Jump to content
IGNORED

The Ear - DSD under Fire


Nikhil

Recommended Posts

 

Interesting in itself, that he seems to have changed his mind a bit in past 10+ years after that post. Since he know has a 1-bit DAC that he pretty much claims to be better than anything else on the market.

 

Have some fun reading some newer stuff:

Blog!

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
I believe Schiit claims they solved the problem by using a closed-form filter on their expensive yggy and gungnir-multibit version.

 

PCM manufacturers have been claiming it for the last 30+ years. And for the same period people have been changing their DACs, and switching from the new filter du jour to NOS playback. Not to mention that you can't remove ringing that was already introduced during downsampling.

 

DSD deals with the problem wholesale (no special filters required) - you simply get the raw output of the delta sigma modulator and send it to the DSD DAC where it's converted to analog at the original sampling rate.

Link to comment
Hi Nikhil. Charles Hansen and DSD is old news. There's a huge couple year old thread on the forums here that's about 50% comprised of his often very heated criticisms of it.

 

So is Putzeys' specific criticism, which has often been quoted in various forums as this argument lumbers onward; and then someone will point out his quote about where it's a good thing, plus the fact he designs A/D converters that use SDM.

 

But I'm sure this will be good for additional discussion. :)

 

Hi Jud,

 

Definitely some good discussion on here so far. Like others I hear what I hear on my setup and I agree that (native) DSD is just special. What I can't understand is that there is significant resistance to DSD from some very technically qualified folks including my own DAC designer here in India who refuses to get on to DSD.

 

The whole issue of SACD and Sony seems to keep this issue hanging. If Sony were to get behind DSD in a big way then there would be some traction in my opinion but the long silence on DSD by Sony is a little perplexing.

 

Regards.

Custom Win10 Server | Mutec MC-3+ USB | Lampizator Amber | Job INT | ATC SCM20PSL + JL Audio E-Sub e110

 

 

Link to comment
The whole issue of SACD and Sony seems to keep this issue hanging. If Sony were to get behind DSD in a big way then there would be some traction in my opinion but the long silence on DSD by Sony is a little perplexing.

 

I believe that PCM manufacturers hell-bent on defending their market of expensive PCM filters are to blame. If SDM direct audio path were pursued more aggressively we would have more editing tools for DSD, and even 5bit DSD DACs by now, which are childishly easy to build.

 

BTW, Sony has just unveiled a DSD128-ready car stereo at IFA 2015...

Link to comment
In your dreams :)

 

If you take the effort to read point 2, you'll see that the comparison was between a 192KHz PCM file, and that same file then converted to DSD. Well duh....

 

I record acoustic music in all three DSD bit rates (now almost exclusively in 256fs) using a Pyramix system. For those recordings recorded with the Grimm AD1 at 64fs as the A/D converter, it's possible with Pyramix to mount the DSD file in a DXD project, and monitor the playout in DXD with desired level changes. Then once the track(s) level changes have been determined monitoring in DXD, the original DSD file can be re-modulated with the desired changes producing a new DSD file, without going through DXD conversion.

 

The point of this is after DSD rendering the new changed DSD file, it's simple to then have a DXD Project alternating with a DSD Project with only a second of switching time to compare the two. They're level matched DXD and DSD playouts of the same file, and it's easy to hear the differences between the two. There's a loss of spaciousness and aliveness of instrument detail, the very low level information that's affected by the DSD > DXD conversion.

 

if you haven't worked and experienced this yourself, I guess all you can do is postulate theories of how stuff should work.

 

Does this allow you to digitally remix 1 bit streams without having to do additional analog conversions with decimation to DXD?

Link to comment
Hi Jud,

 

The whole issue of SACD and Sony seems to keep this issue hanging. If Sony were to get behind DSD in a big way then there would be some traction in my opinion but the long silence on DSD by Sony is a little perplexing.

 

Regards.

 

Hi Nikhil. Not too perplexing, I think what Sony really wanted to see was SACD take off with Sony's proprietary hardware encoding/decoding as part of it. (This seems to me to play a significant role in Charles Hansen's and others' antipathy to DSD.) Sony is only now beginning to come off that mindset and very tentatively "stick their toes in the water" with regard to commercialization of DSD files vs. SACD discs.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment
PCM manufacturers have been claiming it for the last 30+ years. And for the same period people have been changing their DACs, and switching from the new filter du jour to NOS playback. Not to mention that you can't remove ringing that was already introduced during downsampling.

 

DSD deals with the problem wholesale (no special filters required) - you simply get the raw output of the delta sigma modulator and send it to the DSD DAC where it's converted to analog at the original sampling rate.

 

Non-ringing low pass filters are possible, but they roll-off all frequencies somewhat, including frequencies that one would like to preserve. The best possible non-ringing filter is a Gaussian filter. If used in such a way that it eliminates aliases then there will be significant audio roll-off at 20 kHz even if the sampling rate is 384 kHz. Doubling the sample rate will reduce the roll-off (measured in dB) by a factor of four, so a sampling rate of 768Khz will probably be OK as far as ringing is concerned. There will be no problem with ringing at DSD sampling rates or above.

Link to comment
PCM manufacturers have been claiming it for the last 30+ years. And for the same period people have been changing their DACs, and switching from the new filter du jour to NOS playback. Not to mention that you can't remove ringing that was already introduced during downsampling.

 

DSD deals with the problem wholesale (no special filters required) - you simply get the raw output of the delta sigma modulator and send it to the DSD DAC where it's converted to analog at the original sampling rate.

 

As far as I know, no one ever claimed they use closed-form filter. So far there are many types of digital filters for PCM.. some uses extreme oversampling, some goes for NOS while others use different ways that 'shape' the ringing. Schiit's claim is definitely not a same claim for the last thirty years.

Link to comment
Does this allow you to digitally remix 1 bit streams without having to do additional analog conversions with decimation to DXD?

 

Hi Tony,

 

In Pyramix, I don't know, for I always track a microphone per channel, therefore never mix in my DSD projects. But recording and post processing in 64fs you can change levels, but I don't think mix channels, without converting to DXD when creating an edited master; selecting DSD Render. The original Philips modulator is still in Pyramix for this level changing function.

 

In Sonoma (with the mixer card) and SADiE you can mix, level change, and apply many other effects processes at 64fs. These devices all use the Sony E-Chip, as did several SACD players for realtime DSD processing at 64fs.

 

The Sony based Sonoma/SADiE and Philips based Pyramix are fundamentally different approaches to mixing and level changes. However with either Digital Audio Workstation (DAW), you create an Edit Decision List (EDL) over the entire sequential post processing effort, without affecting the original DSD file. You build additional PCM monitoring files while adding instructions to the EDL. Once all edits and sweetening have been decided, a one time implementation of the EDL instruction list configuring the modulator(s) or DXD conversion is executed on the original file(s) when creating the edited master. In Sonoma/SADiE, all of these functions are performed in one pass through the modulators. In Pyramix, except in simple level change case on 64fs only dff data, like the other DAW's, the entire accumulated EDL is performed only once at the edited master creation stage. Therefore only one DSD > DXD > DSD conversion regardless of the post processing complexity.

Link to comment
Has anyone tried to prove what DSP operations are and aren't mathematically possible on DSD in a reversible matter, in comparison to PCM? I mean, either they are or they aren't, the math surely says either "yes" or "no".

 

It's not a subject of much interest, I would imagine, because there is hardly an operation of any consequence on either one that is mathematically reversible. That is a question totally aside from the one I'm concerned about, which is how to get the sound most faithful to the event or the artist's vision regardless of format.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment
It's not a subject of much interest, I would imagine, because there is hardly an operation of any consequence on either one that is mathematically reversible. That is a question totally aside from the one I'm concerned about, which is how to get the sound most faithful to the event or the artist's vision regardless of format.

 

I relaxed my original question after I posted... I guess the issue of "reversible" isn't so interesting, what's more relevant is what DSP operations are even possible in DSD without conversion to PCM/analog realm. Surely that involves interesting and very relevant mathematical questions. It seems people keep hoping for holy-grail software to do such things, but the math could tell us if we hope in vain.

Link to comment
Has anyone tried to prove what DSP operations are and aren't mathematically possible on DSD? Or if possible, are they reversible in cases where the corresponding operation in PCM is reversible? I mean, either they are or they aren't in both cases, the math surely says either "yes" or "no".

 

You can easily prove that two PCM signals can be mixed losslessly, provided there is sufficient headroom to avoid clipping. You add a known signal to the original to get the mix. Then you subtract the known signal from the mix and get back the original. The arithmetic is done sample by sample. The theory works in practice with any audio editor. It will not be possible to do the same thing with DSD, because when you add the +1's and -1's you get other values, namely -2, 0 and +2. To get back a one bit signal you have to remodulate the bit stream and this necessarily adds noise.

 

If you do a gain change with PCM it will be almost transparent, again assuming there is no overload. The error in the final conversion will be +- 2 least significant bit after adding in dither noise. (The RMS error will be about -136 dBfs.) Again, this theory can be demonstrated with any audio editor. With DSD, there will have to be a remodulation. This will add new modulation noise to the signal, and this will be at a much higher level than -136 dB in the relevant portions of the audio band. It seems likely that with DSD256 there will be enough of a separation from the audioband and the noise that the extra noise can be safely filtered out.

 

Sample rate conversions using PCM can be done losslessly, subject to dither noise, provided there aren't headroom problems and provided that the PCM original file has no energy at Fs/2. (If it does, the amount of headroom required can be infinite with pathological signals, plus it will not be possible to resolve what the DC level of the original signal was, etc...) The theory here also plays out in practice when using commercially available software. With DSD, sample rate conversion issues involve remodulation and this will add additional noise. Again, with high enough rate DSD signals, the noise can be effectively filtered out. This is not the case with DSD64 because the noise begins in frequency ranges that overlap signal energy from musical instruments.

Link to comment
I relaxed my original question after I posted... I guess the issue of "reversible" isn't so interesting, what's more relevant is what DSP operations are even possible in DSD without conversion to PCM/analog realm. Surely that involves interesting and very relevant mathematical questions. It seems people keep hoping for holy-grail software to do such things, but the math could tell us if we hope in vain.

 

DSD is mathematically a form of PCM. DSD64 is 2822.4/1 PCM. "DSD" is a Sony trademark.

 

The theoretical benefits of DSD come from its high sampling rate. The practical benefits of DSD include the high sampling rate, low bit budget for a given sampling rate, and the use of a single switching element which makes low-level linearity possible without precise tolerances on physical components.

Link to comment
DSD = SDM/PDM not PCM.

 

The fact that most PCM converters also use Delta Sigma Modulation and noise shaping makes them more DSD-like than the other way around.

 

Your first comment is incorrect. The mathematical analysis of DSD signals is identical to that for other PCM formats. The signal consists of a set of coded pulses (impulses) at the sampling rate, with the coding using a mapping between the bit(s) assigned to each sample which controls the amplitude of the pulse. The resulting sequence of pulses is then low pass filtered to produce the desired continuous analog signal. The difference in playback between 1 bit and multibit formats concerns the particular mapping between the coded bits and the amplitude of the resulting impulses. However, this is true as well for other forms of PCM, e.g. different rounding methods and ranges are used, e.g. 2's complement vs. 1's complement arithmetic.

 

SDM is a mechanism for performing noise shaping. It is not part of the definition of a format. Indeed, it is not necessary to use a traditional sigma delta modulator to generate signals that can be played back by a DSD DAC.

 

The essential point of this discussion concerns the difference between "function" and "mechanism". The format of DSD as described mathematically defines the function provided by an ADC or a DAC. A sigma delta modulator is one mechanism for generating a one bit signal that will decode back into a suitable analog signal.

 

Your comment about PCM converters also confuses function with mechanism.

Link to comment
Your first comment is incorrect.

 

If I haven't made myself clear... please read Tailspn's post below:

 

By definition, PCM is digitized samples represented as absolute 2's complement binary word VALUES. There are no represented absolute VALUES in DSD (1-bit two level Pulse Density Modulation), only change of relative values represented through the density of bits in the bit stream.

 

As for...

 

SDM is a mechanism for performing noise shaping. It is not part of the definition of a format.

 

The inventor of DSD, Prof. Yoshio Yamasaki calls the format Delta Sigma Direct, so Delta Sigma is definitely a fundamental part of it.

Link to comment
Sample rate conversions using PCM can be done losslessly, subject to dither noise, provided there aren't headroom problems and provided that the PCM original file has no energy at Fs/2. (If it does, the amount of headroom required can be infinite with pathological signals, plus it will not be possible to resolve what the DC level of the original signal was, etc...)

 

I think mixing math and practicality here can be confusing. When you use the term "losslessly" above, you are no longer speaking in terms of a mathematical operation to perfectly reverse the conversion and obtain the original samples.

 

Along analogous lines, it is never the case that a real world signal can be perfectly bandlimited, but it may be close enough for practical purposes.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment
I think mixing math and practicality here can be confusing. When you use the term "losslessly" above, you are no longer speaking in terms of a mathematical operation to perfectly reverse the conversion and obtain the original samples.

 

Along analogous lines, it is never the case that a real world signal can be perfectly bandlimited, but it may be close enough for practical purposes.

 

The problem here is that there are people in this thread who are approaching the problem from different perspectives.

 

It is possible to characterize the error bounds in PCM sample rate conversions and gain changing operations precisely and when done digitally in an audio editor theory and practice can coincide. The essential nature of binary encoded PCM is that its steady state and instantaneous resolution increases exponentially with the bit depth. This makes it economical to treat these operations on PCM as if they are "perfect", e.g. adding a few guard bits increases the size of a 20 bit word to 24 bits and allows for dozens of signal processing operations without loss of practical resolution. This can not be done with DSD. Mathematically, (n-1)^k increases exponentially with k when n > 2. It increases linearly with k when n = 2.

 

Mixing theory and practicality may be confusing to a layman, but it should not be confusing to a competent practicing engineer. Also, the essential difference between function and mechanism is one that will be understood by lawyers who have familiarity with patent law.

Link to comment
Then you subtract the known signal from the mix and get back the original. The arithmetic is done sample by sample. The theory works in practice with any audio editor. It will not be possible to do the same thing with DSD, because when you add the +1's and -1's you get other values, namely -2, 0 and +2. To get back a one bit signal you have to remodulate the bit stream and this necessarily adds noise.

 

Using the same logic, it is not possible with PCM either. When you add two 24-bit values you get 25-bit result. In order to get back to 24-bit signal you need to do rid of the LSB, this adds noise and/or distortion depending on how you do it.

 

So if you mix values 3 and 5 you should get (2 + 3) / 2 = 2.5, now the closest to correct integer is either 2 or 3, both having same 0.5 quantization error. It becomes much worse when you mix eight channels, like for example (2 + 3 + 5 + 1 + 7 + 9 + 5 + 11) / 8 = 5.375 now 5 is closest integer and still leaves error of 0.375. Since adding eight 24-bit samples results in 27-bit result and you need to get rid of three bits of precision. IOW, you lose 18 dB.

 

When you mix eight channels in DSD, the noise drops 9 dB and even in worst case remodulation would only put 3 dB worth back, so the SNR would improve still 6 dB.

 

If you do a gain change with PCM it will be almost transparent, again assuming there is no overload. The error in the final conversion will be +- 2 least significant bit after adding in dither noise. (The RMS error will be about -136 dBfs.) Again, this theory can be demonstrated with any audio editor. With DSD, there will have to be a remodulation.

 

Again with PCM, when you multiply two 24-bit numbers, the result is 48-bit and to go back to 24-bit you need to get rid of bits. In addition, when you reduce volume of 24-bit PCM by 6 dB you lose one bit, so the result has at most 23-bit worth of relevant data. Reducing volume by 12 dB you lose two bits so now you have just 22-bit worth of relevant data.

 

However, worst part in PCM is that you may have lot of redundant value space. For example if you have a song that has only single peak reaching 0 dBFS and all other samples are below -6 dBFS you are not utilizing half of the value space! IOW, you may have the MSB constantly 0 except for single sample. This is not the case with SDM which is always utilizing all the value space!

 

This will add new modulation noise to the signal, and this will be at a much higher level than -136 dB in the relevant portions of the audio band. It seems likely that with DSD256 there will be enough of a separation from the audioband and the noise that the extra noise can be safely filtered out.

 

Certainly not. Even with DSD64 you can get much better SNR in audio band than you get with 24-bit PCM. At DSD128 you can already significantly exceed SNR of 32-bit PCM in audio band.

 

Sample rate conversions using PCM can be done losslessly, subject to dither noise, provided there aren't headroom problems and provided that the PCM original file has no energy at Fs/2.

 

Little too many conditions... Now here's you first problem, design a pure PCM ADC-DAC chain that has your >136 dB dynamic range and won't have both time/phase distortions vs analog input signals.

 

PCM may seem all nice and dandy when you look at it in pure digital domain. But when you begin to involve analog domain filters and AD/DA converters things change and DSD becomes much better. So it is much better to do everything in SDM.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
The problem here is that there are people in this thread who are approaching the problem from different perspectives.

I always think that for solving problem effectively, approaching from different perspectives is the best way to solve problem.

Sorry, english is not my native language.

Fools and fanatics are always certain of themselves, but wiser people are full of doubts.

Link to comment
If I haven't made myself clear... please read Tailspn's post below:

 

 

 

As for...

 

 

 

The inventor of DSD, Prof. Yoshio Yamasaki calls the format Delta Sigma Direct, so Delta Sigma is definitely a fundamental part of it.

 

Tailspin's comments about two's compliment arithmetic apply to the form of PCM commonly used in digital audio. There are other forms of PCM used for other applications. One of the earliest commercially practical PCM systems was 8 bit PCM sampled using piece wise linear encoding at 8 kHz and was the basis of the T1 digital telephony system introduced in the 1950's. https://en.wikipedia.org/wiki/Pulse-code_modulation

 

The use of Sigma Delta modulators to create one bit pulse streams has a long history going back to the 1950's. The only thing that Yamasaki "invented" was the idea of not doing conversions to/from multi-bit formats. All of the elements of his invention existed previously. In any event, a trade name does not characterize a mathematical format.

 

One example of a 1 bit format is the audio generate by the Apple II computer. Here a bit stream was created by software control of a flip-flop. One output of the flip flop was wired directly to a small speaker. This generated arbitrary sounds, for example it was possible to generate the pairs of tones used in phone phreaking "Blue Boxes" by software and many phone hackers did this around 1980, starting with Wozniak and Jobs. One way of generating the necessary three levels corresponding to the addition of two square waves (+2, -2 and 0) was by representing 0 as a rapid toggling back and forth. There was no sigma-delta modulator and no use of feedback. The decoding was a simple low pass filter implemented by the inertia of the speaker cone.

 

The following reference shows a short history of sigma delta modulators in analog to digital converters:

http://www.analog.com/media/en/training-seminars/tutorials/MT-022.pdf?doc=cn0354.pdf

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...