Using the most common definition* of dBFS, a -0.1dBFS sine wave (by definition) does not have any inter-samples that exceed -0.1dBFS. Stereophile does 0dBFS testing at 1kHz and/or 50Hz for digital products. Almost all products easily do >-80db@50hz. The only one I've seen that sucked bad was the Dragonfly near full scale.

*AES17-1998 3.3-3.4 (page 5)

This will get you started. It is a real problem mastering engineers face.

http://www.solid-state-logic.com/music/X-ISM/

https://www.gearslutz.com/board/showwiki.php?title=Tips-and-Techniques:Intersample-peaks

http://www.digido.com/forum/mfit/77-intersample-peaks.html

http://www.askaudiomag.com/articles/why-you-need-an-intersample-meter-when-mastering

http://www.musictech.net/2012/09/10mm-no211-inter-sample-peaks/

I appreciate the detailed response. Yes, I am aware signals are reconstructed and reconstruction of some signals can generate "overs". But a sine wave <=0dBFS is not such a signal. The definition of 0dBFS sine wave is one that kisses the positive max, but never touches the negative max. A full scale sine wave does not "over" in reconstruction.

Did you notice section 8.5.1 in AES17-1998:

8.5.1 Total harmonic distortion and noise versus frequency
The measurement should be conducted with a sine wave at – 1,0 dB FS and repeated with a sine wave at – 20
dB FS. The test signal present in the output should be removed by means of a standard notch filter and the
remaining signal bandwidth limited to the upper band-edge frequency or 20 kHz, whichever is lower. The
level of the filtered signal should be measured and reported as a ratio to the unfiltered signal level. The
measurement should be repeated at each octave frequency from 20 Hz to one-half the upper band-edge
frequency.

There is a good reason the spec was written to use -1.0 dBFS.

Ok, here is another look at the UHA-6SmkII THD

Test setup:
1. Cool Edit Pro DAW with -0.1 dBFS and -0.2 dBFS 100 Hz and 1000 Hz sine 44KHz 32 bit wavefiles as signal generator
2. Vaunix Lab Brick powered USB hub
3. UHA-6SmkII USB connected
4. picoscope 5243b - see screen shots for setup parameters
5. USB cable provided by Leckerton
6. Radio Shack nothing special 3.5 mm TRS to stereo RCA
7. RCA to BNC adapters from Markertek / Allied Electronics / Digikey etc.

The tests are performed at 240 mVpp, which is a level consistent with my listening levels of 70 to 77 dB SPL for an HD600 or HD650
The UHA-6SmkII is capable of producing greater than 9 Vpp without clipping so we are well below any analog section limitations

UHA-6SmkII 1000 Hz -0.1 dBFS sine 240 mVpp




UHA-6SmkII 100 Hz -0.1 dBFS sine 240 mVpp




UHA-6SmkII 1000 Hz -0.1 dBFS spectrum THD = 0.023%




UHA-6SmkII 100 Hz -0.1 dBFS spectrum THD = 0.101%




UHA-6SmkII 100 Hz -0.2 dBFS spectrum THD = 0.025%




UHA-6SmkII 100 Hz -0.1 dBFS spectrum THD = 0.102% - expanded spectrum x-axis




UHA-6SmkII 100 Hz -0.2 dBFS spectrum THD = 0.025% - expanded spectrum x-axis




Different system setup, similar results to the dScope.

I'm led to suspect that the digital signal being fed to the DAC has itself been distorted by a conversion error, a level error, or some other OS/driver processing that has kicked in.

Yes, you are correct. It is an inter-sample over as I stated, but it is coming from the windows driver, not the DACs. I thought I had this solved with the appropriate configuration for exclusive mode of the WASAPI, but I have since discovered that the configuration reverts for reasons I don't understand. This leaves the driver behaving as if the KMixer were engaged, running MS idea of sample interpolation rather than sample accurate output. AAAARRRRRGGGGGGHHHHHH!

I am in the process of rectifying those measurements affected. I will be spending the next several days performing the measurements again and posting corrected results.

^Yes!  Anax did the legwork on the ADA4627 and we later bought ADA4627-BRZ uprated versions that we all like.  Nick sold enough of them to put them up on his site as an option.  I think they are +$15 to put in the unit.  For my iems, I also had Nick adjust down -6 from stock.  I tried -12, then stock, then -6 for the best range on low gain.  At times with very quiet recordings, I need high gain.  I do get hiss in high gain with UERM but have it if I need it.  Low gain works great 90% of the time.