28 May
2010
28 May
'10
10:07 p.m.
On Fri, May 28, 2010 at 07:22:24PM +0200, Robin Gareus wrote:
OT: I would not trust any consumer-grade
audio-interface for signal
differences larger than 80dB anyway. I'm amazed that card A shows such a
low noise-floor in the loop-back test.
Not OT at all. Don't mistake the -140 dB noise line on that plot
for the S/N ratio... changing the analyser bandwidth (i.e. the FFT
length) in JAAA you can put this line almost everywhere you want it.
The more frequency bins you have, the less noise there is in each of
them, and that is what the plot shows.
The really interesting value is the noise density shown in the upper
left corner (for the marker at 1100 Hz): noise density is -137.9 dB/Hz
Add 43 dBHz (for 20 kHz), and the resulting S/N ratio is -94.9 dB.
Which is 16-bit performance, even if the card is 24-bit.
Ciao,
--
FA
O tu, che porte, correndo si ?
E guerra e morte !