[time-nuts] from Sputnik to CD
Henk ten Pierick
henk at deriesp.demon.nl
Sat Oct 6 07:27:17 EDT 2007
On Oct 6, 2007, at 3:01, Magnus Danielson wrote:
>> If the sample rate is 44.1 kHz then the adev below 1 s, even
>> down to 10 us, or jitter, or the phase noise above 1 kHz are
>> they key parameters; not long-term drift. In this case I'd take
>> a nice 10811A over a typical Rb any day. Does anyone know
>> for sure, or am I off base here?
> No, I meant to say the same. The phase-noise in the 10-1000 Hz
> region is
> surely what will mostly affect the sound, so we are looking at side-
> phase-noise rather than Allan dev plots. However, slow modulations
> even up to
> a minute is known to be perceptible, but the tolerance levels is
> not flat.
I think this is correct. Is there a figure in dBc/Hz at say 10, 100
or 100 Hz.
When measured with a Wavecrest 2075 there is a clear difference in
an oscillator with -100 dBc and -96dBc both at 10Hz. The listening
difference is in the sound stage accuracy and
acoustics of the concert hall.
I do not have (access to) instruments that measure phase noise
directly to say -120 dBc at 10Hz.
> If anyone is interested in learning about serious reasoning behind
> high dynamic digital audio, look up Julian Dunn, who has written a
> few very
> good papers. He also did work for the AES/EBU jitter limits and
> also wrote some
> stuff for Audio Precision. Great stuff and a sensible reasoning on
Agreed, but Julian does not give phase noise figures in dBc/Hz figures.
> It should be noted that for lower modulation frequencies, the side-
> band will
> be so near that the ears masking-effect will cancel it out unless
> it is
> strong, so tolerance levels rises for lower frequencies.
I doubt if this is true. I have never found an agreement to known
psycho-acoustic masking figures and
own jitter measurements. It is amazing what can be heard with jitter
values of pico-seconds.
There must be a relation between these low values and something else,
but I am not aware of it.
The phase noise statement of Magnus can be correct, I think it is,
but the side bands resulting from
these values are within the masking curves.
Experiments with a modulated clock show that large modulation can not
be heard. However, noise modulation much smaller can destroy the
sound stage. I am very interested in a good explanation of this.
I you do not believe me you are welcome for a demonstration.
More information about the time-nuts