[time-nuts] Re: Thunderbolts ...

[Stewart Cobb]

A few days ago, Charles Steinmetz mentioned that even the "good"
Thunderbolts with the 37265 OCXOs fall into two groups, one 100x worse than
the other. Perhaps that difference is not due to the OCXOs themselves but
to other parts on the board.

We know that, during the production run, the high-resolution board
temperature sensor was replaced by a software incompatible chip which
resulted in much coarser temperature resolution. We don't know how that
measurement was used in the control loop.

I gave a writeup of the OCXO control DAC here:

<https://www.febo.com/pipermail/time-nuts/2013-November/081058.html>

I forgot to mention in that writeup that the 74AC174 flip flop that
implements the PWM output gets its power from a precision voltage reference
chip. This stabilizes the DAC voltage over temperature. That reference, or
the precision opamp that serves as a level shifter for the DAC, might have
been substituted for parts with worse tempco at some point in the
production run. Either of those could cause an apparent increase in
observed tempco for the system as a whole.

It's even possible that the "good" units had offsetting temperature
coefficients between the OCXO and the DAC, and that insignificant parts
changes destroyed that favorable offset, even though all the parts remained
within spec.

The only good way to find out would be to examine all the date codes and
part numbers on the relevant parts, and to compare a group of "good" units
with a group of "bad" units and look for correlations.

Still if someone wants a project ...

Cheers!
--Stu