Soundcard calibration look ok?

[atharva]

Hi,

I did my first sound card calibration for the UR22C audio interface using a dual-channel setup with the output looped back from the amp to the input.

My settings are..
Sample rate: 96K
Buffer : 128K (in preferences)
Sweep Length : 512K (in measurement window)

1. Are those wiggles ok? The REW documentation says to expect "variations much less than 1dB". Each time I run this I get slightly different results. The worst I got was +1.8dB -1.9dB.
2. Are the settings fine? Should I try different settings?
3. Is it ok to have the timing reference plugged in while doing this or should that be disconnected?

Thanks.

 

[John Mulcahy]

Shouldn't usually be any need to increase the REW buffer settings. The measurement is OK, though a bit noisy. Bear in mind that graph only spans +/- 3 dB, so it isn't as bad as it looks. The noise appears mostly mains-related, odd harmonics of 50 Hz. Moving the cables away from anything carrying mains might help, but a little smoothing (1/48 octave) would also pretty much remove the noise. The timing ref is ignored when making a soundcard cal measurement, though the connection itself might be contributing to the noise.

 

[atharva]

John, you're observation is spot on. I should've caugtht that! Moving the cable away from the power cable improved it a lot. I reduced the buffer size to 64K. Thought I'll post some screenshots in case it helps others. Also see the bottom screenshot where disconnecting the timing reference made it even better.

Blue : noisy with 50Hz odd harmonics
Red : moved loopback cable away from power cables. 3dB points almost same but substantially reduced noise. The timing reference cable is still connected.

Here is a comparison with the timing reference cable removed. What this means is that we just have loopback from the amp back to the input where the mic goes in.

Red : same as previous with timing reference connected.
Orange: with timing reference cable removed. Even those small ripples are gone! So the connection itself does seem to impact. I've scaled the response by -0.1dB to show the difference else it's difficult to see the difference.

As John said, it's all acceptable within +- 1dB 20-20K. So this is probably just informative than critical.