There is a guy on DIY audio stating that jitter in SPDIF digital audio causes distortions. OK. He also claims to have jitter in the 100 fs area. Is he blowing smoke about the jitter or can he really achieve sub pico second jitter? It seems to me that jitter in that system is similar to Time Base errors in video systems. In analog video 1-2 nS was achieved in both analog and digital Time Base Correction (TBC) systems but I'm having problems getting my head around 100 fS and - is it REALLY audible?
100 fs RMS cycle-cycle jitter is possible from a very good, pretty expensive crystal oscillator. Of course it's not audible.
Any signal processing, including quantization, causes distortion. So do microphones, speakers, and ears. People who worry about femtoseconds are lunatics.
Sound doesn't travel very far in 100 fs. The performer's guitar, the recording microphone, the speaker, the walls of your listening room, your eardrums... all are moving around constantly much, much more than the distance sound travels in 100 fs. One *microinch* of vibration of any of them corresponds to 30 ns of jitter.
Of course. Though sadly I have the same problem that others have mentioned in this group; relieving fools of their cash is all very well, but I need to be able to look myself in the mirror :(
The jitter in the clock does not really add distortion. It adds a signal dependent noise. The noise in the data depends on the clock jitter and the slew rate of the input.
The jitter than really matters is the portion that makes things that can be heard in the output. This biases it towards the near carrier noise where the noise is usually the worst. I know from experience that in the near carrier case, an OCXO can be over 180dB down.
A coworker has been concerned about jitter-induced degradation in a data acquisition system and has been trying to find a simple equation like this one. Can you give me a reference I can pass along to him? He's an analog-circuits guy like myself, and not terribly familiar with the digital literature.
This seemed to be a popular subject for more than a decade ago, someone are still trying to make money on this :-).
Apparently some early external DACs took the self clocking Manchester code SPDIF signal, passed it through a comparator and use the transitions to kick a multivibrator, which was divided down to directly drive the DAC sample clock.
Of course, the previous data contents in not so linear phase transfer path or some added hum in coaxial versions shifted the threshold level of the digital data, causing some timing jitter. Even after dividing down this to the audio sampling rate, the sample pulse timing contains some jitter.
The sampling pulse jitter will cause some amplitude distortion in the presence of high amplitude high voltage signals (high slave rate). For audio DACs the largest slave rate is during the zero crossing of a full amplitude 20 kHz sine wave.
From the slave rate it can be calculated, how much the sample point must be off to produce say 1 LSB error. This still might be audible, since low frequency (< 1kHz) tones can be detected down to -110m ..
-120 dB levels in fade to noise tests due to dithering, even if the 16 bit CD system can go only down to -96 dB for any frequency.
Thus -120 dB or 1/16 LSB (of original 16 bit signal) should be quite safe, especially when 20 kHz 0 dB signals are _quite_ rare in any actual recordings :-), reducing the maximum realistic slew rate with an order of magnitude.
In any modern SPDIF receivers, some kind of PLL is used and the loop filter removes any SPDIF receiver comparator threshold related jitter, so the external DAC jitter should not have been an issue for a very long time.
I can't recall any book from the top of my head; however it is not difficult to do the numbers. Jitter is a small angle phase modulation of the signal; therefore the effect of jitter is essentially the same as the effect of the AM. That creates noise sidebands on both sides. The rms amplitude of the sidebands is proportional to the signal frequency. Now integrate the noise through the bandwidth of interest and you have the answer.
Vladimir Vassilevsky DSP and Mixed Signal Design Consultant
I found a nice treatment for this in Microwaves & RF, no less. Here is the reference:
B. Drakhlis, Calculate oscillator jitter by using phase noise analysis ? Part 1, Microwaves & RF, Vol. 50, No. 1, January 2001, pp. 82?90, 157.
Presumably the article should be floating on the web somewhere. Failing that, I summarily reproduced it in a course on AD conversion I gave some time ago. The course text is at:
Yes, good crystal oscillators can have sub 1ps RMS jitter. e.g.
formatting link
Nope. Anyone that thinks so is delusional.
I had some guy who makes and sells low jitter SPDIF audiophool RCA cables take me to task over my Audiophoolery Blog, claiming he knew it all, and has measurements to prove it etc. And he tried to impress me by saying he's good buddies with Bob Pease, so I'd better shut my mouth, yarda yarda... Hilarious. I'll have to dig up the link...
Dave.
--
================================================
Check out my Electronics Engineering Video Blog & Podcast:
http://www.eevblog.com
Slowly but surely i am becoming convinced that it is my civic and moral duty to remove cash from such pinheads before they do something really dangerous with it, like give it to a political campaign. =20
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.