Keithley 2700 DCV suddenly failed

[jp430bb]

I've got a Keithley 2700, bought used several years ago, and it's worked fine, for occasional hobby use, up until today.   

The meter was measuring about 5VDC sourced from a wall wart, and when I looked at the display, it showed something like 315V then displayed ----.-- (or something like that).  I power cycled the meter, and now it reads something (but wrong) in DCV mode.  For a 1V input, it reads about 0.2V, and for 2V input, it reads about 0.4V.  On power-up, it does not display any errors.  From what I can tell, this unit doesn't have any self-test other than for EPROM, RAM, display, and front panel buttons.  Those all pass.

I have the case opened up.  There aren't any obviously bulged/leaking capacitors or burnt components.  The date codes look to be 2004. 

Where should I start with troubleshooting this thing?  We're getting into ESD season here in the US, so does that suggest any starting point?

[The Soulman]

How are all the other functions behaving?

[jp430bb]

I don't have a good test source for ACV.  2-wire resistance shows 1 MOhm with open inputs and 10 KOhm with inputs shorted.  DC and AC currents both read near zero with open inputs. 

[jp430bb]

In series with a Fluke 77 IV, the DC current reading on the Keithley 2700 for a 125mA signal agrees to the last digit with the Fluke. 

[giosif]

Is there something specific to that 125 mA DC current value you mention, or it just happens that you tested with that?
In other words, do the Keithley and the Fluke agree on DC current for other (e.g. higher) values as well?

Assuming the answer to the 2nd question above is affirmative, then this would indicate that the ADC in your Keithley meter is working.
And, based on your description of the present behaviour of the Keithley on DCV readings, I would guess something on the meter input side has failed (specifically, the Volts & COM side).
Could you provide a close-up picture of the section of the board where the input terminals connect?

[jp430bb]

125 mA DC just happened to be what I tested with.  I checked a few other values, for example right now 225.2 mA is showing on the Fluke 77 while 225.2 mA is also showing on the Keithley.  So DC current is behaving OK.

Here's a picture of the input area of the board.  More input components may lie under the metal shield, but I have not removed that yet.

[E-Design]

Under the metal shield are the ACV circuits... not likely related to your DCV issue.

Other initial suggestions..

Test power supply rails?

Input a known voltage and with a meter, measure along the input pathway until you find it going wrong.

Question: When you put a voltage in, does it look fairly stable but wrong or are the readings flying around all over the place?

[giosif]

Could you please take another photo of the same area, but from a different angle?

Also, you may want to read the service manual, particularly page 4-9 (or page 89 of 174).
The manual can be located here (courtesy of xdevs.com web site): https://doc.xdevs.com/docs/Keithley/2700/2700-902-01%20%28B%20-%20Feb%202000%29%28Service%29.pdf
As a test, if you feed the meter with, say, 20 VDC and go to manual ranges and set the meter to the 100 V or 1000V range, do you start to see the proper reading?

[chekhov]

Hi,

I don't own 2700, but according to your description, meter was not connected to some high-power input, input section should not have any visible damage.

Have you checked ranges manually (for volts, ohms, or both ?) Or do you hear that meter at least trying to do auto-ranging ?
Since current is functioning, it must be the path from input to ADC that fails for voltage and resistance.
Maybe some analog switch (usually ICs with names DGxxxx or MAXxxxx)  or relay driver has failed, usually this means you have to track the input path down (for voltage for example) to see where meter loses signal. Also verify capacitors, they like to leak or short themselves.

[jp430bb]

With 20V DC at the inputs, the Keithley reads 9.42 V on a 10V range, 1.13 V on a 100V range, and 1.13 V on a 1000V range.  The voltage reads between 9.42 and 9.52 V on the 10V range over a few minutes with a 20 V input from my (stable) Agilent 66332A supply.

I'll try probing around R113 and R107, but not really sure where to head after that.

[giosif]

Thanks for the photo!
That provides a better view.

Could you also measure resistance between High and Low input terminals, while switching between different DCV ranges?
You should get OL (on the meter you are measuring resistance with) for 100 mV, 1 V and 10 V ranges, and 10 MOhms for the 100 V and 1000 V ranges.

[The Soulman]

Have you tried cycling the front/rear terminal switch?

From your measured values it looks like something went high impedance early on in the signal path.

[jp430bb]

Measuring resistance across HI and LO input terminals with my Fluke meter, with the Keithley set for DCV, I see 27.68 kOhm (very wrong) on the 0.1 V, 1 V, and 10 V ranges.  Interestingly, if I swap the leads on that 2-wire resistance measurement across the Keithley inputs, I get 27.85 kOhm instead, consistently.  I see 9.9 MOhm (a good value) on the 100 V and 1000 V ranges. 

I cycled the front/rear terminal switch about 10 times.  2-wire resistance with the rear inputs selected shows a stable 1.0527 MOhm on the 1 MOhm range, but 0.000 MOhm on the 10 MOhm range. 

[jp430bb]

With 1 V applied across the Keithley inputs, and the range set to 10 VDC, I probed some voltages with respect to the Keithley low input. 

R107 input side (top): 0.999 V, R107 output side (bottom): 0.997 V
R196 input side (right): 1.000 V, R196 output side (left): 0.946 V
Q101 and Q102 tabs: both 0.946 V
R134, both sides: 0.214 V (matching the Keithley displayed voltage)
TP104: 0.215 V (matching the Keithley displayed voltage)

[giosif]

With your Fluke meter in Diode mode (and the Keithley meter powered off), can you test the two MOSFETs Q101 and Q102 and see if they are shorted?

[jp430bb]

With the Fluke in diode mode, and Keithley meter powered off, both Q101 and Q102 showed no conduction from drain to source, and they showed a 0.451 V drop from source to drain.  I didn't touch the gate lead.

[giosif]

Ok, so those appear to be fine.
How about Q104, Q105, Q108 and Q113?

[jp430bb]

With the Fluke in diode mode and the Keithley powered off, Q104, Q105, Q108, and Q113 all showed no conduction from S to G and from D to G, and they showed 0.78 V to 0.81 V forward drop from G to S or G to D. 

[doktor pyta]

Common failure of Keithleys 2000 and 2700 series which I repaired is LM339 quad comparator.

It is used to drive gates of JFET switching input circuitry.
I'd suggest to replace it and not to waste time.
P.S. I hope You checked supply voltages.

[Kleinstein]

The LM339 comparator to turn off the FET switches are good suspects. One could measure the voltage at the outputs - for the off state they should go to a fixed negative voltage, like some -20 V.


The errors like there is some connection not turning off - as the current ranges still seem to work it could be the current ranges not disconnected - either from the LM339 or from the JFET used as a switch.


Edit: the current ranges are go a different path, only after the input buffer, so no JFET for the shunts.

[jp430bb]

The LM339D looks suspect.  With the Keithley powered on, and 1V applied to the inputs, here is what I see on the LM339D:

VCC: 5.041 V, GND: -15.65 V (both stable)
OUT 1, 2, 3, 4: all around 0.213 V, which is also about what the Keithley displays as a measurement

Do people replace an SOIC-14 like this with the board still in the meter, or is it better to remove the board?  Removing the board seems safer and doesn't look too complicated. 

[doktor pyta]

Here is a tip to do it without removing the board.

Use kapton tape to protect nearby area against flux, fingerprints etc.
Then use sharp blade to gently cut IC pin by pin.
Next step is using solder wick to clean pads.
Clean pads with IPA.
Solder new IC also pin by pin.
Clean using IPA.
Remove kapton foil.
Allow few minutes for IPA to dry off.
Finished.

This way You will avoid overheating other components and the repair should not affect calibration.

[Kleinstein]

The voltages at the LM339 are indeed suspect it would point to all 4 JFETs on. One could still check the inputs, just in case there is something wrong with the digital input side: There should be 4 pins (pin 8,10, ...) at some 2.5 V and 4 digital control signals from the shift register.

[jp430bb]

I have a few new LM339Ds arriving tomorrow. 

Meanwhile, I did check the inputs, and they're not what I would expect, but it could be that a failed LM339 is pulling them up or down.  The -0.54V seems like a diode drop.  The positive inputs coming from the shift register look correct, based on the table of JFET states in the service manual.

Keithley inputs open, DCV function, 10V range selected, measured with respect to Keithley low input:

VCC 5.04 V
GND -15.65 V
1IN- -0.54 V
1IN+ 5.04 V
2IN- -0.54 V
2IN+ 0 V
3IN- -0.54V
3IN+ 0.071V
4IN- -0.54V
4IN+ 0V

[Kleinstein]

The negative input values look odd. There should be some divider, to something like 2.5 or maybe 1.5 V, but nothing negative.  So chances are the LM399 is indeed pulling that voltage low. There may be also a LM393 as a dual comparator also used for JFET control and this may share the same ref. level.