• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Review and Measurements of Purifi 1ET400A Amplifier

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
My g-d, this is a place full of mental people. I just asked for a proof that passing those tests are sufficient to prove faithful reproduction of all signals. Just one paper. No, I get a barrage of "you do not understand signal processing" with no arguments.

Calm down and explain. Because I haven't found any valid argument, unless one assumes that the behaviour of physical devices is somehow ideal. Which you should be able to admit, it is a bit of a loop. I am sincerely ready to learn, but curiously nobody has provided the material.

What you have said was far more than that, with a bunch of incorrect assertions and started with some pretty rude comments about Amir. As I said right at the start it would be very difficult to cover the topic here considering where you are starting from.
 
Last edited:

VintageFlanker

Major Contributor
Forum Donor
Joined
Sep 20, 2018
Messages
4,940
Likes
19,692
Location
Paris

murraycamp

Senior Member
Forum Donor
Joined
Jul 10, 2019
Messages
421
Likes
647

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Any chance of seeing a bit of a screen grab of what a typical wave form is like for these signals? Of course, it is varying with time. Can one get a good idea of the range of variations seen in one screen? That is, can you capture enough time in a screen or two, with adequate resolution, to get a sense of the range of 'shapes' it has?

a few screen grabs at various time bases tested with a motu 8A in loopback. Its anything but a static signal.

1578450444011.png


1578450349208.png


1578450493846.png


1578450542528.png


index.php


1578450624103.png
 

Attachments

  • 1578450575765.png
    1578450575765.png
    300.1 KB · Views: 1,496

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Seems like this never gonna end... The last time Purifi is mentioned stucks two pages behind.

@mocenigo, @March Audio, @JohnYang1997... Would you please end this by private conversation or something? Thanks. ;)
...
Agreed, but these sort of misconceptions about measurements and testing are very common and worth tackling when they crop up.
 
Last edited:

PaulD

Senior Member
Joined
Jul 22, 2018
Messages
453
Likes
1,341
Location
Other
Agreed, but the sort of misconceptions about measurements and testing are very common and worth tackling when they crop up.
Absolutely! But perhaps we need another thread titled something like "Why Measurements Capture Everything We Can Hear", or "What Measurements Are Valid & Better than Casual Listening" etc etc. Then when a thread starts to be derailed we can point someone to a thread with a bunch of information in it already. I'm pretty sure all of the info is already sitting around on ASR...
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,497
Location
Monument, CO
In my world there are static signals, like DC; steady-state signals, like sine waves (singly or in combination) that are after starting time-invariant in the sense that they look the same over a period; and, transient signals that change over time like the starting ramp ("attack") of a percussive sound.
 

mocenigo

Major Contributor
Joined
Dec 8, 2018
Messages
1,274
Likes
1,034
Also you quote the wrong phrase of mine. I never said you don't understand signal processing. I'm thinking of either you lack the knowledge or you lack the critical thinking. Those "think about"s are just directing you to find out the answer. If you know what you are talking about, you will get the answer very quickly.

I can assure I do not lack the critical thinking. All I am asking is evidence that simpler tests cover more complex situations. So, yes, my knowledge is lacking and I would like this to be remedied. As Alan said, yes "we do IMD" and other tests. I have not seen evidence these tests are sufficient. I did not say they are not sufficient - I want to know.

Regarding my rude comments to Amir, well, on that I was out of line and I apologise to Amir, but speaking in a more polite way, the fact remains that the validity of his methodology is debated, and inconsistencies have been pointed out here as well. So, please show me papers that prove that simpler tests allow us to bound distorsions in all cases. I am eager to learn.

I am a professional mathematician and I can read those papers. I am not mentioning this to brag, but it was just to mention that you can point to me to real studies. Taking about "32 waves that go up and down", as Alan did, is puerile at the very least. My current job is easy to find, and I have a global responsibility where, if I do a mistake, literally billions of people may be affected. In the past a slip of mine led to a potential 900 million devices being vulnerable. I have experience being wrong and owning it. What I do not accept is "you have no clue", a salad of buzzword (all of which I know well) and then being dismissed.

Returning to critical thinking, using (1) measurements that prove that devices are not ideal, and then (2) claiming that you can extrapolate from these measurements in an ideal way is a logical fallacy, and not one I have done. [OTOH I can accept even heuristic arguments that combined distortion byproducts are so low that they fall under the noise floor, and therefore they can be ignored.]

I started on the wrong foot. Mocking (as Alan did) repeats my mistake, I gave a bad example.

Roberto
 

mocenigo

Major Contributor
Joined
Dec 8, 2018
Messages
1,274
Likes
1,034
Absolutely! But perhaps we need another thread titled something like "Why Measurements Capture Everything We Can Hear", or "What Measurements Are Valid & Better than Casual Listening" etc etc. Then when a thread starts to be derailed we can point someone to a thread with a bunch of information in it already. I'm pretty sure all of the info is already sitting around on ASR...

I agree. And we could report literature there. It would be a fantastic resource
 

mocenigo

Major Contributor
Joined
Dec 8, 2018
Messages
1,274
Likes
1,034
In my world there are static signals, like DC; steady-state signals, like sine waves (singly or in combination) that are after starting time-invariant in the sense that they look the same over a period; and, transient signals that change over time like the starting ramp ("attack") of a percussive sound.

Well said. It was Alan that spoke of "32 waves that go up and down" as transient signals. Whereas they are steady state. So much for somebody claiming I am clueless.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
Well said. It was Alan that spoke of "32 waves that go up and down" as transient signals. Whereas they are steady state. So much for somebody claiming I am clueless.
In a sense, yes multitone is still periodic. The point is that each sine is riding on each other. There are essentially countless combination of phase and amplitude modulation. So in a sense it's also as good as not static state. It's almost enough for everything. The extra thing you can add is measuring multitone at different amplitudes or adding a sweeping signal at high frequency. ESS hump is revealed with low amplitude two tone.
As for transients, our hearing is band limited. So anything produced higher than what we can hear means nothing (almost there are different ways that may affect but not discussing here). If you consider the combined transfer function of the whole system, adding some extra response at high hundred kHz doesn't make a lot of sense. Let alone that R-2R should also have proper lpf, so at the end you should get same waveform. The ess's white paper you were talking about doesn't make sense to us. There is no reason for a dac which has better transient at equivalent frequency of tens of megaherts that plays 44.1kHz music better. The transient is limited in the music file. Higher rate does give a tiny bit improvement from phase but you can still only hear the frequency that you can physically hear.
There are two types of distortion.
Linear distortion and non-linear distortion (nonlinearities)
Linear distortion is like frequency response.
Non-linearity contains harmonic distortion, random(non source related)noise, intermodulation distortion.
That's basically it. I don't think we miss anything by making these measurements. Anything different from the source should be included in these. There is a thing called null test. It basically tells you any difference between the source and the output. I can guarantee you if all measurements above are perfect, the residue will be minimal.
That's all I can say. Hope this helps.
 

mocenigo

Major Contributor
Joined
Dec 8, 2018
Messages
1,274
Likes
1,034
That's basically it. I don't think we miss anything by making these measurements. Anything different from the source should be included in these. There is a thing called null test. It basically tells you any difference between the source and the output. I can guarantee you if all measurements above are perfect, the residue will be minimal.
That's all I can say. Hope this helps.

Splendid, thank you for your answer! As I said, I do not doubt much that a null test will give you (at least "almost always") minimal residues once the other tests are good. I could even write down a simple heuristic argument, but I do not want to bore anybody further. However, I have not seen much evidence of that, and this may even be missing. And I am very curious to see whether there may be surprises. I cannot exclude these a priori and, since (1) transients already revealed surprise (that's why we have TIM) and (2) DS has already shown in the past that the representation of transients can be in theory arbitrarily long, there might be interesting byproducts in some architectures. If this has been properly addressed, show me the evidence (peer reviewed papers), because all I have see so far is ESS claiming older DS had issues with trasients and then that they solved it - without saying why, because it is their secret sauce. Similarly, some amplification topologies may also exhibit weird behaviour. Is this not interesting and worthy of discussion in the name of audio science?

Roberto
 

Julf

Major Contributor
Forum Donor
Joined
Mar 1, 2016
Messages
3,004
Likes
3,998
Location
Amsterdam, The Netherlands
In a sense, yes multitone is still periodic. The point is that each sine is riding on each other. There are essentially countless combination of phase and amplitude modulation. So in a sense it's also as good as not static state. It's almost enough for everything. The extra thing you can add is measuring multitone at different amplitudes or adding a sweeping signal at high frequency. ESS hump is revealed with low amplitude two tone.
As for transients, our hearing is band limited. So anything produced higher than what we can hear means nothing (almost there are different ways that may affect but not discussing here). If you consider the combined transfer function of the whole system, adding some extra response at high hundred kHz doesn't make a lot of sense. Let alone that R-2R should also have proper lpf, so at the end you should get same waveform. The ess's white paper you were talking about doesn't make sense to us. There is no reason for a dac which has better transient at equivalent frequency of tens of megaherts that plays 44.1kHz music better. The transient is limited in the music file. Higher rate does give a tiny bit improvement from phase but you can still only hear the frequency that you can physically hear.

Exactly. The steepest rise time (a definition of transient) you can get is with a square wave, and that is a perfectly repetitive, steady waveform.

Of course a square wave (or impulse for that matter) requires infinite bandwidth that doesn't exist in nature, and is meaningless to our ears. Thus the steepest rise time we can heat and that matters is that of a full-amplitude 20 kHz sine wave.
 

Julf

Major Contributor
Forum Donor
Joined
Mar 1, 2016
Messages
3,004
Likes
3,998
Location
Amsterdam, The Netherlands
I cannot exclude these a priori and, since (1) transients already revealed surprise (that's why we have TIM)

TIM (really SID - slew-rate induced distortion) was caused by slew rate issues, and slew rate has already been discussed.

DS has already shown in the past that the representation of transients can be in theory arbitrarily long, there might be interesting byproducts in some architectures.

This is where I will again offend you by asking if you are familiar with the work or Monsieur Fourier. A representation of a transient can only be arbitrary long if it contains arbitrary low frequencies.

If this has been properly addressed, show me the evidence (peer reviewed papers), because all I have see so far is ESS claiming older DS had issues with trasients and then that they solved it - without saying why, because it is their secret sauce. Similarly, some amplification topologies may also exhibit weird behaviour. Is this not interesting and worthy of discussion in the name of audio science?

Not really, because all you are saying is the audio equivalent of Russell's Teapot. "Show me the evidence that there is not a china teapot revolving about the Sun in an elliptical orbit between the Earth and Mars".
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
Splendid, thank you for your answer! As I said, I do not doubt much that a null test will give you (at least "almost always") minimal residues once the other tests are good. I could even write down a simple heuristic argument, but I do not want to bore anybody further. However, I have not seen much evidence of that, and this may even be missing. And I am very curious to see whether there may be surprises. I cannot exclude these a priori and, since (1) transients already revealed surprise (that's why we have TIM) and (2) DS has already shown in the past that the representation of transients can be in theory arbitrarily long, there might be interesting byproducts in some architectures. If this has been properly addressed, show me the evidence (peer reviewed papers), because all I have see so far is ESS claiming older DS had issues with trasients and then that they solved it - without saying why, because it is their secret sauce. Similarly, some amplification topologies may also exhibit weird behaviour. Is this not interesting and worthy of discussion in the name of audio science?

Roberto
One thing, we have to comprehend, is that as devices, components, science progress our ears don't progress(at least for now).
The TIM issue is repeatable with full amplitude 20khz sine + other frequencies. As we reach the limit of our hearing, more progress doesn't make it better. There will always be faster devices, but we don't have faster hearing.
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,497
Location
Monument, CO
The issue with DS converters and transients has more to do with the digital filters before or after the modulator. Since DS designs are oversampling by nature that is not an issue in the real world -- a DS DAC with net 20 kHz output bandwidth has the same transient response as any other DAC architecture with the same bandwidth, and least in terms of rise-time. There are secondary and below (tertiary, quaternary, etc. -ary) effects that can impact settling and all that jazz (jazz, get it? ;) ) <Started to discuss transients in filters and quit because (a) it's complicated and (b) ain't my field; I have done filter designs analog and digital but it's been a while and I don't claim great expertise. I do deal with CTLE and DFE all the tie in my day job but at much higher frequencies than audio and generally far fewer taps.>

In-band multi-tone testing can be revealing and @amirm is already doing that for us now (thanks Amir!) His 32-tone test is IME/IMO a good test for what might happen with real music or other "busy" signals.
 

Koeitje

Major Contributor
Joined
Oct 10, 2019
Messages
2,292
Likes
3,880
TIM (really SID - slew-rate induced distortion) was caused by slew rate issues, and slew rate has already been discussed.



This is where I will again offend you by asking if you are familiar with the work or Monsieur Fourier. A representation of a transient can only be arbitrary long if it contains arbitrary low frequencies.



Not really, because all you are saying is the audio equivalent of Russell's Teapot. "Show me the evidence that there is not a china teapot revolving about the Sun in an elliptical orbit between the Earth and Mars".
Important to note that you should state the teapot is too small to see.
 

Julf

Major Contributor
Forum Donor
Joined
Mar 1, 2016
Messages
3,004
Likes
3,998
Location
Amsterdam, The Netherlands
Important to note that you should state the teapot is too small to see.

I was waiting for "but you could look", so I could play audiophile and keep moving the target. :)
 

mocenigo

Major Contributor
Joined
Dec 8, 2018
Messages
1,274
Likes
1,034
Not really, because all you are saying is the audio equivalent of Russell's Teapot. "Show me the evidence that there is not a china teapot revolving about the Sun in an elliptical orbit between the Earth and Mars".

No. I am asking for studies showing correlation between standard measurements and more elaborate ones. I am not that naive. If discrepancies are below a certain level then this is proved.
 
Top Bottom