Review and Measurements of PS Audio PerfectWave DirectStream DAC

[AudioSceptic]

You can have THD without N, though I agree with Amir they probably mean THD+N.

THD is adding together all the harmonics. Actually a RMS value of all harmonics compared to the input signal. To add N, the way it is done is taking the value of the input signal, notching out the input signal and measuring the value of what is left vs the input signal. So you are including all harmonics plus the noise. Sometimes noise is the biggest item, so THD is swamped. More commonly THD is the largest part with a little contribution by N.

Yes, I get that THD is total harmonic distortion, but in Amir's dashboards the SINAD figure often looks like the worst of any of the single harmonics in the FFT, unless N dominates. Am I way off here?

 

[Krunok]

Yes, I get that THD is total harmonic distortion, but in Amir's dashboards the SINAD figure often looks like the worst of any of the single harmonics in the FFT, unless N dominates. Am I way off here?

You are. That is because of the way THD is calculated - small values of higher harmonics get even smaller when squared so the largest harmonic values dominates.

 

[carlosmante]

For the sake of argument assume that there's an error in the measurements. For instance, the measured performance is 15% than in reality. Since the methodology is the same across DAC's, A will be better than B, even in the presence of such an error in measurement.

Results would not be informative if they were completely random. However, it's very unlikely that they are random as they often approximate official specs and specs posted in other sites.

So, a probabilistic statement that you can make is that with a fair degree of confidence ASR measures are informative, and in this particular case the DAC underperforms alternatives.

You are dismissing the sampling bias. If you want any results to be informative they should be under a random sampling. Remember that
Var (p̂) ≈ σ2/n

 

[pkane]

I have no problem with these measurements being used as engineering goals, real or virtual for yourself. Consumers in the real world don’t have engineering goals (except for a niche tech-nerds). They have audibility and price goals and perhaps reliability goals related to engineering. If you don’t agree with this premise, then there is no debate since we have very different perspective of what the real world is. We will agree to disagree.

If a 100db SINAD device costs $1500 and a 90db SINAD device costs $500 and the features are identical and aesthetically acceptable, should one buy the more expensive one? This is where the engineering goal criterion breaks down.

Nobody except perhaps the tech-nerds here will argue for the more expensive unit because it satisfies the engineering goals better. Good for them.

But how do you answer the question from a user I posted earlier. Try answering that seriously and see what makes sense. No hemming and hawing. May be you can and may be you cannot but hey I like the better numbers because 24-bits, jargon, jargon, dynamic range, jargon, jargon, controlled tests, no difference, jargon, jargon, etc.

It is fine if a bunch of similar minded people made the engineering goal their criterion and created an echo chamber for themselves here but that is not a solution to anybody outside. The conflict happens when these reviews don’t stay within the choir here but pretend to have real world, consumer implications outside it and solicit comments from outside or pretend to have implications on the industry that caters to the outside.

That is where the problem is. Instead of looking for ways to solve the problem by identifying the limitations here and its applications in the real world of people, I see people repeating the same talking points like the Bogleheads use the “active management bad, indexing good” mantra over and over again. That is a cult.

It has nothing to do with psychology of hearing at the base level to hide under that nebulous term.

Almost every consumer outside this echo chamber will want to know things like -

Does the higher SINAD position in the ranking table mean that the sound will be better than the one below it and so should I buy the one higher up at a higher price? Should I not buy this unit at the lower end of the scale because I am going to be unhappy with that sound? If two units have the same SINAD ranking, are they going to sound the same? I liked the sound of my old amp from brand X but I need a replacement. What can I buy that will sound similar?

If those are banal questions or questions that are based on subjective preferences and something that we should not bother with, then let us sit in this artificial world for ourselves with a groupthink that we have the best goal. Or we could, in the tradition of science, try to study what bridges the gap between the measurements and those needs.

I would say the engineering approach at the moment utterly fails to answer any of the above in any satisfactory way. And that is what I meant by solving real world goals.

But if it is fine to have an echo chamber here in that elusive goal of perfection where the chase is more important than the result, fine by me.

Proper engineering is not a specious goal. DACs and electronics are designed and built for a purpose. They don't grow on trees. We don't discover them and pick the one that tastes better: we design them based on knowledge, science and engineering to faithfully reproduce recorded sound. If one is designed with a specific taste, then it should be advertised as such. The most transparent DACs will have no taste at all.

The logic is pretty simple: start with something that's provably transparent and clean. Then, add some DSP to taste, if you prefer. No cult, no echo chamber, just a logical approach that makes sense to people here. If your choice is to buy a $6k DAC to add some designer coloration to all the recordings, that's your choice. But don't claim that this is the best approach. It isn't when the same can be achieved with about 1/10th worth of equipment.

 

[AudioSceptic]

You are. That is because of the way THD is calculated - small values of higher harmonics get even smaller when squared so the alrgest harmonic values dominates.

Thanks.

 

[AudioSceptic]

Conversely, a company producing gear that regularly falls short piques our interest. We'd secretly like to see real improvement but an extrapolated trend line prepares us for failure. It does not lead us there. That is entirely a matter of how the gear performs under lab conditions.

It's hard not to look at a car crash even if we think someone's badly hurt or worse (in the UK this is called "rubbernecking"). There's also a reason disaster movies make so much money. There's a perverse streak in us that can't resist looking at things gone or going badly wrong.

 

[AndrovichIV]

This is probably both the best post on ASR and the most satisfying and revealing.

God knows they gave you a kicking but you showed all your best qualities and were vindicated. It was hard at the time but it's was the best thing to happen for us.

For me it was total DAC

 

[AudioSceptic]

I would argue that the design goal of audio equipment should be not to have a sound as such but to faithfully extract, amplify and play source material accurately. Any sonic signature should form part of the source material. If people like coloured sound then use EQ or DSP, one of the great mysteries of audio is why so many audiophiles have a pathological aversion to tone controls and DSP then buy deliberately coloured equipment. Why not buy accurate equipment and then colour it however you like with EQ and DSP. If it was a cheap and cheerful product sold honestly it would be different (for example the Polaris headphone amp reviewed elsewhere on this website is very open about reproducing a valve sound rather than measuring well and is modestly priced) but this thing is a $6000 DAC sold as hifi.

The bridge analogy is interesting. In a mechanical structure there will be thousands of calculations forming part of the design and an awful lot of measurements but it is not about creating perfection, it is about meeting a specification and building something which is fit for purpose. The problem for audio companies in the high end is that if you consider the specification as being to deliver audibly transparent performance, which does not need SOTA measured performance, then this has been commoditised up to the speakers and can be found for low cost. That doesn't make expensive audio irrelevant as things like industrial design, build quality, through life support, durability, functionality, measured performance and just plain wow factor still matter to many buyers. However a lot of high end equipment equipment is not particularly well made or blessed with good industrial design and if it is not about the SQ and it does not tick the boxes on build and design and measures badly then it is hard to see why some gear should exist.

What's worst about coloured audiophile gear is that, other than a non-flat FR, the distortions can not be corrected by anything else in the chain, IOW you are stuck with them for everything, not just for the few pieces of music that they might benefit (being generous here!). Something transparent, OTOH, gives you a much bigger choice in tailoring the sound if you wish to do that. It's strange that we've got to this. You don't see equivalent products in cameras or video displays/TVs. Yes, there are premium brands but they are superior products or ones made in countries with higher costs. They still "follow the rules" and are objectively as good or better than the mass market ones.

 

[GrimSurfer]

What's worst about coloured audiophile gear is that, other than a non-flat FR, the distortions can not be corrected by anything else in the chain, IOW you are stuck with them for everything, not just for the few pieces of music that they might benefit (being generous here!).

This is an excellent point that justifies bringing various pieces of familiar music to a sound auditioning [and choosing the component with excellent measurements that make the least sonic coloration on them all].

 

[AudioSceptic]

Well, that is exactly what "High Fidelity" is supposed to mean. Looks like a ton of people forgot about it...

Am I wrong, or Chord Electronics is the only exception known on ASR?

I think we need to define "boutique". Does it just mean expensive and low volume? Does it include pro gear?

 

[scott wurcer]

What surprises me is how much these results are tainted by transformer artifacts (BTW no firmware upgrade can fix that). Many of the classic recording mics had matching (SE to Diff) transformers in fact I have several NOS Nakamichi CM300 mics that are highly sought after by DIY concert tapers and are capable of making some fine recordings. They have a tiny low cost 2000:200 Ohm transformer in the body of the mic.

The multitone plot looks similar to the one SY got from that L/R/C phono stage. I can only imagine the transformer is being used outside it's normal recommended range. I suspect there is also some of the level dependent frequency response going on.

Before I retired we measured some differential headphone reference designs for Chinese cell phone manufacturers, they were quite respectable at <$2 total BOM.

 

[HammerSandwich]

The logic is pretty simple: start with something that's provably transparent and clean. Then, add some DSP to taste, if you prefer.

And if EVERYBODY would do this, the Circle of Confusion would shrink tremendously.

 

[Veri]

About transformer coupled XLR output. My Audial AYA TDA1541A DIY DAC outputs single ended out to jensen transformers so that it can benefit from common mode rejection over XLRs:

Audial has graphs too on the effect of Jensen transformers on this 16-bit DAC.

Capacitor-coupled output:

Transformer-coupled output:

I think one can reasonably say these Jensens fare a whole lot better than whatever PS Audio is using. If one would use Jensen transformers in something like Matrix X-Sabre or the Okto DAC though, the third harmonic distortion / odd harmonics will definitely cause performance loss. One would better avoid transformers in the output stage. A modern opamp-based balanced output stage will result in far better transparency.

But with my old 16-bit DAC, it is a reasonable trade to gain common mode rejection. It does not sound bad at all and will use it until the dreadful day it might crap out

 

[GrimSurfer]

Great point. The industry pushes this component over that component but the real trick is to match components. Toroids work fine in some devices, whereas EI cores do better in others. It depends on a lot of things like orientation, available space, weight considerations, susceptibility of a circuit to DC or magnetic fields etc.

 

[murraycamp]

How is the psychological aspect of how it sounds to any individual -- an aspect that will probably vary greatly depending on mood, physiology, environment, what sort of day they are having -- supposed to be anymore "real world" than what measurements show? If anything, the measurements allow us to rule out bias and listening idiosyncracies. Our ears are not the great arbiters of audio truth you seem to imply, unless it's a phenomenological truth you are searching for.

Word. I notice my system sounds different on some days depending on things as pedestrian as pollen count (I have mild seasonal hay fever) or other sinus issues, or whether I am all torqued up from a bad day (probably blood pressure increase).

 

[pozz]

@audimus The DAC SINAD table in this thread is correlated to a DAC's ability to reproduce a signal equivalent to CD quality (16 bits or 96dB), a format that's been around for 30 years.

Consider that the oldest DAC measured is the Arcam Black Box 3, released around 1991, produces a SINAD of only 84dB and still outperforms the PS Audio DAC. The top range of the SINAD chart (>120dB) is correlated with the extremes of engineering quality, corresponding to a noise and distortion-free dynamic range that exceeds the limit of human hearing (there are nuances here to do with equal loudness curves and sensitivity).

Under many circumstances the distortions produced by the bottom range of DACs will be inaudible because of quiet listening volumes, noise floors of typical rooms and psychoacoustic masking. The distortions produced by DACs are also very minor compared to those of recording microphones and speakers.

Keeping that in mind, the way to understand the SINAD chart is, as the measured value increases, so does, generally speaking, the quality of engineering. This is not true in all cases and no engineer would tell you that a single number is sufficient to describe the functioning of a product. The important thing, really, is to use the chart as a reference. The reviews that Amir and others write help interpret measurements in terms of the functioning of the given unit, which can be deficient or excellent in certain physical areas (power supply, output stage, grounding, D/A conversion, filtering, routing, component selection and so on), or in terms of functioning (interface, overall build quality, connectivity) or price.

The point is that you can't know how well the product was put together without measuring it. ASR has shown conclusively that you can put together an excellent dedicated product for a very affordable price, and that the state-of-the-art is within reach.

 

[amirm]

Audial has graphs too on the effect of Jensen transformers on this 16-bit DAC.

The transformer non-linearities show the most in low frequencies. So those tests at 1 Khz while useful to look at, don't show the full picture.

 

[617]

The transformer non-linearities show the most in low frequencies. So those tests at 1 Khz while useful to look at, don't show the full picture.

Amir - perhaps off topic, but do you have access to a Pacific Microsonics One/Two DAC/ADC? It's considered pretty legendary in some circles and I think a lot of people would be curious to see how it does compared to contemporary offerings. I think I read you had some connection to them.

 

[amirm]

Amir - perhaps off topic, but do you have access to a Pacific Microsonics One/Two DAC/ADC?

Unfortunately no. I led the acquisition of Pacific Microsonics into Microsoft and at the time I thought about getting one but I though, "why would I need one?" I regret that decision to this day!

 

[Veri]

The transformer non-linearities show the most in low frequencies. So those tests at 1 Khz while useful to look at, don't show the full picture.

Ahh, thank you for pointing that out Amir. Definitely something to keep in mind here.