I would have to use foobar's upsampling. If the jitter would be induced into the audio signal, would it be "interpolated", and therefore possibly less audible when upsampled?
Does all this, Arnold, means that ordinary people can't do ABX test?
It does sound much more complicated than it really is. All points Arnold mentions can improve the significance of ABX results. I would rather call them recommendations than rules, though. It's not that your results are necessarily worthless if you don't stricly follow this 3 page ruleset.
Much is, for example, already automatically taken care of when you just use Foobar's ABX component. Should that show with a very high probability that you can hear a difference, there are good reasons to assume that you really do, without any necessary additional effort. When you can't hear a difference, though, it get's more complicated. To demonstrate that differences are inaudible you must rule as many external factors as possible.
Comments and suggestions welcome please!
Jitter ABX:
...
This would certainly be interesting, but I'm not sure though what it would actually demonstrate. Can jitter be made audible? Sure it can, we don't need ABX tests to answer that question. It all depends on the
amount of jitter you add to a signal.
Showing that jitter still is an audible real world phenomenon with the current (low-cost) state of the art would be the much more interesting demonstration!
That's were you are heading here:
CD player ABX:
Capture the output of a CD player using a "jitter free" A/D with precisely normalized gains. Capture the output of the same CD player through a "jitter free" D/A feeding the same A/D. Publish original track and captured CD player track, and the captured "clean track" for download and ABX comparison. All three tracks would be 44.1 kHz 16-bit.
Play a 10 kHz TPDF -20 dB FS tone through both chains and capture the results for FFT analysis (to confirm presence of sidebands).
I would love to see those tests conducted. I would also prefer employing not some single CD player that most of us don't own, but common audio IC's found on several million PCs like the Realtek ALC line from inside a HF noisy PC case. You could also choose popular devices as a Mac mini (analog out vs. the integrated S/PDIF into a jitter tolerant DAC) or Creative's sound cards.
Oh boy, jitter audibility test! I am looking forward to it. Does all this, Arnold, means that ordinary people can't do ABX test?
And other thing, if the samples aren't 48 kHz or 96 kHz, I wouldn't be able to do the test, since I have Audigy 2 ZX Platinum card, and I would have to use Foobars upsampling. If the jitter would be induced into the audio signal, would it be "interpolated", and therefore possibly less audible when upsampled?
Foobar's upasmpling should not be a problem. We have measured it here and I personally have good confidence in its performance. I cannot say the same for the Audigy 2 ZX Platinum card.
You should be able to hear the differences with a training track (where jitter is added at very high levels), but your sound card may limit your ability to to hear the differences between the output of a CD player and the output of on outboard D/A. To hear the differences reliably, your sound card (and playback chain) will need to be significantly better than the CD player. In general, sound cards are one step below most CD players.
Can jitter be made audible? Sure it can, we don't need ABX tests to answer that question. It all depends on the amount of jitter you add to a signal.
Showing that jitter still is an audible real world phenomenon with the current (low-cost) state of the art would be the much more interesting demonstration!
That's were you are heading here:
CD player ABX:
Capture the output of a CD player using a "jitter free" A/D with precisely normalized gains. Capture the output of the same CD player through a "jitter free" D/A feeding the same A/D. Publish original track and captured CD player track, and the captured "clean track" for download and ABX comparison. All three tracks would be 44.1 kHz 16-bit.
Play a 10 kHz TPDF -20 dB FS tone through both chains and capture the results for FFT analysis (to confirm presence of sidebands).
Yes, but ...
The CD player will have many defects other than jitter that are at audible levels. The D/A converters and output stages in most CD players are surprisingly bad. Inexcusably bad! It is not at all uncommon to see non-harmonic distortion artifacts that are only 40 or 50 dB below the level of a test tone.
The differences are easy to hear but the differences are only partially due to jitter. IMD is a major issue, hum is an issue, microprocessor crosstalk is an issue, and random noise is an issue, and yes, jitter is an issue. The player may boast "96 kHz, 24-bit D/A" but deliver 12 to 14 bits of performance. I like to call the other 8 to 10 bits "marketing bits".
Having a "jitter-only" ABX test will allow us to hear jitter in isolation. I think both tests are valuable.
Does all this, Arnold, means that ordinary people can't do ABX test?
It does sound much more complicated than it really is. All points Arnold mentions can improve the significance of ABX results. I would rather call them recommendations than rules, though. It's not that your results are necessarily worthless if you don't stricly follow this 3 page ruleset.
They are 'rules' for maximizing the discriminative power of the test.
If you are a scientist trying to determine JNDs for jitter, they apply.
If you are testing some particular audiophile's claim to ALREADY HAVE HEARD A DIFFERENCE UNDER MUNDANE CONDITIONS, I would say they can be relaxed; all you need to is 'blind' the listener and let them try to repeat their 'sighted' performance with the same gear, music, etc.
What audiophiles tend to do, disingenously IMO, is react to the often-negative results of the latter kind of ABX by claiming that the full rigor of a scientific ABX was not applied. THey become conveniently evangelical for scientific rigor when it suits them.
The CD player will have many defects other than jitter that are at audible levels. The D/A converters and output stages in most CD players are surprisingly bad. Inexcusably bad! It is not at all uncommon to see non-harmonic distortion artifacts that are only 40 or 50 dB below the level of a test tone.
The differences are easy to hear but the differences are only partially due to jitter.
Then why have the 'easy to hear' differences not been demonstrated easily with extant publicized DBTs?
Really, please, settle this longstanding matter. Give us data on exactly which CDPs display this easy-to-hear behavior, under what conditions, so it could be replicated.
The CD player will have many defects other than jitter that are at audible levels. The D/A converters and output stages in most CD players are surprisingly bad. Inexcusably bad! It is not at all uncommon to see non-harmonic distortion artifacts that are only 40 or 50 dB below the level of a test tone.
The differences are easy to hear but the differences are only partially due to jitter.
Then why have the 'easy to hear' differences not been demonstrated easily with extant publicized DBTs?
Really, please, settle this longstanding matter. Give us data on exactly which CDPs display this easy-to-hear behavior, under what conditions, so it could be replicated.
Try this explanation of jitter and its audibility:
Jitter White Paper (http://www.positive-feedback.com/Issue43/jitter.htm)
Steve N.
Give us data on exactly which CDPs display this easy-to-hear behavior, under what conditions, so it could be replicated.
1) Noise Floor:
The most easy-to-hear behavior is the elevated noise floor of consumer CD and DVD players. Very few of these devices deliver the SNR that is possible with a 16-bit PCM TPDF dithered system. Many have SNR numbers around 80 dB. Furthermore, the noise floors of these devices are not white, but include servo noise, CPU noise, and AC line noise. To her the differences, the playback system must be adjusted to a level high enough to make the noise floor of the CDP audible. The SNR of the playback system must exceed that of the CDP. Under these conditions the noise issues are very obvious.
Other issues include:
2) IMD and THD+N:
Some CDPs measure up to 1% THD+N (including significant IMD components). This energy is only 40 dB below peak audio levels and may be audible under the right conditions.
3) Jitter-induced sidebands:
Jitter sidebands 40 to 60 dB below audio levels. Again these sidebands are high-enough that they may not fall under masking curves, and may be audible under the right conditions.
It is hard to verify the audibility of these other issues (items 2 and 3) when the differences in the noise floor of the D/A converters is so noticeable. The elevated noise floors of the CDPs were a dead-giveaway in most of the informal ABX tests we have conducted. The anecdotal evidence that I have for (2) and (3) is an apparent difference in frequency response (when no differences in frequency response could be measured). In attempting to test for (2) and (3) we had to reduce the playback levels to make the noise floor differences inaudible. Obviously this reduced playback level will make it harder to hear (2) and (3). This is why I am proposing the creation of a "jitter-only" ABX test.
Obviously we need to take a step beyond anecdotal accounts of differences, and measurements that suggest audible differences, and verify audibility (or lack thereof) with ABX tests.
I have some FFT plots of the outputs of CDPs that I can post if there is interest, but I do not want to imply that these plots verify audibility. They simply indicate that there are defects that may (or may not) be audible.
John, how well can this issue be modelled as straight-up sampling jitter?
The CD player will have many defects other than jitter that are at audible levels. The D/A converters and output stages in most CD players are surprisingly bad. Inexcusably bad! It is not at all uncommon to see non-harmonic distortion artifacts that are only 40 or 50 dB below the level of a test tone.
The differences are easy to hear but the differences are only partially due to jitter.
Then why have the 'easy to hear' differences not been demonstrated easily with extant publicized DBTs?
Really, please, settle this longstanding matter. Give us data on exactly which CDPs display this easy-to-hear behavior, under what conditions, so it could be replicated.
Try this explanation of jitter and its audibility:
Jitter White Paper (http://www.positive-feedback.com/Issue43/jitter.htm)
Steve N.
Interesting but by your own admission, still quite inconclusive. You 'believe' that the human sensitivity to jitter is grossly underestimated, butstill haven't got the robust listening test evidence -- just a lot of systems and test signals that are apparently 'not resolving enough'.
Once again, I would say from such work that the differences COMMONLY REPORTED by audiophiles whenever they compare two digital sources -- they almost always claim to hear a difference --are UNLIKELY to be due to jitter.
This does not rule out the possibility that some people have heard bad jitter. But audiophiles probably don't even know when they're hearing it. "White hats' need to start telling audiophiles (and some professinals) that 'jitter' can't just be handwaved as an explanation whenever they try to explain why they 'know' two CDPs or DACs sound different from a sighted comparison.
John, how well can this issue be modelled as straight-up sampling jitter?
Separating artifacts into different categories (noise, jitter, distortion, etc.) can give us the ability to determine audibility of individual defects (in this case jitter). The problem is that the poor SNR performance of CD players makes identification easy in an ABX test. If SNR differences make identification easy, then it is impossible to determine if the other artifacts are also audible.
Give us data on exactly which CDPs display this easy-to-hear behavior, under what conditions, so it could be replicated.
1) Noise Floor:
The most easy-to-hear behavior is the elevated noise floor of consumer CD and DVD players. Very few of these devices deliver the SNR that is possible with a 16-bit PCM TPDF dithered system. Many have SNR numbers around 80 dB. Furthermore, the noise floors of these devices are not white, but include servo noise, CPU noise, and AC line noise. To her the differences, the playback system must be adjusted to a level high enough to make the noise floor of the CDP audible. The SNR of the playback system must exceed that of the CDP. Under these conditions the noise issues are very obvious.
Let me explain where I am coming from: familiarity with an audiophile world where
every digital player is CLAIMED to have its own sound, by the editors and writers of the print and internet high-end press. These are , for better or worse (mostly the latter), the mainstream 'authorities' on 'high end' sound -- the people the popular press will go to, for example , when they need a quote or sound bite about sound quality.
Differences revealed under conditions one would not encounter normally while listening to music, do not strike me as convincing sources of such reports. Meyer and Moran found that the only audible difference in their blind tests between DSD and Redbook audio was encountered using abnormally high output level while playing 'silence' -- a predicted noise floor difference. Listening to music at that level for a lenght of time would be uncomfortable and possibly dangerous.
Other issues include:
2) IMD and THD+N:
Some CDPs measure up to 1% THD+N (including significant IMD components). This energy is only 40 dB below peak audio levels and may be audible under the right conditions.
3) Jitter-induced sidebands:
Jitter sidebands 40 to 60 dB below audio levels. Again these sidebands are high-enough that they may not fall under masking curves, and may be audible under the right conditions.
It is hard to verify the audibility of these other issues (items 2 and 3) when the differences in the noise floor of the D/A converters is so noticeable. The elevated noise floors of the CDPs were a dead-giveaway in most of the informal ABX tests we have conducted. The anecdotal evidence that I have for (2) and (3) is an apparent difference in frequency response (when no differences in frequency response could be measured). In attempting to test for (2) and (3) we had to reduce the playback levels to make the noise floor differences inaudible. Obviously this reduced playback level will make it harder to hear (2) and (3). This is why I am proposing the creation of a "jitter-only" ABX test.
Again, it is not unusual to be able to hear or see a difference *easily* when it is *purposely magnified*. But that doesn't mean it's perceptible at '1X' magnification.
Were the playback levels used in these ABX tests, levels that the subjects would use to listen to music for enjoyment?
Were the playback levels used in these ABX tests, levels that the subjects would use to listen to music for enjoyment?
That's what I asked myself, too. You can always show that bit depth
n is inferior to
n-1 when you just raise the volume level high enough. It never stops.
Were the playback levels used in these ABX tests, levels that the subjects would use to listen to music for enjoyment?
Yes, at normal playback levels of 90 to 100 dB SPL A-weighted slow.
At normal playback levels, the noise of a 16-bit TPDF dithered 44.1 kHz PCM system can exceed audibility thresholds. Many CDPs are 10 to 20 dB noisier than CD quality. Yes, this noise is audible at normal playback levels, especially when playing music that has little amplitude compression applied.
Here is a link to a list of CD players that measured "Less than "CD quality"":
"A Case for the Jitters" - stereophile.com (http://www.stereophile.com/features/1208jitter/index1.html)
The first player on the list has a noise floor that is elevated by 12 dB relative to a perfect CD system.
Were the playback levels used in these ABX tests, levels that the subjects would use to listen to music for enjoyment?
Yes, at normal playback levels of 90 to 100 dB SPL A-weighted slow.
At normal playback levels, the noise of a 16-bit TPDF dithered 44.1 kHz PCM system can exceed audibility thresholds. Many CDPs are 10 to 20 dB noisier than CD quality. Yes, this noise is audible at normal playback levels, especially when playing music that has little amplitude compression applied.
That would seem to suggest a concrete counterexample to Meyer/Moran, except for the fact that 100dB A-weighted slow seems
extremely loud to me. You could only play 2 CDs at that loudness a day before exceeding OSHA occupational noise limits.
If used with "uncompressed" music, suggesting a crest factor of 20db, you're looking at peak SPLs in the 120-130dB range. I'm not really sure that is a normal playback level to begin with.
Here is a link to a list of CD players that measured "Less than "CD quality"":
"A Case for the Jitters" - stereophile.com (http://www.stereophile.com/features/1208jitter/index1.html)
The first player on the list has a noise floor that is elevated by 12 dB relative to a perfect CD system.
... Right, but a freaking
iPod tested 6db better than that. I think it's fair to suggest that the McIntosh tested represents a pathologically bad (and perhaps defective) design, that is not representative of what exists in even low-end consumer audio today. And that's not even getting into audibility considerations.
Separating artifacts into different categories (noise, jitter, distortion, etc.) can give us the ability to determine audibility of individual defects (in this case jitter). The problem is that the poor SNR performance of CD players makes identification easy in an ABX test. If SNR differences make identification easy, then it is impossible to determine if the other artifacts are also audible.
I ask this because, among the items in my bag of tricks, is a program that can simulate sampling jitter. Given a numeric description of a real-world sampling jitter spectrum, and a test music signal, I might be able to provide some simulated jitter files to facilitate ABX testing of that specifically - assuming that the obvious issues regarding interaction of playback device jitter interaction with simulated jitter are taken care of, of course.
Right, but a freaking iPod tested 6db better than that. I think it's fair to suggest that the McIntosh tested represents a pathologically bad (and perhaps defective) design, that is not representative of what exists in even low-end consumer audio today. And that's not even getting into audibility considerations.
I have several DVD players here that don't look much different than the McIntosh. Players this bad do exist and they are being cranked out in great quantities. The problem is that nobody has bothered to measure them.
I might be able to provide some simulated jitter files to facilitate ABX testing of that specifically - assuming that the obvious issues regarding interaction of playback device jitter interaction with simulated jitter are taken care of, of course.
Interesting. This could eliminate the D/A and A/D process that I would use to apply the jitter modulation using jitter signals supplied by an Audio Precision test station. I have good confidence in the transparency of the D/A and A/D that we would use, but my jitter test signals are somewhat limited. The AP is limited to sinusoidal, square wave, or wide band random jitter signals. I can apply jitter over a frequency range of 10 Hz to 5 kHz with this test setup.
I suspect you may be able to do more than this if you can phase modulate the audio in a DSP. Tell us more!
Regarding the playback D/A:
I have playback D/A converters here with jitter-induced sidebands that are below -135 dB FS under all input conditions that we can use for testing here. Obviously these tests cannot be replicated if played back from a computer sound card that has jitter levels reaching or exceeding the test levels. However, files with exaggerated jitter content could be instructional for those who do not have low-jitter playback equipment. It would also be useful for training to listen for specific jitter-induced defects.
I have several DVD players here that don't look much different than the McIntosh. Players this bad do exist and they are being cranked out in great quantities. The problem is that nobody has bothered to measure them.
OK, that's news to me. And I wholeheartedly agree with the notion that the fact that nobody measures such devices is a problem - and I suppose that not having decent sideband measurements with tools like RMAA contributes to the problem. Still... aren't there significant shades of gray here? You clearly show that the McIntosh is in fact representative of low-end audio, but is it representative of mid-end?
Interesting. This could eliminate the D/A and A/D process that I would use to apply the jitter modulation using jitter signals supplied by an Audio Precision test station. I have good confidence in the transparency of the D/A and A/D that we would use, but my jitter test signals are somewhat limited. The AP is limited to sinusoidal, square wave, or wide band random jitter signals. I can apply jitter over a frequency range of 10 Hz to 5 kHz with this test setup. I suspect you may be able to do more than this if you can phase modulate the audio in a DSP. Tell us more!
It's nothing particularly magical - as far as this is concerned, it's an implementation of a jitter simulator according to eg Hawksford, except without the optimization assuming modulation amplitudes much smaller than the sampling period. It's actually using a variable-rate resampler under the hood, with pre/post up/downsampling of arbitrary factor. So as long as the resampling time offsets don't go
backwards, I should be able to simulate modulation at any amplitude and from DC to, well, light.
The simulation is about automodulation right now *cough*, but it could very easily be supplied from an external file or generated from a formula or spectrum. I wrote it up in conjunction with something I'm presenting at the October AES convention. (I'm a noob, wish me luck).
In terms of accuracy, on one null test I've determined the peak error of the simulation with truly ludicrous amounts of modulation to be ~~ -111dbFS vs theoretical results. I haven't tested this particular configuration yet, though.
Regarding the playback D/A: I have playback D/A converters here with jitter-induced sidebands that are below -135 dB FS under all input conditions that we can use for testing here. Obviously these tests cannot be replicated if played back from a computer sound card that has jitter levels reaching or exceeding the test levels. However, files with exaggerated jitter content could be instructional for those who do not have low-jitter playback equipment. It would also be useful for training to listen for specific jitter-induced defects.
True - but before we get into trying to find the best converters, I think it's important here to delineate what the meaning of a given transparency threshold is, relative to a known quantity of playback jitter. I mean, without that knowledge, in theory, one could dismiss all ABX tests by claiming "but you could have used a better converter to get a better result!"
More specifically: is it safe to argue inaudibility purely on the basis of masking?
Could I, with genuine respect, make a small observation here. People are wanting John_Siau to engage in a debate on audibility thresholds. He has said earlier that he has the [a href='index.php?act=findpost&pid=113']'luxury'[/a] of not having to bother too much about thresholds, he wants to make stuff with a safety factor. This is entirely rational, as a luxury: it would seem that if one used that sort of equipment, one could kick out the jams and play music at unusual, or even unwise, levels and still be sure of not hearing artifacts.
I say that because there's a slight tendency to attribute to him scepticism about ABX tests. That is not at all the case, it's just that, for him, audibility in a blind test is not the design target, but rather the certainty of inaudibility even in extreme cases.
Equally, it's a bit unreasonable to expect him to campaign against audiophile errors. He makes genuinely good stuff, which is advertised with honest information (which in itself has an educational value); that seems not merely ethical, but honorable. OTOH, a lot of his market will be rich people with funny ideas, and it is a bit unreasonable to expect him to go around saying "Hey, f'wits, you believe a load of garbage. Buy my stuff."
Sorry to interfere, but I've seen discussions get unnecessarily heated, and I'm enjoying following this one.
What do you want? Ending the discussion, because luxury must stay exempt from reason? That would be both boring and inconsistent.
A sheik doesn't put golden screws under his rear seat, where nobody would ever be able to see them. He puts them where a difference is visible. Some Porsches can be called over engineered, great cars! But they did not waste a hundred million to increase the rear mirrors' reflectivity from 99.96 to 99.9999999 percent, just because they could.
I own a DAC1 myself for many years, and I have always been very satisfied with it. It's like putting an "end of story" on the question which DAC to choose and wether it matters. Still I couldn't answer, if spending $1000 was necessary. I find the question wether it is the equivalent of an over-engineered rear-view mirror or actually able to make an audible difference highly interesting. I have never actually recommended the DAC1 to anyone for exactly that reason. I can't guarantee that it's not a waste compared to a good, modern sound card.
And I agree, ABX playback levels should not be constrained to only usual cases. Especially when I get intimate with Beethoven at 100-105 db every once in a few years, I don't want to hear any additional hiss from my playback chain. So ABX tests regarding high end DACs should not prohibit unusual levels, only those where it gets pathological (like >106 db for more than 30 minutes).
Indeed, I thought I made very similar points to MichaelW's earlier (http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=73201&view=findpost&p=649312)in the thread...
He has said earlier that he has the [a href='index.php?act=findpost&pid=113']'luxury'[/a] of not having to bother too much about thresholds, he wants to make stuff with a safety factor. This is entirely rational, as a luxury: it would seem that if one used that sort of equipment, one could kick out the jams and play music at unusual, or even unwise, levels and still be sure of not hearing artifacts.
Thank you!
My goal is to design products where all of the artifacts produced by the product fall below the threshold of hearing in a silent room. In order to guarantee inaudibility, I am consciously ignoring masking effects and designing as if the only sound being played was that of the artifacts. To me this means keeping the sum total of all artifacts at a level that is at least 110 dB below the peak audio levels. This goal has been achievable in the analog domain for at least 20 years, and for the last 10 years has been achievable in the digital domain. Achieving these goals requires more engineering, careful circuit layout, and a few more components. We deliver products that meet these goals in a price range that may seem extravagant to the average consumer ($1000 to $2000). But, these are very reasonable prices for professional products that will be used in a recording studio on a regular basis. The hi-end hi-fi enthusiast looks at our products and says "how can it be any good if it only costs $2000"?
My approach would be entirely different if I were designing perceptual encoders and decoders. My hat is off, and I take a bow to the engineers who design these systems to ride the hairy edge between audibility and inaudibility. In may ways their job is much more difficult than mine. The data that drives their design decisions has been derived from many carefully designed listening tests and ABX tests. The success of their efforts can and should be evaluated with carefully designed listening tests. On the other hand, if I achieve my very conservative design goals, and validate them with extensive measurements, the listening tests simply serve as a final check to confirm inaudibly.
If we can show that an artifact is 135 dB below the level of the music, most reasonable people will agree that a listening test is not required to prove inaudibility. If an artifact is 70 dB below the level of the music, we should consider testing for audibility. These two numbers -135 dB FS and -70 dB FS represent the level of jitter artifacts in good professional gear and common consumer products respectively. If we measure noise the numbers are -117 dB FS and -70 to -80 dB FS. If we measure THD+N the numbers are -110 dB vs -60 to -40 dB. The large discrepancy in performance between professional and common consumer equipment should make some wonder if the differences are audible!
The average consumer has no idea how bad their CD player or sound card really is. There often are no published specifications.
The average consumer has no idea how bad their CD player or sound card really is. There often are no published specifications.
Or one could say "The average consumer has no idea how
good their CD player or sound card really is. There often are no published specifications."
As already stated, audibility and safety factor are not the same things. From what I've seen, the DAC1 does measure better than most (if not all) stuff out there. But would I be able to pass a blind test between it and a basic (less than $100) CD/DVD player or sound card from an local electronics store?
In my experience with volume controlled tests, no. Granted, I cannot generalize my findings to the population at large, but I think a number of people on Hydrogen-Audio would side with me regarding how good even entry level stuff is. Your statement makes it seem like there is a lot of junk out there, but I won't believe such a statement unless some measurements and blind tests results are posted.
The average consumer has no idea how bad their CD player or sound card really is.
How bad can it really be if they don't know?
Your statement makes it seem like there is a lot of junk out there, but I won't believe such a statement unless some measurements and blind tests results are posted.
OK, fair enough. I like the fact that you need to see evidence.
I can post the measurements and will do so shortly. I need to put up a web page where interested persons can download PDFs of our tests of CD and DVD players. I hope to have the page up by the end of the day.
I will also post tests of computer sound cards within a few days.
Remember, if you are going to make claims about audible differences in sound quality, they must be accompanied with double-blind test results.
Could I, with genuine respect, make a small observation here. People are wanting John_Siau to engage in a debate on audibility thresholds. He has said earlier that he has the [a href='index.php?act=findpost&pid=113']'luxury'[/a] of not having to bother too much about thresholds, he wants to make stuff with a safety factor. This is entirely rational, as a luxury: it would seem that if one used that sort of equipment, one could kick out the jams and play music at unusual, or even unwise, levels and still be sure of not hearing artifacts.
I say that because there's a slight tendency to attribute to him scepticism about ABX tests. That is not at all the case, it's just that, for him,
audibility in a blind test is not the design target, but rather the certainty of inaudibility even in extreme cases.
No one would have an issue if he said his design goal was 'bullet proof' performance for its own sake and left it at that.
But he's also the one making claims edging towards *routine* audible difference in CDPs under normal listening conditions (Although whether the conditions cited are 'normal' is also under dispute; I've seen no address yet to this point of Axon's:
That would seem to suggest a concrete counterexample to Meyer/Moran, except for the fact that 100dB A-weighted slow seems extremely loud to me. You could only play 2 CDs at that loudness a day before exceeding OSHA occupational noise limits.
If used with "uncompressed" music, suggesting a crest factor of 20db, you're looking at peak SPLs in the 120-130dB range. I'm not really sure that is a normal playback level to begin with.
)
There are no free passes, nor should there be. Free passes for audio claims have done enough damage to the hobby already. As tiptoe eloquently noted on the 'do we need audiophiles' thread:
There is an entire ecosystem of false information, over-priced products, magazines that pontificate on said information and products, and then tons of forums where the audiophiles all sit around and agree with each other. Once you buy into it, you get constant reinforcement. The more strongly you believe, the more of it you will take in and believe.
But he's also the one making claims edging towards *routine* audible difference in CDPs under normal listening conditions (Although whether the conditions cited are 'normal' is also under dispute; I've seen no address yet to this point of Axon's:
That would seem to suggest a concrete counterexample to Meyer/Moran, except for the fact that 100dB A-weighted slow seems extremely loud to me. You could only play 2 CDs at that loudness a day before exceeding OSHA occupational noise limits.
If used with "uncompressed" music, suggesting a crest factor of 20db, you're looking at peak SPLs in the 120-130dB range. I'm not really sure that is a normal playback level to begin with.
Movie theaters and movie post-production facilities are calibrated to deliver 85 dBc at -20 dB FS. At this level, the -93 dB noise floor of a 16-bit system will be reproduced at a level of 12 dBc. This means that the noise floor of a 16-bit digital system should be audible in these calibrated facilities. We build converters that are used in these applications so it is important for us to occasionally calibrate them to these levels for short intervals of critical listening. Systems with a 70 to 80 dB SNR are not acceptable in an environment that is calibrated for film. What constitutes normal levels an acceptable performance in the living room is another matter altogether. Some home theater systems are calibrated to match the levels of public theaters. These home theater applications put significant demands on the SNR of the system.
Free passes for audio claims have done enough damage to the hobby already. As tiptoe eloquently noted on the 'do we need audiophiles' thread:
I agree there should not be any free passes!
There is an entire ecosystem of false information, over-priced products, magazines that pontificate on said information and products, and then tons of forums where the audiophiles all sit around and agree with each other. Once you buy into it, you get constant reinforcement. The more strongly you believe, the more of it you will take in and believe.
I wholeheartedly agree.
Remember, if you are going to make claims about audible differences in sound quality, they must be accompanied with double-blind test results.
My test plots will not confirm audibility but they may suggest the need to investigate audibility with the double-blind tests that I proposed earlier in this thread.
What do you want? Ending the discussion, because luxury must stay exempt from reason? That would be both boring and inconsistent.
So ABX tests regarding high end DACs should not prohibit unusual levels, only those where it gets pathological (like >106 db for more than 30 minutes).
You really do need to learn to read. I was not saying that ABX tests are in any way wrong. What I am saying is that there is no law in nature or ethics which says it is wrong to overdesign anything, to build to way beyond the threshold of audibility, as long as you know what you're doing, and acknowledge it is a luxury. It's only wrong if you then claim there are clear audible differences.
What I'm worried about is the zealotry, that says it is somehow a betrayal of reason to build anything with a healthy margin. What John_Siau's stuff is doing is the same as someone who knows that they can only ABX LAME V5, but encodes at V2 to be on the safe side, because disk is cheap. Do you see anything immoral in that?
I am NOT wanting to end the discussion. Did you fail to read my statement that I found the discussion interesting? What I did not want was for it to get diverted from the refinements of audibility into yet another bit of audiophile bashing. Yes, they can be fools (satisfied?), but there are other things to talk about, and when we get someone who both builds extravagant stuff and knows what he's doing, it would be a pity to start bashing him as though he were purveying pixie dust. Which I saw a danger of happening. And your response is just what I'm worried about. See something that on a sloppily careless reading can look like questioning reason, and flame away.
Reading is also to do with thinking, just like mathematics.
But he's also the one making claims edging towards *routine* audible difference in CDPs under normal listening conditions
That, I think, is your suspicion talking. I have not so read him. It is common to acknowledge, here, that a difference is audible between 16-bit and 24-bit
when played at unusual (but not fatal) levels. As I said, there may well be people who are prepared to risk a bit of hearing loss to, just occasionally, really crank it up. It might be unwise, but as long as it's their decision, why not include this unusual case in the design targets.
I am NOT questioning the value of all the careful work that has been done on the limits of hearing. (Also, I myself have crappy hearing and don't ever intend to spend real money on audio equipment--I'm more interested in music). But there is a tendency to be really fundamentalist about it. I can not see that it is any kind of crime against reason to design to better than audibility threshold, and to say so. That is what I mean by luxury. Except, of course, that in professional use, it might not be luxury. Not even the fundamentalists question using 24/96 for recording and processing.
Also, a manufacturer of high quality equipment is NOT responsible for the garbage put out by scam artists. Another point that fundamentalists find hard to accept.
That, I think, is your suspicion talking. I have not so read him.
This is a quote from post 113.
"We have a QSC ABX tester that we used to compare consumer CD players to prototypes of our DAC1 converter. It was fairly easy to score perfectly on the ABX tests. The CD players with modulation problems sounded like they had more midrange when played through their internal D/A converters than when played through the external DAC. This was a rather surprising result given that both devices had nearly identical frequency responses."
This is from post 154
"The average consumer has no idea how bad their CD player or sound card really is. There often are no published specifications."
I interpreted all of these as audible differences. Don't get me wrong, John_Siau hasn't made the blanket "Night and Day" difference comments you hear on many audio forums, but a claim to an audible difference, even a minor one, is still a claim to an audible difference. If the claim is a difference "at untypically loud volumes" that's fine as that tells me I can still buy my el-cheapo DVD player from a local audio store and get a negative blind tests at subjectively (to me) sane volume levels.
But at this point I'm not sure if these claims are for ear damage inducing volumes or not.
But he's also the one making claims edging towards *routine* audible difference in CDPs under normal listening conditions
It is common to acknowledge, here, that a difference is audible between 16-bit and 24-bit when played at unusual (but not fatal) levels. As I said, there may well be people who are prepared to risk a bit of hearing loss to, just occasionally, really crank it up. It might be unwise, but as long as it's their decision, why not include this unusual case in the design targets.
You need to add another condition. In order to hear a difference beteween 16-bit and 24-bit
when played at unusual (but not fatal) levels you need some special program material. Your average orchestral/pop recording has no more than about 65-70 dB dynamic range, which means that no matter whether you listen to a 16 or a 24 bit recording of it, the noise floor is built into the recording, typically at the outputs of the microphones. The widest dynamic range commercial recording I've ever found had more like 85 dB dynamic range, but that still means that the noise floor is set by the recording, not the CD medium nor the player.
But he's also the one making claims edging towards *routine* audible difference in CDPs under normal listening conditions
That, I think, is your suspicion talking. I have not so read him. It is common to acknowledge, here, that a difference is audible between 16-bit and 24-bit when played at unusual (but not fatal) levels. As I said, there may well be people who are prepared to risk a bit of hearing loss to, just occasionally, really crank it up. It might be unwise, but as long as it's their decision, why not include this unusual case in the design targets.
AIUI, the difference is audible on low-level content , played at a high output level. There may be people 'cranking up the fadeouts of Metallica songs specifically, but I'd guess more typically they're cranking up the loud parts -- parts where the 16 vs 24 difference is not going to be audible even when 'cranked'.
I am NOT questioning the value of all the careful work that has been done on the limits of hearing. (Also, I myself have crappy hearing and don't ever intend to spend real money on audio equipment--I'm more interested in music). But there is a tendency to be really fundamentalist about it. I can not see that it is any kind of crime against reason to design to better than audibility threshold, and to say so. That is what I mean by luxury. Except, of course, that in professional use, it might not be luxury. Not even the fundamentalists question using 24/96 for recording and processing.
Actually some 'fundamentalists' do question the 'need' for 96 kHz SR. It doesn't impart inherently better sound (as audiophiles tend to believe), it just makes it EASIER to get it. And high-bit recording and production is typically used NOT to 'increase resolution' in the audiophile vernacular parlance, but to increase headroom (during recording) and prevent audible artifacts from being introduced (by production).
The implication: one could 'overbuild' a 44/16 chain, and with carefully monitored recording, and minimal production, make a recording that doesn't 'need' either higher SR or higher bitdepth.
Also, a manufacturer of high quality equipment is NOT responsible for the garbage put out by scam artists. Another point that fundamentalists find hard to accept.
All the more reason to very, very carefully observe the boundary of 'we do it because we can' versus 'we do it because it makes stuff sound better'.
Btw, your use of the label 'fundamentalist' is hardly conducive to the amity you're advocating.
Btw, your use of the label 'fundamentalist' is hardly conducive to the amity you're advocating.
Oh, I'd quite got over amity. I was pissed off at the suspicion and hostility that was starting to show towards a rational contributor.
"Fundamentalist" was mostly trolling, but not entirely, because there is a mindset around that says because 44/16 is enough, it's somehow a betrayal to use more (more headroom is a kind of concession to human weakness, or something).
There's even sometimes a suspicion of using measurements. Surely the great achievement of the last 50 years or so is to discover what actually makes a difference to perception, and what instrumental measurements are relevant, and what numbers. Of course, DBT is vital to validate and calibrate this, but surely engineers find it much easier to design and develop to criteria that can be measured instrumentally? Check, if necessary, by DBT at the end. But what some people might want to do (e.g. John_Siau) is to say something like "<defect> may be audible at, say, -70 db, so I'll design to -100, because if I designed to -75, I'd have to run a DBT." And what is fundamentalist is to take a statement like that and challenge it on grounds of suspicion of heresy--"How do you know it might be audible at -75? How dare you act on that untested assumption?".
The problem with fundamentalism is not what is believed, but how it is believed. And I've seen before that someone with a rational and perhaps verifiable point gets attacked because it looks like it challenges the established view. I'm thinking of the thread on the possibility that a cartridge has a compensating error in stereo reproduction which might mean that you can get a better stereo image with vinyl. That idea got flamed because people thought it was just another vinyl-is-better rave. That failure to actually listen, and the manichaean division of the world into the enlightened and the phools, seems reasonable to call fundamentalist.
Btw, your use of the label 'fundamentalist' is hardly conducive to the amity you're advocating.
Oh, I'd quite got over amity. I was pissed off at the suspicion and hostility that was starting to show towards a rational contributor.
Rational?
I'd say somewhat rational.
I'd also say someone who is grasping at straws.
Take for example, this statement:
"The average consumer has no idea how bad their CD player or sound card really is. There often are no published specifications."
Today, the inverse of that is also true.
The average consumer has no idea how good their CD player or sound card really is. There often are no published specifications. However, if you put even a $39.95 optical disk player or a $29.95 sound card on the test bench, it performs well enough for sonically transparency to be extremely likely. If you listen to it and get past any biases you may have, it usually sounds great.
"Fundamentalist" was mostly trolling, but not entirely, because there is a mindset around that says because 44/16 is enough, it's somehow a betrayal to use more (more headroom is a kind of concession to human weakness, or something).
It's not a betrayal, its just a waste.
In this day and age, building a DAC with 100 dB dynamic range is almost a slam dunk. But, building a true 24 bit DAC is mission impossible. Why put a ton of effort into doing something that's going to be difficult and costly?
There's even sometimes a suspicion of using measurements. Surely the great achievement of the last 50 years or so is to discover what actually makes a difference to perception, and what instrumental measurements are relevant, and what numbers. Of course, DBT is vital to validate and calibrate this, but surely engineers find it much easier to design and develop to criteria that can be measured instrumentally? Check, if necessary, by DBT at the end.
OK.
But what some people might want to do (e.g. John_Siau) is to say something like "<defect> may be audible at, say, -70 db, so I'll design to -100, because if I designed to -75, I'd have to run a DBT."
You are distorting reality. Current reality is that just about the cheapest DAC that's worth the trouble to stick on a circuit board has maybe 90 dB or better dynamic range. Nobody is using 70 dB DACs as their baseline any more, except maybe people building $20 or cheaper portable CD players.
And what is fundamentalist is to take a statement like that and challenge it on grounds of suspicion of heresy--"How do you know it might be audible at -75? How dare you act on that untested assumption?".
It's 2009, and your argument is off by over 20 dB. That makes it a straw man argument whether you know it or not.
The problem with fundamentalism is not what is believed, but how it is believed. And I've seen before that someone with a rational and perhaps verifiable point gets attacked because it looks like it challenges the established view. I'm thinking of the thread on the possibility that a cartridge has a compensating error in stereo reproduction which might mean that you can get a better stereo image with vinyl. That idea got flamed because people thought it was just another vinyl-is-better rave. That failure to actually listen, and the manichaean division of the world into the enlightened and the phools, seems reasonable to call fundamentalist.
Yet another straw man argument. The LP format has a ton of grossly audible things wrong with it, poor and variable imaging being just one of them. The *real* situation is that the people who prefer vinyl are using a biased criteria to judge sound quality. They don't want the best possible imaging, they want the sort of garbaged-up imaging that you get from vinyl. They like that classic nasty LP sound, along with the grit, the tics, the other noises, and all the rest. If what they hear has the audible flaws of vinyl, its what they like.
Pursit of purer sound that more closely resembles a live performance has nothing to do with it, their rhetoric notwithstanding.
You are distorting reality. Current reality is that just about the cheapest DAC that's worth the trouble to stick on a circuit board has maybe 90 dB or better dynamic range. Nobody is using 70 dB DACs as their baseline any more, except maybe people building $20 or cheaper portable CD players.
Yes, but are these circuit board delivering 90 dB to the analog outputs? Do you have measurement data that show that they do? If so, please provide this information, it would be helpful to this discussion. If measurements are not available, lets take the first step and make a few measurements.
I will measure the CD, DVD, and sound cards that we happen to have in our facility here. I will test these devices for SNR, THD+N, and frequency response. I will also capture an FFT of the output. These tests may (or may not) show that audible differences
could be possible. If the tests show a
possibility of audible differences, it would seem that the next logical step would be to conduct some DBTs.
And what is fundamentalist is to take a statement like that and challenge it on grounds of suspicion of heresy--"How do you know it might be audible at -75? How dare you act on that untested assumption?".
It's 2009, and your argument is off by over 20 dB. That makes it a straw man argument whether you know it or not.
OK, how bad does a device need to measure before we should suspect audible differences? Or, how good does a device need to measure before we can say "there should be no audible differences"? And how large is that gray area in between?
This forum demands DBTs for all claims of audibility. I am a firm believer in the need for DBTs and in the need to confirm audibility of measured defects. The industry is full of ridiculous claims that cannot be verified with DBTs, and those making these claims frown upon DBTs. I enthusiastically commend this forum for its strong stance on DBTs.
But lets not be so focused on DBTs that we ignore other evidence. DBTs are not the only tool at our disposal: We also have a great deal of psychoacoustic data that was generated from well-designed listening tests that give us guidelines for what may be audible or inaudible. We also have excellent measurement equipment that is widely available.
Surely we cannot ignore what good psychoacoustic data can tell us about the significance of properly executed measurements. Measurements do have some value when properly interpreted in the light of well-established psychoacoustic data.44.1 kHz/16 bit PCM was designed to be good enough to be inaudible at normal listening levels. Is it too much to ask that our CD players and sound cards at least deliver this level of performance? Is 44.1 kHz 16-bit performance is a luxury in a typical home environment? If so, is there a lesser standard that we should expect from our equipment? What is this lesser standard (if someone wishes to propose one and defend it)?
If we agree that "CD quality" is a reasonable expectation wouldn't it be interesting to know where the average home CD or DVD player falls in relationship to this? If they are usually significantly worse that CD quality, perhaps some DBTs are needed. If they are significantly better than CD quality, many in this forum may feel that these DBTs would be a waste of time.
... there is a mindset around that says because 44/16 is enough, it's somehow a betrayal to use more (more headroom is a kind of concession to human weakness, or something).
It's not a betrayal, its just a waste.
In this day and age, building a DAC with 100 dB dynamic range is almost a slam dunk. But, building a true 24 bit DAC is mission impossible. Why put a ton of effort into doing something that's going to be difficult and costly?
Regarding the issue of "Luxury":
It may be a luxury to drive a 4X4 on the paved roads of suburbia, but a 4X4 is not a luxury for the farmer that needs to drive out to distant fields on his farm to check on crops or livestock. What is a luxury in the living room may be a necessity in professional audio applications. Let me give an example:
We build line-level analog to analog audio distribution amplifiers with a 500 kHz bandwidth (they measure 0.1 Hz to 500 kHz +0/-3 dB). Immediately most sane people would assert that this is "overkill", and a "luxury", and "20 Hz to 20 kHz is more than adequate". Or they might say "if professional users are willing to pay for this then let them, but what a waste".
Actually the professionals that buy this equipment, specify it because they need it. They need the extended bandwidth (not because they claim to have golden ears, but because they have a legitimate verifiable need). These 500 kHz audio devices are used in television broadcast facilities where it is not uncommon to find 50 or more audio stages in the signal chain between the microphone and the TV transmitter. If each stage was 3 dB down at 20 kHz, we would have 50 cascaded 20 kHz low pass filters which would yield an overall system bandwidth of less than 1 kHz! With the 500 kHz devices they can achieve 10 to 15 kHz system bandwidths in typical applications. Believe it or not, it took the television industry a number of years to figure out that they needed 500 kHz audio distribution amplifiers. Nobody had taken the time to measure the end-to-end performance of the television network, and nobody had done the math.
We have a similar situation: Nobody has taken the time to measure the end-to-end performance of audio delivery to the home. We have a significant body of psychoacoustic data that suggests what we can and cannot hear. What we don't have is a good grasp on how well the delivery quality matches the design goals of the 44.1 kHz 16-bit system.
If we want to deliver a CD product that approaches the limits of the format, we need to do substantially better than this in the recording studio. Again it took the industry a while to figure this out. "CD quality" can and is being done, but such recordings are still few and far between. Each element in the studio chain must significantly outperform the CD system in order to deliver a "CD quality" product. If you do the math it can be shown that most of the devices in the chain need to be about 12 dB quieter than the noise floor of the CD system. Likewise bandwidths must exceed that of the CD system if the goal is a product with "CD quality".
The tools that are required to deliver an end product with "CD quality" are now readily available to recording studios. We are one of the companies that specialize in building these tools for professional applications. When these professional tools are brought into the home environment, the quality of the recording and the limitations of the CD format become the sole performance limitations. When consumer grade products are used in the home, the end results may still be limited by the quality of the recording, but in other cases the CDP will be the limiting factor. Ultimately it would be nice to see CDPs that deliver "CD quality". I suspect that we are not there yet. Let's find out where we are!
Are the current CDPs DVDs and sound cards good enough? It all depends on how they are being used. Are you driving through suburbia, or through 2 feet of mud? If you are using a digital volume control that is not set near 100% at normal listening levels, you may be driving through 2 feet of mud! There are three valid solutions:
1) Stop driving through the mud - (set the digital volume near 100% and turn down an analog volume control - if one exists)
2) Buy the 4x4 - (use a "luxury" D/A converter so that you can enjoy the convenience of a digital volume control and still achieve CD quality playback)
3) Get stuck once in a while - (tolerate the degraded performance introduced by the less-than-ideal gain staging caused by the digital volume control)
All three choices are valid. Most consumers will pick option 3. Option 3 is also a very good choice when other factors such as speakers and amplifiers may be limiting the system performance.
Has anybody besides John Siau access to a high end ADC? For example from Benchmark or Apogee. I would love to hear some common gear's outputs (Mac minis, Intel HD audio compliant onboard chips, Creative's cards, etc.) recorded at high resolution. It would be interesting to see wether those were ABXable at all vs. the original files, when you play them back through a good DAC.
While I'm looking forward to John Siau's "jitter only" tests, the former would be much easier to conduct and still have much practical relevance for many of us. It would show if and if yes how much of a difference a good DAC can make for (non professional) home use.
Of course, also a regular ADC would work, but it could add its own artifacts to the result and thereby mask a tested DAC's signature.
How about starting with the EBU SQAM (http://www.ebu.ch/en/technical/publications/tech3000_series/tech3253/index.php) Harpsichord sample?
There are a gaggle of people here with Transporters. Certain mid-fi interfaces like the 0404 USB might qualify.
In other news, is anybody able to make actually good use of my jitter simulator, if I packaged it up?
I don't know about CD players, but measurements for onboard and dedicated sound cards are all over the web.
Some onboard chipsets - Bit-tech.net (http://www.bit-tech.net/hardware/motherboards/2009/01/21/gigabyte-ga-ex58-ud4p-and-ds4-review/6)
Some Discrete Cards - Firingsquad (http://www.firingsquad.com/hardware/asus_xonar_essence_stx_review/page5.asp)
Other Various Cards - Firingsquad (http://www.elitebastards.com/cms/index.php?option=com_content&task=view&id=318&Itemid=27&limit=1&limitstart=2)
I found all of that in 10 minutes. If I had specific cards to look for and a few hours...
As you can see, even onboard chipsets exceed 75 db for dynamic range, stereo crosstalk etc.
Edit: The following is inaccurate. Please read my [a href=\'index.php?act=findpost&pid=651262\']follow-up[/a].Even without access to a good ADC I have just successfully ABXed a Benchmark DAC1 vs. a Macbook Pro's (early 2008) onboard audio (Intel HD audio compliant Realtek codec) - both level matched and time synched!
Procedure (Recording 2x):
- Connect the MBP's line-out into its line-in with a loopback cable (1. run) / connect the MBP via Toslink to the DAC1 and its outputs to the MBP's line-in (2. run).
- Set Mac OS X's audio parameters to output 44.1 kHz/24 bit, input 96 kHz/24 bit.
- Playback fullscale 1 kHz sine wave to adjust input gain to just below clipping.
- Record while playing back tec_sqam_40a test sample.
- Normalize recorded sample to -1 db.
Procedure (ABX):
- Normalize tec_sqam_40a to -1db and convert its sample rate to 96 kHz (Audition 3, highest quality, without pre/post filtering).
- Synchronize recorded samples' time/offset to tec_sqam_40a's with Synchrotron.
- Invert/restore phase of DAC1 recording.*
- Set Mac OS X's output parameters to 96 kHz/24 bit.
- Plug headphones into DAC1.
- ABX with Foobar in VMware Fusion virtual machine.
It was quite easy, I focused onto the range 0:07-0:09.
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/12 00:57:02
File A: Z:\rpp3po\Downloads\rec_dac1_24_inverted.wav
File B: Z:\rpp3po\Downloads\rec_mbp_24.wav
00:57:02 : Test started.
00:58:02 : 01/01 50.0%
00:58:08 : 02/02 25.0%
00:58:21 : 03/03 12.5%
00:58:32 : 04/04 6.3%
00:58:41 : 05/05 3.1%
00:58:50 : 06/06 1.6%
00:58:59 : 07/07 0.8%
00:59:14 : 08/08 0.4%
00:59:26 : 09/09 0.2%
00:59:42 : 10/10 0.1%
00:59:43 : Test finished.
----------
Total: 10/10 (0.1%)
The MBP's output fades into noise of a rather awkward type, the DAC1's is of lower volume and has also a more pleasing character.
The DAC1 sample is also still ABXable vs the original file, but how could it have been better than the MBP's ADC? So that was to be expected in the light of the MBP's DAC results.
*One curious thing I have noticed: The DAC1 record comes out with inverted phase. This can either mean that both the MBP's DAC and ADC invert phase or they work correctly and my DAC1 is inverting. For the sake of this ABX test the phase could be corrected/inverted easily with Audition, but I'll open another [a href=\'index.php?showtopic=73999\']thread[/a] to follow up on this.
- Set Mac OS X's audio parameters to output 44.1 kHz/24 bit, input 96 kHz/24 bit.
AFAIK OSX can't serve two different sample rates at the same time without applying SRC somewhere in the path. Since your rec_mbp_24.wav sample shows some >23kHz components I get the impression that the MBP DAC output wasn't properly low pass (anti-image) filtered.
May I suggest that you redo the test:
-either completely at 24/44.1 or
-at 24/96 by (carefully) converting the test sample to 96kHz first.
SRC is a very carefully designed issue in OS X. The only limitation is that you can't playback two sample rates into the same (virtual) device at once without SRC, which makes sense. Different rates for input and output or completely different devices is no problem (as long as your hardware supports this). You can even choose the clock source (when available) on a per device level through the standard OS dialogs.
I think the >22 kHz content you are seeing is HF noise, not aliasing. To be sure I am going to repeat the test later with everything at 96 kHz. Initially I had chosen the approach to playback at 44.1 kHz because the DACs' Redbook performance is what I was mainly interested in. Recording was chosen to be 24/96 to minimize the ADC's influence on the results.
ABX results
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/12 09:29:00
File A: \\bmsbs\redirectedfolders\JSiau\Desktop\ABX Tests\DAC1 vs MacBookPro\rec_dac1_24_corrected.flac
File B: \\bmsbs\redirectedfolders\JSiau\Desktop\ABX Tests\DAC1 vs MacBookPro\rec_mbp_24.flac
09:29:00 : Test started.
09:32:33 : 01/01 50.0%
09:33:04 : 02/02 25.0%
09:33:18 : 03/03 12.5%
09:33:27 : 04/04 6.3%
09:33:46 : 05/05 3.1%
09:33:57 : 06/06 1.6%
09:34:06 : 07/07 0.8%
09:34:22 : 08/08 0.4%
09:34:30 : 09/09 0.2%
09:34:41 : 10/10 0.1%
09:34:50 : 11/11 0.0%
09:34:57 : 12/12 0.0%
09:35:09 : 13/13 0.0%
09:35:19 : 14/14 0.0%
09:35:35 : 15/15 0.0%
09:35:44 : 16/16 0.0%
09:35:54 : Test finished.
----------
Total: 16/16 (0.0%)
The rec_mbp_24 file sounds like it was truncated to 16-bits somewhere in the processing chain. This truncation causes noise pumping and distortion on the fade out of the final note.
Used a DAC1 USB for playback using the USB input and headphone output. Headphones were closed-back Sony MDR-V600.
Would it be fair to say that you turned the volume up to listen to the fade out?
I had the volume up quite far, but I took care that the whole normalized track was still listenable without getting too uncomfortable and I did not turn up the volume higher for the 0:07-0:09 position.
Why there should be 16 bit truncation going on, is a riddle to me. * From the point of any user controllable parameters I can rule it out, but I can't guarantee the same for the hardware and driver level. I wish I had better equipment available right now. I'm going to prepare a second round of testing in a minute and will upload a new set of samples afterwards.
PS * Maybe the input level gain control is digital instead of analog. I'll try identical input gain levels in the second round.
Would it be fair to say that you turned the volume up to listen to the fade out?
Yes. The first test just confirms that an artifact exists. It does not address the audibility in normal listening.
I am just about to re-run the test at a lower volume using speakers. I will do two tests: The first test will be in my office where the AC noise is measuring 39 to 41 dBA slow. The second test will be in our listening room where the ambient noise is below the range of my B&K 2219 Sound Level Meter. I will measure and document the SPL produced by a 1 kHz -20 dB FS test tone (at the same settings used in the tests).
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/12 11:58:23
File A: \\bmsbs\redirectedfolders\JSiau\Desktop\ABX Tests\DAC1 vs MacBookPro\rec_dac1_24_corrected.flac
File B: \\bmsbs\redirectedfolders\JSiau\Desktop\ABX Tests\DAC1 vs MacBookPro\rec_mbp_24.flac
11:58:23 : Test started.
11:58:47 : 01/01 50.0%
11:58:58 : 02/02 25.0%
11:59:05 : 03/03 12.5%
11:59:11 : 04/04 6.3%
11:59:19 : 05/05 3.1%
11:59:24 : 06/06 1.6%
11:59:30 : 07/07 0.8%
11:59:36 : 08/08 0.4%
11:59:42 : 09/09 0.2%
11:59:48 : 10/10 0.1%
11:59:53 : 11/11 0.0%
11:59:58 : 12/12 0.0%
12:00:03 : 13/13 0.0%
12:00:15 : 14/14 0.0%
12:00:20 : 15/15 0.0%
12:00:28 : 16/16 0.0%
12:00:34 : Test finished.
----------
Total: 16/16 (0.0%)
Ambient noise at listening position 37 dBA (at my office desk)
Listening level: -20 dBFS test tone plays at 76 dBA at the listening position - both speakers driven
Test file measures 76 to 79 dBA slow at loudest section
D/A = DAC1 USB
Speakers/amplifier = Dynaudio BM5A (powered speakers)
Notes: I am hearing a difference in the noise floor. The MBP sample has higher noise and noise modulation.
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/12 14:47:44
File A: \\bmsbs\redirectedfolders\JSiau\Desktop\ABX Tests\DAC1 vs MacBookPro\rec_dac1_24_corrected.flac
File B: \\bmsbs\redirectedfolders\JSiau\Desktop\ABX Tests\DAC1 vs MacBookPro\rec_mbp_24.flac
14:47:44 : Test started.
14:48:13 : 01/01 50.0%
14:48:21 : 02/02 25.0%
14:48:28 : 03/03 12.5%
14:48:35 : 04/04 6.3%
14:48:41 : 05/05 3.1%
14:48:48 : 06/06 1.6%
14:48:55 : 07/07 0.8%
14:49:07 : 08/08 0.4%
14:49:17 : 09/09 0.2%
14:49:26 : 10/10 0.1%
14:49:37 : 11/11 0.0%
14:49:43 : 12/12 0.0%
14:49:51 : 13/13 0.0%
14:49:59 : 14/14 0.0%
14:50:06 : 15/15 0.0%
14:50:12 : 16/16 0.0%
14:50:20 : Test finished.
----------
Total: 16/16 (0.0%)
Ambient noise at listening position less than 30 dBA (in our listening room)
Listening level: -20 dBFS test tone plays at 64 dBA at the listening position - both speakers driven
Test file measures 69 to 71 dBA slow at loudest section
D/A = DAC1 USB
Speakers/amplifier = Klien and Hummel O 300D (powered speakers) - using analog inputs
Notes: I am hearing a difference in the noise floor. The MBP sample has higher noise and noise modulation.
I recall all posted samples and claims! Turns out, the difference was the ADC and test setup, not the DACs!
The DAC1 has much higher output volume in its standard "calibrated" setting than the MBP. Because of that I had adjusted the input gain setting in OS X's audio control panel in the first round until I had about equal values for both inputs on the level meter.
My ABX results were honestly much better than I ever would have expected, so I wanted to make sure that I hadn't overlooked anything for the second round. So this time I set the input gain to "0db" for both. I put the DAC1 into "variable" mode so that I could use its analog gain control to adjust its volume to exactly the same level as the MBP's output. I also converted the tec_sqam sample to 96 kHz
prior to to playback as suggested (Izotope intermediate phase) and applied 24 bit dither after normalizing.
The results are like night and day now and both very good - only ABXable vs. the original at insane volume levels to me.
Two findings for MBP owners:
- Better leave the "input gain control" untouched. It added considerable amounts of noise in my case.
- Even if OS X allows different settings for input and output sample rate in the audio/midi control panel for the MBP, you get aliasing if you actually use it.
Sorry for the stir.
I recall all posted samples and claims!
Turns out, the difference was the ADC and test setup, not the DACs!
Two findings for MBP owners:
- Better leave the "input gain control" untouched. It added considerable amounts of noise in my case.
- Even if OS X allows different settings for input and output sample rate in the audio/midi control panel for the MBP, you get aliasing if you actually use it.
Thanks for the correction.
What you have demonstrated is that digital volume control can greatly reduce the performance of computer sound cards. I discussed this at the end of a prior posting: "Regarding the issue of Luxury" (http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=73201&view=findpost&p=650859)
Higher D/A (and A/D) performance is needed when digital volume controls are used. For example, if 20 dB of digital attenuation is applied at normal listening levels, the D/A would need to have a signal to noise ratio (SNR) of about 116 dB to deliver CD quality. Obviously a simple solution is to reduce the digital attenuation and apply some analog gain reduction. Unfortunatly many consumer audio systems have no gain control in the analog-domain. As a result, these systems often deliver audio that cannot approach CD quality.
rpp3po took great to try to get the best possible quality for the ABX source files. The results were not what was expected because of the presence of the digital volume control. This demonstrates just how difficult it can be to get good audio out of many consumer devices.
The ABX tests in the prior postings demonstrated that the digital volume control in the Mac Book Pro degraded the performance of its A/D converter to the extent that it was easy to hear differences at normal listening levels. This degradation was significant enough to allow me to successfully ABX the two test files using the speakers and sound card built into my HP Mini 2133 netbook while listening in relatively high ambient noise levels produced by a nearby AC ventilator (44 dBA at the listening position):
foo_abx 1.3.4 report
foobar2000 v0.9.6.8
2009/08/14 10:25:42
File A: \\bmsbs\redirectedfolders\JSiau\Desktop\ABX Tests\DAC1 vs MacBookPro\rec_dac1_24_corrected.flac
File B: \\bmsbs\redirectedfolders\JSiau\Desktop\ABX Tests\DAC1 vs MacBookPro\rec_mbp_24.flac
10:25:42 : Test started.
10:26:15 : 01/01 50.0%
10:26:21 : 02/02 25.0%
10:26:27 : 03/03 12.5%
10:26:33 : 04/04 6.3%
10:26:39 : 05/05 3.1%
10:26:51 : 06/06 1.6%
10:26:59 : 06/07 6.3%
10:27:13 : 07/08 3.5%
10:27:28 : 08/09 2.0%
10:28:06 : 09/10 1.1%
10:28:12 : 10/11 0.6%
10:28:33 : 11/12 0.3%
10:28:49 : 11/13 1.1%
10:29:25 : 12/14 0.6%
10:29:34 : 13/15 0.4%
10:29:42 : 14/16 0.2%
10:29:54 : Test finished.
----------
Total: 14/16 (0.2%)
Ambient noise 44 dBA slow
Peak levels in sound track 79 to 81 dBA slow
Listening level: -20 dBFS test tone plays at 76 dBA at the listening position - both speakers driven
Volume control on HP Mini set to maximum
Playback through HP Mini internal sound card and internal speakers.
The problem I generally see with externally generated jitter, either through up-/downsampling or electronically, is the applicability of those results. You can take it so far, that you can hear at least something, i. e. what jitter of model m at gain g does or does not sound like. After that you will have several pairs m, g that seem to be relevant. But those then have to be translated and tested against real world implementations.
The way I see to address that situation is to simply know what kinds of jitter exist in the real world, and duplicate them. I learned quite a bit about jitter by testing equipment for my now-defunct www.pcavtech.com web site. The most common kind of really heavy jitter is related to the power line frequency. Frame rates also show up fairly frequently. While quite a bit has been written about self-jitter, it is not all that common at levels that are likely to be heard.
The problem I generally see with externally generated jitter, either through up-/downsampling or electronically, is the applicability of those results. You can take it so far, that you can hear at least something, i. e. what jitter of model m at gain g does or does not sound like. After that you will have several pairs m, g that seem to be relevant. But those then have to be translated and tested against real world implementations.
The way I see to address that situation is to simply know what kinds of jitter exist in the real world, and duplicate them. I learned quite a bit about jitter by testing equipment for my now-defunct www.pcavtech.com web site. The most common kind of really heavy jitter is related to the power line frequency. Frame rates also show up fairly frequently. While quite a bit has been written about self-jitter, it is not all that common at levels that are likely to be heard.
That's part of the problem though: it's hard to know what jitter really does to the signal. I have read a few papers on modelling jitter in the last days, and yet am not exactly sure what is going on. I suppose because it presumably depends on the DAC, which is hard to model. Most papers I've looked at do give explicit models, but they appear to have assumptions on jitter that I find it hard to take on trust (not really, I am sure it is accurate, but with no personal experience of what it "sounds like", they just seem to me to be assertions; I'd like to find a way to avoid those).
I'm running some numerical experiments right now but am not at all convinced by my techniques (lots of perhaps unwarranted assumptions) so will shut up for a while.
The problem I generally see with externally generated jitter, either through up-/downsampling or electronically, is the applicability of those results. You can take it so far, that you can hear at least something, i. e. what jitter of model m at gain g does or does not sound like. After that you will have several pairs m, g that seem to be relevant. But those then have to be translated and tested against real world implementations.
The way I see to address that situation is to simply know what kinds of jitter exist in the real world, and duplicate them. I learned quite a bit about jitter by testing equipment for my now-defunct www.pcavtech.com web site. The most common kind of really heavy jitter is related to the power line frequency. Frame rates also show up fairly frequently. While quite a bit has been written about self-jitter, it is not all that common at levels that are likely to be heard.
That's part of the problem though: it's hard to know what jitter really does to the signal.
It is hard to know the details of just about anything. That's why education can take so long and require so much effort, There's another post active here that wants to know the flow of data from the CD to the headphone jack, including all data busses, all buffers, all data formats. Not a bad task for a thesis project.
I have read a few papers on modelling jitter in the last days, and yet am not exactly sure what is going on.
That would be your problem!
I suppose because it presumably depends on the DAC, which is hard to model.
Modeling DACs is very easy compared to modelling say, a vacuum tube or magnetic tape.
I don't think that a person needs to model DACs to understand jitter. Jitter is FM distortion or FM modulation. What do you know about FM modulation, or modulation in general?
Most papers I've looked at do give explicit models, but they appear to have assumptions on jitter that I find it hard to take on trust (not really, I am sure it is accurate, but with no personal experience of what it "sounds like", they just seem to me to be assertions; I'd like to find a way to avoid those).
The way to avoid many kinds of assertions is to gather evidence. To paraphrase what I said above, I have gathered a certain amount of evidence. I gathered evidence about modulation and I gathered evidence about the results of modulation in real world audio equipment. I have gathered information about how the ear perceives the results of modulation on audio signals. So much for many spare moments and also dedicated moments during maybe 15 years of my life.
You lookin' for a rose garden, kid? ;-)
I'm running some numerical experiments right now but am not at all convinced by my techniques (lots of perhaps unwarranted assumptions) so will shut up for a while.
If you have questions, then by all means don't shut up about your questions. Ask them. I had virtually nobody to talk to when I was gathering my evidence. Very few people would even tell me where or how to gather evidence. I spent a lot of time and money measuring things and then analyzing what I measured. I had to buy expensive equipment. I had to buy expensive books. Some people lied to me, outright. Others belittled me in public for asking questions. Other people shared what I now know, based on superior evidence, to be wrong ideas. I probably have a few things wrong, too.
Welcome to the real world!
At least Amazon sells books to everybody with money!
My goal is to design products where all of the artifacts produced by the product fall below the threshold of hearing in a silent room. In order to guarantee inaudibility, I am consciously ignoring masking effects and designing as if the only sound being played was that of the artifacts. To me this means keeping the sum total of all artifacts at a level that is at least 110 dB below the peak audio levels. This goal has been achievable in the analog domain for at least 20 years, and for the last 10 years has been achievable in the digital domain. Achieving these goals requires more engineering, careful circuit layout, and a few more components. We deliver products that meet these goals in a price range that may seem extravagant to the average consumer ($1000 to $2000). But, these are very reasonable prices for professional products that will be used in a recording studio on a regular basis. The hi-end hi-fi enthusiast looks at our products and says "how can it be any good if it only costs $2000"?
I don't see any compelling logic here.
If your goal was to design products where all of the artifacts produced by the artifact fall below the threshold of hearing in a silent room, any reasonable person would say that you are tilting at windmills, given that there are no silent rooms.
A reasonable goal is to design products where no real world customer will hear artifacts in the most silent room that he is likely ever to encounter. A resasonable goal is to design products where no real world customer will hear artifacts while wearing the most isolating earphones in the best room that he is likely ever to encounter.
Once you've adopted a reasonable goal, then you should use a reasonable means to define the details of that goal.
Saying that your goal is "A", and then immediately and without any other evidence spouting some arbitrary technical specifications is again, not at all reasonable.
Once you have stated your goal, a reasonable way to proceed would be to model the performance of a product that is just barely noticable in terms of all known audible artifacts. Once you have produced that model, pick a reasonble but somewhat arbitrary number, and set your goal to produce a product that measures that good. For example, if you know all about the audibility of noise and distoriton, then set a goal that is X dB better than that.
The average consumer has no idea how bad their CD player or sound card really is. There often are no published specifications.
This statement shows obvious prejudice and bias. An unbiased person would say:
The average consumer has no idea how good or bad their CD player or sound card really is. They don't know what kind of specifications are required for total freedom from audible artifacts under critical but real world listening conditions, and most product spec sheets are so incomplete that they don't know how the product compares to the actual requirements.
The average consumer has no idea how good or bad their CD player or sound card really is. They don't know what kind of specifications are required for total freedom from audible artifacts under critical but real world listening conditions, and most product spec sheets are so incomplete that they don't know how the product compares to the actual requirements.
I absolutely agree with you - well said!
My goal is to design products where all of the artifacts produced by the product fall below the threshold of hearing in a silent room. In order to guarantee inaudibility, I am consciously ignoring masking effects and designing as if the only sound being played was that of the artifacts. To me this means keeping the sum total of all artifacts at a level that is at least 110 dB below the peak audio levels. This goal has been achievable in the analog domain for at least 20 years, and for the last 10 years has been achievable in the digital domain. Achieving these goals requires more engineering, careful circuit layout, and a few more components. We deliver products that meet these goals in a price range that may seem extravagant to the average consumer ($1000 to $2000). But, these are very reasonable prices for professional products that will be used in a recording studio on a regular basis. The hi-end hi-fi enthusiast looks at our products and says "how can it be any good if it only costs $2000"?
I don't see any compelling logic here.
If your goal was to design products where all of the artifacts produced by the artifact fall below the threshold of hearing in a silent room, any reasonable person would say that you are tilting at windmills, given that there are no silent rooms.
A reasonable goal is to design products where no real world customer will hear artifacts in the most silent room that he is likely ever to encounter. A resasonable goal is to design products where no real world customer will hear artifacts while wearing the most isolating earphones in the best room that he is likely ever to encounter.
Recording studios have many audio devices in cascade. The system performance is limited by the combined performance of all of the devices. My goal is to provide tools that don't get in the way (by causing audible defects). A comfortable margin of safety is appreciated in a professional environment. The goal of the recordist is to capture a musical performance. If he is distracted by the poor performance of his equipment chain, he may miss the opportunity to capture that very special musical performance.
Obviously most consumers do not need or demand this sort of performance.
My goal is to design products where all of the artifacts produced by the product fall below the threshold of hearing in a silent room. In order to guarantee inaudibility, I am consciously ignoring masking effects and designing as if the only sound being played was that of the artifacts. To me this means keeping the sum total of all artifacts at a level that is at least 110 dB below the peak audio levels. This goal has been achievable in the analog domain for at least 20 years, and for the last 10 years has been achievable in the digital domain. Achieving these goals requires more engineering, careful circuit layout, and a few more components. We deliver products that meet these goals in a price range that may seem extravagant to the average consumer ($1000 to $2000). But, these are very reasonable prices for professional products that will be used in a recording studio on a regular basis. The hi-end hi-fi enthusiast looks at our products and says "how can it be any good if it only costs $2000"?
I don't see any compelling logic here.
If your goal was to design products where all of the artifacts produced by the artifact fall below the threshold of hearing in a silent room, any reasonable person would say that you are tilting at windmills, given that there are no silent rooms.
A reasonable goal is to design products where no real world customer will hear artifacts in the most silent room that he is likely ever to encounter. A resasonable goal is to design products where no real world customer will hear artifacts while wearing the most isolating earphones in the best room that he is likely ever to encounter.
Recording studios have many audio devices in cascade.
I addressed this issue in a portion of my post that you didn't quote for some reason. Didin't read it? I didn't make it clear enough?
Once you have stated your goal, a reasonable way to proceed would be to model the performance of a product that is just barely noticable in terms of all known audible artifacts. Once you have produced that model, pick a reasonble but somewhat arbitrary number, and set your goal to produce a product that measures that good. For example, if you know all about the audibility of noise and distoriton, then set a goal that is X dB better than that.
I have no problem with intelligent overkill. I get the requirements of the production environment because I do production all the time.
My other problem is with the way you seem to be doing overkill, and I addressed that too, and you didn't quote or respond to that either:
Saying that your goal is "A", and then immediately and without any other evidence spouting some arbitrary technical specifications is again, not at all reasonable.
The point is that spouting some arbitrary technical specs isn't a logical way to proceed. The logical way to proceed is to determine the actual thresholds by means of reliable subjective tests, come up with a reasonable overkill factor for each of them, and built to that balanced overkill design.
I've done a fair amount of regression testing to determine how various forms of distortion build up when euqipment is cascaded. Have you?
The point is that spouting some arbitrary technical specs isn't a logical way to proceed. The logical way to proceed is to determine the actual thresholds by means of reliable subjective tests, come up with a reasonable overkill factor for each of them, and built to that balanced overkill design.
Since this thread is about jitter, let's have a look at a Benchmark statement about jitter (http://www.benchmarkmedia.com/appnotes-d/jittercu.html):
Jitter can only be considered totally inaudible if the worst case jitter induced sidebands are at least 23 dB below the A-weighted system noise. Above this level jitter may be audible or it may be masked by the program audio. At Benchmark our goal is to achieve totally inaudible levels of jitter.
[/size]Any comments on that design goal ?
Jitter can only be considered totally inaudible if the worst case jitter induced sidebands are at least 23 dB below the A-weighted system noise. Above this level jitter may be audible or it may be masked by the program audio. At Benchmark our goal is to achieve totally inaudible levels of jitter.
[/size]Any comments on that design goal ?
Not compliant with TOS8. Here, jitter can only be considered totally inaudible if someone fails an ABX test searching for it. But you can't prove a negative, so the closest we can come is finding the threshold at which jitter becomes audible and assuming that it is inaudible below that threshold.
The point is that spouting some arbitrary technical specs isn't a logical way to proceed. The logical way to proceed is to determine the actual thresholds by means of reliable subjective tests, come up with a reasonable overkill factor for each of them, and built to that balanced overkill design.
Since this thread is about jitter, let's have a look at a Benchmark statement about jitter (http://www.benchmarkmedia.com/appnotes-d/jittercu.html):
Jitter can only be considered totally inaudible if the worst case jitter induced sidebands are at least 23 dB below the A-weighted system noise. Above this level jitter may be audible or it may be masked by the program audio. At Benchmark our goal is to achieve totally inaudible levels of jitter.
[/size]Any comments on that design goal ?
Besides the obvious TOS8 violation, it fails reasonableness checks. By setting the acceptable level for jitter based simply on a level, it ignores the fact that masking is strongly dependent on the spectral distance between the spurious response(s) due to jitter and the masking signal(s). It also ignores the fact that the human ear is far more sensitive to FM distortion at some frequencies than others.
To put the audiblity issues related to jitter, we must remember that both the LP format and analog tape are simply full of jitter, and that the entire high end industry has nothing to say about it. I don't think there ever has been a halfways decent CD player that had even 1/100th the jitter inherent in LP playback or analog tape. Furthermore some of the early CD players that high end audiophiles love to hate such as the CDP 101, actually have very low jitter.
Obviously, jitter is one of high end audio's more obvious red herrings.
the closest we can come is finding the threshold at which jitter becomes audible and assuming that it is inaudible below that threshold.
From reading the Benchmark paper I had the impression that it was exactly that what they did. They didn't specify their sources (scientific papers, their own DBT's?) but the assumption is that when tones (in this case clock jitter distortion) are at least 23 dB below the system (in this case ADC) noise, they can be assumed inaudible.
Do you have any evidence that this assumption is incorrect ?
Obviously, jitter is one of high end audio's more obvious red herrings.
Your TOS8 violation indicates that you have doubts about the Benchmark claim of inaudibility of the jitter effects. On the other hand you state that any halfway CD player has much better jitter performance than LP or analogue tape, which is concidered (proven?) to be no audible problem.
In that case it's probably not unreasonable to assume that the Benchmark jitter performance is far below audibility thresholds ?
Obviously, jitter is one of high end audio's more obvious red herrings.
Your TOS8 violation indicates that you have doubts about the Benchmark claim of inaudibility of the jitter effects. On the other hand you state that any halfway CD player has much better jitter performance than LP or analogue tape, which is concidered (proven?) to be no audible problem.
In that case it's probably not unreasonable to assume that the Benchmark jitter performance is far below audibility thresholds ?
I'm pretty sure that the jitter in just about anything halfways decent is inaudible.
There's a JAES paper that covers this:
"Theoretical and Audible Effects of Jitter on Digital Audio Quality", Benjamin, Eric and Gannon, Benjamin, 105th AES Convention, 1998, Preprint 4826.
I don't know if the samples mentioned here are still acessible:
HDDaudio jitter test files link (http://hddaudio.net/viewtopic.php?id=63)