Keithley 2015 high input current at the end of the 10V range

[diodak]

I have a problem with the input current in the 10V DC range with high input resistance mode. In this range, the indications quickly drop down (without the input connected) to about -11.8V. This is possible due to the high input resistance and the input begins to flow in one direction. However, I did a voltage change measurement when I turned on the 100M resistor in series with the voltage source. From the calculations (delta U / R) it follows:
+ 10.5V - current 147pA
+ 11.5V - current 180pA
-10.5V - current 1570pA
-11.5V - current 9900pA
The meter had the A / D MUX (U163 - DG408DY) replaced - it did not pass the self-test. Even then, there was an input current symptom, but I ignored it. Now I have decided to face the problem. So far, I have replaced Q108 with MMBF4393LT1G (switch for the 1: 100 divider). The effect is the same with or without this transistor. I would not like to replace everything one by one. How to select a damaged / degraded element?

[doktor pyta]

Front end circuit of K2015 seems to be similar to K2000.
Download the K2000 schematic and keep tracking.
https://www.eevblog.com/forum/testgear/keithley-2000-full-schematic/

Often I've repaired Keithley 20xx series with faulty LM339, LM393 ICs.
It is possible that some JFETs are not driven properly to be dissconected from HI node.

EDIT
I hope You have checked ALL power rails.
Possible cause might be also components of the bootstrap circuit (around U114).

[diodak]

I have a diagram for the K2000, so I gave the markings of the elements. The power supply is ok. Inputs through Q105,113,108 are blocked - the gates at the level of -15.78V.
The power supply for the U113 is 5.26V, which suggests a little bit. The VR107 and 108 diodes are marked "8A" which means they have 3.3V, which should give a difference of> 6V. When measuring -11.8V on pin 4 it is -14.4V.

EDIT
Since about 1.66mA can flow through R150 (25.5k) and there is 0.61V on R151, it means that the current is limited. 4mA then flows through R151 and R149 (150R). More or less, it means that 4-1.66 = 2.34mA is consumed by U113 (no voltage drop on R314 means no current consumption from U113 output). LTC1050 is supposed to take 1.5mA max. So I guess he is to blame for the large input current.

[Kleinstein]

It is quite normal that the 3.3 V zeners will show less than 3 V, as they are operated at a current well below the nominal current. So the supply to U113 looks perfectly OK.

A way to detect leaky JFETs, FET OPs or similar can be local heating by some 20 C or so. In may cases leakage gets significant worse if a leaky part is warmed up. It does not always works, but if the leakage goes up significant this is a strong indication and it is a relatively easy test.
A slightly hidden path is through Q114+Q136

[diodak]

So can VR107 and 108 draw 0.8mA at a voltage of 2x2.6V?

[Kleinstein]

The 3.3 V zener diodes are usually pretty poor regulation. So the voltage can drop quite a bit with a lower current like 0.8 mA.
The current to get the nominal voltage can vary between the different zener types. Chances are the nominal 3.3 V are for some 5-20 mA. 

The 2x2.6 V should be Ok for the OP - so no problem there. It should work from 4.8 V on. The supply to the LTC1050 would also not change much with the input voltage.

[diodak]

Desoldering pin 3 of the U113 buffer did not change the input currents. Similarly, replacement of Q120 (switching on the current source for resistance measurement) and Q104 switching the measurement up to 10V without a divider. The measurement of the reference resistance 100M (+/- 1%) gives the result of 99.9M, therefore the leakage current from Q109 does not occur. Disconnecting the protection consisting of the series VR106, CR118, VR105 and U107 eliminates the problem of leakage current at the input. However, it does not change the drift towards negative voltage. This explains why the voltage thresholds are asymmetrical. I have to replace the VR106 / 105 zener diodes (11V preferably with a tolerance of 2%). I don't know about the CR118 diode (marked C2V), is there any replacement available?

[Kleinstein]

I would not expect CR118 to fail. I would not consider it critical and some like a dual switching diode (similar to 1N4148) should be OK.

The slightly critical part may be U107.
If VR105,VR106 and CR118 work should be visible from the voltage for that string. It should relatively close follow the input voltage / U114B up to some 11-12 V.

For VR106 / VR105 the tolerance should not be that critical. It is more about not having too much leakage at some 9-10 V. 

[diodak]

My guess is that the zener diode changed the threshold of operation. Maybe when this protection worked. Unfortunately, the HI input connects to U107, VR106, CR118, VR105, so the sum of the voltage thresholds of these diodes affects the leakage at the 11-12V limit. I am waiting for the diodes and I will replace them.

EDIT
Diodes on 11V replaced. No improvement. I wonder if I should worry about this leakage current in the range of 11-12V.
I made measurements after, and I have:
-10V -> 90pA
-11V -> 560pA
-12V -> 43.6nA (!)
0V -> 50pA
10V -> 90pA
11V -> 180pA
12V -> 360pA
I changed the measurement method. Now I measure with a multimeter with an internal resistance of 11.178M connected in series with the voltage source.