Just as a sanity check, I ran the OH300 overnight without any firmware in the ATTiny841. This gives the ADEV of the oscillator without any discipline, which should be a good baseline for what can be achieved.

The oscillator's specification for ADEV at tau 1s is 1E-11, so I'm within spitting distance of that. The measurement and reference timebase instability is somewhere in the low 2E-12 range or so, so that may be a small part of the difference, but even if it's all the fault of my design, I'm happy with this result. The oscillator's own stability needs to carry out to tau 10E3 or so, and then the GPS discipline takes responsibility from there. Comparing this graph to the other graphs I've gotten with the older firmware running have me satisfied that low-tau performance is where it should be.

Now it's just a matter of getting the phase discriminator to make a good contribution.