The tests conducted in part 1 and part 2 was run on my personal 53230A ("53230A-1135"). It is not impossible that this counter has some malfunction or defect that cause the bias. I have since had the chance to borrow another 53230A ("53230A-2985") from a calibration laboratory, and rerun all the tests.
I have also been in contact with an individual who has a 53230A and is intimately familiar with the workings of the instrument - he observed the same bias. I do not have his raw measurement data.
All tests were run with the counters referenced to a 10 MHz signal derived from the hydrogen maser as described in part 1. Input impedance was set to 50 ohms,
SAMPle:COUNt was set to 25 and the data collected with 1000 individual calls to
READ. The tests were run on both counters in parallel.
The temperature on the room was kept at better than 22°C +/- .5°C for the duration of the tests.
The data presented below. Red is 53230A-2985, blue is 53230A-1135.
Both counters show the bias in varying degrees. Both counters seem to show bias of roughly the same order of magnitude, 53230A-1135 performing perhaps sightly worse.
The 33510B was clocked from the same 10 MHz distribution amplifier, set to produce a 4Vpp sinewave output on both channels. Channel 1 was connected to Channel 1 on 53230A-1135, channel 2 was connected to channed 1 on 53230A-2985.
The 33510B was configured and connected as in the previous test, gate times set as multiples of 3. 3 second gate time was not tested.
Unprocessed data used to create these plots available here
I also tested input 2 on the counters, with the same results. Data is included in the datafile above.
The bias is not a results of a single defective 53230A. The magnitude of the bias varies between counters, perhaps a result of manufacturing process differences.