[time-nuts] 5370A vs 5370B

[Magnus Danielson]

Mark Sims wrote:
> 99.95 ns is a typical number for the period reading at minimum gate time.   Set the gate time to 1 sec.  My 5370A shows 99.999 999 9650 +/- 50 at 1 sec gate time.

Measuring it's own time-base is expected to give value not exactly 
on-beat. Internal cross-talk with the internal distribution of time-base 
scews the response somewhat. Also, differences in start and stop 
channels is achieved, and since those deviation patterns relate to the 
time-base it will be a particular point that will be viewed statically 
rather than being smoothed out over time. Fiddle with the trigger levels 
can remove the bias. It is also worth looking at the RMS jitter number.
Ultra-sensitive fingers and a big bag of patience can find down into the 
valeys of the jitter numbers.

A trigger voltage offset difference between start and stop channels will 
crunch out as a time-difference on the slopes of the waveform. The 
number of events (edges) or alternatively the overall time will control 
how bit error contribution it effects to.

> From my experience it can take at least a couple of months of continuous operation for the
> oscillator in an unused 5370 to stabilize its drift rate.   Once stabilized,  the 10544A's
> in the 5370A seem to be much LESS drifty than the 10811's in the 5370B.

This is of importance only if you rely on the internal OCXO for your 
measurements. If the basic plan is to have it hooked to an external 
source, then the OCXO is muxed out and is unused.

The SR-620 chooses the PLL approach instead.

> There are three mods that I do to all the 5370s that cross my path.

> First I replace the fan with a 12V 120 ma brushless DC fan driven by the 10V supply.

What is the rational for doing that? Is your goal only to get a quieter fan?

> Second I drill a hole in the top of the case over the oscillator tuning control.

To maintain the thermal conditions while tuning it's frequency I 
presume. This is meaningful if you intend to use the internal OCXO as 
reference and care about the calibration, then it is a very good idea.
If always externally fed, then skip this step.

> Third,  I add a switch to the oscillator buffer board to disable that STUPID oscillation detected
> LED that spew put 5 MHz noise all over creation.

Very annoying indeed. I was happy that I discovered that one myself. I 
don't know if others found it and approached it before me, but a 
spectrum analyser on the 10 MHz output gave me a surprise. Why is there 
spurs inbetween the 10 MHz and overtones? Look at the schematic... 
think.. oh no... A switch or as in my case... just solder a blob once 
and for all.

The really bad thing was that the detector is built in ECL with all the 
extra UMPF in the risetime that causes a true EMI/SI issue. Well built 
isn't always a good thing.

> I also disassemble and clean those slide switches on the front panel.

Sounds like a good idea. I fail to do that... I guess I am lazy.

Cheers,
Magnus