Beware of silver mica capacitors in old equipment!

[PM3295]

I want share my experience fixing a problem on a Fluke 8506A bench DMM. The unit started to develop a fault that caused the full-scale voltage on the 20 V range to drift down over time. When switching to the next range it will be correct. I pinned this down to the DC conditioner board where the DC range switching happens.
According to the service manual, drifting readings could be due to range switching JFET’s going leaky. I replaced all the JFET switches. Problem remained. Started to look at the transistor clamps end basically tested everything I thought could affect it. Even lifted up the one side of the smaller value caps, which measure open circuit on the ohm meter. As another precaution I changed all the tantalum caps as well. No luck. Since the card is a pain to work on without the correct extender, I decided to power it up on the bench with external supplies.  I isolated the input range switching relays as well as the other range switching JFET’s to eliminate them. So, the unit was hard strapped for X1 range. And the test voltage fed direct into the unit via 150k input resistor, similar to that in the circuit.
I was really surprised that there was no sagging with the test voltage up at 20 V. I left it on for some time and it was rock steady. Checked a couple of suspect old solder joints and tried it back in the meter. Started sagging in voltage after several seconds! I could not figure out why this was happening. I unplugged the output filter board in case it was causing a load in the DC output. Still nothing.
Then I accidentally found something that led me to finding the culprit. I measured for resistance between various test points while the unit was plugged in, and saw a short between TP1 and TP8. These two points (common ground and analog ground) are only connected with the card plugged in!
Back on the bench, I connected these two points and observed the sagging. Great! I was confident now that there is some capacitor leaking, SO again I started lifting up one side of the silver mica caps with the board power on. The moment I lifted C15 near U3, the voltage came back up. Took C15 out. Checked it once more with ohm meter and it still read open circuit. So, I used my Keithley 237 source meter to apply voltage and measure the current. No leakage current till about 9 V. Then as soon as I increased the voltage, the leakage current will be around 4 uA at 20 V. What is even more surprising is this 47 pF cap is rated at 500 VDC. In the circuit it never sees more than 22 V over it. The lesson is: beware of old silver mica capacitors, it can cause very confusing faults. People fixing old tube radios often talk about silver-mica disease, but it goes short most of the time.  

All the modules in this instrument uses lots of SM caps, so it may not be the end of my headaches. Time will tell.

[optikon]

interesting troubleshooting story..

wonder what internal degradation occured in that cap causing it to become leaky like that.. also only leaking past a certain voltage is particularly deceptive.

[PM3295]

interesting troubleshooting story..

.. also only leaking past a certain voltage is particularly deceptive.

i was ready to start looking for another parts unit on eBay with a working DC board.     

[PM3295]

So, there was another failure with this meter. Yesterday the meter started reading low on DC but on all ranges. Luckily, this was easy to diagnose as a leaky input capacitor. The cap measured about 84 k resistance. It is C6 (4.7nF, 200V) marked on the diagram.

[PM3295]

I have discovered another problem that may have been there before this last failure. I can't be sure since the meter is hardly used to measure AC voltages.

The problem shows up on the ACV 100 mV range in the form where the reading is exactly 50% of the input signal. I made some initial measurements that caused more questions than answers. When the meter is in ACV mode, but the DC coupling is enabled (feature of the 8506) as well, it reads the correct  values on the 100 mV range. Looking at this function, there are coupling capacitors that gets bypassed when the DC mode is selected.

The puzzling part is that when in the AC mode with a jumper wire across these coupling caps (effectively activating the DC coupling) the meter still reads 50% low. It reads correct if the AC + DC function is selected. So, it appears that the AC + DC function activates something else that is not that obvious from the circuit operation described in the manual or the circuit diagrams. I suspect an issue somewhere in the RMS/amplifier block. Without some extender board, it is really difficult and time-consuming to perform measurements on this module.