World’s First Valid Comparison of PCM versus DSD?

[Teresa]

It's OK, John Swenson, who actually knows something about these matters, provided helpful explanation a few comments later.

 

Yes, I too found John Swenson's post most helpful, he cleared up a few questions I had.

[JohnSwenson]

John,

These two statements seem at odds with each other. If a lowpass filter is involved there is no instant change of state within 1Bit, the filters settling time must be considered.

 

Yes the maximum slew rate of the output of the system is the slew rate of the analog filter. I was refering to the input to the analog filter. In delta modulation there is a maximum slew rate that is dependant on the size of the "difference" value. If you choose a difference value to be small to get decent dynamic range, you limit the slew rate to one of those differences per clock, so what is going into the analog filter is already severely slew rate limited. With DSD the input to the filter can go from full negative to full positive in one clock cycle.

 

Further, if as you say, the bits reflect the Error then a continous string of 1's in effect represents the system slewing if a dynamic signal is involved (read for example a fullscale step). At least if the system involved was an Amplifier we would call it slewing. So one would have to either accept this slewing limitation and the resultant distortion (which we call TIM in amplifiers) or filter the input to the system same as for feedback amplifiers.

 

So I still wonder, what is the slew rate of DSD (that is fastest change of the input from fully negative to fully positive that does not cause the modulator to saturate)?

 

The slew rate of the system is the slew rate of the analog filter whatever that may be.

 

John S.

[magnum innominandum]

Miska,

 

Now to the correct thread...

 

Here's 20 kHz sine at DSD128 with 60 kHz reconstruction filter:

[ATTACH=CONFIG]9598[/ATTACH]

 

Second harmonic is -93 dB and third harmonic at -92 dB level from analog stages.

 

I had to look up picotech. Looks like they sell glorified external Soundcards to me.

 

Do you have an actual oscilloscope (LeCroy, Tek, Agilent)?

 

BTW, what you show has significant distortion (e.g. visible) near the sinewave peaks.

[magnum innominandum]

John,

 

Yes the maximum slew rate of the output of the system is the slew rate of the analog filter.

 

I should have thought so. So, the key is the analogue filter.

 

Now what is this filter? What is the standard. I have searched long and hard. In fact any hard information on DSD seems near impossible to come by. I knew not much about DSD until recently, so Itriedtoreadup...

 

Whereas the performance,operations and behaviour as well as the limits of PCM are well documented, I observe that anything relating to DSD seems hazy, often contains obvious misinformation or misinterpretations of common terms in ways not common and generally boil down to being extremely light on hard data or tech and full of sayings "Don't worry, it is way better than anything else, just believe us...".

 

Reviews in Stereophile of DSD replay stuff often show a quite poor measured performance, making one wonder if this is because the designers are not that competent or because the format is the limitation?

 

Given that it would be trivial to actually demonstrate the real performance of DSD compared to PCM and to illustrate any claimed superiority, I find the absence of such at the very least suspicious.

[magnum innominandum]

Miska,

 

PCM has much more clearly defined limited slew rate since it is heavily band-limited system by definition. Nyquist frequency clearly defines the maximum slew rate. PCM has same speed limitation regardless of signal level.

 

Forgive me for being rude,but are you sure you use the term slew rate in the same in which it is commonly used?

 

Frequency response and slewrate have a link, but it is not direct. One may argue that for PCM the analogue lowpass filters (recording and playback) determine the slewrate. Given that most filters are digital these days things are not as easy.

 

With DSD the amount of time required to slew a signal depends on signal level (just like analog), lower the swing, higher the available bandwidth is (and also better the linearity becomes). At 0 dB level DSD64 slew rate is equivalent of 117.6 kHz PCM. At -6 dB level it is equivalent of 235.2 kHz PCM and so on...

 

This would be true if there were no lowpass filters involved.

 

Later in the thread you show a'scope trace (likely using heavy averaging) with a 60KHz lowpass filter.

 

So surely, this particulat DSD system has a fixed slewrate determined by this lowpass filter?

 

Incidentally, I found some more oscilloscope shots for DSD which are, shall we say interesting:

 

http://www.craigmandigital.com/education/PCM_vs_DSD.aspx

 

 

I guess that person also had a very poor SACD player, right?

[esldude]

magnum innominandum,

 

The following shows about the same.

 

Archimago's Musings: MEASUREMENTS: PCM to DSD Upsampling Effects (JRiver MC19 Beta).

 

Some scope traces of DSD vs PCM. Show pretty much the same thing with the ultrasonics making for fuzzy waves. Though things look much more similar for DSD128 vs PCM than does DSD64.

[Miska]

Forgive me for being rude,but are you sure you use the term slew rate in the same in which it is commonly used?

 

Well, RedBook PCM slew rate is limited by the 22.05 kHz pass-band. Or other way around, it cannot change faster than 1/44100th of second.

 

Frequency response and slewrate have a link, but it is not direct. One may argue that for PCM the analogue lowpass filters (recording and playback) determine the slewrate. Given that most filters are digital these days things are not as easy.

 

RedBook PCM ADC is usually oversampled and brickwall filtered in digital domain with about 1 ms long digital filter. If you don't manage to bandwidth-limit the input properly you get really nasty effects due to aliasing.

 

Since DSD64 has Nyquist frequency of 1.4 MHz, there's no need for steep anti-alias filters.

 

Later in the thread you show a'scope trace (likely using heavy averaging) with a 60KHz lowpass filter.

 

So surely, this particulat DSD system has a fixed slewrate determined by this lowpass filter?

 

Yes, naturally. DSD is supposed to be filtered for signal reconstruction. So in this case there's ~8th order 60 kHz low-pass (combination analog FIR and Butterworth), compared to brickwall at 22.05 kHz needed for PCM.

 

I guess that person also had a very poor SACD player, right?

Looks like result of DSD-to-PCM conversion. I guess this is from one of the multi-format players based on Mediatek chipset?

 

If I compare scope output of for example PCM1795-based DAC playing 20 kHz sine wave in PCM and in DSD128, the scope output looks exactly the same to me. This is because it is a delta-sigma DAC...

 

Here's ESS Sabre-based DAC playing 1 kHz square wave in DSD:

and here it is in frequency domain:

[Miska]

Though things look much more similar for DSD128 vs PCM than does DSD64.

 

And I bet you are going to have hard time reliably telling difference when playing sine wave between the two...

[Miska]

Reviews in Stereophile of DSD replay stuff often show a quite poor measured performance, making one wonder if this is because the designers are not that competent or because the format is the limitation?

 

Can you point us at some specific measurements and we can try to figure out?

 

Now in the most recent one there's a review of dCS Vivaldi.

[Le Concombre Masqué]

Well, RedBook PCM slew rate is limited by the 22.05 kHz pass-band. Or other way around, it cannot change faster than 1/44100th of second.

 

 

 

RedBook PCM ADC is usually oversampled and brickwall filtered in digital domain with about 1 ms long digital filter. If you don't manage to bandwidth-limit the input properly you get really nasty effects due to aliasing.

 

Since DSD64 has Nyquist frequency of 1.4 MHz, there's no need for steep anti-alias filters.

 

 

 

Yes, naturally. DSD is supposed to be filtered for signal reconstruction. So in this case there's ~8th order 60 kHz low-pass (combination analog FIR and Butterworth), compared to brickwall at 22.05 kHz needed for PCM.

 

 

Looks like result of DSD-to-PCM conversion. I guess this is from one of the multi-format players based on Mediatek chipset?

 

If I compare scope output of for example PCM1795-based DAC playing 20 kHz sine wave in PCM and in DSD128, the scope output looks exactly the same to me. This is because it is a delta-sigma DAC...

 

Here's ESS Sabre-based DAC playing 1 kHz square wave in DSD:

[ATTACH=CONFIG]9618[/ATTACH]

and here it is in frequency domain:

[ATTACH=CONFIG]9619[/ATTACH]

 

When I first entered computer audio and CA forums soon after, I posted that I preferred to feed my TEAC 501 with dsd converted to pcm on flight by A+, especially since I can then EQ. Listening to Charles Hensen's samples, I preferred to send DoP, for I love the speed, lightness, livelihood dsd conveys with those files (and preferred to listen that way over any breed of PCM, converted dsd or CH's straight pcm, with or without eQ up sampling...). I discover that I use a PCM1795-based DAC, thus converting anyway... I yet have tried and preferred Korg's Audiogate (with 0 gain, let it clip mode) over A+ and I can EQ. I understand that HQPlayer can sort of EQ DSD ; but is it worth it if there's conversion anyway? Should I rather refine my Korg (other suggestion ?)settings to keep what I like in unconverted DSD ? can HQPlayer convert offline and better ?

[magnum innominandum]

Esldude,

 

magnum innominandum,

 

The following shows about the same.

 

Archimago's Musings: MEASUREMENTS: PCM to DSD Upsampling Effects (JRiver MC19 Beta).

 

Some scope traces of DSD vs PCM. Show pretty much the same thing with the ultrasonics making for fuzzy waves. Though things look much more similar for DSD128 vs PCM than does DSD64.

 

Thank you. Hmm, it seems we get a fair bit of the same results with a 'scope on output, no matter if the source is an unnamed "very poor DSD DAC" or a "Universal player with a Mediatek Chipset" according to Miska (when in fact it was stated in the original article that it was a professional DSD ADC/DAC setup) or a "High End Teak DAC".

 

I wonder what this tells us?

[magnum innominandum]

Miska,

 

Well, RedBook PCM slew rate is limited by the 22.05 kHz pass-band.

 

And 768KHz 24 Bit PCM is limited by the 384KHz passband. And?

 

(Naim's DAC handles 768KHz/24Bit BTW)

 

Since DSD64 has Nyquist frequency of 1.4 MHz, there's no need for steep anti-alias filters.

 

So why filter DSD at all, if it does not need filters.

 

Yes, naturally. DSD is supposed to be filtered for signal reconstruction. So in this case there's ~8th order 60 kHz low-pass (combination analog FIR and Butterworth), compared to brickwall at 22.05 kHz needed for PCM.

 

PCM does not need a brickwall filter at 22.05KHz, only 44.1KHz sample rate PCM needs that.

 

An 8th order lowpass filter at 60KHz is certainly going to limit slew rate some...

 

Looks like result of DSD-to-PCM conversion. I guess this is from one of the multi-format players based on Mediatek chipset?

 

You could read the article that was linked which would give the required clues. Given the time the article was written it would have been likely either a Meitner or dCS DSD ADC & DAC Chain, compared to a PCM ADC/DAC chain, all pro audio gear.

 

If I compare scope output of for example PCM1795-based DAC playing 20 kHz sine wave in PCM and in DSD128, the scope output looks exactly the same to me. This is because it is a delta-sigma DAC...

 

Miska, how do you then explain the results posted in the Web article by Archimago, which are based on a DAC using this exact chipset, made obviously with areal 'scope?

 

Could it be your picotech 'scope is actually not telling the truth, the way a LeCroy, Tek or HP/Agilent would?

 

Here's ESS Sabre-based DAC playing 1 kHz square wave in DSD:

 

I remember reading that ESS DAC converts DSD to PCM and applies digital filtering (ESS Whitepaper). In that case we would be seeing the not DSD, but something else. What, well there are no datasheets for the ESS Parts that tell anything, so I don't know.

[magnum innominandum]

Miska,

 

Can you point us at some specific measurements and we can try to figure out?

 

Not an DSD DAC:

 

Ayre Acoustics QB-9 USB DAC Measurements | Stereophile.com

 

"Overall, Ayre's QB-9 is well engineered, offering excellent performance in both the analog and digital domains, and is not compromised by its having just a USB data input.—John Atkinson"

 

Not an DSD DAC:

 

MSB Technology Platinum Data CD IV transport & Diamond DAC IV & D/A converter Measurements | Stereophile.com

 

"In most ways, MSB Technology's Diamond DAC IV offers the best measured performance in the digital domain that I have encountered.—John Atkinson"

 

An SACD Player:

 

Playback Designs MPS-5 SACD/CD player Measurements | Stereophile.com

 

"So while I was impressed by the player's standard of construction, I can't say the same about its technical performance. The relatively high level of background noise limits the MPS-5's resolution with SACD and external 24-bit data to not much better than 16-bit CD. I am puzzled, therefore, why Michael Fremer liked the sound of this player so much.—John Atkinson"

 

Another SACD Player:

 

Luxman DU-50 universal player Measurements | Stereophile.com

 

"it does offer a prematurely rolled-off top octave, as well as slightly higher noise levels and only modest suppression of ultrasonic image energy. The mystery to me was the DU-50's relatively poor performance with SACD.—John Atkinson"

 

Another SACD Player:

 

 

dCS P8i SACD player Measurements | Stereophile.com

 

"...the noise floor with both CD and SACD playback was a little dirtier than I have found with other high-end disc players.—John Atkinson"

 

Now in the most recent one there's a review of dCS Vivaldi.

 

Which uses a multibit hybrid (aka "Ring DAC"), not pure delta sigma or Sigma Delta.

[JohnSwenson]

John,

 

 

 

I should have thought so. So, the key is the analogue filter.

 

Now what is this filter? What is the standard. I have searched long and hard. In fact any hard information on DSD seems near impossible to come by. I knew not much about DSD until recently, so Itriedtoreadup...

 

Whereas the performance,operations and behaviour as well as the limits of PCM are well documented, I observe that anything relating to DSD seems hazy, often contains obvious misinformation or misinterpretations of common terms in ways not common and generally boil down to being extremely light on hard data or tech and full of sayings "Don't worry, it is way better than anything else, just believe us...".

 

Reviews in Stereophile of DSD replay stuff often show a quite poor measured performance, making one wonder if this is because the designers are not that competent or because the format is the limitation?

 

Given that it would be trivial to actually demonstrate the real performance of DSD compared to PCM and to illustrate any claimed superiority, I find the absence of such at the very least suspicious.

 

DSD is a compromise (as is pretty much everything else). It is a 2 level system with a fairly low clock rate, the result is massive quantization noise (which you can see in the scope images). To make this work at all heavy noise shaping is used to push the noise which would show up in the audio band (defined as under 20KHz) up above 20KHz. The result is a severely rising noise floor above 20KHz.

 

There are several approaches to deal with this ultrasonic noise. I think the original intent was to use a 50KHz analog filter to filter the 2.8MHz carrier and git rid of some of the ultrasonic noise. It certainly does not get rid of all of it and doesn't get much of the noise close in to the audio band. The idea is that you can't hear this ultrasonic noise so don't spend a lot of effort trying to get rid of it. With a DAC that uses this approach scope traces are always going to look "messy" because of this ultrasonic noise riding on top of everything.

 

Various DAC implementations have tried to improve on this by attenuating more of this noise using all kinds of various approaches. Unfortunately most of these wind up affecting the analog band to some degree in their attempt at attenuating the ultrasonic noise.

 

Others have listed some of these other approaches. Some of these run the DSD signal, which is the result of a specific DSM configuration and running it through another DSM in the DAC chip. This can produce a different noise spectrum. Most of these approaches have an analog filter after whatever scheme they use to try and attenuate the noise.

 

Because of all these different approaches it's hard to make any hard and fast rules about what is coming out of the analog out jacks. It IS going to vary significantly from DAC to DAC. About the only thing you can say is that they are all going to have some amount of ultrasonic noise.

 

John S.

[Hiro]

DSD is a compromise (as is pretty much everything else). It is a 2 level system with a fairly low clock rate, the result is massive quantization noise (which you can see in the scope images). To make this work at all heavy noise shaping is used to push the noise which would show up in the audio band (defined as under 20KHz) up above 20KHz. The result is a severely rising noise floor above 20KHz.

 

There are two ways to deal with the rising noise floor above 20kHz, one is multi-speed DSD (double/quad DSD), another multi-bit DSD.

 

BTW, the LPs that were used for the "world's first valid comparison of PCM vs DSD" are probably far noisier in the audio band than single speed DSD in the 20-50kHz region.

[Hiro]

You are not alone and I think there have been several posts in this thread that were uncivil. I've noticed that pretty much any DSD-related thread seems to bring out the crowd that is mostly interested in tearing someone down.

 

Charles just can't stop himself from throwing mud at his friends at Sony and Super Audio Center. I noticed that too.

[beetlemania]

Charles just can't stop himself from throwing mud at his friends at Sony and Super Audio Center. I noticed that too.

 

A much better example:

 

"I was not attacking Ayre, nor Charlie Hansen, . . .

 

But I can see clearly now from your system, you are an Ayre user...

 

While I Sleep, get a better DAC."

[Hiro]

A much better example:

 

"I was not attacking Ayre, nor Charlie Hansen, . . .

 

But I can see clearly now from your system, you are an Ayre user...

 

While I Sleep, get a better DAC."

 

While not the best choice of words, admittedly, the competition on the DAC market is huge, and everyone has right to choose a DAC that meets their needs.

 

I for one find the QB-9 DAC too expensive for what it has to offer; no headphone amplifier, coax, optical inputs, off-the-shelf DAC chip inside...

[beetlemania]

While not the best choice of words, admittedly, the competition on the DAC market is huge, and everyone has right to choose a DAC that meets their needs.

 

I for one find the QB-9 DAC too expensive for what it has to offer; no headphone amplifier, coax, optical inputs, off-the-shelf DAC chip inside...

 

And I, for one, find DSD to be well overblown hype. You might say: "Much ado about nothing".

[Hiro]

And I, for one, find DSD to be well overblown hype. You might say: "Much ado about nothing".

 

To each his own. Whenever I listen to a native DSD recording such as this all I can do is ask for more of this stuff!

[One and a half]

And I, for one, find DSD to be well overblown hype. You might say: "Much ado about nothing".

 

Hype usually dies down after the initial hysteria after a short time, or the advertising campaign stalls. The first DSD Dac that really fuelled the active interest in DSD was the Mytek in 2010. Nearly four years later, there's still great interest in DSD Dacs and the source material, so not sure what you consider as hype. Examples are from RMAF this year with a few A/B comps winning more ears for DSD.

 

I, for one, find DSD far more engaging and listenable to music than hi-res PCM from hdtracks. There are very small exceptional recordings with PCM, with pretty much the rest of offerings being a crap shoot. With DSD, I haven't found a really sour quality download yet, whether it's a recording from analog tape, native direct to DSD, or an sacd derived recording.

 

If PCM only is good for you, go for it, it's one person less competing with download bandwidth for DSD tracks.

[PopPop]

The cattle drive hasn't turned into a stampede yet, eh?

[Miska]

And 768KHz 24 Bit PCM is limited by the 384KHz passband. And?

 

Yes, although good luck for finding material with that sampling rate. Files would be pretty big, and FLAC refuses to encode sampling rates higher than 384k.

 

So why filter DSD at all, if it does not need filters.

 

At ADC side you should always have anti-alias filter before the converter, regardless of sampling rate. Otherwise you get nasty effects due to aliasing of any RFI/EMI.

 

PCM does not need a brickwall filter at 22.05KHz, only 44.1KHz sample rate PCM needs that.

 

That right, PCM needs to be anti-alias filtered by Nyquist frequency fs/2.

 

An 8th order lowpass filter at 60KHz is certainly going to limit slew rate some...

 

Sure, but less than filters needed for PCM at 192k or lower rates.

 

You could read the article that was linked which would give the required clues. Given the time the article was written it would have been likely either a Meitner or dCS DSD ADC & DAC Chain, compared to a PCM ADC/DAC chain, all pro audio gear.

 

I've seen it long ago, but not really interested. Something is wrong there because of symmetric overshoot of the square wave.

 

Miska, how do you then explain the results posted in the Web article by Archimago, which are based on a DAC using this exact chipset, made obviously with areal 'scope?

 

My measurements have been made with 100% real scope. Mine has more resolution and features than the one used by Archimago. If you compare the PCM and DSD128 results of Archimago, the differences are not big. One difference on my side is that I use my own modulator to create the DSD stream while Archimago uses the one in JRMC. Archimago didn't post 20 kHz sine wave.

 

What is there to explain?

 

Could it be your picotech 'scope is actually not telling the truth, the way a LeCroy, Tek or HP/Agilent would?

 

Why are you comparing my results with those posted by others, different DAC and different test signals - different results. Notice also that my square wave is produced from full 2.8 MHz bandwidth while the Archimago's square wave is upsampling from 44.1k PCM.

 

I remember reading that ESS DAC converts DSD to PCM and applies digital filtering (ESS Whitepaper). In that case we would be seeing the not DSD, but something else. What, well there are no datasheets for the ESS Parts that tell anything, so I don't know.

 

No it doesn't convert it to PCM, it converts it to multi-bit SDM at same sampling frequency, there is no sampling frequency down-conversion at all. And of course it is DSD because that's what the DAC receives.

[Miska]

Which uses a multibit hybrid (aka "Ring DAC"), not pure delta sigma or Sigma Delta.

 

No, it is as pure delta-sigma DAC as one can be. With 24 equally weighted current sources.

 

I am using 32 weighted current sources in my DSD DAC prototype.

[Miska]

Not an DSD DAC:

 

Ayre Acoustics QB-9 USB DAC Measurements | Stereophile.com

 

"Overall, Ayre's QB-9 is well engineered, offering excellent performance in both the analog and digital domains, and is not compromised by its having just a USB data input.—John Atkinson"

 

This is non-PCM pure delta-sigma DAC based on Sabre chip. Sabre has 64 1-bit DACs run in parallel.

 

But overall, this particular implementation has too high distortion for my taste and the "listen" filter leaks like hell.

 

And what did JA say about dCS Vivaldi in January 2014 issue:

"Overall, the dCS Vivaldi measured superbly well.

 

...there was an increased sense of ease to Vivaldi's sound quality, particularly when the Upsampler transcoded the data to DSD, that proved addictive. In this respect, it was even better than the MSB it replaced in my system. For a third time: Wow!"