This may be a bit long. I'm new to using REW, my compliments to JohnM for the extent and quality of REW.
I do have a curious issue in testing sound cards on loopback. I've spent a fair amount of time the last couple of weeks in setting up a hardware test rig and learning the ins/outs of REW and other software. So far REW looks to be my go-to for that. However, recent tests provided conflicting results. I'm fairly sure I'm using REW appropriately at this point. I'm testing both a new Focusrite Scarlett Gen 4 and an old M-Audio Delta 410 (1010LT in the queue). This is on Windows 10 Pro on an older PC. Last evening I ran a series of tests for distortion on the 410. I'm trying to characterize sound card(s) and an amp for raw driver distortion to ensure that those aren't polluting the driver results. I've found considerable issues with cabling for one, that's a separate issue.
The issue is that in the series of tests last evening I went from 0dBFS input and dropped the level sequentially. The SNR was far lower what I expected for this card. As I dropped the loopback output setting I noticed that the background noise dropped as well, but the SNR remained roughly the same The individual harmonic levels appeared correct and changed as I expected. At the lowest output setting the noise floor was much lower. Microsoft forced an update/reboot of that PC overnight. When I resumed testing today the noise floor results changed. It was in a more expected range and remained roughly constant with output voltage setting changes as one would expect. I've noticed in the past (when learning REW) that overdriving an input that stopped REW seemed to result in further test results questionable. I would stop/restart REW. BTW, I had a similar issue with Arta in trying it. I'll attach window captures of REW.
I think I've seen similar with the Scarlett, both in this old PC and a laptop, but that testing has been more limited. I attributed it to my errors at first. But nothing changed in my test rig from last night to this morning other than the reboot and restarting REW. I can only surmise that it may have something to do with the sound card buffer(s) somehow. I'm using ASIO @48kHz, no secondary output enabled. I also have a calibration file for this card at this sampling rate. The measurement is flat with it while the 410 has an excellent response without cal file. Lock frequency to RTA FFT is set using a Rectangular window.
I only recently saw mention of ENOB. It changed considerably from last night to today's measurements. Maybe that will provide a clue.
I do have a curious issue in testing sound cards on loopback. I've spent a fair amount of time the last couple of weeks in setting up a hardware test rig and learning the ins/outs of REW and other software. So far REW looks to be my go-to for that. However, recent tests provided conflicting results. I'm fairly sure I'm using REW appropriately at this point. I'm testing both a new Focusrite Scarlett Gen 4 and an old M-Audio Delta 410 (1010LT in the queue). This is on Windows 10 Pro on an older PC. Last evening I ran a series of tests for distortion on the 410. I'm trying to characterize sound card(s) and an amp for raw driver distortion to ensure that those aren't polluting the driver results. I've found considerable issues with cabling for one, that's a separate issue.
The issue is that in the series of tests last evening I went from 0dBFS input and dropped the level sequentially. The SNR was far lower what I expected for this card. As I dropped the loopback output setting I noticed that the background noise dropped as well, but the SNR remained roughly the same The individual harmonic levels appeared correct and changed as I expected. At the lowest output setting the noise floor was much lower. Microsoft forced an update/reboot of that PC overnight. When I resumed testing today the noise floor results changed. It was in a more expected range and remained roughly constant with output voltage setting changes as one would expect. I've noticed in the past (when learning REW) that overdriving an input that stopped REW seemed to result in further test results questionable. I would stop/restart REW. BTW, I had a similar issue with Arta in trying it. I'll attach window captures of REW.
I think I've seen similar with the Scarlett, both in this old PC and a laptop, but that testing has been more limited. I attributed it to my errors at first. But nothing changed in my test rig from last night to this morning other than the reboot and restarting REW. I can only surmise that it may have something to do with the sound card buffer(s) somehow. I'm using ASIO @48kHz, no secondary output enabled. I also have a calibration file for this card at this sampling rate. The measurement is flat with it while the 410 has an excellent response without cal file. Lock frequency to RTA FFT is set using a Rectangular window.
I only recently saw mention of ENOB. It changed considerably from last night to today's measurements. Maybe that will provide a clue.
