Ayre's PCM and DSD Comparison

I think its fair to say that Charlie Hansen, Ayre's Founder and Designer, is not exactly a fan of DSD. Thankfully, that didn't stop him from offering DSD-capable versions of Ayre's QB-9, DX-5-DSD, and now the QA-9 thanks to AlpineSoft's VinylStudio DoP vinyl ripping software. What the guys at Ayre have gone and done is ripped parts of three tracks to single-rate DSD and 24/192 using the QA-9 so we can compare DSD to PCM. But there's more...

From the Ayre website:

Ever since Sony’s early (convoluted and self-contradictory) marketing claims about the performance advantage of DSD, backed by the fact that their multi-million dollar budget was generally able to make better sounding products than the disorganized renegades that released DVD-Audio, many, many audiophiles have believed that DSD is inherently superior to PCM.

At the highest levels of audio engineers there has been little consensus as to which format is better, but most audiophiles have had first-hand experience with SACDs generally sounding better than DVD-Audio disc. So the prospect of being able to play downloadable DSD files via computer has led to a great deal of excitement, even controversy in the audiophile community. This article aims to sort out some of these claims.

You'll want to read the rest of this article "DSD vs. PCM Recording" as its got a lot of historical perspective and general information (and opinions) about DSD that you won't find anywhere else:
DSD vs. PCM RecordingCharles Hansen, Ayre Acoustics, Inc.
I plan to give these a listen and report on my findings since I happen to have the Ayre QB-9 DSD here for review. Here's why that's important according to Charlie Hansen:
When switching between recording DSD or PCM, all of the internal hardware is identical. All of the analog circuitry, power supplies, digital circuitry, converter chips, PCB material , et cetera is 100% identical. The only difference in the algorithms used to process the raw data from the Arda AT1201 ADC chip (6 bits at 256 Fs) either into DSD or into PCM.

If you are fortunate enough to own either an Ayre DX-5-DSD or Ayre QB-9-DSD, then the same is true for the playback also. There are no differences in the signal path whatsoever except for the algorithms used in the ESS ES9016 Digital-to-Analog converter chip. There is no way to make a more fair comparison.

If you'd like to download the samples and hear 'em for yourself, here they are (there's about a 9db difference between the dsf and wav files so make sure you adjust accordingly):

Ella Fitzgerald, "Black Coffee" from Let No Man Write My Epitaph (Verve - Classic Records Reissue)

PCM Version
DSD Version

Crosby, Stills, and Nash, "Helplessly Hoping" from Crosby, Stills & Nash (Atlantic - Classic Records Reissue)

PCM Version
DSD Version

Beethoven "Symphony No.7", Otto Klemperer (Decca - HiQ Records Supercuts)

PCM Version
DSD Version

nick's picture

interseting article, but it seems that there is a big presumption in it. it's great how the author levelled the playing field as much as possible, but isn't he assuming that the particular chip in that dac is equally good at converting dsd & pcm files? i've read elsewhere that dac chips that are optimized for dsd can't do pcm justice, and vice versa.

michael, what are your thoughts on that?

Charles Hansen's picture


You raise an interesting point. However if there is an advantage, it would be to the DSD side. The only "true" PCM DAC chips are "ladder" (sometimes called "R-2R") types and the only "true" PCM ADC chips are "flash" or "SAR" (Successive Approximation Register) . All of these are virtually out of production these days and haven't been widely used in audio for at least fifteen years. So if there is any inherent advantage here, it would be to the side of the DSD playback.

Charles Hansen
Ayre Acoustics, Inc

nick's picture


thanks for the response. i'm guessing you're "team pcm", right?


Charles Hansen's picture

I think that it is safe to make at least two sweeping generalizations about the two formats:

1) It is a huge pain in the ass to make a modern multi-track recording in DSD.

2) Anybody that would to the the lengths required to make DSD recording is so dedicated to their craft that they are probably going to make a great recording even with tin cans and a piece of string.

Then there is the sound of the files themselves. When I listen to them, I hear more textrual differences in the midrange with the PCM recordings. Conversely, on the DSD recordings there is a certain "sameness" to the high frequencies that is almost certainly an artifact of the particular noise-shaping curve that Sony chose. It's a sort of pervasive coloration that makes all DSD recordings made with any specific brand of recorder sound very similar in the high frequencies.

The only real justification for DSD at all is if one can attain a higher level of sound quality that is possible with PCM. I am not hearing that when I listen to the files. But you tell me what you hear. It's your money and your purchasing decision.

And one last thing is that this is certainly the first time that PCM has been done properly in the real world. Everybody else uses super-steep and/or brickwall filters that ring like a bell and kill the music. The moving average filter used in the QA-9 has no overshoot, undershoot, ringing, or any type of time smear at all. That was one of the lessons we learned from DSD and applied to PCM so that we can have the best of both worlds.


Charles Hansen
Ayre Acoustics, Inc.

Wavelength's picture

Nick, Charlie;

The problem is that there are no native DSD dac chips anymore. The last one was the TI/BB DSD1700 as I remember.

All current DAC chips that can handle DSD do so by converting the data to a managable form that is compatible with their PCM interface.

Between that hurdle that DSD goes through on the output and what I would call a poor DoP specification... I would say we are not hearing what the capabilities of DSD could be.



labjr's picture


I thought DoP specification was designed to playback native DSD? And you worked on it? Is it not complete or did everyone give up?

As far as DAC chips go, what's the difference between playing back a DSD file with current DAC chips or playing PCM files with with a sigma delta chip? Neither would seem to be ideal?



Wavelength's picture


As we found out in Computer Audio over the years is that making things as simple as possible and less time consuming in processing yields better sonic results. To best understand my frustration with DoP, let's see how PCM is transmitted.

If the file is not a true flat PCM file (AIFF, WAV) then it is converterd to flat PCM data. This is one of the reason's why AIFF and WAV files sound better than others. There is less EMI, RFI radiation as well as power consumption (more processing more consumption and higher radiation). The now flat PCM data is in memory and gets applied to the device driver (I write these so I know) which copies the data to an area where the data is directly DMA (Direct Memory Access) mapable to the output device (i.e. USB, Firewire... other). Depending on the device some (XMOS and others) copy the received USB frame into an output buffer and then this data is strung out to the DAC chip. Some other processors have a circular buffer and the processor does absolutely nothing in streaming out the data to the DAC chp.

Since DoP packs DSD information into PCM frames the software has to first create packeted frames of data. This is pretty labor intensive and therefore requires more processing. Every 24 bits of data for both channels has a 8 bit header on it to indicate it is a DoP frame. This means there is an extra 33% of data has to be transmitted for each frame. The data is now passed to the device driver and all that is the same. On the DAC side the controller has too look at each sample coming in to see if it is DoP or not. If it is DoP then the data has to be reconstructed into DSD frames and sent to the DAC chip. Which by funny will then convert the DSD information to quasiPCM format because there is no DSD available chips.

Another fall out of the DoP spec is the dacs now have to support PCM at higher rates. So DSD1 uses 176.4, DSD2 uses 352.8... so basically you have to support 384K to also support DSD2.

Please note... Just because you can do something doesn't mean it's right or correct. All of you are asking for higher and higher rates. REALLY!!! When does it get back to what sounds best? That is really what you should be asking yourselfs.

There were several approaches much better than this, like indicating that DoP was an 8 bit format with higher sample rates that would work just like straight PCM without the need for packing frames or anything else. The nice thing about the 8 bit idea was that when the sample rate/bit rate changes you would know it's DSD and prepare for that. You would not have to look at each sample coming in to see if it is DoP or not.


labjr's picture


This is obviously more complex than it appears to the average consumer.

Is there still a group working on improving DoP?

Or Why not just use an ASIO DSD driver? Then at least part of the problem is solved Then worry about the sound of DAC chips.

BTW, have you heard properly done DSD?  Do you share the same opinion as Charlie Hansen? If you and everyone who was working the DOP spec does, then I can see why interest in DSD seems to have slowed. What would be the point of going any further?





Charles Hansen's picture

Hello Gordon,'

'While a "true" DSD DAC chip would be a one-bit affair, it would still be a delta-sigma chip. I think that modern multi-bit delta-sigma DAC chips (which is pretty much all that is available) have much more in common with a "DSD" DAC chip (single bit delta-sigma) than it does with a true "PCM" DAC chip (ladder or R-2R type).

So while it is obviously true that anything  can be made better than it is currently, it seems to me that if the comparison favors anything, it is favoring DSD, as the "ideal" DAC chip would also be a delta-sigma type like the ESS we currently use.


CG's picture

What we need is for the MRI world to demand faster DACs.  If true 20+ bit ladder chips were available that would run at 3+ MHz sample rates, the same device could be used for true native DSD as well as multibit PCM decoding.  Way easy.

We're actually close now, but the parts aren't cheap.

Bruce B's picture

I thought the "World's First Valid Comparison of PCM vs. DSD" was done by you Michael, where you compared master tapes that were digitized by the same converter at various sample rates November of last year?

PCM vs. DSD Comparison:

Michael Lavorgna's picture

Here's the link - PCM v DSD Comparison: 16/44.1, 24/96, 24/192, 64x DSD, 128x DSD

I don't want to speak for Charlie but my guess is he would say there were too many variables involved in those comparisons. In this "Ayre Comparison", if we use the Ayre QB-9-DSD to play back the files, the signal path from A to D and D to A are nearly identical, leaving us with just the difference between file formats. Or something along those lines.

(I also found your comparison of A/D converters very interesting and informative - Bruce Brown at Puget Sound Offers the "DSD Battle Royale!)

I have not had time to compare these "Ayre" downloads more than a few times but I hope to get to them later tonight or tomorrow (I hate snow shoveling). My initial impression is Ella Fitzgerald has such a lovely voice regardless of the format.

Charles Hansen's picture

Much of the point of this exercise was to make as humanly sure as possible that every single variable was controlled between the DSD and PCM versions of the recordings. I think we did as good or better of a job doing that than anybody has done before, and I am certain that nobody has done that part better.

The other thing to highlight was the reason for many peoples' preference for DSD. Specifically, 99.9% of all PCM converters (A/D or D/A, professional or consumer) have used brickwall filters. This unavoidably smears the time domain and also causes ringing at the cutoff frequency.

The same exact thing happened when metal dome tweeters were released. People comlained that they sounded metallic. And they did. But it had nothing to do with the metal dome and everything to do with the little plastic "phase plug" in front of the dome that created a Helmholtz resonator at around 18 kHz and made everything sound "metallic". But metal dome tweeters with small rods to protect the dome do not have this coloration.

Well when there is a brickwall filter that rings at 22 kHz, guess what?

It makes the music sound metallic and harsh.

Nothing to do with PCM per se,  and everything to do with the implementation. So when you listen to the downloads on an Ayre or a Wadia without a brickwall filter they will sound quite nice. And in my opinion slightly better than the DSD. YMMV.

But even though they were recorded without a brickwall, you need to listen to them without a brickwall. These are very rare birds. So Michael, when you get your Ayre QB-9 DSD connected up, set the switch on the rear panel to "Listen" and you will be hearing a PCM recording/playback chain with no brickwall filters.

The moral?

It's not the format. It's the implementation. And since we can get sound as good or better that DSD when properly implemented, why mess around with DSD???


Michael Lavorgna's picture

...why mess around with DSD???

Because it typically sounds better than PCM in my experience.

The other thing to highlight was the reason for many peoples' preference for DSD. Specifically, 99.9% of all PCM converters (A/D or D/A, professional or consumer) have used brickwall filters. This unavoidably smears the time domain and also causes ringing at the cutoff frequency.

I also think that the A/D converter plays a part in these preferences but even if what your saying here is the case, the practical result is the same - DSD will sound better than PCM on the majority of DACs  (and everyone can't own the Ayre QB-9-DSD ;-).

Michael Lavorgna's picture

I've now had a chance to listen through these samples a few times using the Ayre QB-9-DSD connected to my Pass INT-30A via Kimber XLRs driving my DeVore Fidelity The Nines.

To my ears they are very close, certainly not a night and day difference between the 24/192 and DSD files. I did notice a slightly softer, more fluid sound to Ella Fitzgerald's voice with DSD and also with the CSN track. I'd also say the 24/192 versions appeared to have a skosh more resolution and bite especially in the upper registers coupled with a slightly flatter presentation. They both sound very good to me, especially the Ella Fitzgerald. I'll have to pick up the LP ;-)

Jason Victor Serinus's picture

I can't comment on the comparison, given that, when I finally reassembled my reference system last night, my dCS Puccini refused to turn on. But I can say that Analogue Productions is reissuing Ella's LP in SACD format, with an anticipated street date of January 10 or thereabouts. I expect it sounds fabulous.

And now for the good news. With a simple fuse replaced, my reference system is up and singing. Hopefully Ella will be singing on it once the SACD is in my eager little hands.

Charles Hansen's picture

Hello Michael,

I agree completely. Both formats are capable of truly excellent sound. That's why I think it is sheer madness to throw away all of the recording, editing, and playback hardware of PCM, for a slight (and debatable) improvement (if any) from DSD.

There is no magic in DSD. Properly done PCM sounds as good or better and is infinitely easier to work with.

Now all we have to do is to get all of the recording studios to use Ayre A/D converters and everthing will have great sound. smiley


Michael Lavorgna's picture

Hi Charlie,

It seems to me that high resolution PCM, and to a lessor extent DSD, have initiated some wonderful remasters. The recent Kind of Blue PCM release is just one example where we now have access to what is arguably a closer representation of what was originally recorded. At this year's RMAF, I got to hear some of Dave Wilson's recordings remastered from the original analog tape to DSD and they sounded wonderful. I've also got some other DSD remasters as well as native DSD recordings that sound great.

My point being, while I agree there's "no magic in DSD", it is simply another option that is responsible for providing us with access to more great sounding music. As far as I know, the majority of DSD releases are also available as PCM releases so people have the option to buy and listen to the format they prefer.

Here's something you said in an earlier comment here (with my emphasis):

"The moving average filter used in the QA-9 has no overshoot, undershoot, ringing, or any type of time smear at all. That was one of the lessons we learned from DSD and applied to PCM so that we can have the best of both worlds."

While we're not talking about magic, this suggests to me that there are aspects of DSD performance that are indeed preferable to typical PCM performance. The fact that they can be applied to PCM is nothing but a good thing.

When all is said and done, I view this interest in DSD as a win win for us listeners.

fmak's picture

I don't believe so.

The Ayre files bear the hallmark of what is to me not very good Analog transcription or A to D conversion, with little depth and width to the sound stages. They are pleasant enough and DSD does sound a ittle warmer than pcm, but they are so so and not demanding music material.

My native DSD recorded and 192/176k recorded music sound much more transparent, with much better sound staging.



jim tavegia's picture

Wouldn't the best measure be to take a small ensemble recording and with 2 feeds. Do a  pure DSD 2 track recording and then do the other at 24/192 and leave those files raw with no mastering? 


I now now have some mastering engineers who I prefer and some now that I avoid. I never considered the importance of mastering in my youth.  Now I do. 

Michael Lavorgna's picture

I've taken part in the following comparisons:

  1. the same tracks converted from the same master tape and conversion hardware/software to PCM and DSD
  2. the same DSD tracks converted from the same master tape using different A/D converters
  3. vinyl rips to PCM and DSD from more than one vinyl setup using different A/D converters and software
  4. master tape compared to a double rate DSD copy
  5. the same tracks in single and double rate DSD
  6. the DSD, PCM, and vinyl version of the same track

And I may have left out a few comparisons and I've heard dozens of systems playing back DSD files. If we add up all of the variables involved in these comparisons and stack them one on top of another, we could touch the moon.

The only logical conclusion one can draw, imo, is simply to find the version you prefer on your own hi-fi.

But I agree that mastering quality trumps file format.

Charles Hansen's picture

.... it's called "The Scientific Method". Change just ONE variable at a time and compare results. I believe this is the first time that this has been done. Everything was exactly the same except for the modulation method. Neither method "destroyed" the other. Both produced excellent sound. But one has 30 years of usage and investment of equipment. To say that we should throw it all out the window for some extremely small (or imaginary) gains is the height of folly, particularly when the practical disadvantages of DSD are real and huge.


Michael Lavorgna's picture

I'm certainly not advocating doing away with PCM, but as a listener I welcome DSD as another option just as I welcome higher resolution PCM formats.

SAS's picture

Actually, nothing about the scientific method prescribes single factor experiments. If you change a single factor, all you have proven is that the factor under test has, or has not, produced a measurable response in the limited inference space of that one experiment.


In this case, the variable is not PCM versus DSD generally, but the algorithms used to make the conversion from the 6-bit data stream. Listening back to these files through any one DAC confounds that variable with the treatment of the data types through that DAC. (The QB-9 is no exception.) As ML points out, matching output levels is just the most obvious sign of such differences.


Where the experiment could start to have some useful conclusions is if the files were played back through multiple DACs, and a consistent preference emerged. Still, that would only be evaluating the algorithms used in the ADC -- not anything inherent about PCM or DSD. And, since we cannot compare to the analog original, we can only express a preference for how the ADC handles these limited dynamic range and limited bandwidth sources, not which method provides a higher fidelity transcription of the original. (These questions are both interesting, but different.)


Lastly, the claim that a 6-bit delta sigma ADC is more akin to DSD than PCM is rather odd. DSD is a 1-bit system -- which, despite all its problems, is inherently linear. No matter what math you do with it afterwards a multi-bit converter at whatever sample rate is PCM.


 Personally, I enjoy listening to both PCM and DSD program material through my Ayre.


S. Andrea Sundaram

SoundStage! Network

2_channel_ears's picture

"...simply to find the version you prefer on your own hi-fi."

Isn't this what it's all about folks?


firedog55's picture

Okay, downloaded and listened. Had this big insight: they both sound really good! Differences are very slight. I'm not sure how to phrase it, the DSD sounds a little more "rounded" and softer to me. The PCM sounds a bit more focused. I don't know if that means one or the other is more or less accurate. 

I'd think the format you prefer might be related to how the rest of your system sounds and just your personal taste. Or it might vary according to a specific recording. 

This test is very useful though: it made me realize I shouldn't worry at all about what format I get. Either one. If one has a better master, that's the one I'll want.

Michael Lavorgna's picture

This test is very useful though: it made me realize I shouldn't worry at all about what format I get. Either one. If one has a better master, that's the one I'll want.

And it applies to the recent Kind of Blue 24/192 and DSD releases (which I bet is one example your thinking about ;-).

JR_Audio's picture

Charles, I am very happy that you released these files for comparison and also have written some explanations about DSD and PCM.

I am a strong believer, that with the actual AD and DA chips (all sort of multi level delta sigma converters) the difference between DSD and PCM lies mainly in the representation of the digital filter. Let us just compare DSD64 with PCM 176.4 kHz 24 Bit.

If you are using the typical digital linear phase school book filter with symmetrical pre and post ringing, then you will hear this typical “PCM” sound. DSD does not have these filters, neither at the AD nor at the DA side, so this is a big advantage to DSD.

But I must say, that this is also the only advantage of DSD, because of the one bit architecture it needs some heavy noise shaping in order to get good dynamic range in the audio band and pushes the noise up above 20 kHz.

So basically above 30 kHz, there is more noise than music content in DSD and furthermore, I strongly believe that this heavy noise shaping is audible, as also heavy noise shaping are audible in the PCM format (when shaping it driven too hard).

As also described by Charles, it is possible in the PCM world, to have acoustically optimized digital filters, that have only very short and some have non pre ringing. I you would have these filters at the AD and DA side, you would not hear the typical PCM sound at all.

Take for example the NAS DACs. Yes I know, they have massive “problems” with Alias distortion, but the timing behavior is ideal, so they sound more similar to DSD as to PCM (but do have still the issue, what filter was used at recording side).

What I am trying to say is that, if you have optimized digital filters at AD and at DA side, then you have best of both worlds (DSD and PCM). You do not have the typical pre ringing artifacts and you do not need heavy noise shaping.

DSD Playback and DSD Streaming, DSD Tagging is in my opinion in some fields much too tricky to be used for the average customer, than PCM. All this Alias or Proxy Files in order to be organized in your music library, just to name one.

When I compare the files, given by Charles Hansen, some say DSD sounds more soften in the treble, but one can also say, it sounds more smeared in the treble (because of the massive noise shaping and more noise than signal in that region).

I would like to see more education about the differences between DSD and PCM to demystify the DSD hype and marketing, because, as I mentioned above, with clever digital filters at AD and DA side, you can beat DSD in every respect.


CG's picture

If anybody cares, I agree with Juergen.

If you look at the web traffic this adventure has generated, it mostly falls into two camps.

One is the group of people who are offended at the test regimine.  Yeah, Ayre could have hired the Brian Eno Orchestra, set them up in a suitable hall, and used two optimized microphones into an optimized microphone preamp feeding two QA-9's in parallel (one for DSD and one for PCM) to record blissful music for these comparisons.  I bet a number of people would still be agitated, unhappy, and dismissive and would take to the bitwaves (whatever) to express their displeasure.

Us regular folks, who might wear a tennis shoe, etc., probably do fine with this comparison.  At least it has been done with music we might be familair with or at least have the opportunity to purchase and become familiar with.  That counts for something.

The smaller camp is the people who actually gave it a listen.

From what I can tell, these folks find small differences between the tracks.  They mostly note differences in the treble.  Not surprisingly, often they have a personal preference for one over the other generally, or maybe specifically in regard to one aspect of the presentation.

Well, I say bravo to these folks.  They made their own choices based on what they like.

(There is a subset of people who reside in both groups, but tend to fall into the first group because they don't like Ayre, don't like Charlie, don't like Colorado, don't like anything, or already have an agenda.  Oddly enough, their opinions of the tracks are pretty one sided.  So, I didn't include their comments in my observations of the second group.)

I don't believe the point of this exercise was to settle forever which is the better modulation scheme for recording audio.  If that actually happened, it was merely a happy accident.

Instead, it offered an opportunity to hear for yourself how good PCM could sound in comparison to what has been proclaimed to be the absolute unsurpassed fidelity champion.

Michael Lavorgna's picture

I don't believe the point of this exercise was to settle forever which is the better modulation scheme for recording audio.  If that actually happened, it was merely a happy accident.

Instead, it offered an opportunity to hear for yourself how good PCM could sound in comparison to what has been proclaimed to be the absolute unsurpassed fidelity champion.

Well put.

fmak's picture

Try the dCS 5 bit Ringdac. There may not be $10 ones.

jim tavegia's picture

It reminds me of the student(s) who struggled in algebra class and attend their first Calculus class. What???????

Louis Motek's picture

Charles, great article, an excellent and sober overview of where we are and how we got here. My favorite part was:

"[...] one of Sony’s more obviously amusing marketing contradictions. On the one hand they say that its [DSD's] extended bandwidth leads to a more natural presentation of high frequencies in the music, but on the other hand, they say that the high amounts of high frequency noise doesn’t matter because it is inaudible. I suppose the answer to that question depends on which marketing person you are speaking to on any particular day..."


Here's to those who still believe in reason!



Louis Motek

JR_Audio's picture

Can't all these Blue Ray posts, that have hijacked this DSD / PCM blog, not moved to an Blue Ray post? I find it annoying to read more about Blue Ray, than what this post should be about.


Michael Lavorgna's picture

I've moved all of the Blu-ray comments to the forum.

junker's picture

Morten Lindberg has a very interesting summary of the formats. He records in DXD PCM and mixes to both DSD and PCM so it's not like he is just converting analog tape straight to DSD128.


Digital reproduction of analogue sound

At venue recording sessions our analogue to digital converters can do both the one-bit DSD and the multi-bit PCM formats. We can also listen directly to the analogue output from the microphones. Digital eXtreme Definition is a professional audio format that brings “analogue” qualities in 24 bit at 352.8 kHz sampling rate. DXD preserves 8.4672 Mbit/s (3 times the data of DSD) per channel. This leaves headroom for editing and balancing before quantizing to DSD for SACD or PCM for Blu-Ray.
All audio formats on The Nordic Sound are sample rate converted from the same DXD master. Comparing them in our studio we find only subtle differences from DXD down to 192kHz and 96kHz. The obvious degeneration is from 96kHz down to 48kHz. We find DSD, as used in the SACD format, somewhat different in colour from PCM; in some mysterious way DSD is softer and more beautiful but slightly less detailed. In DXD we find the shimmering brilliance from the original analogue source as directly from the microphones. Linear PCM is offered in addition to DTS HD Master Audio on this Blu-ray with the purpose of convincing audiophiles of the true lossless qualities of commercial encoding. The stereo layer of the SACD and the LPCM 2.0-stream on the Blu-ray are both full resolution mix from the original microphones. Mostly we find that the microphone placements used for the surround make a fine stereo. Occasionally we put up extra microphones dedicated for the stereo stream. 
I personally prefer extremely high resolution PCM over the DSD and I would claim that DSD is not transparent. But it all comes down to what the sound from your speakers can do to your body and mind. I find that the placement of microphones has an infinite more important role in the final experience of music, than the difference between HiRes PCM and DSD. Sometimes a lie can be more beautiful than the truth!


barrows's picture

Junker: Thanks for posting ML's comments from 2L.  BTW, my understanding is that 2L records at 24/352.8 using no filtering whatsoever, so no filter artifacts are present in the initial capture.  If one has a 24/352.8 (or confusingly known as DXD) capable DAC, it can be quite nice to hear to hear 2L's recordings at their native rate, although the file sizes are a little large.  And then, for kicks, try comparing the 24/352.8, 24/192 and DSD versions of the same recordings.  This will give a pretty good barometer on whta format might sound best with your own DAC/system.  

Of course, the software 2L uses, and its filters, for conversion are also a variable.  But I am willing to assume, for myself, that the original AD at 24/352.8 is virtually transparent, considering no filters at all, and such a high sampling rate. 

Also interesting to read CG's comments on the future promise of much higher speed ladder style chips...  That is something to keep an eye on.

Louis Motek's picture

The legendary Burr-Brown PCM1704 parallel resistor R-2R ladder DAC has for the longest time converted 24 bit PCM audio at 352.8 kHz, 384 kHz, 705.6 kHz and even 768 kHz. But oh! Let's just go faster and faster and faster and faster, why don't we! 

People think high speed equates to better resolution but they are wrong. There is a subtelty here. What applies to the number crunching world of fast computation does not apply to better sound quality through a faster conversion process. 

Better resolution in a DAC results only from less Jitter during the process. You already have the capability of playing 20 kHz sine wave notes at full output level into your ear, even from 44.1 kHz. This is already a pretty much inaudible freqency which can damage your hearing just by being there too loudly. 

All this talk about better resolution in "new formats" is due to better A/D in terms of the Jitter incurred then and there. This means that for the new version releases, there is less Jitter in those new recordings. The enjoyment people perceive has precious little, if anything, to do with sampling rate. But sampling rate sells. 

The reason that for new releases people enjoy the higher sampling rate versions and perceived "degredation" from the lower sampling rate versions concurrently released is because those making the recordings made them initially at high sampling rate, and mathematically, artificially derived all the other lower rates through number crunching mathematical algorithms, akin in a numerical sense to digital volume control.

If the new releases had been made in the same old boring RedBook 44.1 kHz sampling rate directly, the same if not even heightened perception of better sonic quality would have been perceived. The "new and improved" A/D processes are done with less inherent Jitter first and foremost. The sampling rate thing is rather a liability when you need to then mathematically produce the Redbook version as well. 

It is unsexy, unpopular, and there are few fireworks, true, but the issue is lower Jitter, not higher and higher numbers in terms of ultrasonics we need to cut out of the signal anyway, lest they interfere with normal low distortion levels in the gear.

We've had 768 kHz for many years now. The issue is Jitter and mathematics. I advocate direct-level RedBook recordings and it can be made to move your soul in all directions. But you then have to work on "boring" Jitter reduction which, sadly, in general, people don't have the patience to examine more deeply. The big fast numbers are more sexy and that's all there is to it.  


Louis Motek

Louis Motek's picture

Oh, wait, no, it will never work! That of course would mean that the musicians would have to practice, the recording engineers choose wisely an input level, be acquainted with the material...

Nahhh, it'll never work. 


Louis Motek

CG's picture

Actually, it's more than just jitter.

Converters are not perfectly linear.  Big surprise.  So, they generate harmonics as well as IM products.  That part is no different from any other circuit that does any linear processing.

What is different is that converters not only work with the "baseband" frequency range - audio in this case - but also at the sampling frequency and at multiples of the sampling frequency.  You get distortion effects around those frequencies as well.

"So what?"  You might say.  "I can't hear that?"  Well, I bet you can.

Take a sampling frequency of 44.1 KHz as a number I'm just pulling out of thin air.  Let's assume that the converter is converting some tone at 10 KHz.  You probably can hear 10 KHz just fine.  OK, not only do you get the 10 KHz out, but you also get 44.1 KHz plus 10 KHz (54.1 KHz) as well as 44.1 KHz minus 10 KHz (34.1 KHz).  This is just how converters work.  (It's far more complicated than that, but go woth me if you will...)

Well, those products may not be so audible.  Whether your amplification stages that follow can do well with these tones - if not filtered - is a different question for a different time.  But, 34.1 and 54.1 KHz are pretty much considered outside the audible range.

BUT...  How linear is the conversion?  Let's say that you also get third harmonics.

OK, for the baseband you get 10 KHz times three or 30 KHz.  Maybe not so bad.  At 44.1 KHz you get 44.1 KHz plus three times 10 KHz or 74.1 KHz AND 44.1 KHz minus three times 10 KHz or 14.1 KHz.  Oops.  How about fifth harmonics?  How about higher order IM products? 

Not many of those are the harmonics your ears are likely to find in a natural setting, like sitting in front of an acoustic guitar or on the savannahs where your ancestors were trying to avoid being eaten.  Not so good sounding.  Maybe even annoying.

Note that you really can't filter out these tones, because they are in the audio band!

But, if you move the sampling frequency up to 96 KHz, for example, these products move way further out of band.  (Math left to the reader...)

The problem is actually more complicated because of the way modern DAC chips work, and because the linearity of the converters is not constant with frequency - they tend to get worse as you go higher toward the sampling frequencies.

There is also the matter that the conversion from digital to analog really relies on filters of some kind to reproduce the sine waves, especially as you get closer to the so-called Nyquist frequency.  That aspect requires too much typing for me to try to explain here.

Bruce B's picture

A file recorded natively in DSD will sound a lot different than a PCM recording that has been upsampled to DSD. So the 2L files are not a good indication of what the formats will sound like.

Besides, Weiss Saracon conversion places a steep filter on any DSD material.

brrb's picture

Isn't DSD meant to be 6dB down on PCM? Either the PCM or DSD files must have been processed after recording to achieve the matched levels of these files surely?

junker's picture

Much ado about nothing IMO...

I have more storage in SACD ISOs than I do with lossless PCM, and don't want to worry about it one way or another.

Either you can make your DAC play ball with both formats, or you can lose me as a customer. It's that simple...

rexp's picture

CG, thanks for explaining some of the benefits/pitfalls of high-res. Personally I think a digital recording on vinyl sounds better than the same recording on cd or download. Which leads me to the conclusion that the reason I don't like digital playback is not down to the digital recording but somewhere else in the chain. Thoughts anyone?

Vigna ILaria's picture

The bottom line here is that Ayre's "definitive test files" do not offer the possibility to compare DSD to PCM at a fundamental level.  They merely offer the possibilty to compare the DSD and PCM outputs of Ayre's ADC.  A nice bit of marketing!

If I read Ayre's article correctly, their DAC is based on an ESS9016, a chip considered to be a significantly lesser performer than its 9018 sibling.  Hardly a comparison of reference quality.  Reference DACs such as Light Harmonic's Da Vinci Dual DAC (and others) employ totally independent DSD and PCM output stages.  And not just because the Da Vinci is a R2R Ladder DAC -  the analog filters in the output stage need to be oprimized differently for PCM and DSD (at least if no-compromise performance is being sought).

If you are looking to make absolute claims regarding DSD vs PCM then you need to be using only absolutely the finest ADC and DAC components available at any price, otherwise all you are doing is comparing known, compromised, hardware.  Even then, the argument that you are only ever comparing hardware remains difficult to entirely eliminate.

jwf2.tao's picture

Question from a simplistic music lover who really doesn't want to get into all the technical arguments. My one take away is the comments regarding quality of master being the defining difference in the quality of playback with these two formats- where does one find information that is accurate and honest(not just bs marketing) about the quality of masters? Mahalo...

JIMIXY's picture

Having listened to this 'Helplessly Hoping' sample as objectively as possible I genuinely think the DSD does have an audible advantage in this comparison. I basically think Michael Lavorgna has it dead right - there is something slightly flatter and more bleached with the PCM in comparison with the DSD. I think there is something more realistic in the low level fret/fingerboard noise from the guitar at the beginning and the dynamic swing when the voices enter seems that bit more full and lifelike (for want of better words). The low level vinyl crackle seems more authentic and the whole sound seems more intimate and somehow gels better together across the stereo field.

Just my thoughts, thanks for the comparison, was a joy to listen.