Once again, you proved you have no idea what you are talking about, and mechanically copy/pasting from Analog's AN-756 is not helping your cause.
You did not even bother to read what you copy/pasted, otherwise you would sense the enormity of considering an "encode rate of 61.44 MSPS" in the context of audio. Here, the "encode rate" was 48KSPS (more than 3 orders of magnitude less).
Sir, I am afraid you don't even deserve the benefit of ignorance.
Kudo's for trying and keeping the hopes up syn08, much appreciated!
A few questions:
1:
A 1 Hz signal goes against our (Chris and me) earlier assessment that higher frequencies (and not infra) should illustrate the problems better.
If, for every doubling in frequency one should add 6 dB of phase noise, then starting from 1 Hz this would mean a significant chance in outcome, for a signal of, let's say 1024 Hz. (Even though I find it hard to believe it would scale that way, I assume you get the idea behind the reasoning).
2:
Despite it's mediocre results, according to that same article, you did took one of the better dac's to choose from (a sigma delta/delta sigma if I'm not mistaken), which probably uses a very high sample frequency.
Do you think that taking a conventional r2r dac would result in substantial different (more problematic) outcome?
These questions arise from the earlier discussed article that Chris referred to and in which those rules were explained.
Do you or don't you agree with that article (why not) and would you like to take these measurements again, with a much higher frequency signal and/or a different dac structure?
It seems either the article is wrong, or the measurement is a bit one sided, which is it?
Yes, it's true dsm dacs are the standard, even in high end dacs.
I don't know the details of this experiment, but usually 48-32 is way too close to be statistically relevant for a sample size of 80, in a binomial test. Otherwise said, the probability of this score being the result of pure guesses is too high to consider the results relevant. Such a result requires either multiple repeats, or increasing significantly the sample size.
1: Signal is 2KHz, not 1Hz. The FFT bandwidth is 10Hz (Hor: 1Hz/division in the screenshot). I was looking at the close-in phase noise level under 1Hz away from the carrier, which is what others are claiming it matters.
2: I don't know if an R2R DAC will provide better or worse results, but this is beyond the point. It is not about the DAC here, it is about the effect of a cheap clock on the close-in noise in the analog domain. The mediocre DAC shows up in the poor SNR, hiding a full oscillator characterization (the phase noise corner frequency is below the noise floor of the DAC). I don't have any R2R DACs to test, but I may eventually test an 18bit LTC2756 multiplying DAC.
Last edited:
SYN08,
now that you painted Andrea Mori as a fraud, as i predicted, can you resume in a single post what exactly do you reproach Andrea, and why , by «vulgarizing» with metaphors of everyday life examples so that everyone, audiophiles or not, can understand the alleged fraud you claim observing.
Andrea mori is a contributor here since a long time, and it should be clear to everyone what pattern dos Andrea uses, and please tell us what you are going to do about it and what should be done.
thank you.
now that you painted Andrea Mori as a fraud, as i predicted, can you resume in a single post what exactly do you reproach Andrea, and why , by «vulgarizing» with metaphors of everyday life examples so that everyone, audiophiles or not, can understand the alleged fraud you claim observing.
Andrea mori is a contributor here since a long time, and it should be clear to everyone what pattern dos Andrea uses, and please tell us what you are going to do about it and what should be done.
thank you.
Do I have to spell everything out for you....
Check the graph of jitter differences between 'Tools' on internal clock and slaved external WC (CC1).
a/ The Tools baseline jitter is a lot higher than CC1 everywhere (around 100x)
b/ The only difference of slaved Tools is below 60Hz or so (CF of Tools PLL)
c/ We are talking about a few PSec of jitter difference.
So it is surprising that they could discern a difference at all. We are discussing discerning small differences, this would have to be minuscule.
WRT your other comment, Bob is referring to a/ modern Pro converters ability to reject WC jitter b/ internal clocks baseline jitter.
IOW, no advantage using external MC in almost all cases - these days.
TCD
Just yourself would do, adding more people who arent as familiar with the test system or the differences from the clocks would likely only pollute the results
Hmm. When I saw the result of Syn8, it dawned on me that this one I can repeat too. Not exactly, because I only have an 3561a, but the core of it, like to have a close look at a single tone signal, with an instrument developed for this purpose, it could be done.
The declared dynamic range is 80dB, and in fact, I can't get a lower floor than ~-85/86dB.
But looking at the peak zone, how much it is enlarging, could be done. The card is my EMU1616m, i'm generating a full scale 17997Hz signal and I'm using one half of the balanced output.
Only ~18kHz, because it drops too much approaching the max freq. output.
I put it here, just so as to disturb the waters..
10dB/div, 6,11mHz RBW, 600mHz span, (0,6Hz the whole screen..) window is flattop
The signal has a cca. 15mHz FullWidth -75dBc down
The declared dynamic range is 80dB, and in fact, I can't get a lower floor than ~-85/86dB.
But looking at the peak zone, how much it is enlarging, could be done. The card is my EMU1616m, i'm generating a full scale 17997Hz signal and I'm using one half of the balanced output.
Only ~18kHz, because it drops too much approaching the max freq. output.
I put it here, just so as to disturb the waters..
10dB/div, 6,11mHz RBW, 600mHz span, (0,6Hz the whole screen..) window is flattop
The signal has a cca. 15mHz FullWidth -75dBc down
Attachments
Last edited:
Here is a loopback measurement of an RTX. It 192 KHz sample rate and 16M points. It took a while but you can see close in with this. Its 1 KHz +/- 5 Hz. There are some windowing issues I'm sure but its something like .01Hz resolution.
I do not have a similar test between two boxes running on separate clocks however i may be able to set that up later this week.
I do not have a similar test between two boxes running on separate clocks however i may be able to set that up later this week.
Attachments
I am not a critic per se.... more of a live and let live person; whatever keeps people happy's fine with me kinda guy... and I sometimes listen to music while standing
but I'll chime in since you asked...I tend to look at the measured performance BUT ALSO I let my ears and heart do the judgment, unrelated to measurements. If there's a clash, the latter gets priority.
I use HQPlayer 4.5.1 running ASDM7EC modulator at 44.1k x256 on NUC11 running Windows Server 2012 R2 OS, linear power supply with very low noise multi-regs working in parallel, and Holo May DAC L2. Bliss
... LNS15 PCM upsampler at 1.536M with poly-sinc-lp filter also sounds amazing (but it takes a bit longer to buffer due to demanding filter)However, 70% of the time I like that R-2R DAC in NOS, with "raw" stream fed to it... no upsampling of any kind, using JRiver as a player.
So... many options to choose from, all sound very nice and I tend to choose one over the other based on the source material' original resolution... and my mood.
I think you already know that I also have Aleph J (my modified MKII version
); Dynaudio speakers.
Demian, not sure if a loopback is correct/relevant. If the DAC and the ADC are running on the same clocks (as I assume they do) then the phase noise is correlated and it averages out. What you got is mostly a windowing effect (use a flat window, BTW, since it preserves the amplitudes) You need an independent ADC, running on its own clock (with very low phase noise) in order to analyze the DAC clock effect.
I don't know if you do but many who say they let their ears be the judge, do the sighted listening. In other words, they let their ears and eyes be the judge, which isn't the same as using ears only.
The cumulative probability to get 48 correct results out of 80 by random guessing is P(X>=48)=0.047 , so asssuming that it was testing what it was supposed to test, certainly a result worth a second look.
We had a thread to tell what kind of evidence would be accepted if contrary to ones own beliefs; afair it was closed after getting ugly, but it was quite obvious that most sceptics were not able (or willing) to state what they would accept.
Btw, something like an "ears only" listening test does not exist .....
We had a thread to tell what kind of evidence would be accepted if contrary to ones own beliefs; afair it was closed after getting ugly, but it was quite obvious that most sceptics were not able (or willing) to state what they would accept.
Btw, something like an "ears only" listening test does not exist .....
You measured the the wideband noise of the clock.
Phase and frequency are different things.
If I haven't demonstrated anything with my listening impressions, you haven't proved anything with your measurements.
But the most important thing is that we have understood the DACs are all the same and therefore they should all cost 47 EURO.
The audio world is full of cheaters.
I believe you listen to your instruments rather than to music.
Last edited:
Really, are you sure?
Yes, the carrier frequency does not matter.
You keep measuring the phase noise converting it in SNR.
I does not work.