waki, the test tone you generated is indeed the J-test signal - 1/4fs and 1/192fs LSB toggle. However, that's not to test the DAC chip, rather to test the PLL of the SPDIF receiver. Its a signal designed specifically for SPDIF transmission.
The Jtest signal (available on this forum since 2006 HERE)
is a tool to test the ability of a Dac to convert signals with the less jitter artifacts possible when fed through SPDIF.
Or am I missing something, and the DCX2496 has some other digital inputs?
What are you using when listening music?
Or you just claim that there is no jitter present in SPDIF?
Would like to point out that the Jtest signal is just an audio signal like any other. The jitter is generated during the SPDIF transfer.
It is special in that it 'excites' clearly the constantly present underlying data correlated jitter mechanisms. But it excites them exactly the same way like a violin would do. The Jtest permits only to observe it better.
Anyway, the green track is just showing how much better it can be done, in the same DCX. So it is relevant, and depends on the DAC itself. The original is bad.
Ciao, George
is a tool to test the ability of a Dac to convert signals with the less jitter artifacts possible when fed through SPDIF.
Or am I missing something, and the DCX2496 has some other digital inputs?
What are you using when listening music?
Or you just claim that there is no jitter present in SPDIF?
Would like to point out that the Jtest signal is just an audio signal like any other. The jitter is generated during the SPDIF transfer.
It is special in that it 'excites' clearly the constantly present underlying data correlated jitter mechanisms. But it excites them exactly the same way like a violin would do. The Jtest permits only to observe it better.
Anyway, the green track is just showing how much better it can be done, in the same DCX. So it is relevant, and depends on the DAC itself. The original is bad.
Ciao, George
Last edited:
Yeah you're right, however this is not the main thrust of my comment, which is that the signal is the J-test signal, and a totally false impression was being created by the fact that JosephK failed to point this out.
w
So, jkeny declines to test his own equipment, let's see if JosephK will put his money where his mouth is.
Where are these results that will convince us all? As usual, you show a picture, but not the picture that really tells the story.
Show an eye diagram of a recovered clock improved to a degree that is credibly audible. Cite a study that confirms the audibility of that level of jitter.
Produce these and you will convince us. Producing results to discredit other tests do not make your case. Comments like this:- '(not cheap chinese sh..)' only lead us to believe you are lacking in objectivism. The price of equipment is irrelevant. The country of origin of the equipment is irrelevant. Only the performance is relevant.
The Benjamin/Gannon study showed a 10-20nS requirement for audibility of deterministic jitter. Ashihara et. al. find a threshold of 250nS for random jitter.
I have seen the Keith Johnson claims that he can hear 8-15pS of jitter. Claims.
Produce an FFT yourself JosephK, showing reasonable evidence of audible differences.
w
Where are these results that will convince us all? As usual, you show a picture, but not the picture that really tells the story.
Show an eye diagram of a recovered clock improved to a degree that is credibly audible. Cite a study that confirms the audibility of that level of jitter.
Produce these and you will convince us. Producing results to discredit other tests do not make your case. Comments like this:- '(not cheap chinese sh..)' only lead us to believe you are lacking in objectivism. The price of equipment is irrelevant. The country of origin of the equipment is irrelevant. Only the performance is relevant.
The Benjamin/Gannon study showed a 10-20nS requirement for audibility of deterministic jitter. Ashihara et. al. find a threshold of 250nS for random jitter.
I have seen the Keith Johnson claims that he can hear 8-15pS of jitter. Claims.
Produce an FFT yourself JosephK, showing reasonable evidence of audible differences.
w
You have seen them.
Or is it possible that there would be more persons using the same avatar?
http://www.diyaudio.com/forums/everything-else/181793-rf-audio-14.html
And I did not test anything which could be connected to Jkeny.
Otherways you are off topic here, and I think the warning is out that expanding on all this would be just forcing on a topic already closed - so no comments from my side, sorry.
Ciao, George
I think it may be time to go back and read the original post.
Ostensibly, the question was whether FFT (meaning spectrum analysis with FFT) is a good metric of sound quality. And is phase also important.
The answer would have to be yes to both. Both are valuable and show different things.
But that has not been the bone of contention. The conflict comes from devices judged to be equivalent because the spectrum analysis showed no significant difference. On the surface, it may seem wise to question this, but that not what is really happening, because the claims were fairly simple.
There were SPDIF devices that were said to have lower jitter than others, or to reduce jitter caused by bad cabling, connections or terminations. This lower jitter is said to result in noticeably better sound. Better sound is a good thing. But can we measure that difference? Of course we can. The SPDIF stream can be analyzed for jitter. It can be measured to be better, worse or the same as some other stream.
But few of us actually listen to the SPDIF stream directly. The electrical analog of what we listen to is the analog signal that exits the receiver chip + DAC combo. For there to be an audible difference, there must be a difference in this analog signal. What happens before the D/A conversion, we can not directly hear. If the jitter or PSU noise does not make it thru into the A/D conversion stage, we will not hear it. It will not be contained in the analog output of the DAC. Of course it often is. But sometimes it isn't.
We know that jitter will manifest itself in FFT spectrum analysis as sidebands around single tone. These are fairly easy to see in the spectrum analysis of the analog signal. They may even be heard if strong enough. Power supply noise can also be seen as sidebands spaced at multiples of the mains frequency. Those also might be heard if strong enough.
Several devices were compared by analyzing the analog signal issued from the A/D conversion, the signal we actually send to the amps and speakers. Some slight differences were found, but thought not enough to be significant. Hearing artifacts 100dB below the signal may be very difficult. The devices were judges not significantly different.
So why were no significant differences found down to a noise floor of -120dB or better? If there were a jitter or noise difference, wouldn't it be seen in the analysis? To be heard, it must show up in the analog signal, no matter what is happening in the SPDIF stream.
In my testing I DID find noticeable differences in some devices when looking at the analog signal with spectrum analysis. I believe (but can not prove) that I also hear these differences. Cleaner sounds better to me, BTW.
But with other devices I could not find much difference in the analog signal, even down to -120dB or more. Interestingly, I could hear no difference, either. Listening tests were done first, measurements second.
Bottom line. The sonic differences are said to be due to lower jitter. To be heard, this must have an effect on the analog signal. Differences in jitter can easily be seen in the analog signal via spectrum analysis, even down to -120dB. If no difference is seen, then is the SPDIF jitter difference actually finding its way into the analog signal? That, my friends, is the question.
Ostensibly, the question was whether FFT (meaning spectrum analysis with FFT) is a good metric of sound quality. And is phase also important.
The answer would have to be yes to both. Both are valuable and show different things.
But that has not been the bone of contention. The conflict comes from devices judged to be equivalent because the spectrum analysis showed no significant difference. On the surface, it may seem wise to question this, but that not what is really happening, because the claims were fairly simple.
There were SPDIF devices that were said to have lower jitter than others, or to reduce jitter caused by bad cabling, connections or terminations. This lower jitter is said to result in noticeably better sound. Better sound is a good thing. But can we measure that difference? Of course we can. The SPDIF stream can be analyzed for jitter. It can be measured to be better, worse or the same as some other stream.
But few of us actually listen to the SPDIF stream directly. The electrical analog of what we listen to is the analog signal that exits the receiver chip + DAC combo. For there to be an audible difference, there must be a difference in this analog signal. What happens before the D/A conversion, we can not directly hear. If the jitter or PSU noise does not make it thru into the A/D conversion stage, we will not hear it. It will not be contained in the analog output of the DAC. Of course it often is. But sometimes it isn't.
We know that jitter will manifest itself in FFT spectrum analysis as sidebands around single tone. These are fairly easy to see in the spectrum analysis of the analog signal. They may even be heard if strong enough. Power supply noise can also be seen as sidebands spaced at multiples of the mains frequency. Those also might be heard if strong enough.
Several devices were compared by analyzing the analog signal issued from the A/D conversion, the signal we actually send to the amps and speakers. Some slight differences were found, but thought not enough to be significant. Hearing artifacts 100dB below the signal may be very difficult. The devices were judges not significantly different.
So why were no significant differences found down to a noise floor of -120dB or better? If there were a jitter or noise difference, wouldn't it be seen in the analysis? To be heard, it must show up in the analog signal, no matter what is happening in the SPDIF stream.
In my testing I DID find noticeable differences in some devices when looking at the analog signal with spectrum analysis. I believe (but can not prove) that I also hear these differences. Cleaner sounds better to me, BTW.
But with other devices I could not find much difference in the analog signal, even down to -120dB or more. Interestingly, I could hear no difference, either. Listening tests were done first, measurements second.
Bottom line. The sonic differences are said to be due to lower jitter. To be heard, this must have an effect on the analog signal. Differences in jitter can easily be seen in the analog signal via spectrum analysis, even down to -120dB. If no difference is seen, then is the SPDIF jitter difference actually finding its way into the analog signal? That, my friends, is the question.
Just to pick up on one issue raised Pano, because there's plenty of misunderstandings of FFTs around and it seems you're perpetuating one of them:
There's no such thing as a 'noise floor' when looking at an FFT plot. That's because an FFT is a histogram, not a graph. Its composed of discrete frequency points. This figure of -120dB which is being bandied around is the noise in one particular bin of the FFT output. The bin bandwidth is relevant and is substantially narrower than the audio bandwidth.
There's no such thing as a 'noise floor' when looking at an FFT plot. That's because an FFT is a histogram, not a graph. Its composed of discrete frequency points. This figure of -120dB which is being bandied around is the noise in one particular bin of the FFT output. The bin bandwidth is relevant and is substantially narrower than the audio bandwidth.
Pano, I appreciate your time in doing this nice summary.
I did start the thread to ask questions about the use of FFT plots as a measurement tool in audio. These questions included but I hope are not limited to:
- how well do FFT plots determine the sonic equivalence of two devices i.e given two FFT plots that are equivalent, can we say that they sound the same?
- Given just the FFT Spectrum analysis FFT plots, is thsi enough - are we missing much in not having FFT phase plots?
As to the bone of contention (which I was warned off from discussing but I'm presuming is OK now as you have raised it?), I have stated that:
- the FFT plots presented in this particular instance are specific only to Sy's equipment & set-up
- Your FFT plots (I don't think you posted them?) are specific to your equipment & set-up.
- This is evidenced by the fact that SY stated his FFT of a particular test using his DCX did not show the same side spurs as your posted FFT of the same data using your DCX.
- I raised the issue of repeatability as being problematic because of the number of variables involved in these measurements. As was stated here, repeatability is a necessary primary criteria for a test to qualify as in any scientific.
- Finally, you have mentioned the bottom line - is the lower jitter (demonstrated in Joseph K's plots & measurements done directly at the SPDIF output)of the device audible on a playback system? So now we enter into some territory that gives rise to contention:
- you might argue that a DAC can level the playing field by being able to remove jitter (or at least make it no-longer audible). I would argue with you that this is not so & would cite even the ESS DAC (recognised to be one of the more sophisticated jitter handling devices) as being audibly sensitive to different levels of jitter
- I would also cite the use of SRCs (the ESS has probably the most benign form of ASRC) to be detrimental to low jitter devices & to cause a masking of the audible benefits of low jitter. You have probably read PS-audio's statement about this (I know they are a commercial entity so not to be trusted
- This is one of the reasons I asked SY to demonstrate that his set-up was indeed capable of resolving what he claimed it could. Stating -120dB is not kosher - see below
- I see that Abrax has corrected the notion that the "noise floor" in an FFT is indeed a noise floor - so saying that we cannot hear below -120dB is a red-herring because this is not what is being shown in a FFT spectrum plot. This is one of the reasons I'm asking questions of those who use FFTs as a means of analysis - to clear up any issues of this nature.
I did start the thread to ask questions about the use of FFT plots as a measurement tool in audio. These questions included but I hope are not limited to:
- how well do FFT plots determine the sonic equivalence of two devices i.e given two FFT plots that are equivalent, can we say that they sound the same?
- Given just the FFT Spectrum analysis FFT plots, is thsi enough - are we missing much in not having FFT phase plots?
As to the bone of contention (which I was warned off from discussing but I'm presuming is OK now as you have raised it?), I have stated that:
- the FFT plots presented in this particular instance are specific only to Sy's equipment & set-up
- Your FFT plots (I don't think you posted them?) are specific to your equipment & set-up.
- This is evidenced by the fact that SY stated his FFT of a particular test using his DCX did not show the same side spurs as your posted FFT of the same data using your DCX.
- I raised the issue of repeatability as being problematic because of the number of variables involved in these measurements. As was stated here, repeatability is a necessary primary criteria for a test to qualify as in any scientific.
- Finally, you have mentioned the bottom line - is the lower jitter (demonstrated in Joseph K's plots & measurements done directly at the SPDIF output)of the device audible on a playback system? So now we enter into some territory that gives rise to contention:
- you might argue that a DAC can level the playing field by being able to remove jitter (or at least make it no-longer audible). I would argue with you that this is not so & would cite even the ESS DAC (recognised to be one of the more sophisticated jitter handling devices) as being audibly sensitive to different levels of jitter
- I would also cite the use of SRCs (the ESS has probably the most benign form of ASRC) to be detrimental to low jitter devices & to cause a masking of the audible benefits of low jitter. You have probably read PS-audio's statement about this (I know they are a commercial entity so not to be trusted
- The DCX has a first generation ASRC, CS8420, I believe, which has been said to sound bad & is buggy?
- This is one of the reasons I asked SY to demonstrate that his set-up was indeed capable of resolving what he claimed it could. Stating -120dB is not kosher - see below
- I see that Abrax has corrected the notion that the "noise floor" in an FFT is indeed a noise floor - so saying that we cannot hear below -120dB is a red-herring because this is not what is being shown in a FFT spectrum plot. This is one of the reasons I'm asking questions of those who use FFTs as a means of analysis - to clear up any issues of this nature.
The relevance of 'Dodge City' here is rather lost on me. Do please elucidate? Its somewhere in USA?
Keny, if your question were a legitimate "is FFT the final judge, or even a good one", that would be OK. It could be answered. The answer would be "it can show us many things, but not all."
But apparently that is not your question. Your question is closer to "Is this or that DAC good enough to show us the differences brought about by low jitter?" Not the same question.
So far you have NOT provided any proof that the DACs or measurements are not good enough.
Add to that the moving target. I measured and listened to a stock DCX, a modified DCX, A DEQ, a Citypulse DAC and my ESS DAC. Apparently none of these are sensitive enough to reveal the claimed differences. Oddly, they all do reveal a difference with an inferior SPDIF stream. And THAT is repeatable.
If these devices and measurements are not good enough, then we should also dismiss the measurements shown by Joseph K. However, if measurements are only deemed valid when they confirm your hypothesis, then this argument need go no further.
But apparently that is not your question. Your question is closer to "Is this or that DAC good enough to show us the differences brought about by low jitter?" Not the same question.
So far you have NOT provided any proof that the DACs or measurements are not good enough.
Add to that the moving target. I measured and listened to a stock DCX, a modified DCX, A DEQ, a Citypulse DAC and my ESS DAC. Apparently none of these are sensitive enough to reveal the claimed differences. Oddly, they all do reveal a difference with an inferior SPDIF stream. And THAT is repeatable.
If these devices and measurements are not good enough, then we should also dismiss the measurements shown by Joseph K. However, if measurements are only deemed valid when they confirm your hypothesis, then this argument need go no further.
Why? That could only follow reasonably if jkeny's contention was 'All measurements are useless'. As far as I can see that's not what he's claiming.
And another red herring to put alongside the 'didn't like the result' one.
- Status
- Not open for further replies.