The DAC acts as a phase demodulator. An input signal with phase noise results in output signal of amplitude noise. The sensitivity of human hearing varies with frequency. I remember the Fletcher-Munson curves or Robinson-Dadson curves which gives the hearing threshold graph of the average human hearing. Maybe those curves could be compared with the phase noise graphs we have seen from Nagaesan and Guido. We know that in a 16 bit system the LSB is -96 dB, and probably we want the noise below within the human hearing range. -100 dB @ 20 Hz seems a good starting value, and both curves seem OK here. We also know that a periodic signal below the noise floor can be heard, and the human ear is more sensitive around a few kHz. Perhaps the demodulated phase noise can be heard/measured with an all-zero (digital silence) signal fed in the DAC. Also, with a real audio signal, not only noise superposition but also modulation could occur. Any thoughts?
Guido Tent said:
I think this shift range spec is not master clock diviation.
16.9344MHz master clock of CD decorder (like CXD2500BQ ).
That is CD format 44.1kHz sampling frequency, it is not
And consumer digital signal transmission format in IEC-60958 recognize +/-1000ppm for PCM Encoder unit 44.056kHz.
Attachments
Terry Demol said:
Those recording engineers are misguided.
Rubidium master clocks are optimised for good long-term stability as time references. Very short-term stability or phase noise that is in the low end of the audio band is therefore not as critical a design parameter.
Current state of the art crystal oscillators are virtually at the limits that physics dictates for the type or resonator used. SC cuts performing better than AT cuts.
Rubidium master clocks use a CRYSTAL OSCILLATOR stabilised to the rubidium hyperfine transition of 6 834 682 610.904 324 Hz. It is that hyperfine transition frequency accuracy that gives long-term accuracy by the use of synchronising to it.
The short-term accuracy is determined by the CRYSTAL OSCILLATOR used in the rubidium standard.
A brief description on that can be found here http://en.wikipedia.org/wiki/Rubidium_standard
A more detailed description here http://www.thinksrs.com/products/PRS10.htm
So to put it in simple terms the Rubidium standard oscillator is superfluous, redundant - lowers performance and adds unnecessary cost when applied to high quality audio/digital conversion.
Redbook vs. Good Sound
The DIN norm also specified the 0.1% THD is less than audible.
I don't know how old these men were, when they wrote the DIN norm, i am thinking a bunch of old guys with cains, and white hair, saying "turn down that noise, we heard enough". Nothing wrong with that, just better let people decide what they can hear or not.
Frequency adjustment of clocks is a cost addition to the production process, therefore some clock producers skip it, as i agree with guido tent, that it is technically unnecessary. And i agree with guido also that jitter is of course the most interesting feature of the clock, and more than the initial accuracy. However leather interior is also technically unnecessary in your new Porsche, but may add a liitle extra feeling of quality ;-)
However i think 100ppm deviation is in fact audible in an A/B test.
Just like THD , jitter performance itself however seems no exact measure of how the clock will sound. I have personally tested clocks, that were measuring similarly, but made the CD player sound completely different.
As an example, you can take a good clock, and use a capacitor as DC blocker to the existing XTAL circuit in the CD player. (Which will btw many times give a more relaxed sound).
If you change that cap from say polypropylene to mica, you can hear a world of difference in the sound profile. However i bet you will not measure any difference in jitter.
Anyone can try it out for himself, please post results here
The point is, that high sound quality has to be designed into the clock circuit, and that is where it gets complicated. When looking at available designs, it's obvious that some of these people don't understand this factor at all. Not speaking about my esteemed clock producing colleagues, but as an example recently i have seen a commercial DAC board using PCM1704-K, and for analog decoupling, the board was plastered with cheap ceramic SMD caps. Why waste money on nice 1704's if you gonna ruin the sound with Z5Y's ??
The DIN norm also specified the 0.1% THD is less than audible.
I don't know how old these men were, when they wrote the DIN norm, i am thinking a bunch of old guys with cains, and white hair, saying "turn down that noise, we heard enough". Nothing wrong with that, just better let people decide what they can hear or not.
Frequency adjustment of clocks is a cost addition to the production process, therefore some clock producers skip it, as i agree with guido tent, that it is technically unnecessary. And i agree with guido also that jitter is of course the most interesting feature of the clock, and more than the initial accuracy. However leather interior is also technically unnecessary in your new Porsche, but may add a liitle extra feeling of quality ;-)
However i think 100ppm deviation is in fact audible in an A/B test.
Just like THD , jitter performance itself however seems no exact measure of how the clock will sound. I have personally tested clocks, that were measuring similarly, but made the CD player sound completely different.
As an example, you can take a good clock, and use a capacitor as DC blocker to the existing XTAL circuit in the CD player. (Which will btw many times give a more relaxed sound).
If you change that cap from say polypropylene to mica, you can hear a world of difference in the sound profile. However i bet you will not measure any difference in jitter.
Anyone can try it out for himself, please post results here
The point is, that high sound quality has to be designed into the clock circuit, and that is where it gets complicated. When looking at available designs, it's obvious that some of these people don't understand this factor at all. Not speaking about my esteemed clock producing colleagues, but as an example recently i have seen a commercial DAC board using PCM1704-K, and for analog decoupling, the board was plastered with cheap ceramic SMD caps. Why waste money on nice 1704's if you gonna ruin the sound with Z5Y's ??
LM329
It seems from your post that you know something about low noise design. We constantly search for new improvements to our low noise designs, and LM329 still seems to be - by far - the option with least reference noise. At least if it has to be available as a component, operating at above 5 deg K. Compared with several new fancy 'ultra low noise references' from leading chip manufacturers, the old LM329 has far less noise. We not only base this on datasheets, but also we use a noise analyser, to check out the final result of a voltage regulator integrated in a clock or other circuit. Also here the LM329 prevails every time.
If you know of anything better, please let us all know about it.
It seems from your post that you know something about low noise design. We constantly search for new improvements to our low noise designs, and LM329 still seems to be - by far - the option with least reference noise. At least if it has to be available as a component, operating at above 5 deg K
. Compared with several new fancy 'ultra low noise references' from leading chip manufacturers, the old LM329 has far less noise. We not only base this on datasheets, but also we use a noise analyser, to check out the final result of a voltage regulator integrated in a clock or other circuit. Also here the LM329 prevails every time.If you know of anything better, please let us all know about it.
Re: LM329
LM329???
A string of green LEDs has an order of magnitude less noise. See Christer's noise measurement document on this forum.
See also:
http://www.diyhifi.org/forums/viewtopic.php?f=36&t=1721
Lars Clausen said:
It seems from your post that you know something about low noise design. We constantly search for new improvements to our low noise designs, and LM329 still seems to be - by far - the option with least reference noise. At least if it has to be available as a component, operating at above 5 deg K
If you know of anything better, please let us all know about it.
LM329???
A string of green LEDs has an order of magnitude less noise. See Christer's noise measurement document on this forum.
See also:
http://www.diyhifi.org/forums/viewtopic.php?f=36&t=1721
Lars Clausen said:
Did you read that link ??
They are discussing which Jocko Homo clock to put on the side of a YM3623. Which should be rather amusing for those who know that chip ;-)
Lars,
I think he was referring to this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=35821&highlight=
cheers
Terry
CraigBuckingham said:
Misguided? They are merely reporting and acting on what they
hear. When a number of engineers report similar findings then it is
worth investigating what mechanisms are actually occuring.
I'm not so much interested in the absolute performance of
Rubidium, more the ADC's behaviour when being externally clocked
by one.
I suggest you go to:
http://www.grimmaudio.com/whitepapers.htm
Read whitepaper titled 'PLL and Clocking'. In particular point 7.
As shown in the graph, the only region an externally clocked ADC
can benefit is below the internal PLL's cutoff point, ie; low
frequencies. (Good paper Grimmaudio folks)
Others are getting benefits from ULN at LF PSU's on their oscillators,
so the finger points at one region - close in (to carrier) phase noise.
Dig?
T
Lars Clausen said:
Did you read that link ??
They are discussing which Jocko Homo clock to put on the side of a YM3623. Which should be rather amusing for those who know that chip ;-)
Jocko is using a very low noise supply in his clock using a few green LEDs as reference.
http://www.diyhifi.org/forums/viewtopic.php?f=36&t=1570
Thanks Terry, yes i am aware of Christer's commendable work, and i love the very practical approach to the search for low noise. We use a similar approach, just using the B&O NM1, which is an instrument similar to Christers LT1115 + soundcard, more or less. We can also add filters, and do direct readings.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.