Skip to main content
Topic: more 192kHz nonsense? (Read 5177 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

more 192kHz nonsense?

I just stumbled into this https://www.youtube.com/watch?v=geaoEt-9V-w
He makes an argument for 192kHz not based on golden ear claims but on the quality of bandlimiting and reconstruction filters.
It doesn't make a whole lot of sense to me, but I'm not an expert, so, maybe some of you can give their thoughts on whether he's got a point or just talking nonsense.
He has a word of praise for MQA at some point in the video, so that kind of raises a red flag for me, but... as I said, I'm no expert.
This doesn't make sense to me, but I'm not sure I can disprove it either.

I'm specially interested in his claim about pre-echo in pulses after reconstruction. That steep raise on the pulse he shows t 5:20, isn't supposed to contain higher frequency components anyway?

Re: more 192kHz nonsense?

Reply #1
I just stumbled into this https://www.youtube.com/watch?v=geaoEt-9V-w
He makes an argument for 192kHz not based on golden ear claims but on the quality of bandlimiting and reconstruction filters.
It doesn't make a whole lot of sense to me, but I'm not an expert, so, maybe some of you can give their thoughts on whether he's got a point or just talking nonsense.
He has a word of praise for MQA at some point in the video, so that kind of raises a red flag for me, but... as I said, I'm no expert.
This doesn't make sense to me, but I'm not sure I can disprove it either.

I'm specially interested in his claim about pre-echo in pulses after reconstruction. That steep raise on the pulse he shows t 5:20, isn't supposed to contain higher frequency components anyway?

The pre-echo oscillation is typically at the Nyquist frequency, which being ultrasonic is inaudible.  So he is just spewing the usual ignorant audiophile garbage with a deceptive twist.

Re: more 192kHz nonsense?

Reply #2
Ah, I see. Makes sense, I probably should have realized xD
So, if I understand this correctly, reconstruction induced pre-echo would only start to potentially cause any trouble at sampling frequencies under the threshold of hearing, like 32kHz, and even there, only for purposely generated synthetic signals with pulses at the very edge of the band, not for signals that have passed through a proper bandlimiting filter, as, then, those pulses would be atenuated before they could cause any trouble.

That is, if you generate, say, single sample full-range (or less, to avoid damages) pulses at 22050Hz sampling, you may hear some pre and post echo, but such a signal is not realistic and won't happen in a properly bandlimited scenario.
At 44.1kHz, even if you had such a pulse, you wouldn't hear the pulse nor either echo.
Is this right?

Re: more 192kHz nonsense?

Reply #3
Notice how he jumps back and forth between DAC and bandlimiting pre-ADC. If 192 kHz makes sense, he is conveniently unclear as to whether it is a good thing to record at 192 kHz or play back. Is this promoting a revolutionary new discovery of something called "oversampling"?


with a deceptive twist.

Yeah, and a "good" one, in a couple of ways. One is that it is a bit like making an ad for floating-point PCM files based on it making sense in processing, but framing it as if it were essensial for an end-user-delivery format.
Memento: this is Hydrogenaudio. Do not assume good faith.

Re: more 192kHz nonsense?

Reply #4
I am not very familiar with this stuff. So my question might be dumb:

If my speakers state something like "38Hz - 23kHz", which is pretty much close to every speaker I ever owned, then how would 192kHz audio even be useful? The speakers can't even reproduce these frequencies. It seems anything above 46kHz audio (23 * 2) would be useless?

Re: more 192kHz nonsense?

Reply #5
The question is not dumb.

There is a "problem" called aliasing. For how the phenomenon works with visuals, see https://en.wikipedia.org/wiki/Wagon-wheel_effect . If you record to digital at 44.1 kHz, then frequencies above 22.05 kHz could show up as if they were frequencies below. So they should not be there, right?

Solution1: Apply a low-pass crossover from somewhere below 22.05 to get rid of them.
Solution2: Quantize at more than 44.1 kHz, and then apply the crossover in the digital domain afterwards.
Solution3: Quantize, distribute and DAC at more than 44.1 kHz. At 176.4, then you have two extra octaves for the filter. (192? That is two octaves above 48.)

Solution2 is widely employed (together with oversampling in the DAC, which means the sampling noise is farther away than just one octave). The video advocates 3 in place of 1. And offers one argument for 3 in place of 2: them oscillations.

Which (cf. Arnold's posting) are ultrasonic. No wonder: 44.1 cannot get stuff above 22.05 right in the first place, so unless the distortions one octave up were significant enough to cause other troubles ... forgettaboutit. That part of the alleged "problem" can be summarized as follows: signals up there are inaudible, so we can remove them; but, the process does not just remove them, but also "replace" by a component one octave up ... so you will get inaudibles in place of inaudibles, and he wants you to believe that it is a baaaaaaaaad thing.

Someone else can probably correct a few details here.
Memento: this is Hydrogenaudio. Do not assume good faith.

Re: more 192kHz nonsense?

Reply #6
Hans is a clueless believer idiot. Pay heed if compatible.
Loudspeaker manufacturer

Re: more 192kHz nonsense?

Reply #7
Maybe I should assume good faith, but: are we talking absence of clue, or - ahem - presence of economic incentives?
Memento: this is Hydrogenaudio. Do not assume good faith.

Re: more 192kHz nonsense?

Reply #8
I'm specially interested in his claim about pre-echo in pulses after reconstruction. That steep raise on the pulse he shows t 5:20, isn't supposed to contain higher frequency components anyway?
He's confusing pre-echo ("a softer copy of the signal before the original signal") with ringing—they're entirely different things.

Re: more 192kHz nonsense?

Reply #9
The video caught me a little off-guard, so I wasn't entirely sure what to make of my impression that none of it made any sense, but it's clear that it's just more audiophoolery.

@Porcus Hans' mention of MQA as a good think would seem to indicate, superficially at least, that there is some economic incentive after all, but I guess he could just be convinced of what he says, who knows.

@bandpss I see. I think I understand the way ringing works, but can you point me to information about pre-echo, how it happens and how it is different than the "front tail" of the ringing arround a pulse?

and, in general to anyone: https://hydrogenaud.io/index.php/topic,115577.msg953539.html#msg953539 could someone answer this? I think I get it, but I'd appreciate confirmation :)

Re: more 192kHz nonsense?

Reply #10
Pre-echo is an artefact of equiripple filters (which are desirable in real applications as they minimise the number of taps needed to obtain specific passband ripple and stopband rejection levels).

Unlike ringing, where the artefact is a pure tone at the cut-off frequency, pre-echo is full-band and very close to being an identical, but quieter copy of the original signal.

Whereas pretty much all introductions to filtering cover ringing, pre-echo is much less commonly covered.  The most famous paper on the subject is here: http://www.nanophon.com/audio/antialia.pdf

It is a real effect, as can be seen in this somewhat extreme example, utilising an equiripple filter with hundreds of thousands of taps:

Re: more 192kHz nonsense?

Reply #11
I didn't realise pre-echo is an issue with digital recording or playback (whether audible or not).  I thought it was more of an issue with analog gear, particularly analog tape.

Re: more 192kHz nonsense?

Reply #12
I didn't realise pre-echo is an issue with digital recording or playback (whether audible or not).  I thought it was more of an issue with analog gear, particularly analog tape.
You are probably thinking of print-through, which is entirely different.

Re: more 192kHz nonsense?

Reply #13
I'm thinking vinyl: when you slightly the hear the next track while the stylus is in the groove before it begins.

What's the right technical term for that one?

(It's probably a fault of non-absurdly-expensive decks, but I always rather liked it)
The most important audio cables are the ones in the brain

Re: more 192kHz nonsense?

Reply #14
I'm thinking vinyl: when you slightly the hear the next track while the stylus is in the groove before it begins.
I've seen that referred to as "pre-echo" yes, but also "groove echo" (I guess it does affect the "next" groove as well as the previous, but I guess there is more music with a sudden increase in volume - and maybe ears don't notice the "post-echo" equally well either).

(It's probably a fault of non-absurdly-expensive decks, but I always rather liked it)
According to that above link, it occurs during cutting.  (But surely, if there was "print through" on the master tape, it would end up on the record as well.)
Memento: this is Hydrogenaudio. Do not assume good faith.

Re: more 192kHz nonsense?

Reply #15
I'm thinking vinyl: when you slightly the hear the next track while the stylus is in the groove before it begins.
I've seen that referred to as "pre-echo" yes, but also "groove echo" (I guess it does affect the "next" groove as well as the previous, but I guess there is more music with a sudden increase in volume - and maybe ears don't notice the "post-echo" equally well either).

(It's probably a fault of non-absurdly-expensive decks, but I always rather liked it)
According to that above link, it occurs during cutting.  (But surely, if there was "print through" on the master tape, it would end up on the record as well.)


In the case of vinyl or tape. pre echo and post echo can be addressed by means of adding spacing. In tapes, a common way to do that was to splice  leader tape between the segments of tape that represented the various tracks on the recording. With disks, you just goosed the lead screw on the lathe.

Re: more 192kHz nonsense?

Reply #16
When watching Hans' youtube video, a few points sprang to my mind immediately:

1. "Nyquist ist just a theory" (around the 4th minute)

This is a very tendentious uttering that reminds me of creationists when they speak about evolution. Or short: It's bullshit.

The contribution of Nyquist ist mathematical in nature, and it is proven as well as anything can be proven. However, it assumes signals that are already bandlimited, and proves that they can be reconstructed perfectly when the sampling rate exceeds the bandwidth by a factor of 2. So any discussion of antialiasing filters in front of the ADC is a different matter from Nyquist. It is a discussion about the transparency and/or artefacts in filters. Keep those two things mentally separate.

He does mollify his earlier point at around 9:00, but by repeating that a signal can't be perfectly bandlimited, he just raises a strawman. Some (artificial) signals are perfectly bandlimited to start with. Even if they aren't, the question isn't about perfection, but about audibility of artefacts. Those who insist on perfection usually want you to believe that anything non-perfect is also audible. That's very far from true.

2. "Pulse response of reconstruction filter" (at about 5:20)

What he means with "pulse" is a nonzero sample embedded in a stream of zero samples. People intuitively take this as what a pulse looks like in digital. Yet it is bullshit, and has been used any number of times to fool the unwary. The question is what such a nonzero sample actually means, when it is embedded in a stream of zero samples. What signal does it represent? What would be its ideal rendering in analog, given sampling theory? The answer may be surprising to many, but it certainly can't be anything that resembles an analog pulse, because that pulse would violate the Nyquist criterion. Hence it can't have resulted from sampling an analog pulse, either.

It turns out that the ringing shown in the pulse response is actually a correct rendering of such a digital sequence, because it is one of the few ways in which the result sticks to the allowed bandwidth. You could say that it only shows how pathetic a signal you fed into the DAC, i.e. something that can't have arisen from sampling a real-world signal.

A real-world analog pulse, bandlimited according to Nyquist, and sampled by an ADC, would never result in a sequence of samples that have all zeros except for one sample. Judging a DAC's sonic performance with such an artificial and pathetic signal isn't going to result in anything useful.

Nevertheless, there is no law that demands that digital filters exhibit pre-echo. The can be designed to only show oscillation after the "pulse", as shown in one of Hans' diagrams. It is a choice of the filter designer which particular charateristic is wanted. The downside of a filter that only has those ripples afterwards is their phase response. So for real life signals, you would compromise phase performance for an imagined advantage when using pathetic test signals. Which one would you choose?

Furthermore, the requirement from Nyquist is not that your filter has 96 dB attenuation beyond half the sampling rate. It merely demands that the signal level beyond half the sampling rate is "low enough". How much attenuation you need is dependent on how much signal you have there in the first place (with real music, not a lot). Different compromises are possible here, if you aren't dogmatic. But Hans is right that there are different filters, and that people may differ in their appreciation, depending on the situation.

3. "Only analog filters" (around 8:00)

Digital filters have won the world by storm after the early DACs that weren't oversampling had such poor analog filter implementations and sounded (as people thought back then) inferior. There's a good reason for that: Analog filters are much more expensive, and much harder to get right, let alone produced reliably and with small tolerances. Noone with a sane mind would want to go back to this state of affairs, but that's exactly what NOS-advocates want. The technical argument is not a matter of taste, even though the "sound" inevitably is.

There may well be a good reason to use 192k sampling, namely in recording and mastering. That has to do with the fact that digital manipulations of the signal (dynamics processing, clipping, and so on) can introduce distortion products that get folded down at the Nyquist frequency, and hence produce a peculiarly "digital" type of distortion. If you have 96 kHz of bandwidth, most of that ends up inaudible, whereas with 22 kHz bandwidth, most will fall into the audible range. But that is a problem specific to a situation where the audio is actively being worked with, not just being played back. Before delivering the result to a consumer for playback, there's no harm in converting down to 44.1 kHz.


Re: more 192kHz nonsense?

Reply #18
@ajinfla
I'm almost speechless that he equated an ABX test to experimenting with raising children without human contact... just 1 minute into the video!!

Re: more 192kHz nonsense?

Reply #19
@ajinfla
I'm almost speechless that he equated an ABX test to experimenting with raising children without human contact... just 1 minute into the video!!

Oh, this one. The thing goes off the rails even before it starts - the lead-in ad was for a health food panaceal that was just as vlid as his auido weirness.  I get healthful eating, but...

The he starts out with this chip fresh off his shoulder: "Double blind tests (are used) to prove a difference between two ‘sounds’ is often seen by people as a way to prove me and my colleagues wrong.". 

This guy has no sense of history.

My associates and I devised the form of ABX testing that he damns long before audio had its current very serious problems with mythology and anti-science, back in the middle 1970s.  In those days he might not have even been a glint in his father's eye.

We devised ABX even before the advent of magic interconnect cables, a myth that I still think was invented when people saw the mass of cables that it takes to actually properly compare two pieces of audio gear with external level matching. Magic speaker cables were even rare. "Cobra Cables" had just been recently been invented and started blowing up amplifiers.  It was before the consumerization of digital audio and CD players. Jitter was not even on the horizon as a serious issue.

Re: more 192kHz nonsense?

Reply #20
I just noticed that Benkhuyzen made another video recently, where he explains the sampling and reconstruction process better than many others:

https://www.youtube.com/watch?v=z8dIzTZaRFY

It doesn't help, however, as he still manages to convey the wrong message in passing. Note, for example, how he creates the impression that time inaccuracies (Jitter) are the predominant source of audio impairment in digital audio (besides filtering). No quantitative information is presented, of course, you are just expected to believe him.

It gets particularly farcical, when he reminds us towards the end, that analog has its impairments as well, for example wow/flutter, yet analog can sound extremely good. Wouldn't it have been useful to add that those analog impairments are orders of magnitude stronger that those from jitter in digital audio?

It goes to show again that you can't judge anything by asking "is it perfect, yes or no?". Meaningful comparisons are quantitative. And for that, you must first ask "how much?". Try to find out how often Benkhuyzen asks and answers this question, compared to dropping the word "perfect".

Re: more 192kHz nonsense?

Reply #21
I don't take Beekhuyzen too seriously. He makes claims and statements that are also a lot over the edge of audiophoolism. Just watch some random video's of him, some make sense, but there are a lot that are just over the edge, especially the jitter claims. You have to do some reality check, but I unsubscribed to be honest. I don't want to be the guy that buys expensive linear adapters for fully digital equipment because he keeps insisting it sounds better.

He is also one that claims difference in the wav vs flac discussion if I recall correctly. I don't take that kind of sh*t anymore.

I have a (for me) fine setup consisting of a i5 Windows HTPC, a Gustard U12 DDC, NAD C388 and a pair of B&W CM10 s2 speakers. Now there a lots of people surfacing that claim any class D amplifier sucks because it's class D. I enjoy my setup and will not let these people influence me anymore. Maybe these people cannot afford them I guess.

A pair of Soundcare Superspikes is the last thing I bought as an upgrade (mainly to protect my wooden floor) to my set and I have nothing to wish no more. :)

Enjoy listening to music instead of talking about shortcomings!

Re: more 192kHz nonsense?

Reply #22
I have just been banned from his channel without notification.
Apparently he doesn't like scientific facts ( completely explained ) disputing his nonsense.

Re: more 192kHz nonsense?

Reply #23
I have just been banned from his channel without notification.
Apparently he doesn't like scientific facts ( completely explained ) disputing his nonsense.

The web is full of anti-science sites.

If you haven't been kicked off an audio site for talking science, you probably haven't been trying. Or not trying and just being your sweet self.

I even get public complaints from anti-science posters on sites where the moderator has specifically encouraged me to post early and often about science. He even ushers these people off if they get too testy about it. Doesn't seem to matter, people want things their way, regardless.  Meanwhile the site has grown by 100% to over 39,000 participants in less than 6 months.  People love science but sticking up for it can be bad for your health, it seems.


The problem with science is that it doesn't care what you want to think or believe. Lots of people have big problems with that.


 
SimplePortal 1.0.0 RC1 © 2008-2018