To create a fully detailed lateral image, the (digital) medium needs to accurately resolve interchannel time differences as small as 2 microseconds. With tonally complex material and with 16 bit linear PCM, for instance, jitter of a few hundred picoseconds or more significantly degrades the imaging accuracy, keeping in mind that the instantaneous amplitude of the reconstructed signal will be incorrect because of the jittered timing. This tends to increasingly mask lower level detail and imaging as the overall amplitude of the digitized signal increases.
This corresponds to 500kHz. I doubt anyone can hear such timing differences.
The fact that you`re talking about the timing of music signals in combination with jitter tells me that you`re not aware of how jitter really influences the music signals. Jitter does not cause timing errors but amplitude errors in form of modulation products! Figure 5 in this paper shows a jitter modulated signal. This means that if your clock is for example modulated by 100Hz hum from the rectified mains, and your test signal is an 1kHz sinewave you`ll get two modulation products besides the 1kHz tone. One at 900Hz and another one at 1100Hz.
Regards
illusionxx
illusionxx said:
That article is very interesting. If I understand it properly, the period of the jitter determines the frequency of the noise generated. And I presume that the amplitude of the jitter determines the volume of the noise?
So does anyone know the formula to calculate the amplitude of the noise induced by jitter from the amplitude of the jitter? At what amplitude does the jitter induce noise equal to the noise floor of a CD?
illusionxx -
Apparently you overlooked this part of my post:
"...keeping in mind that the instantaneous amplitude of the reconstructed signal will be incorrect because of the jittered timing."
then you repeat the same thing in your own words as if I had not already established that.
So, in effect you actually agree with me on this point but want to make it appear to be some sort of dispute.
You also seem not to be aware that lateral localization differences of as little as 1 degree are distinguishable by the average listener. If you do the math, that is a result of a differential timing change of the incident waveform between the left and right ears of about 2 microseconds. It's simple algebra, and has nothing whatsoever to do with 'hearing 500 khz' or any similar false conclusion
Apparently you overlooked this part of my post:
"...keeping in mind that the instantaneous amplitude of the reconstructed signal will be incorrect because of the jittered timing."
then you repeat the same thing in your own words as if I had not already established that.
So, in effect you actually agree with me on this point but want to make it appear to be some sort of dispute.
You also seem not to be aware that lateral localization differences of as little as 1 degree are distinguishable by the average listener. If you do the math, that is a result of a differential timing change of the incident waveform between the left and right ears of about 2 microseconds. It's simple algebra, and has nothing whatsoever to do with 'hearing 500 khz' or any similar false conclusion
preiter said:
Well I was thinking about this, and if the amplitude of the noise induced by jitter is a simple ratio between the wavelength of the frequency being modulated and the amplitude of the jitter...
If the frequency being modulated is 20k (50 uS) and the amplitude of the jitter is say 50pS for easy calculation.
That's 1 / 1,000,000 which is -120dB. Overkill, but not nearly the wild overkill I thought.
The noise floor of a CD is -96dB which is 1 / 65,536 = 50uS / 65,534 = 762pS.
Does that analysis make any sense?
thoriated
In a typical setup you have a stereo DAC so both signals are advanced/delayed by the same amount of time and the (stereo) image stays the same. Beside that there`s also another mechanisms involved in lateral localization. It`s the diverse distortion of the soundwave by the pinna and the whole head for soundwaves comming from different angles.
I guess what you`re speaking about are tests with "click tones" over headphones. I wasn`t aware that timing differences that small can be percieved. But we must take into account that there is absolutely no problem in keeping jitter far below the mikrosecond range.
preiter
Actually it`s a little worse than that. I`ve found this formula for the amplitude of the modulation products D relative to carrier in an AES paper by Bruno Putzeys and Renaud de Saint Moulin:
D=20*log(tjit*fc*pi*sqrt2) [dBc],
with tjit = RMS value of the jitter signal
and fc = modulated frequency.
So we get -107dBc in your example.
Guido
Why should especially low frequency jitter components be heard down to the ps range? The lower the frequency, the nearer the modulation products are located around the carrier. At some point they can`t be distinguished anymore.
I´m interested in which test equipment you use to determine your jitter values.
In a typical setup you have a stereo DAC so both signals are advanced/delayed by the same amount of time and the (stereo) image stays the same. Beside that there`s also another mechanisms involved in lateral localization. It`s the diverse distortion of the soundwave by the pinna and the whole head for soundwaves comming from different angles.
I guess what you`re speaking about are tests with "click tones" over headphones. I wasn`t aware that timing differences that small can be percieved. But we must take into account that there is absolutely no problem in keeping jitter far below the mikrosecond range.
preiter
Actually it`s a little worse than that. I`ve found this formula for the amplitude of the modulation products D relative to carrier in an AES paper by Bruno Putzeys and Renaud de Saint Moulin:
D=20*log(tjit*fc*pi*sqrt2) [dBc],
with tjit = RMS value of the jitter signal
and fc = modulated frequency.
So we get -107dBc in your example.
Guido
Why should especially low frequency jitter components be heard down to the ps range? The lower the frequency, the nearer the modulation products are located around the carrier. At some point they can`t be distinguished anymore.
I´m interested in which test equipment you use to determine your jitter values.
preiter said:
That's why they invented single bit DACs, and dacs > 16 bit (so you can use just the more linear upper bits), and stacking multiple DACs to average the response.
But that's an issue I can understand and accept
Its just as much of an issue on higher n multibit DACs.
And audiophiles seem to run screaming away from single bit DACs for some odd reason.
Jitter in CD-Players:
I'm confused on the issue of how the master clock stability helps in jitter reduction. What matters is how clean and jitterfree the DAC-clock is. But the DAC-clock is a slave clock that's derived from a data stream and recovered via a PLL. The master clock (i.e the crystal clock) is used at the drive to read out the bits from the CD.
So what is the benefit of replacing the drive clock with a low-jitter aftermarket circuit, when the DAC clock is a still a slave clock derived from the same PLL? Do I miss something here? Guido Tent?
In a proper design, the DAC-clock should of course be the master, and the drive clock should be slaved to it, not vice versa. Only then, would a well designed, low-jitter master clock circuit really make sense. With jitter discussions going on now for many, many years, it is a complete mystery, why cd-players are still designed the other way around. Does anybody know why this is so? Or are any of the newer cd-players doing it right now? If so, which ones?
I'm confused on the issue of how the master clock stability helps in jitter reduction. What matters is how clean and jitterfree the DAC-clock is. But the DAC-clock is a slave clock that's derived from a data stream and recovered via a PLL. The master clock (i.e the crystal clock) is used at the drive to read out the bits from the CD.
So what is the benefit of replacing the drive clock with a low-jitter aftermarket circuit, when the DAC clock is a still a slave clock derived from the same PLL? Do I miss something here? Guido Tent?
In a proper design, the DAC-clock should of course be the master, and the drive clock should be slaved to it, not vice versa. Only then, would a well designed, low-jitter master clock circuit really make sense. With jitter discussions going on now for many, many years, it is a complete mystery, why cd-players are still designed the other way around. Does anybody know why this is so? Or are any of the newer cd-players doing it right now? If so, which ones?
I still have an older Carver single bit DAC vacuum tube output stage player that works fine. However, as big a fan of vacuum tube sound as I am, it is bettered by a number of competent solid state CD players and falls far below the potential CD SQ of all three of the Panasonic and Toshiba DVD players that I have.
My theory is that 95% of price point designed audio equipment is sonic crap, being designed to function to spec as cheaply as possible, but the clocking requirements for DVD generally result in superior CD audio reproduction at a low price point, and some even sound relatively satisfactory to an analog-o-phile when playing certain DVDA and SACDs.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.