[volt-nuts] Fluke 5200A instability... A resolution...

[Chuck Harris]

Back in July, I posted this note about a stability problem I
was having with my Fluke 5200A.  I didn't get much response,
so I dove in with both feet.

The symptom was viewed on a trio of HP3456A 6-1/2 digit DVM's,
when measuring the 10V 60Hz output of a Fluke 5200A.

The 3456A's showed a 20mv offset, and an instability in the
tree least significant digits.  The instability was cyclical,
with about a 1/3Hz period.  The instability was gone when
the range switch was changed from 100Hz to 1KHz.  It clearly
diminished as the frequency rose farther and farther above
60Hz.

My first stop was the power supply, and checking and replacing
about 1/2 of the electrolytic capacitors.  They had ESR's on
the high side of good, but still seemed to be functioning.

Result:  No noticeable difference.

The next stop was to check all shields, grounds, isolation
between various guards and isolated grounds, etc...

Result, nothing found, no noticeable difference.

I then went to the AC/DC board, which housed an AC to DC
rectifier, various feedback paths from the sense terminals,
the integrator, and the control amplifier... It should be
accountable for any feedback irregularities.

I found numerous carbon composition resistors that were between
20% and 50% high of their intended values, and replaced them.

I checked, and found the 2uf mylar 100Hz range integrator
capacitor had some leakage, and a bit of dielectric absorption,
so I replaced it.

Tested all the tantalums, they were good...

Result, no noticeable difference... Hmmm?

I took a little time out to make an extender board using my
PCB mill.  I couldn't find the appropriate connector, so I
cut the middle out of a 100 pin 0.1" spacing card edge connector,
and mounted the two halves to the extender.

I tested the reference board, the oscillator, and the oscillator
control, and found minor offset issues, a pernicious bit of
digital ripple on the -15V(f) supply... that seemed like it should,
but didn't really matter... and a couple of more carbon composition
resistors that were well out of tolerance.

Result, no noticeable difference... Hmmm?  Hmmm?

I checked out the power amplifier board, and found several
electrolytics that were high ESR, loose screws on the output
transistor's heatsink.  Offset in need of adjustment, but
nothing much else.

Result, no noticeable difference... What's with this thing?

After concluding that the instability had something to do with
the 100Hz range, and checking all of the circuitry associated
with the 100Hz range selection signals, and finding nothing,
I finally did an experiment I should have done before:

I made some test measurements at 110Hz on the 100Hz range, and
then made the identical test measurements at 0.110KHz on the
1KHz range, and.....

They were exactly the same!  I can rule out anything to do with
the 100Hz range selection, and just look at the output frequency
as being important to the failure.

I was puzzled by how I was supposed to measure mv signals on
top of a 10V 60Hz sinewave, and got the vague idea to use my
dusty old highly neglected Fluke 893A AC/DC differential
voltmeter.  I figured that it could show the instability in the
galvanometer meter used for nulling the bridge....

On trying, I found that I could not see the instability on the
893A's null meter, and further, the Fluke 893A agreed with the
Fluke 5200 calibrator as to the AC values being measured....
everywhere, every voltage, and throughout the differential
voltmeter's frequency range!

This can't be a coincidence.  A differential voltmeter uses
a completely different method of measuring AC and DC from a
DVM, and it says the 5200A is working properly.

So, on advice from my friend Tom B., I set my HP3336B synthesizer
up to output 60Hz, and 7dBm, and plugged that into the HP3456A
DVM, and there it was:  The DVM can read voltages at exactly
60Hz pretty well, but fractionally higher, or lower, it shows
a beat note in the reading... exactly what I was seeing on the
Fluke 5200A.

Having had problems with HP3456A power supplies in the past, I
did a quick check on its filters, and found them all to be just
fine...

Doing something that comes hard for me, I pulled out the
instruction manual for the 3456A, and read through the operating
instructions.

The HP3456A is not specified to operate below 400Hz AC unless
the filter button (that I never use) on the front panel is pressed.

I pressed the filter button, and the meter fell into line, and
read a consistent, and fairly stable value.

Wow!  Such a ride to take for a simple case of operator error!

An additional bit of insight was garnered, when I read the AC
specs section of the HP3456A, and tried to understand them:

    6-1/2 digit (>=1PLC)
    Filter OFF:
    400Hz-20KHz: ±(0.07% + 730)

    Filter ON:
    10Hz-30Hz: ±(0.47% + 450)
    30Hz-20kHz: ±(0.07% + 730)

    Where PLC = Integrator's Power Line Cycles


Given a 10.0000V AC 60Hz signal, the HP 3456A could read it
as anywhere between 10.1430V to 9.8570V, and still be within
specifications.  Further, it could count wildly in that range,
and still be in specification.

Out comes the Fluke 540B, and a recalibration for my poor
maligned 5200A.

On the bright side, it was full of parts that had degraded over
time... not that it seemed to matter...

An so it goes...

-Chuck Harris