Purifi Class D

[mocenigo]

3 hours ago, Allan F said:

 

I guess there are some people who just like to argue with anyone, notwithstanding the futility in doing so. 🙂

I worked incredibly hard for the last three months and now I am on vacation, I was just drunk from the feeling of freedom ;-)

[mocenigo]

2 hours ago, Axiom05 said:

So, nothing more to say about Purify amps? Anyone have a pair of Apollon PET950 mono amps?

 

What can we add? They are great, provided you have a good buffer in front of the modules. The modules themselves have an excellent rejection of power supply noise, so you do not need fancy "audiophile" power supplies.

 

 Roberto

[mocenigo]

1 hour ago, PYP said:

 and power conditioning.  the power conditioning I had previously for an A/B amp was not the right choice for class D, at least in my experience.  

 

what do you mean?

 

It is obvious that a properly designed audio component does not need a power conditioner or special power cords: if they do, it is just their own power supply that is badly designed, and this happens often, and intentionally, in many extremely expensive amplifiers, with just a huge transformer and some large capacitors, in order to make then “revealing.” (Please do not argue what I just say, I will not reply because those are facts than anybody with a limited understanding of electrical circuits can verify.)

 

However, a power conditioner having an adverse effect on the sound a class D amp is weird, unless the power supply is switching (and in fact the class of the amplifying circuit is irrelevant), the filter itself is highly inductive or contains ferrites (again, this can be confirmed by designers of such power supplies), and therefore there can be some trouble in the power supply.

 

But if it is a normal hifi or hiend power conditioner, which is usually almost empty and copies the internal of a Schaffer filter, I cannot fathom how this can happen.

 

 

1 hour ago, PYP said:

The other quality of a quality class D amp is that it will simply pass through what precedes it (to an astonishing degree).   One of the reviews of my own amplifiers had me scratching my head:  "How could anyone hear THAT from these amps?"  Look to your source(s) for some answers.  

 

if you are speaking about the source material and not the power supply, the most recent Class D designs surpass almost everything else, except for some AB designs which are roughly on par.

[mocenigo]

2 hours ago, PYP said:

play it continuously for one month and let us know what you hear.  These kind of amps take a very long time to break in.  

if you refer to high quality class D, they do not take much time. Maybe some caps in the power supply need forming, and there have been cases where some took several days to filter properly, but nothing more. It is the brain that takes long time to “break in” after listening to something that has extremely low levels of noise and distortion. It one then moved from an excellent class d amp to a second one (also excellent) that has had at least a few days of “break in”, nobody would probably notice a difference.

[mocenigo]

3 hours ago, leManu said:

Sadly the Nilai500 doesn't offer any replaceable opamp, so that's not an option. 

 

to be honest, replacing opamps only makes a difference if they are used in extreme conditions or you are using one well within its operating range, vs barely within (that’s why replacing the 1612 with a 1656 in the Purifi buffers may have an effect) vs outside it.  I infer that the Nilai500 has a very integrated design and in that case replacing the opamp would probably decrease performance.

 

3 hours ago, leManu said:

 

I feel like one of the main difference in sound that I experience is the lack of decay.  Notes just disappear so quickly that the sound is dry and lack texture, thus warmness.

 

Interesting, but this should never happen unless one of the two amps is broken by design. The disappearance of clearly audible sounds- as opposed to very subtle cues - is just not possible. So it was probably your previous amp that added some midrange bloom. It is legit to prefer it, it just is not hifi in my book. Or no longer, as the bar has been raised significantly.

 

3 hours ago, leManu said:

 I was never a fan of tubes, but again I was wondering if adding a tube pre amp or external tube buffer would give me back some of the warmth by adding decay to the sound.  And better imaging too.  But maybe I should just sell it and try another Hypex with opamp options...

SS opamps will not give you more decay and bloom, AFAIK - I do not know any that does that intentionally.

 

Or you could do what I have done. When I know (numbers do not lie - and they go way beyond the reductive SINAD, mind you) that a device is superior (and the numbers are not those on the price tag!) I let my brain burn in. In fact, now I cannot stand amplifiers with too much midrange bloom, and anything with more than a tiny little of warmth added will in fact make me sick. Clarity and purity of sound is still an acquired taste, nowadays.

[mocenigo]

1 hour ago, leManu said:

If my old amp was adding bloom in the midrange, which is very possible, would it also explain the superior soundstage I got from it?

 

Well, there are two things here.

 

A bit of H2 with inverted polarity can give the impression of a deeper soundstage (Nelson Pass even designed a toy circuit to play with that).

 

The second aspect is that IMD will increase the amount of apparent information associated to the signal, also slightly increasing the separation between two channels, resulting in a wider soundstage.

 

So, one may perceive a deeper and wider soundstage as a side effect of that touch of warmth.

 

 Roberto 

 

 

 

[mocenigo]

8 hours ago, Allan F said:

 

For those interested in the Conclusion, it can be found at 1:39:13

Sadly the conclusion are invalid. It does not seem to be a controlled test: not blind, volume not matched. So, they just were more impressed by the latest technology, and I guess they will be more impressed by the 1ET9040BA, and then by the next Hypex iteration and so on. 

[mocenigo]

54 minutes ago, Jud said:

Sadly your conclusion is invalid.

 

- The latest technology is the Purifi, not the Hypex.

 

- Their conclusions happen to be in line with the measurements (though the noise and distortion figures are so low for both Hypex and Purifi they may well be inaudible).

 

Edit: BTW, are you certain volume wasn’t matched? I believe there was a reference to the bother they went through to match volume.

*cough* the Nilai design comes AFTER the Purifi they tested.

 

Second, at that level the absolute levels of distortion and noise are only for bragging rights, they are not a meter of sound quality. In fact, the Purifi addresses also hysteresis distortion, and the Nilai documentation does not mention that, so they did not (otherwise they would mention it).

 

Third, Level matching procedure did not look well documented

 

And anyway this is a sighted test, so completely irrelevant for the tiny differences that there may be between these amps.

[mocenigo]

39 minutes ago, Jud said:

Yes, I understand the Nilai iteration of the Hypex design came later, but Putzeys designed Purifi after Hypex. So I suppose one could think of either as the latest technology.

 

in Italy we call this way or arguing “trying to climb a mirror” (with your bare hands, is implied) but you have my sympathy.

 

 

39 minutes ago, Jud said:

Do I suppose the amps really sound that different without hearing them for myself, and even then would I necessarily trust my conclusions? No.

 

But I believe there is a tendency to dismiss these listening impressions on the basis of a laundry list of the “usual suspects” (latest shiny thing, no volume matching), which in this case aren’t correct (volume matching) or arguable (latest tech). I just think, without arguing for the correctness of others’ sighted listening impressions, or even my own, that at least we should be careful about evaluating them on an accurate basis.

 

What seems clear to me is that Hypex put a discrete, optimised input buffer in the Nilai. If the test was conducted with the stock EVAL input board, performance may be different just because of that. Frankly, I did not double check which implementation of a Purifi amp they listened to, but from the images it is clearly the EVAL1. And there is anedoctal evidence that just replacing the OPA1612 with an OPA1656 on that board improves subjective impressions. I use a neurochrome Universal buffer and I can testify that it made a difference over the EVAL1z

 

I am quite sure that IF they can still hear a difference in a double blind test, then it is due to the input buffer.

[mocenigo]

1 hour ago, Allan F said:

 

But, Putzeys left Hypex before he designed the Purifi so, presumably, he did not participate in the design of the Nilai. If so, the latter would be the latest technology.

He confirmed that he did not work on that design, and has also provided some criticism of it (of course)

[mocenigo]

8 hours ago, davide256 said:

I'm getting the impression that you do not play/esteem any sort of wind or string instrument. Which escalate in price based on the strength of that warmth/timbre depth you are eschewing as much as technical capability.

I am not sure what this has to do with the subject at hand (also, I think you are misusing the verb “to eschew”, which means “to deliberately avoid using”). Any I do play a few instruments, including woodwinds and strings. Soooooooo, what are you trying to say? 

[mocenigo]

2 hours ago, Jud said:

Can’t find the criticism. Link?

He spoke about it at private conversations, and the criticism had been also mentioned to me by some folks at Purifi. It was about the shape of the distortion curve that in their opinion showed that they were optimised by trial and error instead of using a mathematical frameworks. In other words, it was only to make numbers better, not to solve a specific problem (which, in Bruno’s approach, is hysteresis distortion). So it is a criticism of the methods and goals, not of the results.

[mocenigo]

7 hours ago, PYP said:

Is the "in other words" part what they said?  I ask because the "only to make the numbers better" hypothesis seems silly to me.  The distortion was already so low that anyone could question whether better numbers = better sound.  And subjectivists use their ears, not the measurements (to the dismay of objectivists).  

 

They said both things, actually. And it is obvious that they cater to objectivists as well.

 

7 hours ago, PYP said:

One could just as likely interpret this theory as -- Mola Mola did less math but a lot more listening and finalized the result based upon listening rather than math. 

You mean Hypex, not Mola Mola. Well, they used measurements, and trial and error, not listening.

[mocenigo]

2 hours ago, PYP said:

Gosh!  Geek Gossip in the audio realm seems rather benign.  

 

I hope that after all the math and measurements there is extensive listening.  Since the Nilai amps include a discrete buffer stage, I'm assuming there was some tuning by ear.  

I do not know. Why design something so precise and then ruin it with some changes done by ear? It is like designing a very precise clock and then deciding on the mood of the day whether it should run faster or slower…

[mocenigo]

1 hour ago, PYP said:

Because the subjective reaction is the final measure of success?

 

Success of what? 

 

1 hour ago, PYP said:

But i would suggest asking the audio designers, including speaker designers who use extensive tests/measures, why they use their ears to determine the final product. 

I did I, and I do. What they do - at least the serious ones - is usually the following:

- They identify one or more problems that are audible for objects within their BOM limits.

- They identify (or try to understand) what measurable quantity is correlated with this problem.

- They choose a specific approach to change that quantity in order to solve the problem(s).

- Then, they verify by listening whether the problem has been solved. If so, then the considered measurable quantity was the correct one.

- In case only some problems can be solved within the budgetary constraints, they decide which ones to address and to which extent.

 

This does not change the fact that listening is a verification step, not properly a tuning step. At least if they are doing their job seriously. It is not “I like this sound more, so let us do it this way.”

 

1 hour ago, PYP said:

I know that the Tambaqui was tweaked according to listening tests.  

 

If they did, then they would have ruined its performance, and Bruno would never do that. The listening tests followed the procedure above. Listening is a business decision aid.

 

Other manufacturers design with a target of getting a very precise distortion profile, for instance. I am not interested and I consider it anachronistic, since it is cheaper to get a “perfect” amp and tweak in the digital domain. Taste can change, why should one go the expensive of trying several amps instead of changing a profile? So, I am not interested in considering a choice of design activity that is in my opinion bordering on intentional disinformation or fraud.

 

You also mention speakers - this is as of today a different matter in my opinion, since the distortions and frequency response variations are various orders of magnitude higher there. So you often have to choose a particular sound. This can be made also by taste or to follow a “company signature sound.” This is a necessity, but for electronics I posit that this should not be any longer the case.

 

I am not a sworn objectivist. I believe objectivism is a necessity to get to affordable, reliable, absolutely neutral and flexible equipment that we can easily tune to our changing tastes by simply tweaking a configuration profile…

[mocenigo]

1 hour ago, barrows said:

I disagree, and most audio designers I hear from do as well.  Bruno Putzeys, for example, has described his design process as doing the maths/simulations first, building a photo, measuring, then listening.  When there are problems found in listening, then he goes to the maths and measurements, and tries to find what the issue(s) are which account for the listening problem.  Then another proto is developed, and so on...  He is never going to "ruin" something by listening, but uses listening to inform his process, and to find new ways to measure for previously less well

understood problems.  Listening is part of the "scientific method" (observation), and never should be ignored.

The "standard set" of audio measurements, such as THD+N at 1 kHz, is woefully inadequate to insure best possible performance with a complex music signal.  So those designers interested in going farther, work with both listening, and measuring, to insure they are getting things right.

Take a look over at ASR measurements, now that they are including the AP 32 tone test.  In many of the components they have measured the SINAD looks great, but the 32 tone test shows a lot more distortion products poking up at audible levels-and even this is simplistic, as the 32 tone test uses evenly spaced tones.

For an example of an advanced measurement, I have heard of some manufacturers who, during development, look at square wave response, at many different frequencies, at many different places in the circuit, to make sure things are operating as predicted in simulation and maths.

For sure all of the above is "engineering", it is not just listening and then making near random changes to try and make it "sound better", but doing the work to get a circuit sounding really good is just not that simple.  These days a merely competent engineer can make a line level circuit look good for THD+N of a single tone, by using a text book IC opamp implementation, but on more complex measurements, that same circuit may exhibit problems.

then we agree 100%. Read my reply to PYP. What I object to is the concept of “tuning by ear,” which can only ruin things. Using the observation to verify whether a problem has been addressed is different and yo go always back to simulations and prototyping and measurements. You do not do changes by ear that go against measurements (unless the differences are in the realm of the inaudible) if your objective is “hi-fi”.

[mocenigo]

2 hours ago, PYP said:

Also, for the Kaluga amps, Kubala-Sosna cable was used to connect the PCB to binding posts.  Why?  Did it measure better?  Or sounded better and measured no worse?

 

There are many reasons for that. It may be that they know that the customers may expect audiophile cables and then they give the customers what they want. I am not saying that Bruno and Co do not believe in cables, but what they use themselves is rather mundane (like the cordial 425 as a speaker cable) and this means something.

 

 

2 hours ago, PYP said:

When one reads about respected designers who carefully choose certain capacitors, for example, the choice is stated not in terms of measured differences (which one assumes were used to narrow the choices, along with considerations of high rate of production meeting the design specifications, cost, availability, etc.) but preferred "presentation."

 

Capacitors are usually chosen for robustness and build quality. On “high end” products they are chosen according to customer expectations.

 

A difference in colour for instance implies a difference in frequency response - thus a measurable one. If the difference is tiny, then it disappears the moment the comparison is done blind. If the difference is large then either the value of the capacitor is slightly different or the cap is defective.

 

A good cap has constant ESR over large periods of time, stable capacitance, forms quickly, and is stable mechanically and does not pick signals from vibration. You do not need to go Duelund for that. The purple clarity caps are perfect.

 

2 hours ago, PYP said:

 

I'm certainly not advocating for designing by ear, nor making changes that measure worse.  But as Bruno has mentioned, sometimes listeners can hear noise (or choose a better word) that initial measurements missed (as @barrows noted).  

 

Aind note that this is exactly what I said: you identify (or believe to identify!l) an issue, you try to determine what measurements are correlated with that issue, then attempt to address that problem and you verify by ear whether you guessed correctly.

 

 

2 hours ago, PYP said:

 

He takes that as a challenge to find the measurements that identify the problem area.   Had he not done so, I don't think Class D would today be considered a topology that works for the high-end market.   He blazed the trail.  His work has greatly influenced Grimm Audio, Hypex, Mola Mola, Kii and Purifi.  

 

At any rate, I think there is more agreement than disagreement in how we perceive these issues.   

Absolutely! I actually think I am 99% in agreement with barrows, for instance.

 

But I want to stress that I firmly believe that everything we hear in a properly controlled setting can be measured. The problem may be to identify the right physical quantity to measure, but our ears are, frankly, crap, so anything we can ear should be easily measurable with standard equipment.

[mocenigo]

5 hours ago, PYP said:

I was nodding in agreement as I read your post until I got this the highlighted part.  Do you mean that gear can be designed that has better measurements than the ear can detect?  If so, what would be the point?  

Well, this is already happening for many years, with DACs, and it started recently with amplifiers. I would say, bragging rights. You have a DAC with a THD+N (A) of 120Db, but I have the next iteration from the same company, with nearly the same circuit, and a THD+N (A) of 121Db. I win. Our ears have an internal distortion that is ludicrously higher, so any distortion level over about 100Db (and probably also much higher) cannot be discerned. Noise is covered by ambient noise except in an anecoic room, but the noise of your own circulatory system is already higher. 

[mocenigo]

14 hours ago, fas42 said:

 

DACs have historically had a type of distortion which masks low level detail - used to be quite common to remark on how the sound would fall into a "black hole", and that decay tails of instrumental sounds would stop abruptly, not be realistic. I used to shake my head, years ago, at how hopeless the replay on CD systems typically was, on disks I knew  well - literally, half of what was going on, musically, just wasn't there.

 

Which was never measured. And still isn't.

Paint me VERY doubtful but does this distortion really exist? The CD was introduced 50 years ago, if that problem really existed it would have been identified and measured.

[mocenigo]

1 hour ago, barrows said:

Signal correlated noise floor modulation, it has been measured, and better DAC designers are well aware of it. There is another form of artifacts produced by DACs which is also a problem, which I cannot remember the description of, but Bruno Putzeys mentions this other artifact in his comments about the Tambaqui DAC review over at ASR.  He also mentions how to measure this other typical DAC artifact problem.  

So, some of the better DAC designers out there, who also rely on proprietary conversion methods rather than OTS DAC chip solutions, have both identified some of these distortions/artifacts, found a way to measure them, and solved them!  But these do not typically show up in the "standard set" of measurements. 

Ok I am aware of these. Fas42 mentioned factors that have never been measured, which is of course nonsense. Today’s DACs can probably pass a zero test with a 20 bit signal.

[mocenigo]

2 hours ago, barrows said:

Signal correlated noise floor modulation, it has been measured, and better DAC designers are well aware of it. There is another form of artifacts produced by DACs which is also a problem, which I cannot remember the description of, but Bruno Putzeys mentions this other artifact in his comments about the Tambaqui DAC review over at ASR.  He also mentions how to measure this other typical DAC artifact problem.  

So, some of the better DAC designers out there, who also rely on proprietary conversion methods rather than OTS DAC chip solutions, have both identified some of these distortions/artifacts, found a way to measure them, and solved them!  But these do not typically show up in the "standard set" of measurements. 

Both signal correlated noise and idle tones are a solved problem in the recent multi-bit delta-sigma DACs. But in 2013, when Bruno started designing his DAC, they were still a problem. 
 

And they show up as distortion (mostly in IM spectra with only a few tones) and spikes in noise, respectively, provided the measurement equipment is sufficiently resolving.

 

[mocenigo]

19 minutes ago, fas42 said:

They are not measured for the benefit of the consumer.

 

This is not what you said. You are changing your story. You said that these are not measured or we do not know how to.

 

19 minutes ago, fas42 said:

Of course there have always been designers and builders of audio components who understand that better techniques are needed to make items like DACs function more accurately.

 

Of course.

 

19 minutes ago, fas42 said:

But the potential purchaser has no idea which is superior in this area, by the way of numbers - handwaving by the company marketing, and review ticks by people who try them is about as good as it gets; you can't decide which is better by looking at a specifications page.

that’s why there have been independent reviewers, and some, like ASR, give you proper noise profiles (then you can see idle noise) and multitone distortion (that contains also the signal modulated noise, so it can be used to bound this).

 

Let us face it, most people would not know how to read those measurements (and that’s why most ASR readers are just fine with the SINAD, that is just ONE data point out of thousands) but the data is there.

[mocenigo]

1 hour ago, fas42 said:

 

Well, we can get into arguments about precision of words used, but what's the point of that? I said "DACs have historically had a type of distortion which masks low level detail" and "Which was never measured. And still isn't."

 

ok, here I can say we agree except for the last sentence.

 

1 hour ago, fas42 said:

The data may be there, but there should be something like at least a single number as part of specs, which gives an indicative measure of how well engineered the product is, in this regard.

well, that value exists, mostly, and is THD+N (aka SINAD) — even when it is given at 1khz only, I would argue that it would be a weird effort to design a DAC that is excellent at 1khz and bad at 200hz or 5khz. Some correlation between these values must exist. Some manufacturers give the value A-weighted, so it is not only at 1khz.
 

Also, that value is an upper bound to distortion and to noise. Even though the first is harmonic distortion, we know that THD and IMD are correlated, so am exceptionally good value of the first implies at least a very good one of the second. Regarding noise, if it is computed over the whole audible spectrum, it covers also idle tones — that in some Δσ DACs of the ‘90s that were probably used also a bit later, were eating up to three least significant bits of a 16bit sample. But if you see a noise value of -120Db you know that idle tones are not audible.

 

I wish companies would agree to show THD+N either as a bound over the audible spectrum or A-weighted. But of course you know that it’ll be precisely the “hiend” audiophile manufacturers that will oppose it.

 

Hence , the question is, how do we get manufacturers to use it consistently?

[mocenigo]

7 hours ago, fas42 said:

 

IOW, all the memories you have of whether rigs that you have encountered in all your years of audio interest count for nothing - that your recollections of them being somewhere in the range from very poor to truly excellent is likely to be completely wrong; you might as well as jumble them up in any order of preference ... yes?

More or less, that’s why subjectivism is ALMOST a random choice. Add to this that it is sufficient to sneeze between two listening sessions to change the response of our internal physical system, or a different quantity of ear wax in two different days, and so on, to influence the ACTUAL listening, even before the memory of it is jumbled by the brain.

 

At most, you can define the character of a system. That’s why the Kroma speakers at Munich hiend were almost universally considered the best of show. There was a very different character and an obviously superior resolution and clarity. And it was all in the speakers, nothing else really mattered. But even that is rare.

[mocenigo]

7 hours ago, fas42 said:

Just found this little snippet of thread, with a notable member of this forum responding, https://www.audioasylum.com/cgi/t.mpl?f=pcaudio&m=126619

 

"Provoking different artifacts" in DACs is also how I see what's going on - it takes attention to detail, in everything, to resolve this ...

You are stating the obvious: it takes attention to detail. In fact the detail is here what AKM and ESS, and probably the other manufacturers as well, do: they increase the resolutions of the samples internally (to even 32 bits) while oversampling, and add both shaped dithering and lower level random dithering, then they modulate using a multibit sigma delta modulator. Finally, if, say, the modulator is 5 bit, they apply load balancing on the 31 switched capacitors (AKM moved to 6 bits). The result has no idle tones and signal correlated noise, or at least pushes them under the residual noise carpet, .