US-16x08 high noise floor on all input channels

[Mihael Bele]

Hi,
I just bought US-16x08 and wanted to do some test 16 channel recording to see how much better signal to noise ratio is compared to my 20 year old PCI soundcards (Sek'd Sienas). After arming tracks for recording with nothing connected to interface inputs, I noticed that monitoring peakmeters are showing very high noise flor. All mic preamp gain pots are on minimum, inputs 9-10 are in line mode, other line-ins are +4dBu. Signal to noise ratio varies, depending on the channel, from -62db to -86db (my PCI cards connected to Spirit live mixer inserts are steady from -78db to -80db). I recorded a short sample of silence in 16bit/44.1kHz and normalized each track - attached screenshot shows how much gain it took to normalize the recording in each channel. Frequency analyser shows peak in very low frequencies, and that may be the culprit (I attached screenshot of one track soloed). I did some more tests with microphone connected (preamp noise swamps mic noise), phantom power on/off, DSP/mixer disabled, 24bit/96kHz resolution, and there's no significant change (except maybe that in 24/96 there are resonances at (4,)8,16,24,32 & 40kHz on channels 13-16 ~12db above noise floor).
Is all this normal for this interface?

 

[Mihael Bele]

Support said that interface is within specs, and that what I was doing was not a proper way to mesure SNR (I agree).

I think that there's DC offset present on every channel except #8 to some degree. Using lowpass filter on card's built-in DSP does not solve the problem since the filter is not steep enough, but maybe it can be fixed in postproduction.
It would not be a problem if resonances end up below microphone noise floor, but I haven't tested that.
Loopback test shows that there's a pre-delay on recorded signal (driver issue I guess) that changes and varies if the buffer size is changed (up to a few ms), but that can also be solved in the post if it's really required (measuring loopback delay prior to recording and not changing buffer size during session). Also, that seems to be happening with other USB interfaces too.

360EUR is great price for this interface features, but for me, it would be too much hassle to work with.