Blue or red pill?

[mansr]

Just now, manisandher said:

My own feeling is that an FFT is not the best way of showing differences in jitter signatures.

Looking directly at the clock signal entering the DAC would certainly be preferable in assessing whether a difference in jitter is present.

[jabbr]

14 minutes ago, manisandher said:

 

What I heard was a difference in sound between 'A' and 'B'. I described this difference earlier in this thread. I even described it to Mans as he listened to A and B, immediately after the A/B/X.

 

Why was it present? Peter puts it down to different "jitter signatures" at the point of d-a conversion. Mans's analysis seems to support this, though the differences show up at low levels in his analysis.

 

My own feeling is that an FFT is not the best way of showing differences in potential jitter signatures.

 

Mani.

 

FFT is not the best way to assess “jitter” however the Fourier domain is apropos. Consider a signal having both voltage noise and phase noise — and the Fourier having a voltage and a phase component. 

 

If you are only able to measure voltage of a pure tone signal then the Fourier will demonstrate widening of the peak as a function of phase variation. 

 

@PeterSt if you question my assertion I refer you to the analog devices white paper you linked to ?

 

If you are are able to measure phase that’s ideal. 

 

In in any case this stuff is math, not merely opinion. ?

[Paul R]

On 4/8/2018 at 8:42 AM, manisandher said:

 

 

I was suggesting that it might be impossible to block certain noise getting into the DAC, without blocking the signal itself. In which case, the only choice is to manipulate the noise getting through to the DAC, in a sonically advantageous manner.

 

Hope that's clearer now.

 

Mani.

 

Not really - are you talking about electrical noise or jitter? If so, nothing new there, just economic and technology limits.  

 

There isn't anything particularly mysterious about this stuff that I can see. Just improved implementations. 

[PeterSt]

9 hours ago, mansr said:

Can you do one with nothing connected to the inputs on the Tascam? Just to get the baseline for noise and spurious spikes added by the recorder.

 

@manisandher, when you're at it anyway, I would do that with shortcut inputs. Or do that with and without shortcuts (not that the difference will tell anything, but so I am not conducting anything Mans wouldn't want.

It is perfectly allowed to apply the shortcut at the end of connected interlinks (more easy to apply it that way).

 

[marce]

10 hours ago, mansr said:

The DAC chip is not only single-ended, it uses a single voltage supply rather than separate supplies for digital and analogue sections like most serious chips. And then there isn't even a case, just a PCB nailed to a plank.

That's really not good.

[manisandher]

9 hours ago, adamdea said:

Thanks. Very helpful.

 

What I'll also do is take a look at the notebook I used to record the two ABXXXXXXXXXX tests we conducted. I scored 4/10 in both. I'll look to see if the 8 correct results correlate with BA or BB in any way, shape or form.

 

Mani.

[manisandher]

1 hour ago, marce said:

That's really not good.

 

I think we can all agree that the Altmann is not a particularly good-measuring DAC, based as it is on the late '80s1543 chip. The key question is: did this compromise the A/B/X in any way? I'd say absolutely not.

 

Mani.

[pkane2001]

16 minutes ago, manisandher said:

 

I think we can all agree that the Altmann is not a particularly good-measuring DAC, based as it is on the late '80s1543 chip. The key question is: did this compromise the A/B/X in any way? I'd say absolutely not.

 

Mani.

 

Mani, if you are still talking about proving that you could hear the difference between  two different SPDIF streams with various amounts of jitter or noise containing the same bits, then I’d say you have proven that. But, that lack of jitter rejection or other issues within your DAC  could have caused you to hear a difference, where a better DAC or a USB reclocking interface would eliminate them, I’m afraid you didn’t test for at all.

[adamdea]

16 minutes ago, manisandher said:

 

What I'll also do is take a look at the notebook I used to record the two ABXXXXXXXXXX tests we conducted. I scored 4/10 in both. I'll look to see if the 8 correct results correlate with BA or BB in any way, shape or form.

 

Mani.

Worth a go, but I'm not sure what one can make of the ABXXXXXXXX results

With ABx, ABx, ABx all one need to know is whether x is the same as B (which one has just heard).

The problem with ABXXXXXXXXXX is that after the first attempt, if one is thinking back to AB then it's too long ago.  Unless one thinks "is X(n )the same or different to X(n-1)",  one is likely to go astray (I think that's probably the only way of doing it within audio memory) . And once you get one wrong the next few choices will go wrong too.

The other thing is that if there is an unconscious switching tell then it might take a while to learn (I know it 's a bit weird to talk about learning something subconsciously). 

My guess is that the 4/10s are probably random.

 

Anyway I'm pretty sure that this type of issue has been well trodden and that someone will have worked out a way to control for it. [I may have to brave HA to ask around, but will have to look in my cupboard for my tin hat]. Obviously it goes without saying that this is just an avenue of inquiry and I am not making any assumptions. 

[acg]

13 minutes ago, pkane2001 said:

But, that lack of jitter rejection or other issues within your DAC  could have caused you to hear a difference, where a better DAC or a USB reclocking interface would eliminate them, I’m afraid you didn’t test for at all.

 

Except the dac that Mani usually listens to is sota in the ways that you mention and SFS changes are audible with it as well... just not proven in this test of course. 

[adamdea]

11 minutes ago, pkane2001 said:

 

Mani, if you are still talking about proving that you could hear the difference between  two different SPDIF streams with various amounts of jitter or noise containing the same bits, then I’d say you have proven that. But, that lack of jitter rejection or other issues within your DAC  could have caused you to hear a difference, where a better DAC or a USB reclocking interface would eliminate them, I’m afraid you didn’t test for at all.

All that this means is that the test may produce results which are dac dependent. But that is inevitable. It certainly would be interesting to see whether it produced the same results with a benchmark dac which is, by conventional measurements, to all intents and purposes jitter immune.  

This is analogous to testing cables using a pre with high output impedance and/or  a power amp with low input impedance.

[pkane2001]

6 minutes ago, acg said:

 

Except the dac that Mani usually listens to is sota in the ways that you mention and SFS changes are audible with it as well... just not proven in this test of course. 

 

Yes, may be true, but was not tested for in this test.

[jabbr]

6 hours ago, adamdea said:

All that this means is that the test may produce results which are dac dependent.

 

Of course the results are DAC dependent. Too much focus here on the DAC IMHO. The test was supposed to show that so-called bit identical streams can have measurablely different SQ. It has and repeating test can increase certainty of this.

 

Next step is to ascertain why and under what circumstances. One would expect that as input isolation improves, these SQ differences decrease.

[manisandher]

On 12/04/2018 at 3:48 PM, mansr said:

Some more minor differences:

 

 

 

 

On 12/04/2018 at 3:59 PM, PeterSt said:

If we look at the more "raspy" expression of the blue / SFS = 0.1 then this is what you can hear.

 

How did I miss this comment?

 

This is the first time I've seen anyone correlate this sort of measurement with sound. Yes, the effect looks very low in level on an FFT, and might therefore be easily dismissed as inaudible. However, if it were possible to measure it in a more appropriate way (see Peter's plots earlier in this thread), I suspect the effect would be far more pronounced, and more easily accepted as being audible by those who remain sceptical that bit-identical playback can sound different.

 

(For anyone coming in late to this thread, the red and blue lines in Mans's plots above are captures of the DAC's analogue outputs being fed two bit-identical data streams.)

 

Mani.

[manisandher]

1 hour ago, jabbr said:

One would expect that as input isolation improves, these SQ differences decrease.

 

In my experience, this is not the case. The Phasure NOS1 DAC is better isolated than any other DAC I'm aware of, and yet it is still massively affected by these sorts of bit-identical changes. Actually, all the DACs I've had here over the years, without exception, are affected.

 

Mani.

[jabbr]

5 minutes ago, manisandher said:

 

In my experience, this is not the case. The Phasure NOS1 DAC is better isolated than any other DAC I'm aware of, and yet it is still massively affected by these sorts of bit-identical changes. 

Mani, we are using a different definition of the term “isolation” then. 

 

To me, by definition, “isolation” means that the DAC is not affected either sonically or electrically by vagarities in otherwise identical or not identical bit perfect streams ?

 

What does “isolation” mean to you?

[psjug]

Are any difference in the analog captures (the music ones) apparent when comparing small sections of time series plots?

[pkane2001]

28 minutes ago, psjug said:

Are any difference in the analog captures (the music ones) apparent when comparing small sections of time series plots?

 

There are in a simple 10KHz sine wave signal capture, so you can assume there will be at least as much, but probably much more, with a more complex signal. Analog captures will never match exactly even under the best conditions.

[psjug]

5 minutes ago, pkane2001 said:

Analog captures will never match exactly even under the best conditions.

I know they won't match exactly.  But maybe obvious and consistent differences can be shown, especially in sections where Mani is most hearing the differences.

[adamdea]

3 hours ago, jabbr said:

 

Of course the results are DAC dependent. Too much focus here on the DAC IMHO. The test was supposed to show that so-called bit identical streams can have measurablely different SQ. It has and repeating test can increase certainty of this.

 

Next step is to ascertain why and under what circumstances. One would expect that as input isolation improves, these SQ differences decrease.

Yes but it’s slightly question-begging, or at least point-missing.  In information-theoretic terms the bitsteams seem possibly to be  identical. Differences in sound quality are (at least arguably) not properties  of the bitstream, they are properties of the dac. 

You could, with equal or greater validity, pose the question “is it possible to design a dac so badly that its output varies audibly according to the non-information properties of the input data? And you may (but only may) have answered that question.

[Andyman]

3 hours ago, jabbr said:

To me, by definition, “isolation” means that the DAC is not affected either sonically or electrically by vagarities in otherwise identical or not identical bit perfect streams ?

 

What does “isolation” mean to you?

 

Hmm. I like your posts jab and you know your stuff, but not sure on this. It’s sort of a circular argument which may well support the exact opposite of its own intent. Funnily enough, my first thought was not too dissimilar from that of Adam (who incidentally I think has been given a bit of an unfair ride of late in this thread)...

 

1 hour ago, adamdea said:

You could, with equal or greater validity, pose the question “is it possible to design a dac so badly that its output varies audibly according to the non-information properties of the input data? And you may (but only may) have answered that question.

 

...or maybe a sufficiently bad dac will render all inputs equally badly. You’re then in GUTB territory, “your dac’s not sufficiently resolving (i.e. expensive enough)”

 

...or yet another way; keep on lowering distortion / noise floor (better dac) and eventually you will reveal (hear / measure) differences between (bit identical) inputs

[Superdad]

12 hours ago, adamdea said:

All that this means is that the test may produce results which are dac dependent. But that is inevitable. It certainly would be interesting to see whether it produced the same results with a benchmark dac which is, by conventional measurements, to all intents and purposes jitter immune. 

 

Well given how many Benchmark DAC owners (and owners of other DACs claiming to be immune from upstream vagaries) have reported improvements with our USB REGEN and ISO REGEN over the years, I’d say that “for all intents and purposes” they are not Immune. 

 

I have read most of this thread and frankly it appears that most here are missing the obvious which is that signal integrity matters and that small variations upstream cause perturbations in ground-plane noise and spikes in current INSIDE  the DAC, causing a concomitant effect on the DAC master clock.

 

Mansr or someone else suggested looking at variations in close-in phase-noise at the clock pin of the DAC chip.  Long ago John Swenson measured variation in ground-plane noise, and he is setting up (building a 32-bit high speed ADC and capture system) for more advanced jitter analysis—both within the DAC and at the output—to prove the effect (with an injected jitter maker pattern that he calls the Golden Gate Bridge) of upstream clocking and other infections.

We certainly won’t make any claims yet, but we are trying to use some new measurement methods (including wavelet analysis s/w) to at last validate some of the things which, while really not hard to hear, seem to have eluded conventional FFT, etc.

[jabbr]

3 hours ago, adamdea said:

Yes but it’s slightly question-begging, or at least point-missing.  In information-theoretic terms the bitsteams seem possibly to be  identical. Differences in sound quality are (at least arguably) not properties  of the bitstream, they are properties of the dac. 

No no no no no.

 

The electrical signals may be bit-identical without being electrically identical — nothing to do with the DAC. 

 

A perfectly isolated DAC, according to my understanding of the term “isolated”, will have the same sound regardless of electrical differences in the “bit identical” incoming bitstream — eg a regen or other isolating device, would, if the DAC were perfectly isolated, not make a difference in SQ.

[Ralf11]

1 hour ago, Superdad said:

...

Mansr or someone else suggested looking at variations in close-in phase-noise ...

 

 

I can't believe that @mansr would trespass on a term trademarked by @jabbr

[jabbr]

1 hour ago, Andyman said:

Hmm. I like your posts jab and you know your stuff, but not sure on this. It’s sort of a circular argument which may well support the exact opposite of its own intent. 

My definition of “isolation” is hardly circular. My intention of providing a definition is so that the term “isolation” can be meaningfully used in discussion — that’s my intent. Do you have a different definition? Please tell?