I respect the findings of BB and Lecroy, but I have learned from practical experience that using these relaxed jitter specs prevent optimal performance for sure. One could argue about the validity of the formula used by Kusunoki, but practical results cannot be ignored.
I wouldn't be too sure of that 🙂
-ecdesigns- said:
They can and often are when they are subjective.
You are free to be as unsure as you like. I'm certain.
Personally I tend to believe the measurements more than human ears when it is cutting it so close of the thresholds. Psychological patterns in the brain can tell us something else, something that we WANT to hear. If it isn't a true double blind test I don't even bother to pay attention to the "findings". And even a true double blind doesn't tell the whole story since it takes the whole audio chain to do the testing, and some masking by other components it will be present always.
I agree, the low freq. components of the jitter are often ignored but that doesn't change the minimum value of the necessary jitter just to compensate for bad measurements in the specs.
I agree, the low freq. components of the jitter are often ignored but that doesn't change the minimum value of the necessary jitter just to compensate for bad measurements in the specs.
SoNic_real_one said:
Speaking of jitter measurements, to reproduce satisfactory a 24/192 track the jitter should be under 10ps, according to a Stereophile article.
Since the best performance of a 24 bit DAC is around 20-21 bit that jitter requirement it is smaller than necessary. If it is correct, that's it.
I had read in CS documentation that switched capacitor DAC's are more immune to signal jitter than ladder type. So even that number could be debatable.
I had read in CS documentation that switched capacitor DAC's are more immune to signal jitter than ladder type. So even that number could be debatable.
SoNic_real_one said:
Well, I do not think that you can get 20-21 bits out the Dac in a real world scenario. Perhaps, in the lab you might, but once you factor the power supply noise plus other sources of radiation, you will be hard pressed to get even 18 bits out. Besides have you tried to even measure 20 bits?
Anyway, one other thing that puzzles me, - although jitter is a stochastic process, still is it not subject to Fourier analysis? I am sure that it is and if so, will the THD+N test not show if jitter is in fact a problem in a given circuit? It seems to me that as long as the THD+N looks good then jitter is not an issue.
Finally, how do you exactly measure jitter in sub 100 ps range? I do not think that direct measurement is even possible with a commonly availably equipment or it can be believed.
Well, the guys from BB give figures like 127dB for a 2V output 127dB, 129dB on 4V output, 132dB on a 9V mono output...
For 20-bit the dynamic range it is 120 dB. For 24 bit it is 144 dB. 21 bit is 126dB... and so on.
They are using NE5534 for I/V and LT1028 for the differential/filter.
I think it is possible practically to go at 20 bit...
Hell, even Creative X-Fi Elite has supposely 116dB. Don't know if it is true 🙂
Jitter can be expressed in the THD+N but... one needs to define the caractesistics of jitter at the input (level, bandwidth) to be a semnificativ measurment.
PS: At the room temperature a 243 ohm resistor has a termal noise equivalent of -117dB from a 2V output. So, the I/V stage MUST have opamps that make a virtual zero impedance for the current input. A resistor converter stage will add the resistor noise to the output.
For 20-bit the dynamic range it is 120 dB. For 24 bit it is 144 dB. 21 bit is 126dB... and so on.
They are using NE5534 for I/V and LT1028 for the differential/filter.
I think it is possible practically to go at 20 bit...
Hell, even Creative X-Fi Elite has supposely 116dB. Don't know if it is true 🙂
Jitter can be expressed in the THD+N but... one needs to define the caractesistics of jitter at the input (level, bandwidth) to be a semnificativ measurment.
PS: At the room temperature a 243 ohm resistor has a termal noise equivalent of -117dB from a 2V output. So, the I/V stage MUST have opamps that make a virtual zero impedance for the current input. A resistor converter stage will add the resistor noise to the output.
Sorry if I am being naive as a newbie, but why is the bit depth being counted in the jitter calculations? I thought jitter was the variation in timing between samples - a sample is 16, 24, 32 bits, or whatever. The variation between bits within a sample is irrelevant, since the DAC reads (ie buffers) the entire sample before latching.
So the degree, and "frequency" of the sample-clock timing irregularity is what matters. I guess if you ran a 1khz tone through it, you'd see some jitter tones, or just noise, overlaid on the plot - ie even with a very small variation in timing, if it's regular, it'll generate a tone, perhaps an audible one, whereas even a much larger variation in timing, if it were random, would just look like white noise.
Or am I being a simpleton?
So the degree, and "frequency" of the sample-clock timing irregularity is what matters. I guess if you ran a 1khz tone through it, you'd see some jitter tones, or just noise, overlaid on the plot - ie even with a very small variation in timing, if it's regular, it'll generate a tone, perhaps an audible one, whereas even a much larger variation in timing, if it were random, would just look like white noise.
Or am I being a simpleton?
- Status
- Not open for further replies.