Keithley 193a multimeter leaky jfet with drain not connected

[Lorisermejo]

Hi everyone
Months ago i bought a keithley 193a and only some days ago i discovered a problem: the meter is drifty and has an offset in almost every range (the 20v one is almost ok) and i traced down to q122 and q123(q123 most probable)jfet but i can't understand what those do on the circuit because looking at the schematic their drain is not connected (nc)
Does anyone knows why are there?
Can i detach the leg that is connected to the input of the dmm and don't replace the component?
P.s. the photo is taken from page 171 of the schematic at https://archive.org/details/manualsplus_01600

[Kleinstein]

The JFETs are used as diodes, looks like as clamps to the positive side.  Using JFETs (specially to the positive side) can help to get compensation of the gate leakage of the JFET switches that are off and thus with a negative gate voltage. extra low leakage diodes could work too, but may not be cheaper than the same fets already used.

At least for a test (with some care to avoid ESD and higher voltages) one could remove these diode used 2 JFETs. A different low leakage diode would be possible too - at least for a test.

[Wallace Gasiewicz]

Had the same problem with an HP 3456 which uses same sort of clamping FETs as diodes but with both drain and source tied together.  Kleinstein helped me understand why these were used rather than just clamping diodes. They theoretically at least would have same leakage (but reverse) as the turned off FETS.  My HP diodes were extremely low leakage.  I did find some new FETs with low leakage that worked, but it was a long search. Did not find any low leakage diodes that I thought were acceptable.     
One FET is clamping POS and other one NEG.     
Disconnecting them one at a time might help narrow  down the problem. Since the FET PN junction is used as a diode, perhaps tie the drain and source together as a test, maybe that might help??   
I do not know the leakage of your FETs and a look at the datasheet might be helpful.   
I assume you tested the FETs grossly (P N Junction test) but testing them for the very small leakage is beyond my capabilities. Mine tested OK but when I replaced them, my problem was solved.

[Kleinstein]

The semiconductor leakage is usually highly temperature dependent. At least for working parts it is usually going up about 2 fold every 10 K more. Often this also applies to the extra leakage of somewhat bad parts. So there is a good chance to identify leaky parts by lokal heating of the parts - the more leaky the more sensitive to temperature. For local heating tests I use a 1/4W THT resistor powered at about nominal power, hold just above the parts as a probe.

The leakage specs with low leakage parts (e.g. < 1 nA) are often just test limits. It just costs extra to test to very low current and the spec limits may thus be far higher than the actual performance.
Low leakage THT diodes are tricky to get, as the normal glass case is not really suitable as it reacts to light.

[dietert1]

Some people use freezing spray for spot-cooling of electronic components during troubleshooting. Of course inside a meter you want to avoid remnants/dirt, so check what it really is.
Dirt might be the issue to start with.
Checking low leakage parts: We have an old HP 3478A meter that goes down so some pA. I use an external voltage source and a 1 MegOhm resistor as shunt to measure current. Since the meter resolves 0.1 uV, resolution reaches 0.1 pA. Of course the meter may produce some leakage of its own. This type of measurement needs care. One may need a good capacitor as well in order to reduce hum.

Regards, Dieter

[Wallace Gasiewicz]

Kleinstein and deitert are have very good points. I did notice that when I touched some of the FETs with my finger, the value on the readout would change. I then changed the FET involved. However I think their hot / cold applications are better since my finger is not a very good heatsink and probably inserts come capacitance / inductance into the circuit.  If you are confident that the problem lies with one of the two FETs you listed, and you have suitable replacements, I would replace both of them. The replacements should match the "switching FETs"  I do not know if you can get the same transistors?         
An Observation: The HP FETs I had in the units I worked on would test as dual diodes on a tester but I do not think this had any bearing on their actual functionality.     

My problem was worse than yours, the overload diode on the resistance input of one unit was broken and many FETs were damaged.  I think there was an accident putting high voltage into the unit before I bought it for $15.   
Another unit just had a few leaky FETS, at least that is what I think. Anyway they both work now and give acceptable readings with a calibrated voltage reference.

[Lorisermejo]

I narrowed down to these two transistor because i disconnected r182 that is a socketed 8 dip short resistor like this one https://www.eevblog.com/forum/metrology/misterious-resistor-network-of-shorts-in-keithley-195-input-multiplexer/ and measured the voltage at each pin with the instrument on and only pin 1 measured 85 mv while all the others are 0.1mv , 0.2mv or just 0mv.
Those 85mv come back at r209 at both sides and disappears but not completely ( i measure around 0.4mv) only with the 20v range because in this range k103 closes
Also now the meter with only pin 3, 4 and 5 connected together (so with only the precharge amplifier, the 2v and the 200mv ranges and the output fet connected) measure almost a stable zero with shorted input on both ranges so those input are good.
I have also tried with all the other ranges, one by one and only pin 1 is not zero and drifty
I measured r181 and r183 and i have no leaking here (i have the same 0.0mv , 0.1mv).
So i ruled out q124
It must be q122 because i have a positive voltage leak and only this fet is connected to a positive rail while the other one (q123) is on the negative rail
I have also tried to put my hand on both jfet but no changes are measurable
Maybe i can use q142 or q143 because they are unused (i have no current or ac option)

[Wallace Gasiewicz]

85 mV is a big leak. I would try removing the transistor and measure.   
Are the transistors available?

[Lorisermejo]

Well i have removed all of them (Q122 q123 q124  and q141 or q142) and now i have 100mv
Progress!
Unfortunately now I'm out of options

[Lorisermejo]

Finally i found it
It was the relay
Luckily there is another one "unused"
But a leaky relay isn't strange?

[Lorisermejo]

Here is the "new" one

[Wallace Gasiewicz]

Congratulations on your hard work.