Keithley 617 Electrometer faulty behavior

[matthuszagh]

I purchased a Keithley 617 electrometer listed as used on Ebay. The power-on self test passes fine. But, the instrument is acting a bit erratically and doesn't appear to be doing anything sensible. Currently, when I perform a resistance measurement (according to "2.7.7 Resistance Measurements" in the manual) it display OL Gohms (zero check off). That's as expected, but previously it showed about 330ohms. The current measurement seems consistent. Previously in the 100s of nA and now in the low pA range. I don't yet have a proper 2-lug triaxial cable, but when I connect multimeter leads to the shield and force (outermost and innermost conductors) of the triaxial connector, I don't see anything sensible. For instance, with the range on auto, shorting the force to the shield with measurement set to ohms it recently showed about 80 kohms. Now it shows about 1Mohm, but it bounces around a lot. Either way, that's way off.

This unit seems faulty, but before I send it back I'd like to make sure I'm not just using it wrong and missing something obvious. Any thoughts? Happy to provide more detail and/or perform various checks if people want. Thanks.

[HighVoltage]

Start with a few simple tests...

Does the output voltage work in the back?
You can set it from 0 to 100 V and measure with a multimeter

You need to get a proper 2-lug Triax cable.
When you short the triax input cable, do you get a stable zero reading?
What does it show on the display, when shorted?

[alm]

What is the position of the V/Ohm guard switch? Is there any connection between the binding posts on the back, like COM and ground?

With guard disabled measurements are between the center conductor of the triax connector and the inner shield. With guard enabled measurements are between center and COM binding post, which might in turn be connected to the shield / ground via a link between binding posts.

See figure 2-3 from the manual (see attachment).

What happens if you supply a small voltage with the 617 in current mode on the right input high and low nodes, supply a current (e. g. DMM measuring resistance) with the 617 set to current, and measure current with your dmm while the 617 is set to Ohms? Note the values on both the 617 and the DMM. Try both guarded and unguarded configuration (observing where input low should go in either case) for volts and Ohms.

[matthuszagh]

Start with a few simple tests...

Does the output voltage work in the back?
You can set it from 0 to 100 V and measure with a multimeter

This works.

You need to get a proper 2-lug Triax cable.
When you short the triax input cable, do you get a stable zero reading?
What does it show on the display, when shorted?

When I set the V, guard switch to off and short the triax connector force to guard (input high and low) I see a reading that is roughly 10s of ohms but moves around a bit.

When I set the V, guard switch to on and short the triax connector force (input high) to COM (connected to input low through 100ohm resistor), I get a reading in the negative Gohms that's very unstable.

[matthuszagh]

What is the position of the V/Ohm guard switch? Is there any connection between the binding posts on the back, like COM and ground?

With guard disabled measurements are between the center conductor of the triax connector and the inner shield. With guard enabled measurements are between center and COM binding post, which might in turn be connected to the shield / ground via a link between binding posts.

See figure 2-3 from the manual (see attachment).

Thanks for pointing this out, I'd been getting the connections somewhat wrong. I redid some basic measurements using these correct connections (see https://www.eevblog.com/forum/testgear/keithley-617-electrometer-faulty-behavior/msg4325992/#msg4325992). COM has a shorting bar to ground.

What happens if you supply a small voltage with the 617 in current mode on the right input high and low nodes, supply a current (e. g. DMM measuring resistance) with the 617 set to current, and measure current with your dmm while the 617 is set to Ohms? Note the values on both the 617 and the DMM. Try both guarded and unguarded configuration (observing where input low should go in either case) for volts and Ohms.

In all of the tests below, when guard is off, high input is made to triax force (center conductor) and low input is made to triax guard (middle conductor). When guard is on, high input is made to triax force and low input is made to COM.


DMM set to ohms, 617 set to current with guard on.

| range (DMM ohms) | DMM     | 617    |
|------------------+---------+--------|
| 10 (10mA)        | OV.LD   | 12u    |
| 100 (1mA)        | OVL.D   | 8.4u   |
| 1k (1mA)         | O.VLD k | 08.4u  |
| 10k (100uA)      | OV.LD k | 08.4u  |
| 100k (50uA)      | OVL.D k | 08.5u  |
| 1M (5uA)         | 1.1M    | 04.96u |
| 10M (500nA)      | 2.4M    | 00.5u  |
| 100M (500nA)     | 2.7M    | 00.4u  |
| 1G (500nA)       | 0.0027G | 00.4u  |

DMM set to ohms, 617 set to current with guard off.

| range (DMM ohms) | DMM          | 617    |
|------------------+--------------+--------|
| 10 (10mA)        | 2            | 9.911m |
| 100 (1mA)        | 2            | .9903m |
| 1k (1mA)         | 0.002k       | .9904m |
| 10k (100uA)      | 00.003k      | 99.12u |
| 100k (50uA)      | 000.003k     | 49.57u |
| 1M (5uA)         | 0.000017M    | 4.959u |
| 10M (500nA)      | 00.00016M    | .4946u |
| 100M (500nA)     | 000.00017M   | .4946u |
| 1G (500nA)       | 0.0000002G   | .4946u |

DMM set to current, 617 set to ohms with guard on.

| range (617 ohms) | DMM     | 617                              |
|------------------+---------+----------------------------------|
| 2k               | 100.03u | . OL                             |
| 20k              | 100.03u | .OL                              |
| 200k             | 9.997u  | O.L                              |
| 2M               | 0.998u  | 200k initial, then 2M (unstable) |
| 20M              | 100.3n  | .OL                              |
| 200M             | 10.0n   | 170M (unstable)                  |
| 2G               | 1n      | .1G (bit unstable)               |
| 20G              | 1n      | 0.1G                             |
| 200G             | 1n      | 00.14G                           |

DMM set to current, 617 set to ohms with guard off.

| range (617 ohms) | DMM     | 617    |
|------------------+---------+--------|
| 2k               | 100.03u | .94k   |
| 20k              | 100.03u | 0.942k |
| 200k             | 9.997u  | 05.46k |
| 2M               | 0.998u  | .0054M |
| 20M              | 100.3n  | 0.045M |
| 200M             | 10.0n   | 00.54M |
| 2G               | 1n      | .0005G |
| 20G              | 1n      | 0.000G |
| 200G             | 1n      | 00.00G |


Function with guard off seems ok. With guard on, current sourced is still ok but readings seem nonsensical. Thoughts?

[alm]

Function with guard off seems ok. With guard on, current sourced is still ok but readings seem nonsensical. Thoughts?

It's clear to me that there's a problem with the unit. Could be something trivial like a bad switch contact, but it's up to you if you want to continue troubleshooting or return it.

Measuring voltage, for example the built-in voltage source, with and without guard might provide a bit more insights.

My first suspect would be the guard switch, since everything is fine without guarding. With the 617 turned off, can you measure between the following points with the guard switch in both positions:
1. Inner shield to COM (for me: guard off: 120, guard on: 149k)
2. Inner shield to preamp out (for me: guard off: 5k, guard on: 154k)

And then measure voltage between the same points with the power on, voltage, zero check off and again the guard switch both positions. Pay attention if the voltage is jumping around a few mV or a solid 0.000 V. For me, with guard off I get a solid 0.000 V between COM and inner shield, and some mV fluctuations between preamp out and inner shield. With guard on, I get mV fluctuations between COM and inner shield, and a solid 0.000 V between preamp out and inner shield.

[matthuszagh]

It's clear to me that there's a problem with the unit. Could be something trivial like a bad switch contact, but it's up to you if you want to continue troubleshooting or return it.

Measuring voltage, for example the built-in voltage source, with and without guard might provide a bit more insights.

This seems to work ok, though with guard on some of the lower digits aren't very stable whereas with guard off all digits are stable.

My first suspect would be the guard switch, since everything is fine without guarding. With the 617 turned off, can you measure between the following points with the guard switch in both positions:
1. Inner shield to COM (for me: guard off: 120, guard on: 149k)
2. Inner shield to preamp out (for me: guard off: 5k, guard on: 154k)

1. ovld in both guard switch positions
2. guard off: 4.94k, on: 155k

And then measure voltage between the same points with the power on, voltage, zero check off and again the guard switch both positions. Pay attention if the voltage is jumping around a few mV or a solid 0.000 V. For me, with guard off I get a solid 0.000 V between COM and inner shield, and some mV fluctuations between preamp out and inner shield. With guard on, I get mV fluctuations between COM and inner shield, and a solid 0.000 V between preamp out and inner shield.

1. guard off: >1V, very unstable. guard on: -16V very unstable
2. off: 8V very unstable (think this is just a ghost voltage since leaving it open produces the same voltage), so I guess it's a complete open. guard on: in the low 10s of mV, bounces around a bit

[alm]

So sounds like there's a problem with the switch, the connection to the COM binding post (loose wire?), or maybe the 100 Ohm resistor is open (overloaded?). Shouldn't be too difficult to repair, I'd think.

That explains why measurements with guard don't make sense: there's no low input connection, so no path for the current to flow into the instrument (other than leakage paths).

[matthuszagh]

So sounds like there's a problem with the switch, the connection to the COM binding post (loose wire?), or maybe the 100 Ohm resistor is open (overloaded?). Shouldn't be too difficult to repair, I'd think.

That explains why measurements with guard don't make sense: there's no low input connection, so no path for the current to flow into the instrument (other than leakage paths).

The 100ohm resistor (R117) is indeed blown (see attached image). I'm not quite sure where that inline 1/16A fuse is, but I expect that's worth checking out too. Actually, I think that may not exist on my board, which is labelled 617-102, rather than the 617-106 presented in the manual I have.

Unfortunately, I think the guard switch also has some issues. With the switch set to on, I see 0.1ohm from common to point B (without the meter leads zero'd out). That seems fine of course. With the switch set to off, I see about 3ohms (not very stable) between common and point A. That seems a bit high, especially if the other connection is 0.1ohm. If anyone knows a compatible replacement for that I'd be interested to hear it. I couldn't find a part number, but it's labelled "CW 3A A.C. 5A D.C. 125V UND LABS INC. LIST". It's a right angle SPDT slide switch. I haven't yet been able to access it sufficiently to make measurements.

[alm]

Looking at the resistor, it was the fuse . Though the parts list does not list it as a fusible resistor. F102 is supposed to be located on the inside of the rear panel according to section 7.3.2 of my manual.

You could try spraying some contact cleaner (careful to keep it away from all the high impedance bits), and/or exercising the switch. Clearly whatever blew the resistor damaged the switch.

I would be worried that whatever blew the resistor and damaged the switch might also have damaged something else. Though so far everything else seems to work fine. Do you see any blown traces or other damaged components, particularly between S103 and the inner shield on the triax connector?

My guess is that someone connected an external voltage source between COM and the inner shield with guard disabled.

[matthuszagh]

Looking at the resistor, it was the fuse . Though the parts list does not list it as a fusible resistor. F102 is supposed to be located on the inside of the rear panel according to section 7.3.2 of my manual.

You could try spraying some contact cleaner (careful to keep it away from all the high impedance bits), and/or exercising the switch. Clearly whatever blew the resistor damaged the switch.

I would be worried that whatever blew the resistor and damaged the switch might also have damaged something else. Though so far everything else seems to work fine. Do you see any blown traces or other damaged components, particularly between S103 and the inner shield on the triax connector?

My guess is that someone connected an external voltage source between COM and the inner shield with guard disabled.

I don't see anything else in that region (or anywhere else) with visible damage.

Using contact cleaner on the switch brought the resistance down to roughly the same as the other switch position. I also replaced the 100ohm resistor (apparently all this needed was a 5% 1/4W carbon composition resistor) and cleaned up the gunk. I couldn't get all of it, but I got the large majority of it.

I reassembled the unit and have performed some basic current (using the DMM method you proposed earlier), resistance, and voltage (using the built-in source) measurements. Everything seems fine with guard on and off. I did run into a little difficulty at first that I think is attributed to the switch. The open resistance measurement was showing 100ohms, but switching it a few times brought it back to OL and it hasn't reappeared since.

Any other tests you think would be worth running to ensure correct operation? The leakage current is presently showing about 500fA (pretty stable), which seems high, but I haven't given it the full warm-up time yet. I'll probably leave this on for a day or so and see how it settles.

[matthuszagh]

After about half a day, the current leakage is sitting stable at about -550fA, which is about an order of magnitude greater than it should be.

I'd originally thought it was only the electrometer board that I needed to be very careful not to touch, spread flux, etc. But, the manual says to be careful with the ranging amplifier section of the motherboard too, which is precisely where I was doing rework. I guess the next step is to perform the cleaning procedure described in the manual in that location and see if that improves the leakage current.

[Kleinstein]

A relatively easy point to test is for a DC offset. A relatively large DC offset (e.g. 10 mV range) could also cause input leakage.

[matthuszagh]

A relatively easy point to test is for a DC offset. A relatively large DC offset (e.g. 10 mV range) could also cause input leakage.

If I place the electrometer in voltage mode and short the inputs I see a reading of about 0.1 mV on the 617 display. If I place the electrometer in current mode and measure the voltage offset at the input with an external DMM I see roughly the same. Did I correctly interpret the test you described?

[Kleinstein]

A relatively easy point to test is for a DC offset. A relatively large DC offset (e.g. 10 mV range) could also cause input leakage.

If I place the electrometer in voltage mode and short the inputs I see a reading of about 0.1 mV on the 617 display. If I place the electrometer in current mode and measure the voltage offset at the input with an external DMM I see roughly the same. Did I correctly interpret the test you described?

That are 2 ways of measuring the amplifiers offset. 0.1 mV should be OK in at least not causing much extra input current. One may be able to adjust the offset even more, but I would not expect this to be needed. The offset is well smaller as the the about +-5 mV that can be applied to the resistor for the bias compensation.

[matthuszagh]

I read in the manual that a dirty input connector can cause leakage current (makes sense). My unit, including the input connector is pretty dirty, including rust on some of the back screws and caked dust on the connectors. I figured it would be easy enough to test if this was causing leakage current by disconnecting the force (center conductor) from the input stage. That had no discernible effect on the leakage current.

[r6502]

hello all,

hello matthuszagh

there is also an interesting thread regrding this instrument with a lot of details, especial about the input stage and how to repair:

https://www.eevblog.com/forum/testgear/fun-with-low-leakagebias-current-femtompere-electrometer-keithley-617/

Guido

[r6502]

in my k617 I had to replace the following parts:

- all electrolytics
- defect -5V regulator U307
- doublle FET in inpt stage Q308 replaced with 2N5909
- OP U309, replaced with OPA177
- NPN Transistors in the output stage Q301 Q304 Q306 replaced with MJE340
- PNP Transistors in the output stage Q303 Q307 Q305 replaced with MJE350
- all high insulating relays K312 K311 K309 K306 K310 K307
- trimpot R348 and R314: after replacing these the output of the input stage was stable

- I added a Heat sink to U306, the +5V regulator for th digital supply in the analog part

Later I exchanged the input stage double FET Q308 and U309 with an ADA4530 as MIDI suggested in this post:
https://www.eevblog.com/forum/testgear/fun-with-low-leakagebias-current-femtompere-electrometer-keithley-617/msg2501178/#msg2501178

I'm just working on the replacement of the relays, if I have any progress here, I will poste the results.

Guido

[matthuszagh]

in my k617 I had to replace the following parts:

- all electrolytics
- defect -5V regulator U307
- doublle FET in inpt stage Q308 replaced with 2N5909
- OP U309, replaced with OPA177
- NPN Transistors in the output stage Q301 Q304 Q306 replaced with MJE340
- PNP Transistors in the output stage Q303 Q307 Q305 replaced with MJE350
- all high insulating relays K312 K311 K309 K306 K310 K307
- trimpot R348 and R314: after replacing these the output of the input stage was stable

- I added a Heat sink to U306, the +5V regulator for th digital supply in the analog part

Later I exchanged the input stage double FET Q308 and U309 with an ADA4530 as MIDI suggested in this post:
https://www.eevblog.com/forum/testgear/fun-with-low-leakagebias-current-femtompere-electrometer-keithley-617/msg2501178/#msg2501178

I'm just working on the replacement of the relays, if I have any progress here, I will poste the results.

Guido

Wow that's quite a repair! Were you able to find suitable replacements for the high-insulation relays? Or, is that what you're working on finding now? Please do post your results when you get them, I'd be curious to see.

[matthuszagh]

I attempted to clean the motherboard around the 100ohm COM to low input resistor where I did rework. I did this by soaking the affected area in 99.9% IPA, tilting the board and blowing it off toward the edge of the board. I think this was at least somewhat effective in removing flux because I could see somewhat shiny areas at the PCB edge where some IPA solution had pooled (I then cleaned this off too). Unfortunately I ran out of canned air before I could do the same area of the back side.

 The 617 has now been left on for another half day or so and is showing a pretty stable -500fA, which is marginally better than the -550fA seen previously. The short term noise is not so far off the 1.5fA p-p noise mentioned in the manual (7.4.9 Input Current Adjustment), but there is longer term drift. In the time I've been writing this the measured leakage has gone from about -510fA to about -490fA.

I'll probably give this one more round of cleaning to get the noise + drift down further. If it gets closer to the target noise I'll probably attempt the input offset followed by input current adjustment.

[Kleinstein]

For leakage from the JFETs and the BJTs used for protection one could test if heating these parts has a large effect.  Semiconductor leakage goes up some 2 times for every 10 K more in temperature. In most cases this also applies to chips with more than normal leakage.   This way one can use carefull local heating (e.g. hold the soldering iron or a similar hot resistor close to the parts in question) as a way to get a hint on leaky chips.

The long term drift can be from something like humidity going down from inside some plastic parts (e.g. the isolating layer in the relays) and this can reduce the leakage.

[r6502]

Hello matthuszagh and all others,

Wow that's quite a repair! Were you able to find suitable replacements for the high-insulation relays? Or, is that what you're working on finding now? Please do post your results when you get them, I'd be curious to see.

Regarding the relays: Yes, I hope so, I made some insolartion measurements on COTO 9002 that  looks promesing, take a look here:
https://www.eevblog.com/forum/testgear/fun-with-low-leakagebias-current-femtompere-electrometer-keithley-617/msg4306690/#msg4306690

Curently I'm doing some measurements with the K617 unit. As soon, as I have robust results, I will poste them in the other K617 thread. I'm just programming in Labview to get the measurement results. Source is a K263 as current source ...

Guido

[matthuszagh]

I tried washing the motherboard again (this time focusing on the backside) and I don't see much difference. The leakage is now sitting at around -540fA. I may try the suggestion of looking for leaky chips by applying a bit of heat to various components, but then again I may just return it since I paid for a working unit.

[r6502]

I tried washing the motherboard again (this time focusing on the backside) and I don't see much difference. The leakage is now sitting at around -540fA. I may try the suggestion of looking for leaky chips by applying a bit of heat to various components, but then again I may just return it since I paid for a working unit.

I think, you mean the smaller board with the shielded preamplifier and where the the input is connected to, correct? The board where the 100R resistor is located, is not really critical regarding leakage.

[Hydron]

You could try hitting up the seller for a partial refund - give them the option of either refunding to where it matches the price that you'd be happy to pay for a non-working one, or dealing with a returned non-working unit and a full refund.