HydrogenAudio

Lossless Audio Compression => Lossless / Other Codecs => Topic started by: tahaa7 on 2011-09-07 23:59:29

Title: DSD frequency range
Post by: tahaa7 on 2011-09-07 23:59:29
Hi everyone,

I have a question regarding the DSD format and its frequency range. I did a research on DSD the other day and read about its alleged 50kHz frequency range. That intrigued me and motivated me to do a little experiment myself.

My research consisted of the following: I found and downloaded a sample DSD (.dff) track online. It was from a classical album usually sold on SACD by the company whose website I downloaded the file from. I used Weiss Saracon software to convert it to 24/192 PCM wav file. Then I opened that wav file with Wavelab to examine its waveform and, more importantly to me for my experiment, its frequency spectrum.

Surprisingly, though I had somehow thought as much, I discovered that there were no audio frequencies beyond 20-21 kHz! And given the proudly advertised (by Sony/Philips) SACD frequency range of around 50 kHz, the logical question is: what about those frequencies beyond 21 kHz?

So the main question is: Could it be that they were "wiped out" during the DSD-to-PCM conversion?

I doubt the company I downloaded the sample from would sell a SACD made from a standard CD master, which would have a cutoff at exactly those frequencies...

Thank you for your help clearing this up.
Title: DSD frequency range
Post by: saratoga on 2011-09-08 00:04:28
Its possible you converted it wrong, but more likely its just a CD master.
Title: DSD frequency range
Post by: tahaa7 on 2011-09-08 00:15:32
Its possible you converted it wrong, but more likely its just a CD master.


Yes, this is what I think too, but it happens on other files I downloaded from elsewhere... Strange...
Title: DSD frequency range
Post by: Wombat on 2011-09-08 00:23:57
Weiss engineering does a low-pass for sure to save you and your system from handling the waste of noise DSD has in higher frequencies. I remember to have read somewhere they low-pass very low but not in the 20-21kHz range. It must be somewhere above 30kHz.
So when you really have converted all correctly the chances are high the recording was done on some 44.1 or 48kHz equipment and then metamorphosed to the spheres of the DSD world
I really wonder how much recordings, now that it seems more and more people can access these DSD files, will show that fooling people is an was important part of the business.
If you read thru some forums there are even real professional recording engineers that swear recording some digital recording played back on an old Sony PCM and recorded with some magic DSD machine improves the sound! No kidding.
So this is what a caring engineer maybe did for you here!


Edit: besides that you should really rethink why you need 24/192, if you really have to 24/88.2 must be enough for archiving and even that with an early low-pass.
Title: DSD frequency range
Post by: tahaa7 on 2011-09-08 00:34:19
Thanks guys.

@ Wombat: yes, they're obviously fooling people. they probably think (all) people are dumb and/or unable to analyze their files to see what exactly is under the hood...

Edit: gonna do another experiment tomorrow.  By chance I have a vinyl record of an album I found a DSD sample of, so I'm gonna rip it to 24/96 (or 192) and see whether those high frequencies exist there. If not, it's possible that the original master never contained them as well, and therefore the DSD track doesn't... Looking forward to testing this.

I will report what I find out.
Title: DSD frequency range
Post by: Wombat on 2011-09-08 00:42:38
@ Wombat: yes, they're obviously fooling people. they probably think (all) people are dumb and/or unable to analyze their files to see what exactly is under the hood...


When DSD selling on SACD started they knew that it won´t be easily possible for the standard customer to access the digital data of it due to the strong copy protection. I bet some didn´t care about what they sell you because of that.
I wonder if some companies did contracts with recording studios and blindely trusted in them. And now that they sell the data they once bought several years back for SACD sellings still believe they sell something special without own checking.
Title: DSD frequency range
Post by: tahaa7 on 2011-09-08 00:44:57
When DSD selling on SACD started they knew that it won´t be easily possible for the standard customer to access the digital data of it due to the strong copy protection. I bet some didn´t care about what they sell you because of that.


You're right. Thankfully today one is indeed able to analyze (though indirectly) a DSD file...
Title: DSD frequency range
Post by: Notat on 2011-09-08 02:45:15
Its possible you converted it wrong, but more likely its just a CD master.

A third possibility is there was no energy above 20 kHz coming from the instruments that were recorded. Most non-percussion instruments are quite bandwidth limited. What are you listening to here?
Title: DSD frequency range
Post by: tahaa7 on 2011-09-08 03:01:25
Its possible you converted it wrong, but more likely its just a CD master.

A third possibility is there was no energy above 20 kHz coming from the instruments that were recorded. Most non-percussion instruments are quite bandwidth limited. What are you listening to here?


The first file is a classical song, which means there should be huge frequency range from instruments like cymbals, violins, cellos, etc. The second one is a rock song.
Title: DSD frequency range
Post by: knutinh on 2011-09-08 05:12:00
Simply passing through air, high frequencies will be attenuated more than mid-range. Only some instruments emit considerable energy above 20kHz. Many professional, good-sounding microphones pass little above 20kHz.

Question is, can you hear the difference? If the recording sounds good, then why care about being "ripped off" by a random spec that has no empirical evidence of being relevant to humans?

-k
Title: DSD frequency range
Post by: Kees de Visser on 2011-09-08 06:56:46
Many professional, good-sounding microphones pass little above 20kHz.
True, but the roll-off is far less steep than a brick-wall filter, so most of them produce frequencies above 20kHz.

Quote
Question is, can you hear the difference?
I thought the question was: Extended bandwidth is a marketing argument for SACD's superior sound compared to CD, but do SACDs really contain energy above 22kHz in practice? (no listening involved here)
IME the main advantage of SACD is the multi-channel format. It's still the most convenient physical media for that, but physical media are becoming less popular.
Title: DSD frequency range
Post by: knutinh on 2011-09-08 13:08:08
Many professional, good-sounding microphones pass little above 20kHz.
True, but the roll-off is far less steep than a brick-wall filter, so most of them produce frequencies above 20kHz.

Yes. So the argument is that if 30kHz is unlikely to be audible for any human, attenuating it by e.g. 20-30dB will make it even less likely.

1. Audio source: most musical instruments have little energy over 20kHz. Cymbals do.
2. Air: HF is attenuated more than LF
3. Microphone: Many/most valued studio microphones start rolling of at 15 or 20kHz
... (various components that may attenuate sound above 20kHz)
5. Loudspeakers typically start to roll of at 20kHz
6. Carefully setup lab-experiments tend to fail to prove that we can hear anything above 20kHz, even when it is there

In sum:
If you are into good sound, there seems to be other things to spend calories on.

-k
Title: DSD frequency range
Post by: tahaa7 on 2011-09-08 13:41:40
3. Microphone: Many/most valued studio microphones start rolling of at 15 or 20kHz
... (various components that may attenuate sound above 20kHz)


I'm not sure about that, because vinyl records, for example, will have a lot of sound well beyond 20kHz

Quote
5. Loudspeakers typically start to roll of at 20kHz


That depends on the manufacturer and the quality, aka price. I've actually seen in person speakers that roll of at about 40-45 kHz...

Quote
6. Carefully setup lab-experiments tend to fail to prove that we can hear anything above 20kHz, even when it is there


Well I think we can definitely hear frequencies beyond 20kHz, but we may think we don't. If those high frequencies exist, they will impact the warmth of the sound, that is, they will make the sound less harsh and "digital" sounding. I have personally had experiences where I could tell the difference between the CD, which doesn't contain those frequencies, and vinyl, which does, every time.

Maybe we can't hear those frequencies individually, but our ears can still feel them, and they all add up to make the warm sound.
Title: DSD frequency range
Post by: db1989 on 2011-09-08 13:51:59
I'm not sure about that, because vinyl records, for example, will have a lot of sound well beyond 20kHz
Musical sound? Sensical sound? Anything besides worthless noise? I hope you’re not about to assert the superior fidelity of vinyl, because that’s not a ‘debate’ we need to have one more time.

Quote
That depends on the manufacturer and the quality, aka price. I've actually seen in person speakers that roll of at about 40-45 kHz...
Any spec-sheets you can link us to? And what would it matter if they could? Oh wait:

Quote
Well I think we can definitely hear frequencies beyond 20kHz, but we may think we don't. If those high frequencies exist, they will impact the warmth of the sound, that is, they will make the sound less harsh and "digital" sounding. I have personally had experiences where I could tell the difference between the CD, which doesn't contain those frequencies, and vinyl, which does, every time.

Maybe we can't hear those frequencies individually, but our ears can still feel them, and they all add up to make the warm sound.
Apparently you missed #8 of our Terms of Service (http://www.hydrogenaudio.org/forums/index.php?showtopic=3974), whose conditions you agreed to upon submitting your registration. Hydrogenaudio is not a forum for “feel[ings]”, sighted tests (I am assuming your hugely successful tests were not double-blind), or anything so subjective, unverifiable, and unrepeatable—especially when it’s couched in the same old (deliberately?) vague language of “warmth”, digital “harsh[ness], and the like”.
Title: DSD frequency range
Post by: andy o on 2011-09-08 13:58:56
IME the main advantage of SACD is the multi-channel format. It's still the most convenient physical media for that, but physical media are becoming less popular.

Huh? I always thought it to be the least convenient. What's its advantage over DVD-A or Blu-ray? Heck, even HD-DVD? Or if you're not considering only lossless formats, DVD-V or DTS CD? (All those formats can be easily read/played/converted from many more devices).
Title: DSD frequency range
Post by: tahaa7 on 2011-09-08 14:01:59
Any spec-sheets you can link us to?


Here's an example: http://www.kef.com/en/loudspeakers/xq/xq40 (http://www.kef.com/en/loudspeakers/xq/xq40). Notice the "Frequency response (+/- 3dB)" row in the table.

Quote
Apparently you missed #8 of our Terms of Service (http://www.hydrogenaudio.org/forums/index.php?showtopic=3974), whose conditions you agreed to upon submitting your registration. Hydrogenaudio is not a forum for “feel[ings]”, sighted tests (I am assuming your hugely successful tests were not double-blind), or anything so subjective, unverifiable, and unrepeatable—especially when it’s couched in the same old (deliberately?) vague language of “warmth”, digital “harsh[ness], and the like”.


OK, sorry, I won't be discussing those "vague" terms again. I was just saying what I thought and what I experienced regarding the benefit of high frequencies in a recording. And no, I won't be starting the CD vs vinyl debate again.

Edit: yes, as a matter of fact, the test were double-blind.
Title: DSD frequency range
Post by: db1989 on 2011-09-08 14:07:59
Edit: yes, as a matter of fact, the test were double-blind.
Wait, this was testing CD versus vinyl? Not a level playing field. I wager most people could detect the idiosyncratic signatures of vinyl with a minimum of fuss. This by no means signifies that vinyl is better at reproducing signals (i.e. attaining transparency) than is CDDA, even if that myth weren’t thoroughly debunked by simple science before it even got off the ground. All this tells us is that you preferred it, for whatever reason; perhaps you’re partial to its particular way of altering the sound of recorded material.
Title: DSD frequency range
Post by: tahaa7 on 2011-09-08 14:13:37
Edit: yes, as a matter of fact, the test were double-blind.
Wait, this was testing CD versus vinyl? Not a level playing field. I wager most people could detect the idiosyncratic signatures of vinyl with a minimum of fuss. This by no means signifies that it sounds better. Neeeext


No, actually most people claim they can't hear the difference between vinyl and CD whatsoever... The vinyls used in the test were all mint, therefore those " idiosyncratic signatures of vinyl" were barely present, if at all. But, as I said, I won't be starting that discussion. Besides, we've gone completely off-topic...
Title: DSD frequency range
Post by: Wombat on 2011-09-08 14:42:32
Well I think we can definitely hear frequencies beyond 20kHz, but we may think we don't. If those high frequencies exist, they will impact the warmth of the sound, that is, they will make the sound less harsh and "digital" sounding. I have personally had experiences where I could tell the difference between the CD, which doesn't contain those frequencies, and vinyl, which does, every time.

Maybe we can't hear those frequencies individually, but our ears can still feel them, and they all add up to make the warm sound.


Funny guy you are. So why do you bother watching some fancy frequency plots? Why don´t you simply listen how the highs of the DSD recording above 20kHz sound? Since you claim you can hear that why not simply listening?
Oh wait, DSD marketing says you can´t hear the noise that is louder as most music information up there but you can hear the extended frequency range... ROFL!
Title: DSD frequency range
Post by: krabapple on 2011-09-08 16:03:13
I did a research on DSD the other day and read about its alleged 50kHz frequency range.

Where did you read about a 50 kHz frequency range?

It's true that most of the added bandwidth of DSD  beyond Redbook is where noise lives.  It's also true that no one can hear anywhere near 50 kHz.  Human hearing tops out around ~24kHz at the absolute very best (in children and very rare adults).

It's also true that most SACD players have a post-DAC lowpass filter set to either 50kHz  or 100kHz.  This is to protect downstream hardware from that ultrahigh-frequency junk.

It's also true that SACDs (and DVD-As) may be sourced from bandwidth-limited digital masters, and therefore contain no frequencies higher than the original sample rate allowed -- they cannot take advantage of the supposed benefits of MHz sample rates.  Stereophile did an article on such releases some years ago, with several examples of steep rolloffs occuring well before the bandwidth limit allowed by the 'hi rez' format.



Edit: gonna do another experiment tomorrow.  By chance I have a vinyl record of an album I found a DSD sample of, so I'm gonna rip it to 24/96 (or 192) and see whether those high frequencies exist there. If not, it's possible that the original master never contained them as well, and therefore the DSD track doesn't... Looking forward to testing this.

High frequencies you see on a vinyl rip may have nothing to do with what's on the master tape.



Edit: yes, as a matter of fact, the test were double-blind.
Wait, this was testing CD versus vinyl? Not a level playing field. I wager most people could detect the idiosyncratic signatures of vinyl with a minimum of fuss. This by no means signifies that it sounds better. Neeeext
No, actually most people claim they can't hear the difference between vinyl and CD whatsoever... The vinyls used in the test were all mint, therefore those " idiosyncratic signatures of vinyl" were barely present, if at all. But, as I said, I won't be starting that discussion.
Besides, we've gone completely off-topic...

If you're saying you carefully digitized a mint LP, then set up a level-matched randomized DBT to determine if you could identify the LP source versus its digital copy, and correctly identified them ~14 out of 16 times, by all means please provide as much detail as you can on the protocol and analysis of outcome.  This would be rather groundbreaking news if true.



IME the main advantage of SACD is the multi-channel format. It's still the most convenient physical media for that, but physical media are becoming less popular.

In terms of popularity the most convenient physical medium for multichannel, by far, is DVD-V.  5.1 AC3 and DTS are common, well-developed lossy perceptual codecs and DVD players that can either decode them, or pass them to an AVR that decodes them, are ubiquitous.

SACD and DVD-A physical media both require special hardware, though at least now both can be passed digitally to an AVR for decoding (something not available when they were both being pushed as the next great thing).  SACD basically still cannot be archived digitally by consumers for networked playback, whereas DVD-A can.
Title: DSD frequency range
Post by: saratoga on 2011-09-08 16:38:23
Simply passing through air, high frequencies will be attenuated more than mid-range. Only some instruments emit considerable energy above 20kHz. Many professional, good-sounding microphones pass little above 20kHz.


Really?  The microphones I've used generally pass out to 25k and beyond with relatively little attenuation, and I've measured 45kHz quite easily with a commercial mic.  They also don't generally have a brick wall drop off, but rather a graduall roll off as the transducer becomes less able to respond. 

How did you measure?  It sounds like you had a low pass filter in there somewhere if you got a 20kHz brickwall.
Title: DSD frequency range
Post by: tahaa7 on 2011-09-08 19:50:35
If you're saying you carefully digitized a mint LP, then set up a level-matched randomized DBT to determine if you could identify the LP source versus its digital copy, and correctly identified them ~14 out of 16 times, by all means please provide as much detail as you can on the protocol and analysis of outcome.  This would be rather groundbreaking news if true.


Well, it's hard to explain how, but I can identify an LP-sourced file over a digital-sourced file about 80 or more percent of the time. Even if the digital source is well mastered (which usually, but not always, isn't because they compress the dynamic range in order to make it louder).
Title: DSD frequency range
Post by: Wombat on 2011-09-08 20:46:43
If you're saying you carefully digitized a mint LP, then set up a level-matched randomized DBT to determine if you could identify the LP source versus its digital copy, and correctly identified them ~14 out of 16 times, by all means please provide as much detail as you can on the protocol and analysis of outcome.  This would be rather groundbreaking news if true.


Well, it's hard to explain how, but I can identify an LP-sourced file over a digital-sourced file about 80 or more percent of the time. Even if the digital source is well mastered (which usually, but not always, isn't because they compress the dynamic range in order to make it louder).

I think you talk about some LP against some other digital master. This is of no use and besides that is nothing special to hear a difference. krabapple did ask "the LP source versus its digital copy"
Title: DSD frequency range
Post by: db1989 on 2011-09-09 09:05:02
That seems to have been the setup. I tried earlier to convey the same sentiment as yourself (not to blow my own horn, but just because it may have been missed due to the editing spree that beget it):
Wait, this was testing CD versus vinyl? Not a level playing field. I wager most people could detect the idiosyncratic signatures of vinyl with a minimum of fuss. This by no means signifies that vinyl is better at reproducing signals (i.e. attaining transparency) than is CDDA, even if that myth weren’t thoroughly debunked by simple science before it even got off the ground. All this tells us is that you preferred it, for whatever reason; perhaps you’re partial to its particular way of altering the sound of recorded material.

Then again, validity or non-validity of the comparison notwithstanding, I’m confused about how confident tahaa7’s conclusion really is regarding such neurologically irrelevant frequencies:
I have personally had experiences where I could tell the difference between the CD, which doesn't contain those frequencies, and vinyl, which does, every time.

Maybe we can't hear those frequencies individually, but our ears can still feel them, and they all add up to make the warm sound.
Title: DSD frequency range
Post by: greynol on 2011-09-09 09:22:59
Not exactly in keeping with TOS #8.

@tahaa7:
Claims such as the one quoted require objective evidence if they are to be made on this forum.
At this point in time, we require that you provide some objective evidence in keeping with the rules and spirit of this forum or withdraw the claim, otherwise it goes in the recycle bin; your choice.
Title: DSD frequency range
Post by: botface on 2011-09-09 10:10:06
What's to get excited about here? Why is it OK for DB1989 to say "I wager most people could detect the idiosyncratic signatures of vinyl with a minimum of fuss." but tahaa7 gets stomped on for saying "I can identify an LP-sourced file over a digital-sourced file" Aren't they both saying the same thing? Is it the difference between "wager" and "can"?
Title: DSD frequency range
Post by: greynol on 2011-09-09 10:34:04
for me, the issue at hand is the statement, "Maybe we can't hear those frequencies individually, but our ears can still feel them, and they all add up to make the warm sound."  However, I find the use of the term "every time" in the previous statement just as troubling, though perhaps an indication of a systemic problem with the vinyl and/or playback hardware if it turns out not to be placebo-based.

This is no longer up for public debate.  Further posts regarding moderation decisions will be removed and warnings will be issued.
Title: DSD frequency range
Post by: db1989 on 2011-09-09 10:45:02
What's to get excited about here? Why is it OK for DB1989 to say "I wager most people could detect the idiosyncratic signatures of vinyl with a minimum of fuss." but tahaa7 gets stomped on for saying "I can identify an LP-sourced file over a digital-sourced file" Aren't they both saying the same thing? Is it the difference between "wager" and "can"?
My statement that you quoted was in response to tahaa7’s: what I meant was that being able to identify vinyl over CD, as tahaa7 asserts, is nothing spectacular if the vinyl/needle is worn (e.g. one can hear crackles), there is some other quirk of the playback process, the sources aren’t level-matched, etc.—and simply being able to discern between the two neither implies the superiority of either one nor necessarily leads to the conclusion that the vinyl version possesses important ultrasonic frequencies (whether heard or ‘felt’).
Title: DSD frequency range
Post by: botface on 2011-09-09 16:07:59
Thanks for the clarification guys
Title: DSD frequency range
Post by: krabapple on 2011-09-09 16:55:58
SACD and DVD-A physical media both require special hardware, though at least now both can be passed digitally to an AVR for decoding (something not available when they were both being pushed as the next great thing).  SACD basically still cannot be archived digitally by consumers for networked playback, whereas DVD-A can.


Self-correction: there were some proprietary connections available (e.g. Pioneer's i-link, Denon's Denonlink) -- I even had a Pioneer setup that could do this.  But again, special hardware...  today, HDMI can handle all of them.
Title: DSD frequency range
Post by: krabapple on 2011-09-09 17:00:28
What's to get excited about here? Why is it OK for DB1989 to say "I wager most people could detect the idiosyncratic signatures of vinyl with a minimum of fuss." but tahaa7 gets stomped on for saying "I can identify an LP-sourced file over a digital-sourced file" Aren't they both saying the same thing? Is it the difference between "wager" and "can"?
My statement that you quoted was in response to tahaa7’s: what I meant was that being able to identify vinyl over CD, as tahaa7 asserts, is nothing spectacular if the vinyl/needle is worn (e.g. one can hear crackles), there is some other quirk of the playback process, the sources aren’t level-matched, etc.—and simply being able to discern between the two neither implies the superiority of either one nor necessarily leads to the conclusion that the vinyl version possesses important ultrasonic frequencies (whether heard or ‘felt’).



And, as Wombat has mentioned, LP mastering is often different enough from CD mastering to be likely to be audible.  There isn't anything special about being able to tell LPs from CDs.  There is something special about being able to tell LPs from digital copies of LPs.
Title: DSD frequency range
Post by: knutinh on 2011-09-09 22:16:36
Well I think we can definitely hear frequencies beyond 20kHz, but we may think we don't.

What makes you think so? Perhaps it is only you who think that you do because you like the idea of being able to?
Quote
If those high frequencies exist, they will impact the warmth of the sound, that is, they will make the sound less harsh and "digital" sounding.

What makes you think so? Perhaps added high frequencies will make the sound more digital sounding?
Quote
I have personally had experiences where I could tell the difference between the CD, which doesn't contain those frequencies, and vinyl, which does, every time.

1. Please provide evidence that vinyl generally have more precise high-frequency rendition than CD.
2. A vinyl pressing and a CD pressing are not suitable for comparing the quality of CD vs vinyl due to mastering.
Quote
Maybe we can't hear those frequencies individually, but our ears can still feel them, and they all add up to make the warm sound.

Sounds like audiophile nonsense.

-k