• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
Try the undocumented "rate -u" which has higher stopband rejection than -v. Yes, I ought to fix the manual.
But the -u flag still doesn't let me use -b less than 74, but even if it can, I suppose I still cannot have flat response up to 20kHz and suppress noise above 30kHz at the same time.
 
Last edited:

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
But the -u flag still doesn't let me use -b less than 74...
Oh, I see what you're trying to do. If sinc removes the noise adequately, you can use "downsample" to drop the unwanted samples. Alternatively, rate followed by sinc will be faster than the other way around.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
The foobar DSDIFF plugin in that article is no longer maintained and the author (kode54) removed it from the foobar website already. Here is another plugin:
https://sourceforge.net/projects/sacddecoder/files/foo_input_sacd/

I took a 32-bit float 352.8kHz RMAA reference file, converted to DSD64 and back to 32-bit float 352.8kHz again with different settings. Multistage and the bundled FIR filters, except the 30kHz one, created too much ultrasonic noise and RMAA is unable to analyze them, so I only analyzed what RMAA can analyze, with an additional 1023 points filter I made (attached). Should be able to avoid further frequency response/ultrasonic noise excuses.
View attachment 84059
View attachment 84060
A reminder for those who downloaded my filter file without reply as I cannot edit my original post. My filter and the bundled 30kHz filter should not be used for more than 32x decimation. That means if you want to convert a DSD64 file to PCM, choose at least 88.2k output. For DSD128, choose at least 176.4kHz. Use Multistage for 64x and higher decimation ratio.

Other bundled filters (40-60kHz) should be used with DSD128 files or above, not DSD64. Again, don't use them for more than 32x decimation.
 

bogi

Active Member
Joined
Sep 16, 2020
Messages
238
Likes
170
Location
Slovakia
Hi mansr, could you please describe recommended usage of DSD conversions (PCM->DSD, DSD->DSD, DSD->PCM) with all the available options regarding conversion quality and filtering? I would be interested for 2 scenarios, a) optimal for speeed/quality and b) the best quality regardless on how much time it will take.
 
Joined
Aug 24, 2020
Messages
86
Likes
47
I've always wanted to see how the waveform (ie time domain) differs between 44.1/16 and DSD, on a thing that has a lot and huge micro dynamics (eg drums, percussions)

There is a huge difference. You have probably seen reports on CD players. Even here, Philips/Sony decided to "cheat" - by giving the frequency of 400 Hz in almost all test CDs - because response to 1 kHz, the standard frequency used troughout the history of analog, really looks gaudy on 44.1 kHz sampling machines - no matter what the filtering. Only Stereophille did issue a test CD with 1 kHz square wave test signal.
Depending how PCM machine works, it CAN introduce interchannel lag - or phase distortion between the channels. There is no way one could test all the hardware and software combinations for this anomally - but, in real life , it could - and still can - rear its ugly head.

DSD is an entirely different animal. Unfiltered, it creates PERFECT square wave - no ringing whatsoever, rise time always faster than PCM of comparable file size. By default, it can not introduce interchanel lag/phase distortion between the channels.
But, DSD shifts it quantization etc noise above 20 kHz - an it is BAD. Only the square wave recorded at 0 dBFS in DSD128 looks "reasonably noise free" on the scope if left totally unfiltered. DSD64 - even at 0 dBFS - looks quite hairy/noisy. At say -30 dBFS, square wave OF dsd64 is no longer discernible from the ultrasonic noise - only by triggering properly noise modulated by 1 kHz square wave can be seen.
The higher the order of DSD, the less noise above 20 kHz - for each increase of sampling frequency, there is 12dB decrease of > 20 kHz noise. That is the reason behind DSD256 - 24 dB less ultrasonic noise than DSD64 aka SACD.
The actual square wave response is determined by the filtering the DSD output - and, as usual, lower slope filters produce better square waves than higher slopes, whereas for the amount of >20kHz noise it is exactly opposite. But, even filtered, it still has the upper hand over any comparable sized PCM file.

Although I have not yet seen a square wave recorded from an analog generator onto any digital gizmo with sampling beyond PCM 192 and DSD128, square wave of PCM/DXD at 384 ( or even twice that ) should come quite close to that of reasonably filtered DSD128. There are FAR too many options/settings in filtering available on already existing machines - but, suffice to say, taken as a group, they make mockery of whatever is possible in 44.1 sampling.

Percussion does go beyond 100 kHz - the most representative is gamelan from Bali. It has to be experienced live, as equipment that could do it full justice still is not available. Up to 100 kHz - yes, beyond not yet.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
You have probably seen reports on CD players.
You are still using early CD players? Then no surprise they can perform worse than some random DACs nowadays. Stop moving the goalpost, it is a topic about PCM vs DSD.
DSD is an entirely different animal. Unfiltered, it creates PERFECT square wave
Of course, unfiltered DSD only has two possible values, so this claim is meaningless.
Percussion does go beyond 100 kHz
If it is the case, DSD512 is necessary to store the signal before noise screwing it up, for PCM you only need 352.8kHz.
 
Joined
Aug 24, 2020
Messages
86
Likes
47
You are still using early CD players? Then no surprise they can perform worse than some random DACs nowadays. Stop moving the goalpost, it is a topic about PCM vs DSD.

Of course, unfiltered DSD only has two possible values, so this claim is meaningless.

If it is the case, DSD512 is necessary to store the signal before noise screwing it up, for PCM you only need 352.8kHz.
Although PCM did improve over early CD players, 44.1 kHz sampling can only go so and so far - even assuming "perfect" DAC. It is not and can not be good enough for music - ever.

You are correct in the other 2 claims.

The second one only partially. PCM 352.8 kHz still needs "almost brickwall" filtering if it wants to have reasonably flat response to say 150 kHz. Please do check the Archimago's report on RME ADI2 PRO (FS?) - where he has been able to verify all the official specs claims. You will see the brickwall filtering in 352.8 - but a FAR more relaxed filtering for the highest sampling frequency at 768kHz.

Although RME unit does look appealing, it is downright apalling how badly its creator(s) treat DSD. Apearently, DSD is only an afterthought and another box ticked as a yet an additional sales point for them. In their official literature, they claim that - among others - Korg Audiogate software can be used for recording and playback. They have written that at face value - and from that, it is clear they never actually tried using their product(s) with Korg Audiogate software. At least not in DSD mode - Korg Audiogate software limits both playback and recording for non - Korg hardware to PCM only. It is not possible ( save for hacking...) to install any other than Korg driver for DSD.
Failing such big issue - what else they are not telling us regarding the use of DSD on their flagship ?

Please note that DAC only machines - without ADC ( recorder ) are of completely no interest to me.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
The second one only partially. PCM 352.8 kHz still needs "almost brickwall" filtering if it wants to have reasonably flat response to say 150 kHz. Please do check the Archimago's report on RME ADI2 PRO (FS?) - where he has been able to verify all the official specs claims. You will see the brickwall filtering in 352.8 - but a FAR more relaxed filtering for the highest sampling frequency at 768kHz.
Guess you didn't read this, see the noise spectrum of the FFT plots:
https://archimago.blogspot.com/2019/03/measurements-look-at-audio-ultra-high.html
Keep in mind that the AK557x ADCs are based on multibit delta-sigma, noise floor above 100kHz is already pretty high. So the performance is already limited by the hardware itself rather than the format. You can see he also used a cheap scope to look further into the MHz range. Guess why he didn't simply use DSD on the RME? Since in MHz range the RME is already flooded by modulator noise, a low bit-depth scope can actually do better in this aspect. So your argument about brickwall filter here has no meaning.
 
Joined
Aug 24, 2020
Messages
86
Likes
47
Guess you didn't read this, see the noise spectrum of the FFT plots:
https://archimago.blogspot.com/2019/03/measurements-look-at-audio-ultra-high.html
Keep in mind that the AK557x ADCs are based on multibit delta-sigma, noise floor above 100kHz is already pretty high. So the performance is already limited by the hardware itself rather than the format. You can see he also used a cheap scope to look further into the MHz range. Guess why he didn't simply use DSD on the RME? Since in MHz range the RME is already flooded by modulator noise, a low bit-depth scope can actually do better in this aspect. So your argument about brickwall filter here has no meaning.
No, I did not read this yet - thank you for pointing it out for me.

But anyway, this confirms there still is no equipment out there (that is commercially available) that could handle DSD above 100 kHz with decent enough dynamic range.

The levels >20 kHz are LOW - and those beyond 50kHz are lower still. Currently, there are two commercially available microphones meant for music recording capable of 100kHz response - Sanken C100K and Audio Precision
https://www.audioprecision.pl/pdf/audio-precision-376M03-mikrofon-eng.pdf
Not many ( if any - besides Telarc using Sanken ) music recordings have been done and released so far in a format that would do them justice. Hence, we can not know how loud are frequencies above approx 50 kHz - and that includes Archimago. Yet, he is constantly trying to downplay the importance of ultrasonics. The only relevant source at the time is https://www.cco.caltech.edu/~boyk/spectra/spectra.htm - where measuring microphones have been pressed into service of obtaining the required data. The measuring mics generally are optimized for as flat frequency response as possible - and dynamic range is rather limited. The same goes for the ADC/DAC used .

Only when the 100 kHz will no longer represent technological "sound barrier", equipment gets adopted at larger scale troughout the industry, etc - will we have the definitive answer. That will take at least decade to become "commonplace esoterica" - at present, such performance is still around the corner.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
Only when the 100 kHz will no longer represent technological "sound barrier"
As long as you don't shift the high frequencies to the audible band, the barrier is on the human race. Genetic engineering is far more practical, though could be unethical. Why there is a need to measure and verify these stuff if one can hear >100kHz (and of course, at low level) without assistance of these equipment?
 

Killingbeans

Major Contributor
Joined
Oct 23, 2018
Messages
4,089
Likes
7,547
Location
Bjerringbro, Denmark.
Percussion does go beyond 100 kHz - the most representative is gamelan from Bali. It has to be experienced live, as equipment that could do it full justice still is not available. Up to 100 kHz - yes, beyond not yet.

Isn't it only the local bats that get the extra satisfaction from the live performance?
 
Joined
Aug 24, 2020
Messages
86
Likes
47
As long as you don't shift the high frequencies to the audible band, the barrier is on the human race. Genetic engineering is far more practical, though could be unethical. Why there is a need to measure and verify these stuff if one can hear >100kHz (and of course, at low level) without assistance of these equipment?
The information contained in ultrasonics has yet to be properly evaluated. But it goes primarily about sound localization, as it does give pretty solid idea about the depth of image - through filtering of high frequencies that occur naturally in air, we can grasp pretty accurately how distant any given sound source is. That depth perception is the first thing to suffer whenever the bandwidth of a recording is reduced.
All that any normal person, without shifting frequencies down or genetic engineering can PERCEIVE - not necessary hear .
It is that thin line dividing "recording" from "real". Musical genres that do provide such content and also require it during reproduction are primarily acoustic. Classical, ethno, jazz - and similar. Anything that requires amplification much less so.
 
Joined
Aug 24, 2020
Messages
86
Likes
47
Isn't it only the local bats that get the extra satisfaction from the live performance?
No.
Try any half decent brass band live - and, immediately after returning home, any brass band recording on CD.
( which made me immediately sad - no such opportunity to listen to a brass band parade or similar at the moment live, in most parts of the world... )
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
The information contained in ultrasonics has yet to be properly evaluated. But it goes primarily about sound localization, as it does give pretty solid idea about the depth of image - through filtering of high frequencies that occur naturally in air, we can grasp pretty accurately how distant any given sound source is. That depth perception is the first thing to suffer whenever the bandwidth of a recording is reduced.
All that any normal person, without shifting frequencies down or genetic engineering can PERCEIVE - not necessary hear .
It is that thin line dividing "recording" from "real". Musical genres that do provide such content and also require it during reproduction are primarily acoustic. Classical, ethno, jazz - and similar. Anything that requires amplification much less so.
Unless you can explain how one perceives ultrasound then your suppositions about it being part of depth perception are ridiculous.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
The information contained in ultrasonics has yet to be properly evaluated. But it goes primarily about sound localization, as it does give pretty solid idea about the depth of image - through filtering of high frequencies that occur naturally in air, we can grasp pretty accurately how distant any given sound source is. That depth perception is the first thing to suffer whenever the bandwidth of a recording is reduced.
All that any normal person, without shifting frequencies down or genetic engineering can PERCEIVE - not necessary hear .
It is that thin line dividing "recording" from "real". Musical genres that do provide such content and also require it during reproduction are primarily acoustic. Classical, ethno, jazz - and similar. Anything that requires amplification much less so.
More than one time I reacted to some interesting or unusual sound when I was on the street. When I heard those sound, I turned my head to what I thought the sound was, but I turned my head to the wrong position. No audio equipment was involved. Actually, I react more correctly in artificially positioned sound, like game or movie sound effects. No hi-res is involved.
 

Killingbeans

Major Contributor
Joined
Oct 23, 2018
Messages
4,089
Likes
7,547
Location
Bjerringbro, Denmark.
But it goes primarily about sound localization, as it does give pretty solid idea about the depth of image - through filtering of high frequencies that occur naturally in air, we can grasp pretty accurately how distant any given sound source is. That depth perception is the first thing to suffer whenever the bandwidth of a recording is reduced.

Do those conclusions come from any peer reviewed studies? For a long time I've been under the impression that super tweeters and ultrasonic information on recordings were complete bollocks, but if you have any solid evidence to suggest otherwise, I'll gladly change my perspective.

Try any half decent brass band live - and, immediately after returning home, any brass band recording on CD.

Of course it's not the same. But what convinces you that utrasonics holds any merit over all the other well known shortcomings?
 

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
3,334
Likes
3,278
Location
.de
Try any half decent brass band live - and, immediately after returning home, any brass band recording on CD.
( which made me immediately sad - no such opportunity to listen to a brass band parade or similar at the moment live, in most parts of the world... )
What leads you to believe that this has anything to do with ultrasonics though? There are about a million issues coming in before we even get to this point - trying to capture an uneven radiation pattern with microphones and then squeezing everything into two channels to be played back by two speakers placed in front of the listener poses a great number of challenges and is by no means lossless, not to mention dependent on playback system quality to boot. Seems you are worrying about a fly while ignoring the elephant in the room.
 
Joined
Aug 24, 2020
Messages
86
Likes
47
I was wondering why @AnalogSteph went off his rocker all of a sudden, but it was actually just @analogSurviver instead. ;)
For the life of me, I do not have the faintest idea who AnalogSteph might be. Never even came across that name - or can not remember. Will have to google that member. I go under the same name ( except mostly as analogsurviver, with all small characters ) - on quite a few forums, with the sole exception of Audiokarma, which for some strange inexplicable reason did not allow that name - forcing me to sail under the analogSVR banner.

I will have to find my avatar - recent digital mumbo jumbo with khmomputers made a bit of a mess and less important data such as avatars is not yet in its proper normal space.
 
Top Bottom