"Soundcard" calibration at 96k distortion measuring issues

[BSL8068]

Hello
I am using REW for speaker tuning, but also to test audio gear. For the audio gear I would like to use 96K to see where the hardware rolls off at the top end.
Attempting to calibrate my Tascam US 2x2 as a interface cal ( I like this box because unlike most it will output a true +20 dBu and accept and true +20 dBu input which is handy for testing pro outboard gear).
All is fine at 48K, distortion low and flat. But when I try to use 96khz sample rate, the frequency response is flat , but when I look at distortion I see a sawtooth pattern above 4Khz. it rises from -100, to about -80 dbu.
It's a great laptop, 12th gen i7 with 32 gigs of ram, and a dedicated Nvidia GPU. So I checked REW on my other system, again a i9 990k with 32 gigs of ram, Nvidia 2060 gpu, with a different interface, Avid Mbox3.
Same deal, smooth low distortion at 48K but sawtooth distortion above 5k when using 96K sample rate. Frequency response again is smooth it only the distortion measurement that changes if I use a different sample rate..
Again, this is not measuring gear, it's just the loopback to cal the interface.

I'm using ASIO drivers on both rigs. Had issues getting Java to work.
I tried lookback for timing. but same result. I assume it some type of timing error , I tried different sample lengths. Two different computers, two different D/A A/D converter, Any thoughts?
I can always test distortion at 48K and frequency response at 96, but would prefer not switching as I am shooting a video while I work.
I'll try to attach a photo, but I am new here.
Thanks in Advance!
Brad

 

[John Mulcahy]

That sawtooth effect is due to each harmonic in turn dropping out of the THD figure as it reaches the bandwidth limit of the measurement.

The latest beta versions have a lower high frequency distortion noise floor which helps with measuring electronic equipment, you can get those builds here.