Review and Measurements of Purifi 1ET400A Amplifier

[March Audio]

Ok, so we have a misunderstanding here.
I never a accept mediocrity. in fact, I do not tolerate it, because in many cases it is just the result of laziness. I meant of course designing a GOOD one. Linear, precise, with low noise, and sounding good.

d.

[/QUOTE][/QUOTE]

R2R dacs are fundamentally limited in their linearity by the tolerance of the resistors. This equates to higher harmonic distortion which is quite evident on your soekris and the board I had.

 

[March Audio]

-
I already described it.

What you described was a fundamental misunderstanding of digital audio.

 

[mocenigo]

Sorry but with respect you have a fundamental lack of understanding which is going to be extremely difficult to cover in the context of this thread. I'm not even sure where to start.

Try me.

The time waveform is the time waveform. All that dynamic change (within the defined limits of sampling theorem and the bandwidth limit) is in the waveform. The dacs accuracy in reproducing that waveform is quantifiable by various measurements, many of which you see here. R2R is not fundamentally better.

You are assuming distortion adds linearly? If indeed all distortions add linearly, then you are right. But, I just do not know. I think it does not, because non-linear transforms cannot add linearly (you can actually prove that mathematically) and I also think (surprise!) that any non-linear combination of harmonic distortion is most likely not relevant or not-audible. But. We. Do. Not. Know. Until. We. Measure. It. Never seen a measurement of, say, several transients close to each other in time, by fractions of a millisecond. I am merely claiming that the fact that for some people there is an audible difference should prompt us to do more in depth measurements. Excluding the need of further measurements because the current ones are "perfect" is not scientific. Even heard of falsifiability?

Never forget that mathematical models are rarely a perfect depiction of reality.

Btw I have blind tested people with a soekris R2R V an Ess based DS DAC and they couldn't tell the difference.

I never claimed otherwise. This does not surprise me for two reasons.
The first one is because distortion and noise are most likely too low to be distinguishable.
The second one because knowing what to hear to distinguish two DACs is sometimes very very hard. In one case I could find only one very specific spot of a work of music in a single recording by which I could differentiate between two DACs. To me this means they were so close as their differences being irrelevant.

"oftentimes"? You are gonna have to provide some evidence to back that up.

Well, I heard reproductions of strings (even of an orchestras I know well, in a venue I know well, recorded by their usual engineers, of a concert where I was present in person) on different DACs, one praised by measurement community and one not, and one shocked me by how much it was just wrong, with the wrong tonal color and with an obvious moire effect on the sound. Just listen to the Sabaj D5. And, by the way, maybe it is not a "ESS" artifact since the litte NuForce DAC5, for all its wrongs, did not exhibit that. Nor did the Gustard X20. I understand that this is difficult or even impossible to accept, since you cannot enter my brain. But claiming "we know everything we need to assess a DAC" is quite arrogant even before being anti-scientific. I am not asking you to prove me god exists (I would never because I am an atheist). I am asking you whether you really believe we can measure everything we hear. I have strong doubts, and, so far, everybody that claimed that were proved wrong.

 

[mocenigo]

R2R dacs are fundamentally limited in their linearity by the tolerance of the resistors. This equates to higher harmonic distortion which is quite evident on your soekris and the board I had.

You have tested the 1541? Ever programmed order magnitude aliasing? (not "sign magnitude"). But, if you say, this is irrelevant since your sample of blind testing did not result in ability of distinguish it from a "perfect measuring" other DAC, how is the DAC I chose not excellent? Read again my first comment on the 1541 being excellent.

 

[Julf]

-80db SINAD?

Not surprising for a R2R.

I never a accept mediocrity. in fact, I do not tolerate it, because in many cases it is just the result of laziness. I meant of course designing a GOOD one. Linear, precise, with low noise, and sounding good.

We indeed have a misunderstanding. R2R DACs are very simple and primitive devices. That is why it is so hard to make a linear and precise one. Why try when there are much better ways to get superior performance?

It is like how it is not hard to design a 3-wheeled car. It is very hard to make one that handles well and is stable at higher speeds. That is why car manufacturers do 4 wheels instead.

Yes I am, and here's your fallacy: When you are performing certain measurements, you are basically measuring both linear distorsions and non-linear ones together. Fine.
The latter ones are the key.
If there is a non-linear interaction, we do not have distortion(signal A + signal B)=distortion(signal A)+distortion(signal B), in fact there is a discrepancy.
So if distortions added linearly, the measurement of a single-point-in-time spectrum change would tell us the whole story.

Here is your fallacy: any decent DAC has very little linear distortion. What we are interested in is non-linear distortion. Measuring THD gived us a pretty good picture of the non-linearity.

Therefore, these measurements do not tell the whole story unless we have proof that the distortion adds linearly. Which we can reasonably assume it does not.

What makes you think that is a reasonable assumption?

Yes, but I also understand how it is applied.

It seems you don't.

 

[Julf]

Well, I heard reproductions of strings (even of an orchestras I know well, in a venue I know well, recorded by their usual engineers, of a concert where I was present in person) on different DACs, one praised by measurement community and one not, and one shocked me by how much it was just wrong, with the wrong tonal color and with an obvious moire effect on the sound.

So not only are you relying on subjective, sighted perception, but also auditory memory.

 

[March Audio]

Try me.



You are assuming distortion adds linearly? If indeed all distortions add linearly, then you are right. But, I just do not know. I think it does not, because non-linear transforms cannot add linearly (you can actually prove that mathematically) and I also think (surprise!) that any non-linear combination of harmonic distortion is most likely not relevant or not-audible. But. We. Do. Not. Know. Until. We. Measure. It. Never seen a measurement of, say, several transients close to each other in time, by fractions of a millisecond. I am merely claiming that the fact that for some people there is an audible difference should prompt us to do more in depth measurements. Excluding the need of further measurements because the current ones are "perfect" is not scientific. Even heard of falsifiability?

Never forget that mathematical models are rarely a perfect depiction of reality.



I never claimed otherwise. This does not surprise me for two reasons.
The first one is because distortion and noise are most likely too low to be distinguishable.
The second one because knowing what to hear to distinguish two DACs is sometimes very very hard. In one case I could find only one very specific spot of a work of music in a single recording by which I could differentiate between two DACs. To me this means they were so close as their differences being irrelevant.



Well, I heard reproductions of strings (even of an orchestras I know well, in a venue I know well, recorded by their usual engineers, of a concert where I was present in person) on different DACs, one praised by measurement community and one not, and one shocked me by how much it was just wrong, with the wrong tonal color and with an obvious moire effect on the sound. Just listen to the Sabaj D5. And, by the way, maybe it is not a "ESS" artifact since the litte NuForce DAC5, for all its wrongs, did not exhibit that. Nor did the Gustard X20. I understand that this is difficult or even impossible to accept, since you cannot enter my brain. But claiming "we know everything we need to assess a DAC" is quite arrogant even before being anti-scientific. I am not asking you to prove me god exists (I would never because I am an atheist). I am asking you whether you really believe we can measure everything we hear. I have strong doubts, and, so far, everybody that claimed that were proved wrong.

It's very simple. You don't understand that the transients are in the waveform. If the Transient is of a lower frequency than the bandwidth limit then it will be captured.

Your comments about distortion linearity are irrelevant. Non linearity will be seen in measurements. Are you under the impression that a sine wave is a constant signal? And yes we test complex signals not just constant amplitude single sine waves.

So if it doesn't surprise you where does that leave your claim that R2R is better than DS?

Recordings do not and cannot sound like what you hear personally in the venue. That's your first error. Also your auditory memory is very poor. Moire effect is visual not audio.

You did not perform your dac comparison blind and accurately level matched. Massive error. Your conclusions will be faulty.

If it's big enough an effect to hear its big enough to measure.

 

[Julf]

It's very simple. You don't understand that the transients are in the waveform. If the Transient is of a lower frequency than the bandwidth limit then it will be captured.

Ah, our French friend Monsieur Fourier again!

 

[March Audio]

Ah, our French friend Monsieur Fourier again!

It's about as fundamental as it gets. I hope he watches the video.

 

[mocenigo]

What you described was a fundamental misunderstanding of digital audio.

Please be more specific. Because I am quite entertained by this discussion. And we are getting off-topic.

So not only are you relying on subjective, sighted perception, but also auditory memory.

I know how strings should sound like. Yes. I can also hear them at home. Live. As in live.

 

[March Audio]

Please be more specific. Because I am quite entertained by this discussion. And we are getting off-topic.


I know how strings should sound like. Yes. I can also hear them at home. Live. As in live.

Erm everything you said in #1127 and 1131. Watch the video. Distortions and non linearities will be captured in measurements. They don't have to be music signals.

As already mentioned, a recording through a microphone does not capture what you hear live. Your brain actively filters room and other sound, reflections etc. It focuses. A microphone does not, so using your echoic memory (which is ver poor BTW) and comparing in this way is quite misleading.

 

[Julf]

I know how strings should sound like. Yes. I can also hear them at home. Live. As in live.

And do you feel they sound the same on a recording? Have you ever done a proper, controlled double-blind test?

 

[mocenigo]

What you described was a fundamental misunderstanding of digital audio.

Please be more specific. Because I am quite entertained by this discussion. And we are getting off-topic.

It's very simple. You don't understand that the transients are in the waveform. If the Transient is of a lower frequency than the bandwidth limit then it will be captured.

I don't understand that the transients are in the waveform?!? What do you think I am? One of those lowlife trolls that have attacked you in the past? Sorry, but have you worked with the Qualcomm DDFA team, just for mentioning something? I was even asked whether I knew what a Fourier transform was, FFS.

Show me where do I doubt this.

But, does this address a non-linear function applied to the whole spectrum? Mind you, with mathematics you have to be careful, I have been a university professor of mathematics (a real one, not a lecturer at a community college in the US), and I can tell you that if F is a non linear function, then F(A+B) in general is not equal to F(A)+F(B), where A and B are spectra, which can be also single frequency spectra. So measuring a sudden single sound change evaluated on F does not tell anything about F in general.

Again, just as Otala's work showed us that IMD is important, I am just asking does anybody care about the influences of several transients close in time to each other, when they may even partially overlap.

Your comments about distortion linearity are irrelevant. Non linearity will be seen in measurements. Are you under the impression that a sine wave is a constant signal? And yes we test complex signals not just constant amplitude sine waves.

Oh god. I already spoke of signal that are constant in the spectrum domain, SO yes, a sine wave in that domain is obviously constant. But it is not constant in the amplitude domain. And if complex signals are tested why are not they often published? (one of the issues is synchronisation, and maybe audio people should learn some science for real and talk to the guys that in my main field, security, do side channel analysis).

So if it doesn't surprise you where does that leave your claim that R2R is better than DS?

To the risk of being banned, where did I f*cking say that R2R is better? I just said that Soekris' DAC 1541 is an excellent DAC (and since nobody has measured it in ASR nobody can disprove it, I was only shown the measurements of a DAC 1421, OTOH the measurements on SBAF of the DAC 1541 show full 21 bits or linearity which is SotA). Then some people jumped at me in a trolling way. Cool down.

The two (subjectively) best DACs I heard are a Sabre based and a Soekris R2R based. Both measure well (the Sabre better). You are just assuming I am saying things I never claimed or even implied. I understand there are trolls here, and many attacked you as well. But, please.

Recordings do not and cannot sound like what you hear personally in the venue. That's your first error.

As a person with a composition diploma that has also done some conducting, I fricking know this well. Have you hands on experience with orchestras? how well do you know what to expect from live sound?

Also your auditory memory is very poor.

I am aware of that. But I am also very accustomed to live sound, and I perceive some characters I expect of sound.

Moire effect is visual not audio.

Wrong. It is also audio. It is basically noise induced by quantisation, it is used in some circles to describe what in the high time of synthetisers was called a "phase" filter.

You did not perform your dac comparison blind and accurately level matched. Massive error. Your conclusions will be faulty.

I did not claim my conclusions were not faulty. I just claimed that measurements are incomplete, and the stubborness in refuting the mathematically obvious shows me that there is some uneasiness in all this. Ok, fine. I have more important things to do in my day job than wasting my time with people that think having an oscilloscope means they can measure everything. If anybody wants to do what I do for a living, come and listen to me at Columbia University tomorrow.

If it's big enough an effect to hear its big enough to measure.

I never said it is not, but how? You guys apparently claim to know how to measure everytying. I am not sure. This is all all I am asking. In my job, that includes hardware security, we are disproved every single day, we think we get it, and somebody finds what we missed. This actually teaches us some humbleness I see missing here.

 

[mocenigo]

Erm everything you said in #1127 and 1131. Watch the video. Distortions and non linearities will be captured in measurements. They don't have to be music signals.

If a distorsion only appears when a spectrum varies widely over a certain threshold, no, those won't be measures by simpler tests. If some types of distortion raise very quickly with the complexity of the signal, then a lot of tests won't detect them, because testing the same condition on temporal or spatial subsets of the spectrum will be under the measurement noise. "Everything you said in #1127 and 1131." If you have arguments you should deconstruct them.

If you do testing on more complex signals, then please tell me what synchronisation techniques you use to to compare the sampled output signals. This is a very serious question. Least squares? Chi method? Really. Comparison of traces is my bread.

As already mentioned, a recording through a microphone does not capture what you hear live. Your brain actively filters room and other sound, reflections etc. It focuses. A microphone does not.

Yes of course my brain filters a lot. It also tells me that the folks that reduced everything to S/N ratio and THD in the 70's were wrong. Now we have SINAD, which is the same mistake.

 

[mocenigo]

I have decided to no longer engage in discussions over measurements since the "science" here is astonishingly primitive. Let us return to the build of a 1ET400A based amp.

 

[Julf]

I have decided to no longer engage in discussions over measurements since the "science" here is astonishingly primitive.

 

[murraycamp]

I have decided to no longer engage in discussions over measurements since the "science" here is astonishingly primitive.

OK Professor

 

[VintageFlanker]

Damn, that was a hell of an Ego trip...

 

[hyperknot]

Hi, I'm about to receive my first Harbeth speaker, an P3ESR. I'll be listening in small room / near-field environment. My question is that is there any point in building a Purify 1ET400 based amp over buying an amp with an NC252MP inside? Currently I have an ICEPower 125ASX2 module based amp, from which I believe the NC252MP would be a clear upgrade. However I'm not sure about the 1ET400. I'll probably not be using any of these modules not even near to their max power output.

 

[March Audio]

I have decided to no longer engage in discussions over measurements since the "science" here is astonishingly primitive. Let us return to the build of a 1ET400A based amp.

And there again you betray your lack of understanding of the subject. As such there is no point in trying to go through your last posts.

Just one thing though.

"To the risk of being banned, where did I f*cking say that R2R is better?"

That would be here;

"they will represent transients more faithfully."