Electronics > Repair

Keithley 148 nanovoltmeter repair

(1/3) > >>

doktor pyta:
I'm pulling my hair out trying to repair Keithley 148 nanovoltmeter from 1970.

After all steps I made it drifts about 100nV per day- not very high but still 10 times above the specs.

What has been done:

1. replaced all electrolytics and tantalums. That helped, meter can be nulled on mV range.
2. power supplies checked for voltages and drift. Result: correct
3. Helipot checked for dirty wiper's contact: Result: correct
4. all switches cleaned: no change
5. The main cause of troubles at that point was located: the borrowed Keithley 1488 nanovolt short was corroded inside and has resistance changing while touching up to 50ohm. A new nanovolt short was done using the 2.4mm copper pipe. The short is covered with a shape made of plastic foam to remove air movement. Added electrostatic shield doesn't helped.
6. ALL copper to copper connections were disassembled (!), gently cleaned with abrasive tool, cleaned with acetone from dust and mounted again. While doing this I found that one of leads of red 1R/1k divider slipped from the crimped lug. It obviously was crimped improperly by the factory. Fortunately it was easy to clean and fix. I checked all other crimped connections and they were fine. This helped: the drift decreased by 2...5 times.
7. frequency and amplitude of the chopper's drive oscillator were adjusted to 94Hz (frequency drifted to 98Hz), then was changed to 93Hz to match the drift of components (resonant tank, phase shift network) and now oscillogram looks as shown in the manual.
8. The manual says that the drift can be caused by the chopper so I disassembled the chopper. For Your info it is "Stevens Arnold DC-AC Chopper CH-1238-94". It operates quietly, waveforms are stable as shown in the manual. I managed to put tiny piece of paper soaked with acetone between chopper's contacts and the contacts were not contaminated. I assembled the chopper back and checked if it works: fortunately I didn't break it. Effects : no change.
9. Degaussing coil's potentiometer was re-adjusted, : no change
10. using my new nanovolt short I attached two wires to it and connected to the current calibrator. I applied +1mA and -1mA and observed tens of  nanovolts of deflection. After the meter has drifted I applied same level of current and obtained same deflection. This means that the gain of the meter is stable, only the offset changes.

Additional observations:
-voltage on the main's rectifier and filter cap is 25V. This seems to be to high. The voltage switch is set to 230V.
-the pitch of the irritating DC/DC converter changes when operated from mains. When a bench power supply is connected in place of battery the pitch is constant. This is natural feature I think as the DC/DC converter uses saturable core oscillator.

Any idea what/how to check will be appreciated.
I need some brilliant test which will answer the question: is the chopper faulty?

P.S. 1. Data logged with input shorted with 2.4mm dia copper pipe covered from air movement by plastic foam. Range: 0.1uV full scale. Zero suppress is off.
P.S. 2. Schematic of the front end for vacuum tube version that I have is below.
P.S. 3. Manganin prepared nice teradown photos some time ago: https://www.eevblog.com/forum/metrology/keithley-148-nanovoltmeter-(a-brief-teardown)/

The drift could come from the chopper part or from the detector part. One could test the detector part without the chopper. So connect a short to ground instead of the chopper, but still keep the chopper running. This way one should see the noise / drift caused by the detector part only.

doktor pyta:
Kleinstein, thank You for the suggestion.
As I understand I should connect both ends of primary winding of T101 transformer to input LO terminal.

I have two doubts regarding this method (correct me if I'm wrong):
-by doing that, the overall fedback loop will be opened
-synchronous demodulator stage will produce zero volts for all DC or low frequency or random phase signals.
So I assume that the test will always show nice reading close to zero volts (with good or bad mechanical chopper).

Opening the feedback loop is indeed a problem. So my suggested test would not work.

A possible test could be using a different type chopper and this way check the rest of the circuit. With no need to have a high input impedance or low input current for a test with shorted input, this might lower the requirements compared to a real instrument. So a reasonable simple modern chopper made from 2 JFETs or maybe a good CMOS switch (like ADG633) might be good enough.

doktor pyta:
Finally, with help from ManateeMafia I bought the electromechanical chopper.
After fitting it into my unit the results are much better and now the instrument is within specs.
Below the measurements taken with shorted input. My observation is that it looks slightly better on analog meter possibly due to additional RC filtering and mechanical inertia.
My only concern is the rated operation time of the chopper (found some old catalogue showing value of 2000h)...

My suggestion for people who operate / repair DC nanovoltmeter is to check the instrument for proper full scale sensitivity.
What's the reason ?
There is a lot of copper to copper crimped connections which tend to fail after many years.
This failure can result in very low noise level being observed (with input shorted). This is a trap! Low noise behaviour is due to much lower gain of the amplifier.

ManateeMafia, manganin, Kleinstein, thanks for help! :-+



[0] Message Index

[#] Next page

There was an error while thanking
Go to full version