Another review of the DS DAC

[tonydennison]

Guilty as charged. My apologizes.

T

[andolink]

@Elk

I am interested in whether the reviewer’s conclusions are valid and, if not, how they are flawed.

Me too!

[tarnishedears]

I’ve met Amir. He doesn’t listen to equipment that he is evaluating, except as an after though. He mostly just measures DACs using this incredibly expensive piece of test equipment which he owns. And he clearly believes that measurements trump everything else.

The fact is that Amir fundamentally doesn’t respect any DAC design which attempts to take a design approach which is not based around todays best measuring sigma-delta DAC chips. If a designer wants to write their own DAC algorithm using an FPGA to try to take a serious stab at advancing the state of the art, but in the process they don’t do it the exactly same way as every delta-sigma DAC chip does, he will call that DAC poorly engineered. But that is not a fact. That is merely an opinion based upon his personal expectations of how a DAC should be built.

His opinion about how a DAC should be properly designed is irrelevant though. Because the goal of these designers was never to produce a DAC which are identical to all of the other chip-based DACs out there. Instead their goal was to advance the state of the art in terms of human sound perception.

If ted had wanted to, he could have easily designed his DAC around any number of common chips, and the result would have been a DAC which sounds the same as most everything else out there. But what would have been the point?

The fact is that this was never Teds goal. The same thing applies to all of those NOS, or R2R DAC designers out there which their relatively poor-measuring DACs which many people love the sound of.

The DS has been designed to measure about as well as a passively-converted DSD DAC can measure with the production compromises which had to be made to build the DS to its price point. But the fact is that this type of design necessarily will inherently have a substantially higher noise floor than will a chip DAC by its very nature. But this does not represent bad engineering in any way! This merely represents a different set of choices in terms of engineering trade-offs.

The TSS DAC will have made fewer of these compromises than has the DS. But fundamentally it is also a passively-converted DSD dac, and as a result, it will have a high noise floor as well (although not as high as the DS). But again, this was a deliberate design choice to produce a truly unique sounding DAC, not “bad engineering” as Amir loves to claim.

[rajugsw]

Time for some “Brown on Brown” comments. I’m not gonna argue about this Amir Guy’s test equipment and how he does what he’s doing. What I strongly disagree with is his listening tests are so flawed, it’s really not funny. He doesn’t site track names, source resolution. and so forth. He’s doing audio through Headphones. Absolutely THE WRONG method of publishing audio quality of a product IMHO.

What this Amir Dude is doing is supposedly debunking “myths” in the audio industry to which I too am skeptical of (don’t get me started). But he has a bad habit of purposely tearing apart High End manufacturer’s online claims by hiding behind measurements that we know maybe correct but measure that way for a reason. SOUND QUALITY !!! Everyone on the PS Audio forums knows that measurements only matter when you are doing product development. So many audio products out there measure perfectly but sound like crap. Now I don’t know PS Audio’s product development methodology. What I do know from work that I’ve done in non-audio related projects is you start off with a reference and then you purposely detract from it to make the thing BETTER. Sometimes this thinking works and sometimes it doesn’t.

Next year, I plan on getting a DSD Sr. used once my Stack is paid off (I’ve mentioned this on other PSA forums already). Thank you to Ted Smith for chiming in.

BTW, the “member” of Audio Science Review that lent him the DSD Sr. What is HIS equipment rig consist of (Speakers, Amps, Cables, room size/acoustics - no photos).

[abeiklou]

Not a good review, quite shockingly poor result. Could someone do a response on WHY there’s such high noise levels, issues with bass, and top end? Also, the claim the issues are caused by output transformer, is that only applicable to DS Sr, or it’s the same with Jr which is using a different output? Saw Ted’s post, but it doesn’t address the issues in details.

Also, in the comments, someone said DS Jr has massive heat generation problems, is that true? I was about to buy a used DS Jr.

[IanB52]

Can’t say if/how the DS Jr differs, but transformers add harmonics, especially in low frequencies creating a bass bump, and saturate at high levels. Besides some noise leakage from the mains and possibly display or FPGA, almost his entire criticism is on the use of output transformers. Now, if you are familiar with the sound of transformer based equipment, it’s not nearly as shocking as what shows on the plots. Tons of preamps and amps have transformers in them and sound great.

[tarnishedears]

The basic design of the DS being a passively converted DSD DAC NECESARILY will have a relatively high noise floor. The laws of physics make this a fact.

But this particular method of D to A conversion was a deliberate engineering design decision by Ted, and it is in fact a large part of what makes the DS a DS.

Also, don’t go thinking that because Amir pushed the transformers too hard in his evaluations that this means that the Jr is a better sounding unit because it uses op-amps instead of transformers. Because while the JR is a great DAC, it definitely sounds somewhat less extraordinary than does the DS.

[abeiklou]

DS Jr, used op-amps, so should have a different measurement I assume and possibly not the same issues Amir has found out. Maybe even lower noise? hmm…

[abeiklou]

Well, if his theory is right, the Jr should have lower noise floor due to not using a transformer on the output stage, IF we believe that’s the main cause.

[stevensegal]

The first time I ever posted on PS Audio was about the my PS Audio PWD Mk2 DAC failing an independent test of this nature.

Signal Test

[Q-copy.jpg]On my speaker supplier forum the focus is on objective performance and we are currently discussing if there is any significant difference between DACs. Some surprising results have arisen with a basic 1Khz peak signal applied to some very cheap and very expensive DACs. (On that forum brand names are rarely mentioned, so cannot say what they were.) However, the cheap DAC performed very well and the expensive one very badly. Another member posted a link to a site that runs tests on a…

Ted replied quite extensively on that occasion. I think we agreed to disagree. It’s the old subjective v objective thing to which there is no resolution of competing views.
It did influence me because instead of ‘upgrading’ to a DSD DAC (cost £6,000), I “upgraded” to an Audiolab MDAC+ (cost £800) from the PWD Mk2 starting point of £2,500.
I just remain firmly of the review that good measurements is a prerequisite before anything else, before any subjective listening observations. That was the prevailing professional approach to consumer audio for many decades, when every review included tests against manufacturer specifications.
That said, I appreciate that valve amplifiers generally measure badly, but the distortion they add can result on a subjectively pleasurable experience. If I want that distortion I can add it to my system using EQ, but my starting point is a flat response.
As noted above, the reviewer pointed to the issue apparently relating to mains noise and the output transformer.
I would add that headphones would seem the best way to do listening tests, and I’ve been told that by speaker manufacturers! Speakers are far to variable and many of them are too bright to start with.

[Dirk]

Personally I couldn’t give a toss about how something measures, all that matters to me is how it performs at a musical level. I own and play a vast LP record collection but if I judged vinyl playback solely by how it measured I would throw the deck in the landfill.

It’s all about musical enjoyment. Similar scenario with the DS; DS lifts my spirit much higher and engages me - takes me closer to the music, a joyful pleasure - better than any other dac I’ve owned. I bought a second DS for the family room which I’ve never done before. DS is in my top 5 products I’ve owned which is no small amount.

I am looking forward to taking delivery of the TSS!

I go by what I hear and feel not by reading a spec sheet.

[tarnishedears]

The JR in fact has a higher noise floor than does the SR.

Amir does not respect the basic design of the DS because it is not simply yet another super-duper measuring Delta-Sigma chip-based DAC. But that was never the design goal of the DS.

[wobblewobble]

@tarnishedears

If a designer wants to write their own DAC algorithm using an FPGA to try to take a serious stab at advancing the state of the art, but in the process they don’t do it the exactly same way as every delta-sigma DAC chip does, he will call that DAC poorly engineered.

This part I’ve quoted is not true.

Review and Measurements of CHORD Qutest DAC

[stevensegal]

Please remind us what the design goal was?

I would be surprised if it wasn’t to take a digital input and turn it into an analogue output with the minimum amount of noise or distortion. I’m not sure what else you want a DAC to do? Just as with an amplifier, I assume the goal is to take an input voltage and output a larger one with the minimum amount of noise or distortion.

Or am I being naive?

[stevensegal]

@wobblewobble

Audio Science Review (ASR) Forum

Review and Measurements of CHORD Qutest DAC

This is a review and detailed measurements of CHORD Qutest DAC. It is on kind short-term loan from Gig Harbor Audio. I previously tested a the older generation, Chrodette and 2Qute DACs so naturally members were curious how the current product line…

He concludes on the Qutest:

“The CHORD Qutest shows that boutique/custom need not come with the heap of distortion as many others I have tested do. On that front, designer Rob Watts needs to be congratulated by not sacrificing measured performance for some unverified audiophile notion. Then again, I wonder how good of a DAC Rob could design using an off-the-shelf DAC chip. Likely would be just as good and cost a lot less…”

The Qutest costs £1,200, less than the Stellar GCD (£1,550) that does not use Ted’s bespoke FPGA processing. The Mojo DAC, than now sells for £300, uses Chord’s FPGA on a smaller processor, so I don’t think many people are complaining that Chord do not do an ESS DAC for £200. The March Audio DAC1 measures brilliantly with an ESS chip and costs £265.

Call me the party pooper, but I was convinced 7 or 8 years ago that digital audio sources would become very good and very cheap to the great chagrin of the posh end of audio, and Class D amplification would eventually do the same to Class A/B.

[wobblewobble]

@rajugsw

He’s doing audio through Headphones. Absolutely THE WRONG method of publishing audio quality of a product IMHO.

Really? There are no room acoustics issues at play… That is a rather large benefit…

If you’re a fan of 2L’s recordings, you should probably let Morten Lindberg know he’s doing it all wrong…

And before anyone jumps in, yes I do know he also uses B&W speakers for monitoring… especially for his multichannel stuff…

[wobblewobble]

@stevensegal

He concludes on the Qutest:

Yes but as I mentioned in my post, I was replying to the following, which is not accurate:

“If a designer wants to write their own DAC algorithm using an FPGA to try to take a serious stab at advancing the state of the art, but in the process they don’t do it the exactly same way as every delta-sigma DAC chip does, he will call that DAC poorly engineered.”

[stevensegal]

@wobblewobble

Yes but I was replying to the following, which is not accurate:

“If a designer wants to write their own DAC algorithm using an FPGA to try to take a serious stab at advancing the state of the art, but in the process they don’t do it the exactly same way as every delta-sigma DAC chip does, he will call that DAC poorly engineered.

I think we agree and are both disagreeing with @tarnishedears, because the Chord FPGA DAC measures extremely well and the PSA FPGA DAC doesn’t.

I think we also agree that headphones are the best way to test, especially for noise and distortion, given the ambient noise level in most rooms.

[B.Gelinas]

Please find below an exchange I had Aug 28, 2017 with Scott Schroeder and Ted:

------ My Question ------------
Hi Scott,
I read a post on a forum (http://www.psaudio.com/forum/directstream-all-about-it/stereophile-review-of-torreys/, January 13, 2017) where Ted Smith mentioned:
“The point is that the changes are all below the noise floor. In particular in earlier measurements JA reports changes in resolution around -120dBFS with earlier software upgrades but still reports that the resolution is “only 17 bits” (-102dBFS). The 2nd is the noise floor and the first is resolution (the ability of the output to show small changes in the input.)”

So, my question: when I buy high resolution 24-bits music files, will the DirectStream DSD DAC effectively only play 17bits out of the 24bits?
------ End of My Question ------------

------ Ted’s answer to Scott --------
Howdy Scott

Everything is a tradeoff: find a PCM DAC that’s linear down to one part in 16million whereas DSD DACs are inherently linear (two points define a line.)

All 24 bits are represented and used, but there is also a higher noise floor than we are used to with PCM dacs. As with dither, the bits below the noise floor do count: e.g. with a steady state tone the noise floor doesn’t affect the long-term average of that tone even when it’s entirely below the noise floor. Said another way if you average over time the noise decreases (on average) leaving the original signal standing out with accuracy to 24 bits. (BTW tho this happens with noise that’s uncorrelated with the signal (as in DSD_, this doesn’t happen if the noise is correlated with the signal.)

Noise is an unavoidable side effect of DSD encoding but the better linearity and simplicity of the output stage (e.g. “simply” a low pass filter) compared to PCM are quite audible. That isn’t to say that one can’t spend a lot of money to get very accurate components for, say, a R-2R PCM output, but even then state of the art is nearer 20 (or 21) bits than 24 bits. For 20 bits you are talking about resistors accurate to 1ppm. (Try to even find resistors with a temperature coefficient near that.)

The ear/brain is extremely good at filtering out uncorrelated noise (if it weren’t, tape, vinyl, and other analog media would sound much worse.) The real trick is to not have any noise correlated with the signal (which is hard in PCM) because the ear/brain is much more sensitive to that.

-Ted
----- End of Ted’s answer to Scott ---------

----- My reply to Scott --------
So,
I will get 17 bits, with the 7 least significant bits below the noise floor, but if the noise is uncorrelated, I may benefit from all 24 bits with the original signal standing out with 24-bits accuracy.

And I do understand from Ted’s answer as well that PCM R2R DAC will not be able to extract all 24 bits (21 at most) because it is not possible to get resistors that will provide the required excessive precision and temperature stability to decode least significant bits.

So in conclusion: unless you are able to pay sky-high prices, no affordable DAC (and I would say that the directstream is certainly near my limit of “unaffordable”) can really claim to be 24 bits.

In other words, marketing is one thing, implementation limitations are another.

Did I get it right?
----- End of My reply to Scott --------

----- Scott’s answer ---------

Yep, you have it correct.

I’m glad you were able to follow the Ted-speak without any trouble—I’m sometimes hesitant opening that font of esoteric knowledge with folks.
---- End of Scott’s answer ---------

[abeiklou]

Hmm, this R2R DAC supposedly does 26-bit, how does that correlate then?

https://www.denafrips.com/terminator