JJAZ skrev:Karsten@StudioSound skrev:Jeg fandt lige denne interessante beskrivelse på Computer Audio Asylum:Super interessant læsning.. og det forklarer jo ret godt hvorfor USB indtil videre er temmelig uegnet til lyd i high-end kvalitet.
Det kan også meget vel forklare hvorfor du oplever stor forskel på forskellige computere hvis deres 1mS pakke-frekvens får lov at bestemme noget som helst. Samtidig vil det også betyde at selvom én stationær lyder godt så er det ikke sikkert at en anden stationær (selvom den ligner en tro kopi af den første) vil lyde lige så godt. Det er ikke lige til den slags transmission at man bruger krystaller med mindst mulig jitter.
Så er det igen at det er lidt svært at lave sort/hvid betragtninger på forskellige grundprincipper. Som det også forklares i nedenstående så er ppm en ret intetsigende værdi for krystallers evne til at holde jitter nede. Faktisk er det "phase noise" der har en egentlig betydning her.
Endnu et citat fra John Swenson (http://www.audioasylum.com/cgi/m.pl?forum=pcaudio&n=23893&highlight=John+Swenson&r=&session):
"I've been designing and building my own DACs and interfaces to computers for a while now, listening and measuring them.
First a few general words on jitter. Using a single "number" for jitter does not correlate well with sound quality. The spectrum of the jitter seems to correlate much better than the single number. Of course its much harder to do a marketing campaign on a series of spectrum plots than a single number, people like single numbers, they can compare them easier!
I've done an experiment where I built two identical DACs, one with USB input (optimized 2706 circuit) and the other with a 8412 receiver. The jitter "number" from the S/PDIF receiver was much lower than the USB, but the USB sounded considerably better. Looking at the spectrums the USB one looks much noiser, the S/PDIF is quite clean in comparison, BUT the S/PDIF has some high level spikes at certain frequencies, the level of these spikes is much greater than any of the "noise" on the USB clock.
I've spent a lot of time looking at clock spectrums and trying to correlate them to sound quality, I've found that the broad
noisy ones generally sound better than ones with sharp spikes.
The marketing material mentioned below about the MSB clock makes me laugh every time I read it. The "ppm" of the tempco of the oscillator has nothing to do with its jitter, again its a low number which marketing departments like. In practive many temperature compensated oscillators actually have higher jitter than the simple ones without the compensation. Why don't they publish the spec for the jitter of their clock rather than its tempco?"
First a few general words on jitter. Using a single "number" for jitter does not correlate well with sound quality. The spectrum of the jitter seems to correlate much better than the single number. Of course its much harder to do a marketing campaign on a series of spectrum plots than a single number, people like single numbers, they can compare them easier!
I've done an experiment where I built two identical DACs, one with USB input (optimized 2706 circuit) and the other with a 8412 receiver. The jitter "number" from the S/PDIF receiver was much lower than the USB, but the USB sounded considerably better. Looking at the spectrums the USB one looks much noiser, the S/PDIF is quite clean in comparison, BUT the S/PDIF has some high level spikes at certain frequencies, the level of these spikes is much greater than any of the "noise" on the USB clock.
I've spent a lot of time looking at clock spectrums and trying to correlate them to sound quality, I've found that the broad
noisy ones generally sound better than ones with sharp spikes.
The marketing material mentioned below about the MSB clock makes me laugh every time I read it. The "ppm" of the tempco of the oscillator has nothing to do with its jitter, again its a low number which marketing departments like. In practive many temperature compensated oscillators actually have higher jitter than the simple ones without the compensation. Why don't they publish the spec for the jitter of their clock rather than its tempco?"
Hilsner,
Karsten
