Jitter to its merits
Right SY,
Jitter at the DAC is the deviation of the exact timing of the conversion from a digital word to an analog value. As pointed out earlier in this thread the output of a DAC is an S&H pulse. The analog low pass filter integrates these output pulses. What makes the total amplitude after the filter is the AEREA of the pulses out of the DAC, thus H(eight) x W(idth).
So what does jitter? It modulates the amplitude of the output signal. Let’s look at a 16 bit DAC and 44.1 kHz sampled signal. Then a timing error of 1 / (44.1e3 x 16^2) = 350 ps gives an equivalent error of 1 LSB. Note that this is at full scale, at lower amplitude the error is also less. For a constant 1 KHz sine signal jitter results in just side band noise. But for music this is quite different. Music consists on smaller signals riding on bigger signals. The error generated by jitter depends on the momentary signal amplitude. With music, when small signals are combined with large signals the smaller signals are much more distorted due to jitter.
With lower jitter the smaller signals are less distorted and this is what we say hey there is more detail to hear. Any one who fiddled with low jitter clocks can confirm this.
Note that jitter does not just add noise to the signal as some say, it modulates the signal with noise and that is quite different. So the lower the jitter, the better.
Cheers
Right SY,
Jitter at the DAC is the deviation of the exact timing of the conversion from a digital word to an analog value. As pointed out earlier in this thread the output of a DAC is an S&H pulse. The analog low pass filter integrates these output pulses. What makes the total amplitude after the filter is the AEREA of the pulses out of the DAC, thus H(eight) x W(idth).
So what does jitter? It modulates the amplitude of the output signal. Let’s look at a 16 bit DAC and 44.1 kHz sampled signal. Then a timing error of 1 / (44.1e3 x 16^2) = 350 ps gives an equivalent error of 1 LSB. Note that this is at full scale, at lower amplitude the error is also less. For a constant 1 KHz sine signal jitter results in just side band noise. But for music this is quite different. Music consists on smaller signals riding on bigger signals. The error generated by jitter depends on the momentary signal amplitude. With music, when small signals are combined with large signals the smaller signals are much more distorted due to jitter.
With lower jitter the smaller signals are less distorted and this is what we say hey there is more detail to hear. Any one who fiddled with low jitter clocks can confirm this.
Note that jitter does not just add noise to the signal as some say, it modulates the signal with noise and that is quite different. So the lower the jitter, the better.
Cheers
Peufeu, as I said, I choose the PCM1704 for is capabilities, not its price or its “sound.” My current implementation involves a bit of custom DSP and, as such, probably sounds different from what you might hear using the chips in the conventional way. You should ask someone else what a “normal” PCM1704 sounds like.
Ulas said:
Hi Ulas,
In stead of all this theory, why don't you do like the rest here in the forum? Publish your design, make your PCB available and we all have the chance to "eat the pudding" (as that is the proof...)
I know you find my design of the dddac1543 childish, but at least I share it, so every one can make up his own mind.
By the way, you are right on the theoretical side of the 4040. Still interesting, that the reclocked position is clearly an improvement over the spdiff generated clock (what was the bet by the way?
) I have measured this as well and the results can be seen at my site: less jitter. Now we can do a lot of speculation why this all is, but this is the same for cables, tubes, direction, etc etc. Still in audio, we seem not to be able to explain every thing straight from the book or datasheet Look forward to your post of your DAC !
Thanks
doede
PS: why challenge us all the time with: "I promise, I won’t interrupt again" We are all just a bunch of enthousiastic DIY trying to help each other, so you are not interupting at all...... you and your ideas are most welcome, may be a little tune on the tone and the music will sound much better
No idea what is inside one, but if you look close at the specs, they say it is for a jitter frequency of less than 1(?)kHz. Also makes me wonder on what type of supply they they used to obtain that number. The ones the gang here might whip up may not be as good.
Maybe I can get tiroth to let me tear one of his up!
Jocko
Maybe I can get tiroth to let me tear one of his up!
Jocko
Yeah, I noticed some weasel words in the spec sheet: "FJ > 1KHz", but also the banner headline "Very Low Phase Jitter at All Frequencies."
I reckon that to realize theoretical peak performance of PCM1704U-K (120dB SNR with a 768KHz, 24-bit input) requires a timebase with jitter less than 414 femtoseconds (that's 414e-15, for those keeping score at home).
To achieve 144dB SNR on some nonexistent perfect 24-bit DAC would require < 25 femtoseconds jitter.
I reckon that to realize theoretical peak performance of PCM1704U-K (120dB SNR with a 768KHz, 24-bit input) requires a timebase with jitter less than 414 femtoseconds (that's 414e-15, for those keeping score at home).
To achieve 144dB SNR on some nonexistent perfect 24-bit DAC would require < 25 femtoseconds jitter.
I would like to see the stuff that can measure that low....so I can steal it.
But back to jitter....I have been making some measurements on the jitter on the recovered clock on two different SPDIF RX chips. Anyone interested in what I found????
(Hint: the loop filter does some strange things...............)
Jocko
But back to jitter....I have been making some measurements on the jitter on the recovered clock on two different SPDIF RX chips. Anyone interested in what I found????
(Hint: the loop filter does some strange things...............)
Jocko
I don't know much about all the digital stuff but feel compelled inqure: Has anyone done a null test configuration on the process? That is, take a test signal - process though a typical ADC - decode through a typical DAC - then compare to the inverse of the orginal signal. You could induce more or less clock error, substitute diferent DACs and so on to get some idea of how sensative the process is to different components and how much digital error actually shows up in the analog output.
Being an ignoramous in this area it oftens seems to me that much of what is debated is far removed from anything one could possibly hear. Being and ignoramous I could be totally off base. A null test of analog in vs. analog out is something I could grasp a little better.
Being an ignoramous in this area it oftens seems to me that much of what is debated is far removed from anything one could possibly hear. Being and ignoramous I could be totally off base. A null test of analog in vs. analog out is something I could grasp a little better.
Null Test
Hi Sam9,
I have not done a complete null test as I lack a AD convertor but I did compare several LP records with the same pressing on CD. Quite revealing!
A firend of mine also has the Mark Levinson organ recording on SACD I have on LP. Looking forward to that one too.
My opinion is that what is debated is quite related to what I hear, but I admit I have quite revealing loudspeakers: Klipschorns. Or better saying no mercy speakers.
sam9 said:
Hi Sam9,
I have not done a complete null test as I lack a AD convertor but I did compare several LP records with the same pressing on CD. Quite revealing!
A firend of mine also has the Mark Levinson organ recording on SACD I have on LP. Looking forward to that one too.
My opinion is that what is debated is quite related to what I hear, but I admit I have quite revealing loudspeakers: Klipschorns. Or better saying no mercy speakers.
Doede, I opened this thread with the question: “Isn’t there more to digital audio than jitter reduction?” I see from the response the answer is a resounding NO! I do not wish to waste my time or this forum’s bandwidth and disk space if there is no interest in any other topics of discussion. There already appears to be enough OT nonsense with pictures of women and mountains. I don’t get a thrill seeing my words posted on the Internet and it is not easy for me to publish schematics because I don’t use them. I use a text-based design system I developed 30 years ago at the dawn of the computer age.
As for my DAC design: It is a tool for my exploration, not a finished product. Perhaps, when I reach the end of that journey, I’ll create schematics and publish them here.
As for your DAC design: Just because you prefer the sound with the asynchronous clock divider doesn’t mean there is less jitter. Why don’t you measure the jitter where it really matters: at the pins of the DAC chip(s). And while you’re at it, why don’t you replace the HC4040 with a pair of VHC161s and report how they measure and sound?
Pjotr informs us that “a timing error of 1 / (44.1e3 x 16^2) = 350 ps gives an equivalent error of 1 LSB.” Very interesting. Many of you claim to hear a few ps of jitter yet can’t hear the 1 LSB error that comes about when the digital data stream is inverted to enable a differential pair of DAC chips. I think you all hear what you want to hear.
As for my DAC design: It is a tool for my exploration, not a finished product. Perhaps, when I reach the end of that journey, I’ll create schematics and publish them here.
As for your DAC design: Just because you prefer the sound with the asynchronous clock divider doesn’t mean there is less jitter. Why don’t you measure the jitter where it really matters: at the pins of the DAC chip(s). And while you’re at it, why don’t you replace the HC4040 with a pair of VHC161s and report how they measure and sound?
Pjotr informs us that “a timing error of 1 / (44.1e3 x 16^2) = 350 ps gives an equivalent error of 1 LSB.” Very interesting. Many of you claim to hear a few ps of jitter yet can’t hear the 1 LSB error that comes about when the digital data stream is inverted to enable a differential pair of DAC chips. I think you all hear what you want to hear.
Your best guess..........
Not quite right.
Yes, lots of peaks.
Where??
Depends on the R-C filter that you select. And that is the rub........which one is best.
(Don't know yet.......)
So........."bub", what time constants did you use?
14 months, eh.........
Any idea how long it took to make the gear to measure it???? And make sure that it is right???
Jocko
Not quite right.
Yes, lots of peaks.
Where??
Depends on the R-C filter that you select. And that is the rub........which one is best.
(Don't know yet.......)
So........."bub", what time constants did you use?
14 months, eh.........
Any idea how long it took to make the gear to measure it???? And make sure that it is right???
Jocko
Jitter Will Give You 'The Jitters'............
Jitter adds non music harmonics related dirt and harshness, and non relaxedness into sounds.
Jitter amplitude is significant, but more so is the spectral/dynamic nature of the jitter.
This spectral characteristic can create patterns of noise in the audio that are revealled in close listening - once noticed these patterns can become obvious and distracting.
Some of these patterns can be sonically quite acceptable too, but less jitter is better - less jitter makes for a cleaner, clearer and more detailed, more relaxed and matter of fact presentation as opposed to a weirdly coloured and dynamically varying presentation.
Vocals and highs become more real, and extended.
Cybals ring all the way out instead of tishing and ending prematurely, and 'anxiety/tremor' and 'zedding' in voices diminishes.
Lows become 'anchored' to the ground instead of floating and bouncing around somewhere above the floor.
Distance imaging is improved - you can start getting sounds from behind too.
Virtually all standard Cd players leave the crystal can floating.
A quick and easy experiment is to carefully solder a thin wire (wirewrap wire) to the crystal can, and try connecting the can to local ground or either of the crystal pin connections, or leave it unconnected - you may need to drill thin holes in the pcb.
This should give you four finely different sounds.
Beware however, once you have heard these four sounds you will have tuned your ear to hearing jitter artifacts and will continue to hear them.
Eric.
Hi Av_phile,av_phile said:
Jitter adds non music harmonics related dirt and harshness, and non relaxedness into sounds.
Jitter amplitude is significant, but more so is the spectral/dynamic nature of the jitter.
This spectral characteristic can create patterns of noise in the audio that are revealled in close listening - once noticed these patterns can become obvious and distracting.
Some of these patterns can be sonically quite acceptable too, but less jitter is better - less jitter makes for a cleaner, clearer and more detailed, more relaxed and matter of fact presentation as opposed to a weirdly coloured and dynamically varying presentation.
Vocals and highs become more real, and extended.
Cybals ring all the way out instead of tishing and ending prematurely, and 'anxiety/tremor' and 'zedding' in voices diminishes.
Lows become 'anchored' to the ground instead of floating and bouncing around somewhere above the floor.
Distance imaging is improved - you can start getting sounds from behind too.
Virtually all standard Cd players leave the crystal can floating.
A quick and easy experiment is to carefully solder a thin wire (wirewrap wire) to the crystal can, and try connecting the can to local ground or either of the crystal pin connections, or leave it unconnected - you may need to drill thin holes in the pcb.
This should give you four finely different sounds.
Beware however, once you have heard these four sounds you will have tuned your ear to hearing jitter artifacts and will continue to hear them.
Eric.
ULESS,
Ok, so show us your test results, and prove to use that jitter does not cause problems in synchronization, which leads to signal degradation. You might not know this but there is plenty of proof in the industry that jitter cause problems in communications equipment problems, video imaging and even audio distortion. Seems like you need to brush up on the subject, let see how about Telecordia GR-1377 or ITU-T O.172 jitter measurement standards, which spells out how to test for jitter and meet performance and design objectives.
Ok, so show us your test results, and prove to use that jitter does not cause problems in synchronization, which leads to signal degradation. You might not know this but there is plenty of proof in the industry that jitter cause problems in communications equipment problems, video imaging and even audio distortion. Seems like you need to brush up on the subject, let see how about Telecordia GR-1377 or ITU-T O.172 jitter measurement standards, which spells out how to test for jitter and meet performance and design objectives.
Isn’t there more to digital audio than jitter reduction?
In this post I made a suggestion that it is not all jitter reduction. There is more to it but I don't know precisely what.
http://www.diyaudio.com/forums/showthread.php?postid=285674#post285674
I made a dedicated supply for the TDA1543 to make it better sounding, more LP like if you want. I don't think the latter is a jitter issue. You are not ready with jitter reduction only. Besides that I believe it is better to prevent jitter rather than to cure it with a jitter reduction device in the DAC as Doede does or my Asynchronous Reclocker, hence my I2S Direct scheme and KWAK-CLOCK.
I apologize for posting off topic pictures of women. Sometimes the off topic issue seems more interesting.
I gave up upon a balanced DAC, as the two halves of the signal were not exactly equal. I found the highs better with a single ended DAC.
In your profile I see that you have built harpsichords and I have the highest esteem for that. You surely can hear things! So I would be interested what you hear in jitter reduction, NON-OS DACs etc. The harpsichord is the most difficult instrument to reproduce through a digital system but I love that kind of music very much. It is my ultimate test of clocks, DACs etc.
In closing, with all due respect to all forum members, I don't believe you will find the ultimate answer to your question on a public forum. That kind of information is tucked well away in the vaults of some manufacturers and I can assure you that I don't post all I know either........
Hello Ulas,Ulas said:
In this post I made a suggestion that it is not all jitter reduction. There is more to it but I don't know precisely what.
http://www.diyaudio.com/forums/showthread.php?postid=285674#post285674
I made a dedicated supply for the TDA1543 to make it better sounding, more LP like if you want. I don't think the latter is a jitter issue. You are not ready with jitter reduction only. Besides that I believe it is better to prevent jitter rather than to cure it with a jitter reduction device in the DAC as Doede does or my Asynchronous Reclocker, hence my I2S Direct scheme and KWAK-CLOCK.
I apologize for posting off topic pictures of women. Sometimes the off topic issue seems more interesting.
I gave up upon a balanced DAC, as the two halves of the signal were not exactly equal. I found the highs better with a single ended DAC.
In your profile I see that you have built harpsichords and I have the highest esteem for that. You surely can hear things! So I would be interested what you hear in jitter reduction, NON-OS DACs etc. The harpsichord is the most difficult instrument to reproduce through a digital system but I love that kind of music very much. It is my ultimate test of clocks, DACs etc.
In closing, with all due respect to all forum members, I don't believe you will find the ultimate answer to your question on a public forum. That kind of information is tucked well away in the vaults of some manufacturers and I can assure you that I don't post all I know either........
Why the preccupation with jitter?
Hi Lars,
I don´t understand why you don´t like
the PCM1704 D/A, the problem is not the money, your referents D/A of Cyrrus and AKM Delta/Sigma convertes these convertes are limites in sound becouse you are having the analog sound of its.
Ok, you can put the analog part of you DAC with more OP but the final result depent of last analog source of DAC.
With the DAC courrent source these problem don´t exist and you can put a discrete source, the OP has limits.
The firs limit is the Vcc and with it the dinamic response, don´t work in clase A
etc,etc....
I´m doing the DAC proyect with 8 PCM1740K and I know that these proyect is expensive but I chouse its for the sound quality.
I´m agree with Ulas
Jesús
Hi Lars,
I don´t understand why you don´t like
the PCM1704 D/A, the problem is not the money, your referents D/A of Cyrrus and AKM Delta/Sigma convertes these convertes are limites in sound becouse you are having the analog sound of its.
Ok, you can put the analog part of you DAC with more OP but the final result depent of last analog source of DAC.
With the DAC courrent source these problem don´t exist and you can put a discrete source, the OP has limits.
The firs limit is the Vcc and with it the dinamic response, don´t work in clase A
etc,etc....
I´m doing the DAC proyect with 8 PCM1740K and I know that these proyect is expensive but I chouse its for the sound quality.
I´m agree with Ulas
Jesús
Hi,
First let me correct a typo error, I wrote “1 / (44.1e3 x 16^2) = 350 ps “. This must be of course 1 / (44.1e3 x 2^16) = 350 ps
Back to:
Apart from theoretical issues, from experience I can say that a jitter reduction (at the clock signal that is reponsible for the actual conversion) from say 200ps to 20 ps is clearly audible. It brings more detail, a more relaxed sound and etc. etc. So the preoccupation with jitter is not overdone IMHO. But I doubt a reduction from 20ps to 2 ps will be a big improvement with regard to the whole DA conversion process, if it is audible at all.
Another important point is PSU noise, or more precisely noise on the DAC’s reference voltage. The output of the DAC is related to its Vref, whether it is a parallel current out DAC or a Delta-Sigma DAC. In some sense noise on Vref has more or less the same influence as jitter, it modulates the output voltage of the DAC. Since with most audio DACs Vref is taken from the PSU we can also put up the question:
“Why the preoccupation with PSU noise?”
Cheers
First let me correct a typo error, I wrote “1 / (44.1e3 x 16^2) = 350 ps “. This must be of course 1 / (44.1e3 x 2^16) = 350 ps
Back to:
Apart from theoretical issues, from experience I can say that a jitter reduction (at the clock signal that is reponsible for the actual conversion) from say 200ps to 20 ps is clearly audible. It brings more detail, a more relaxed sound and etc. etc. So the preoccupation with jitter is not overdone IMHO. But I doubt a reduction from 20ps to 2 ps will be a big improvement with regard to the whole DA conversion process, if it is audible at all.
Another important point is PSU noise, or more precisely noise on the DAC’s reference voltage. The output of the DAC is related to its Vref, whether it is a parallel current out DAC or a Delta-Sigma DAC. In some sense noise on Vref has more or less the same influence as jitter, it modulates the output voltage of the DAC. Since with most audio DACs Vref is taken from the PSU we can also put up the question:
“Why the preoccupation with PSU noise?”
Cheers
If you have listened to lots of analogue recordingds certainly can find out that the attitudes of digital audio were and are present in analogue recordings (dinamics , channel separation ) . Is not that I ear what I want to ear , but the fact is I use this " microscopic mediums" and try to enjoy the music as best as they can reveal to me . And there are lots of potentialities .
For example :
I ear the difference between the single ended version of the nonos dac from a right DATA and an inverted DATA from a flip flop (inverting the absolute phase at the speakers too , of course ) nevertheless i prefear the BALANCED sound over the SE.
The results from a balanced gear depends heavily from the PS ,
For example i do not like at all the triple darlington PS provided by ELSO ( thanks Elso , your circuits are a very good points to start for a DIYer) and I used it for 3 months .
I enjoy the differences , and I search for balance in audio reproduction, not absolute truths.
Also a good clock is a starting point after all , and when you do extensive listening sessions and lots of iron solderings it is normal that some preoccupations occurs. Music is the victim , not a pair of off topic posts.
For example :
I ear the difference between the single ended version of the nonos dac from a right DATA and an inverted DATA from a flip flop (inverting the absolute phase at the speakers too , of course ) nevertheless i prefear the BALANCED sound over the SE.
The results from a balanced gear depends heavily from the PS ,
For example i do not like at all the triple darlington PS provided by ELSO ( thanks Elso , your circuits are a very good points to start for a DIYer) and I used it for 3 months .
I enjoy the differences , and I search for balance in audio reproduction, not absolute truths.
Also a good clock is a starting point after all , and when you do extensive listening sessions and lots of iron solderings it is normal that some preoccupations occurs. Music is the victim , not a pair of off topic posts.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.