Please, Syn08, your criticisms can be more severe than that! How can I answer you with 'sound and fury' if you appear to meet me 1/2 way?
By the way, I am having lunch with Jack Bybee this day. He just got back from Macau, and other points in China. I hope he brings the Bentley, I get tired of the Mercedes.
By the way, I am having lunch with Jack Bybee this day. He just got back from Macau, and other points in China. I hope he brings the Bentley, I get tired of the Mercedes.
That was not what I was asking, though I was actually looking for more degeneration in this application to get better open-loop linearity. For the different Idss there would be different values. The point was the circuit I usually see (right) is difficult to trim for seconds if the gm's on the N's and P's are not the same (inevitable).
I'll be in China in November with Jimmy Carter.
I'll be in China in November with Jimmy Carter.
@ syn08,
jitter is a perfect example for the mechanism PMA mentioned; if jitter spectra exist well above the audio band, it depends on DAC structure and signal content within the audio band what (and if) additional new signal content will be produced.
The new content could be inharmonic, but is otherwise strongly correlated to the signal.
Assumed that it is not random jitter, which would only raise the noise floor as you noted.
jitter is a perfect example for the mechanism PMA mentioned; if jitter spectra exist well above the audio band, it depends on DAC structure and signal content within the audio band what (and if) additional new signal content will be produced.
The new content could be inharmonic, but is otherwise strongly correlated to the signal.
Assumed that it is not random jitter, which would only raise the noise floor as you noted.
I've lurked some of the digital threads and still wonder why there isn't more attention to known (not even that sophisticated) techniques to reduce jitter to meaningless levels. Then again there's the 1-1 != 0 stuff🙁
Jakob2 said:
Ok, so we brought the three examples to two wrong and one potentially good (the non-random jitter associated with digital audio).
Jitter issues are well beyond my expertise, but I know that much:
a) It belongs to the digital world. Deterministic jitter may originate in many parts of the digital system (duty-cycle distortion, intersymbol interference, word-synchronized distortion, etc...) and it's not necessary signal correlated. But as such, I fail to see the relevance in the context of amplifiers and the analog world.
b) In particular for audio, the correlation between low levels of deterministic jitter and subjective sound quality is worse than the correlation between THD and subjective sound quality. I have not seen any systematic study addressing this, but this certainly doesn't mean it doesn't exist.
Finally, as Scott mentioned, eliminating deterministic jitter is relatively easy. However I'm pretty sure that, following the THD model, as soon as the deterministic jitter would be eliminated, some voices from the subjective camp will claim that there isn't actually any correlation between jitter and sound quality. Some high deterministic jitter CD players will be created and sold for $150,000. Of course, they'll all have unbleached cotton silver wiring and an integrated Bybee device.
That's very interesting, thanks for the input!
Again I would be interested what 'very little' means and what the AP1 has to say about that 😀
Have fun, Hannes
"I believe a major source of these is due to widespread use of op-amps under heavy negative feedback."
Bill Whitlock
President & Chief Engineer
Jensen Transformers, Inc.
(and recent AES Fellow)
Bill Whitlock
President & Chief Engineer
Jensen Transformers, Inc.
(and recent AES Fellow)
scott wurcer said:
Yes, it´s surprising, considered that measurement of jitter related distorsion is quite difficult to measure with usual diy equipment.
@ syn08,
in the digital domain jitter doesn´t harm, if it doesn´t cause data errors, therefore tolerable jitter levels in pure digital systems are quite generous.
Otoh, requirements on jitter during digital-analog conversions are tough as even quite low levels of jitter already worsen a (theoretical) 16-bit resolution system.
Jitter itself might be correlated to the signal, but quite often is not, but the effect of deterministic noncorrelated jitter is correlated to the signal as it depends on the higher frequency content inside the audio band.
And it is possible to create IM-Products that fall in the most sensitive region without any helping masking tone around.
Studies are rare; i remember just Benjamin/Gannon and Ashihara on this topic- both not really helpful for various reasons.
in the digital domain jitter doesn´t harm, if it doesn´t cause data errors, therefore tolerable jitter levels in pure digital systems are quite generous.
Otoh, requirements on jitter during digital-analog conversions are tough as even quite low levels of jitter already worsen a (theoretical) 16-bit resolution system.
Jitter itself might be correlated to the signal, but quite often is not, but the effect of deterministic noncorrelated jitter is correlated to the signal as it depends on the higher frequency content inside the audio band.
And it is possible to create IM-Products that fall in the most sensitive region without any helping masking tone around.
Studies are rare; i remember just Benjamin/Gannon and Ashihara on this topic- both not really helpful for various reasons.
scott wurcer said:"I believe a major source of these is due to widespread use of op-amps under heavy negative feedback."
Bill Whitlock
President & Chief Engineer
Jensen Transformers, Inc.
(and recent AES Fellow)
Nothing to worry about as he writes next:
"It is one very good reason for passive bandlimiting at the input of every active amplifier stage. Otherwise, the effect is cumulative - and demonstrably ugly sounding. "
🙂
Jakob2 said:
Follow this gedankenexperiment:
Here's a DAC with adjustable jitter on the digital source side; connect a spectrum analyzer on the analog side (say, with whatever small RBW you like) and begin adjusting the jitter at the input.
If the jitter variations are random, we already agreed that only the noise floor seen on the analyzer will be affected.
Are you telling me that any signal correlated jitter will not affect the measured spectra?
If it does, then signal correlated jitter has an harmonic impact, which may or may not impact a global variable as THD but certainly contradicts PMA's statement "non-harmonic, but signal correlated distortion".
If it does not affect the spectra, then you tell me what is the analog measurement that would reveal a known/measurable signal correlated jitter variation.
BTW, IM products are definitely harmonic in nature, so I still fail to see your point regarding this mysterious "non-harmonic, but signal correlated distortion".
Jakob2 said:
We have measured content dependent clock skew even inside a DAC (serial to parallel registers). One must be careful, the ability to measure time is several orders of magnitude better than the best THD systems i.e. you will always find jitter.
Jakob2 said:
So you're saying that jitter before the DAC makes has no effect on distortion? If that's true then digital cables should have no influence on sound. Any cable would work?
I tried different wires with my DACs, and found some made the sound softer and others would make it intolerably bright sounding. I also had a digital interface processor that made every DAC I hooked up to it really irritating. So, I got rid of it. Maybe it was introducing some noise, and not jitter.
Isn't the timing information contained on the disc. Unless of course the data is buffered and reclocked.
Do most DACs buffer and reclock the digital signal? I don't know much about this topic, just curious.
Good point, Scott. I might try it, myself, BUT not for the input stage, since that is usually OK with Toshiba devices, as the Gm of the complementary devices is usually very close, but for the second stage, which is more problematic in this regard.
syn08 said:
Isn't that what the "J-Test" stuff is all about? The jitter can be nonsynchronous with the audio content but synchronized the the sample rate or even something else in the system. In PC audio systems and many earlier CD players there were several unrelated clocks running and trying to work together.
A spectrum analyzer will only look at content present long enough for it to be seen by the analyzer's technology. Its conceivable that transients could pass through that the analyzer doesn't see. This trick is used a lot to get FCC approval of products that should never have been shipped. An experiment I should try is to see if an audible difference is present on a PC based system with the spread spectrum clock on vs. off. And see if the J-Test can show it.
Related, I use the spectral contamination test often for transducers. Its very sensitive and shows mechanical issues very well.
syn08 said:
Hey i do like gedankenexperiments. 🙂
As 1audio already pointed out, jitter has not to be signal correlated as the sources for jitter are manigfaltig and we are thinking about the more general cases.
So let´s assume that a deterministic jitter exists in a device well above the audio band and the question would be what effect this jitter could have.
Of course any effect will depend first on the jitter amplitude and second on the signal content in the audio band (as fast rate changing signals are more affected than low frequency signals).
It is possible that for example an audio signal of 19 kHz produces a distorsion product of 3 kHz if an "appropriate" deterministic jitter (above the audio band) is present.
This 3 kHz product is correlated to the signal (as it wouldn´t be produced if no 19 kHz signal were existing), but is otherwise nonharmonic as no masking fundamental (for example) is present.
- Status
- Not open for further replies.