I got the new ref last week. Here are first impressions:
To start with I powered it with Keithley 2450 @ 12.00V and measured with Keithley 2002. 2002 was calibrated by CalMachine 9 months ago and read 0.1ppm high on 10V range at the time of calibration. It does not get much better than this. (Thanks CM).
I used 2450 to precisely track power consumed by reference. On the start it took 25mA and then consumption slowly drifted down to 22mA, where it reached thermal equilibrium. After getting stable it would fluctuate very slightly with room temperature.
Unfortunately I could not get enclosed temperature monitor to work. There is some sort of issue with Arduino drivers on my computer. My lab temperature is very stable though varying less than 0.5C around 21C.
Average reading from K2002 was 7.0456737. This is about 4ppm high comparing to other results reported. I assume my Keithley 2002 could have drifted that much in 9 months.
Noise levels coming from K2002 were quite high. At times reaching 0.8ppm or about 5.5uV peak to peak. At this time we should consider noise coming from 2450 and also large amount of equipment concentrated in a same space.
For the second log, I used 2 of 6V SLA battaries to power reference, used 3458a for logging (TiN logging parameters:
https://xdevs.com/article/hp3458a_gpib/) and relocated reference and 3458a away from other equipment. This time noise improved a lot to about 0.08uV peak to peak over a course of few minutes. I would say that this is a remarkably low.
I am considering using Datron 4910 (buffered out) with SR1010 (1kOhm per step) as a voltage divider to create very stable 7.0000V and then use that in series opposition with the PX reference to measure noise directly against Datron. More results and upload of files is to follow.
Cellular, this is an excellent reference, especially considering the size and weight.