Quote:
Originally Posted by yourmando /img/forum/go_quote.gif
We can see total jitter immunity to the limits of the Audio Precision System 2 Cascade test unit, up to 12.75 UI (2075 ns) of jitter.
|
I find it strange that the highest jitter frequency tested was 9 kHz, when jitter effects increase with frequency and is very significant even above 20 kHz. Take a look at this article by Lavry:
http://www.lavryengineering.com/white_papers/jitter.pdf
Of course, the nature of the effects going through an ASRC would be different, but is no reason not to consider such test cases.
I'm also interested in the level of jitter after the ASRC, which is mostly dependent on the local clock. Dunn's paper from the 93rd AES convention derives a 20 ps maximum allowable:
http://www.nanophon.com/audio/jitter92.pdf
Quote:
As for improving the quality of the conversion |
What do you mean? No current DAC chip manages actual 24 bits resolution; about 21 is the highest I've seen. More can be done with multiple DAC chips and scrambling. If you mean the quality of the IV conversion, I suggest you check out Hawksford's current-steering transimpedance amplifiers paper. In simulation, Fig.4-4 is the best performing IV I've seen, and from those that have built it I've not seen anything to make me doubt those results. Of course, tuning a discrete design where parts must be matched may not be economic for certain price ranges.
Finally, you quoted me mentioning the thermal memory distortion, yet you did not address that point. This issue is most significant with chip amps (as even a skeptic like D.Self will point out), which is another reason to prefer discrete designs. This doesn't really show up in THD measurements. It would be interesting if someone would perform the memory measurement described in the paper I quoted before on various SS stages, including this one.
Quote:
(Rant: often, only a few specs are provided, sometimes without stating measurement conditions, and it seems that some manufacturers simply list the specs of the dac chip they use, or misleading specs such as SNR that includes the output mute circuit.) |
That is unfortunately the case. Summary metrics like THD/IMD have limited utility as they're not perceptually weighted. Some distortions are inaudible to a couple of percent (i.e. low order even harmonics), others in the parts per million (B/AB class crossover distortion). Some have audibility established but the ear's exact sensitivity to them not determined (such as the thermal memory issue I mentioned above).
Quote:
In terms of measurable accuracy, the DAC1 is up there with any dac at any price |
I've mentioned some measurements that I don't see there. I'm not singling out the DAC1, I'm talking about DACs in general, and I'm glad the DAC1 at least shows the amount of detailed measurements that it does; hopefully it will set a trend. But I would not say they are sufficient to justify your claim.
Quote:
and that we are really limited by the best A/D converter |
As I pointed out above, we are not. You can use a bunch of them to get an improvement.
Quote:
we are probably already well past the audibility threshold for something at this stage in audio reproduction signal path |
That's an unwarranted statement. The ITU recommends double-blind triple-stimulus with hidden reference testing; I'll believe it when I see the results.
There are many other things wrong with sound reproduction besides the electronics, such as that all speakers and headphones have tremendous distortion (well, with the exception of Alan Hill's helium plasma device in the 1 kHz and up range). Even worse is the recording and playback geometry. Several things like binaural, Ambisonics, etc., but none of them work very well. The single thing that will make most of an improvement in audio is having one's personal HRTF. It's already well established that HRTF differences between individuals are very significant. I find it quite amusing when I read reviews about how some amp makes the imaging clearer and such nonsense, when it has nothing to do with that. HRTF measurement needs an anechoic chamber and expensive equipment, but in recent years the alternative of laser scanning and finite boundary method simulation allows one to compute the HRTF. I wonder how long until hi-end audio stores start carrying laser scanners so one can just go and get a scan, then they compute the HRTF, for a fee, and you can use it in your favorite DSP convolver to make those binaural recordings made with the pinna-less dummy head
really give the right positional sound. Anyway... getting too offtopic here with my rant
![Stick Out Tongue :p :p](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)