Electrometer output stage , keithley 6517

[harrimansat]

 

Please is there anybody familiar with current electrometer output stage? Some schematic and photos will be really appreciated.
Any body knows how works outupus stage? 4 HV mosfets, 2SK1412, lm334 current source as a mosfets drivers.
Positive output works well, but negative output only reach about -700V


Thanks
Regards

[codeboy2k]

Here's the service manual.  It has no schematic.

http://exodus.poly.edu/~kurt/manuals/manuals/Keithley/KEI%206517A%20Service.pdf

Go through section 5, there's tables of voltages to be checked.  Check them first.

Keithley might mail out a schematic if you ask them.

[macboy]

Here's the service manual.  It has no schematic.
...
Keithley might mail out a schematic if you ask them.

[TiN]

We want mooore photos , please!
Did you check -1100V rail?
Does it fail any selftest?

Also please try secret menu like this one, does 6517 have it?

[codeboy2k]

codeboy flings poutine @macboy...

[Smith]

Might be farfetched, but have you tried calibrating the output? You can calibrate the output voltage via the calibration menu with just a decent multimeter.

[harrimansat]

codeboy flings poutine @macboy...

Yes, all supply voltages are ok, no error message. All works well, but not reach -1000V, only arrives to -721 volts. I will try to make a little schematic, is like a op with opto feedback. TLP590B.
Thanks to all for answers, more photos

[TiN]

Ha, ADC module is from Keithley 2001. Nice.

You may want to draw simplified signal diagram for high voltage path, to find suspected parts which can cause your issue. Its likely something on output stage.

So you check CR304,CR305,C311-313,C315-C317 and they have -1200V/+1200V on VR303,VR304?
Check if VR311 anode have -1200V too, when set voltage source to 100V and operate is ON. U308 pin 4 should have 4.81V. Measurements against V Source Lo.

[Smith]

Is this thing any different from a 6517A? In the 6517A SM they mention DAC U308 and ranging circuit U209 controlling the output stage. Might be the same as yours.

[harrimansat]

Is this thing any different from a 6517A? In the 6517A SM they mention DAC U308 and ranging circuit U209 controlling the output stage. Might be the same as yours.

It seems that is an error in the block diagram, ranging circuit is U307, AD548KN

[Smith]

I am surprised it uses a 2001 ADC module. 2001 is 7,5digit, 6517x is 5,5 digit if i remember. Although as i remember correctly the settings menu has a 6,5 digit setting.

This ADC might be a good thing, as I am trying to repair a 6517A with lots of issues on the input circuits, and there isn't much info to be found on the 6517x themselfs.

[TiN]

Well, if there is no much info, then its time to start posting, and I am sure we will end up having other people with 6517's join as well. I would love to see some nice photos of more gear from Kei, since its usually not easy to find, likely due to pricing outside of most hobbyists even for very old gear.

[Smith]

We might be lucky as the Keithley 6517 and 6517A are only being repaired on a best effort repair, as some spare parts cannot be obtained no more. So people might as well put them on Ebay 
One thing to mention is the display. You cannot obtain the FVD from Keithley like most other Keithleys. You  can only obtain the whole display module, wich sets you  back about 420 euro/500dollar ex vat if I remember correctly.

[TiN]

6517 are for sale from time to time on ebay, but usually rather expensive for repair-at-weekend project.
As of displays, there is always alternative way, to buy Keithley 7001 and salvage VFD from it. That's exactly how I repaired my 2400.
Sometimes 7001's go as low as 50$USD on bay... And you will get extra goodies, like nice chassis for DIY projects, SMPS, processor and VFD DC-AC module (same as used in 2001/2002).

[Smith]

As far as i know there are 3 types: 6517, 6517A and 6518B. There was an official upgrade from Keithley to make a 6517A out of a 6517.  This 6517 looks a lot like the 6517A I had a look at. Unfortunately I couldn't take our 6517B apart

[free_electron]

That driver stage with the four transistors looks identical to the older electrometers.

Find a service manual for an older electrometer and look how they do it there.
The principle will be the same. Keithley rewashed that circuit over and over.

Most likely you will find one of the fets is permanently 'on'. You can only get to -719 because one of thets in the positive section cannot come out of conduction.

This thing is a push pull amplifier you go full negative by turning the drivers to negative full on , and the drivers to positive full off.  (It is a two quadrant output and can both sink and source current at the same time. No matter if you sink or source current the system keeps the ouput voltage fixed

So if you cant go full negative, and the -1000 volt rail is ok, it means something is injecting too much 'positive' in the output node.

So that is where you need to start looking. My 2 cents says the fault is in the transistor ladder from +1000 volt to output.

[harrimansat]

Fake mosfets from ebay, from UTSOURCE, seller UTSOURCE STORE, hkutsource, max.V DSS 900, this electrometer needs 1500V mosfets.
Now is working with good mosfets.

Those ones are fake transistors:

http://www.ebay.com/itm/5PCS-2SK1412-Encapsulation-TO-220-High-Voltage-High-Speed-Switching-/131213219499?pt=LH_DefaultDomain_0&hash=item1e8cead6ab


FAKE cheap transistors from ebay are not always good!!!  because normally I don't have problems with ebay semiconductors I don't check it before to install.  Two days working for nothing, but now is working.

I attach some photos with measurements made with my curve tracer. Thanks to all!!!

ORIGINAL 1500V

FAKE 900V

ORIGINAL HIGH RDS

FAKE LOW RDS

WORKING AGAIN!

ORIGINAL LEFT / FAKE RIGHT

That driver stage with the four transistors looks identical to the older electrometers.

Find a service manual for an older electrometer and look how they do it there.
The principle will be the same. Keithley rewashed that circuit over and over.

Most likely you will find one of the fets is permanently 'on'. You can only get to -719 because one of thets in the positive section cannot come out of conduction.

This thing is a push pull amplifier you go full negative by turning the drivers to negative full on , and the drivers to positive full off.  (It is a two quadrant output and can both sink and source current at the same time. No matter if you sink or source current the system keeps the ouput voltage fixed

So if you cant go full negative, and the -1000 volt rail is ok, it means something is injecting too much 'positive' in the output node.

So that is where you need to start looking. My 2 cents says the fault is in the transistor ladder from +1000 volt to output.

[harrimansat]

That driver stage with the four transistors looks identical to the older electrometers.

Find a service manual for an older electrometer and look how they do it there.
The principle will be the same. Keithley rewashed that circuit over and over.

Most likely you will find one of the fets is permanently 'on'. You can only get to -719 because one of thets in the positive section cannot come out of conduction.

This thing is a push pull amplifier you go full negative by turning the drivers to negative full on , and the drivers to positive full off.  (It is a two quadrant output and can both sink and source current at the same time. No matter if you sink or source current the system keeps the ouput voltage fixed

So if you cant go full negative, and the -1000 volt rail is ok, it means something is injecting too much 'positive' in the output node.

So that is where you need to start looking. My 2 cents says the fault is in the transistor ladder from +1000 volt to output.

Thanks, but this is not a usual push pull amplifier, 4 power HV mosfets N driven with lm234 z6, one opto and one 2N7000. It works in two ranges, with 3 electronic relays, 120v and 1200V.

[/quote]

[TiN]

Was looking thru C05 firmware for 6517, and did found secret menu strings in it.
May you help to check if secret menu opens up in MENU/GENERAL if press combinations like OPER+TRIG during power on?
I would like to update secret menus article, if possible
Thanks!

[Smith]

I will be looking around for some 4mbit EPROMs to flash the 6517A C05 firmware.  Maybe I'll flash the original EPROMs, but I'm not quite sure how the UV lamp performs (I've never used it before, just like the programmer ).

[TiN]

It takes about 20min to erase chip with DIY UV barbarian setup..

[Smith]

I used the original EPROM's to flash the C05 firmware, but it didn't work. It just hangs at Checking ROM. Maybe I did something wrong converting the .x to bin files. Guess I have to go back to the C03.

[TiN]

I think ending got you  
Did you try big-endinan binary?

[Smith]

No I didn't. It was ages ago when I last programmed an EPROM. Could you convert the .x to the ODD and EVEN binary for me?

UPDATE: The only difference between the normal and big endian type would be eliminated when both files are dissected aren't they?

[TiN]

Yes, endianess would be matter which file would be called even, which one odd.

Bodged a quick tool to do splits, opposite to merger one I did long ago for 2001's.

Here's result:

6517 Firmware C05, EVEN Binary
6517 Firmware C05, ODD Binary

I just realized, that I can give it a try on my 2001 board

[Smith]

I did swap the even and odd chips, it just gave an no com link message, nothing else. I think I will request the firmware again, but with an checksum to make sure it copied ok.

I reflashed with the C03 firmware wich worked. So erasing and flashing works. By the way, I could't find any secret menu in the C03 version.

[TiN]

C03 does have same secret menu according to strings in firmware, maybe different keys.

I tried C05 on 2001 and had no much luck, same as you.
Meter just waits with message "Keithley 6517A Electrometer, Checking ROM, please wait...".
I waited 5 minutes, it's long enough. Swapping EVEN and ODD gives No comms, as that's standard reply of MCU on front panel when it cannot talk to main CPU.

Noticed ending of file had 0xFF's instead of 0x00's like in C03 or 200x series DMM meters.
So here's EVEN and ODD modified with zero's to give it a try. Same stuff, no go

Pulled C05 from site catalog first, till we clarify working binary, to avoid people bricking meters.

[Smith]

Turns out C03 was the last official release. Too bad the C05 didn't work, the firmware still has some issues. No further info was available at Keithley.

[plesa]

Did anyone successfully upgraded 6517A to C05 firmware?
I'm just thinking if the firmware problems cannot be related to memory selection
https://doc.xdevs.com/doc/Keithley/6517/fw/C03/M27C4001%20EPROM.jpg
On Keithley forum it looks like released and working firmware
http://forum.keithley.com/phpBB3/viewtopic.php?t=15543

[Smith]

Nope, I had contact with Keithley, and C03 is their final FW. C05 didn't work and was never meant to be released as far as they know. The device still has some problems, but it doesn't look like they will ever be solved

[plesa]

Thanks for update.

[plesa]

I was thinking about starting related to repair of 6517A, but I rather use this thread.
Unit does not power up, which is quite often failure of 6517 (based on in formations on forums).
So I started to reading service manual and found there is lot of various power rails.All were OK and within spec.
Another step was connecting GPIB to check if the problem is not only related to display ( there is no beep after power on, only relay click).
On GPIB no communication at all, so I tried RS232 with same results.
What I remember from repairing K2002 is tat without display unit does not power up due to display self test).
Fortunately the display board in 6517A is the K2002 display board
It was question of two minutes to disassemble K2002 and hook display to faulty 6517A and bingo - unit is working.

Checked Voltage source - OK
Current measurement - OK

Firmware is C04 which seems to be latest working. Just uploaded to TiN Keithley repository.It does not use the ST Micro 27C4001 UV EPROM but OTP from Atmel 27C040.

Now I needs to check if I will found some cheap Keithley 7001 as a display donor.

[Smith]

Won't the display be repairable? Is doesn't have that much components. Maybe it's a bad tantalum cap, or the flash/eeprom in U902?

I'm  hoping FW C04 will work, Keithley told me C03 was the latest release.

[plesa]

I was also thinking that it is something easy to repair, but it is for 98% vacuum in VFD.
Maybe there are some minor changes or firmware is for factory only. Many times the release is only due to the obsolescence.

[Smith]

These devices have some mayor problems. I hope this will be resolved someday. Sometimes ranges show wrong values instead of overflows. This makes the auto ranging completely unreliable and impossible to use. I have seen this issue with multiple 6517A's. Even the 6517B I have seen has these issues. I have asked Keithley about it, but I never had any reaction.

[plesa]

Similar issue with not correct values I encountered on Keithley picoammeters 648x series. You needs to press some button to resolve it.
In meantime I purchased 7001 mainframe as display donor, but I noticed from TiN photos that it is not compatible ( 7001/2001 VFD has one contact row and 2002/6517 has two rows).
Display is soldered and with board only the meter is working. What I found recently that on positive voltage is dead one 2SK1412 transistor.
So when I exceeded output voltage above 680V the red LED on front panel is blinking and on output is only 680V.
Based on failure from beginning of this thread it looks like the weak point are the 2SK1412 transistors.
Maybe time to replace them with some recent one the most suiteble what I found is NDFPD1N150CG
http://www.farnell.com/datasheets/1792666.pdf

[Smith]

I have used multiple Keithley 485 and 6485 picoammeters, but I have never had this issue before.

[TiN]

Display is still compatible, it's just leads forming. I just stretched pins far enough to reach second row, and it worked a treat (see my Kei2400 repair, where glass was broken).

[plesa]

Display is still compatible, it's just leads forming. I just stretched pins far enough to reach second row, and it worked a treat (see my Kei2400 repair, where glass was broken).

Excellent, I just scored on almost new 7001 for $35 Thanks!!

[plesa]

Repair continues. The 1000V range indicates overload (front red LED blinking) After replacing 2SK1412 MOSFET diode is blinking.
So I inspected rest of circuit and found and more faulty components
LM234 Current source
1N970B 24V Zener
V11713 Transistor, which seems to be 2N7000 ( according to service manual)
2SK1412 also replaced by NDFPD1N150CG

[plesa]

Repair continues. The 1000V range indicates overload (front red LED blinking) After replacing 2SK1412 MOSFET diode is blinking.
So I inspected rest of circuit and found and more faulty components
LM234 Current source
1N970B 24V Zener
V11713 Transistor, which seems to be 2N7000 ( according to service manual)
2SK1412 also replaced by NDFPD1N150CG

Repair finished.
Found that two 24V zeners VR305/VR308 were damaged (not shorted) also one of 8,6V zeners VR307/309/310/311/312 was damaged (6,1V instead of 8,6V).
And Q316 2N3904 transistor was damaged.
Now is unit completely working, and I just needs to take thermal image to make sure it is not overheating.
Waiting for display donor (Keithley 7001).

[plesa]

I just replaced display from 7001 unit. I noticed that display firmware in Motorola MC68HC705 is same for 7001 and for 2002 unit, but boards are not interchangeable. 7001 board does not work with 2002 unit.
Removing display was quite easy and fast. I used solder extractor and remove solder from all pins and peel of display with plastic knife.
I was thinking that's a pity to destroy almost new 7001 with 7057A scanner card. But for $35 it will be difficult or impossible to buy new display or display board ( someone mentioned that it is about 500 EUR).
Good news is that the 7001 scanner is working with display board disconnected completely or without VFD display. So unit is still usable with remote interface.
6517A is now with new VFD without any dim pixel

[Smith]

The complete module with the vfd is 419 euro excluding vat at Tektronix.

[plesa]

The complete module with the vfd is 419 euro excluding vat at Tektronix.

Thanks.You have some catalog of spare parts with price list? Can you check if the VFD is available? I remember that for 2000 series meter someone mentioned that it is possible to order for 60 USD.

[Smith]

No I don't have a catalog, I just contacted keithley/Tektronix. It took some mails before I even got an answer (took a full year and multiple mails to get a quote on a Keithley 2000 vfd). FYI, the 2000/2015 uses the DD-52, goes for about 60 euro's, the 2700 use the DD-61 (if I remember corectly) which goes for about 70 euro's.

Most or all single line vfd's can be bought separately from the complete front PCB. As i remember they are between 60 and 70 euro's, so it's quite a difference. I have asked multiple people from Keithley why the dual line vfd isn't sold separately, one said it was sold separately years ago. He also told it was to difficult to solder (let me judge that BTW).

Keithley also offers a line of nice connectors, cables and accesoiries. You can also buy their triax cable per inch if you ask them.

[plesa]

No I don't have a catalog, I just contacted keithley/Tektronix. It took some mails before I even got an answer (took a full year and multiple mails to get a quote on a Keithley 2000 vfd). FYI, the 2000/2015 uses the DD-52, goes for about 60 euro's, the 2700 use the DD-61 (if I remember corectly) which goes for about 70 euro's.

Most or all single line vfd's can be bought separately from the complete front PCB. As i remember they are between 60 and 70 euro's, so it's quite a difference. I have asked multiple people from Keithley why the dual line vfd isn't sold separately, one said it was sold separately years ago. He also told it was to difficult to solder (let me judge that BTW).

Keithley also offers a line of nice connectors, cables and accesoiries. You can also buy their triax cable per inch if you ask them.

Thanks. Maybe the availability is related to fact that original manufacturer of Keithely 2000/2010/2015 series VFD called DD-52 was Noritake Itron.
Recent units are using VFD from http://www.newhavendisplay.com/
You are right, Keithley 2700/7700 series is using DD-61.

Difficult part is removing VFD, that's a big difference between single line VFD where is much larger are of double side tape. On dual line is used only 7mm wide tape, which is easy.

I'm familiar with their accessory including their low noise cable with graphite coating ( this seems to be custom made, not comparable to Belden 9220 yellow triax used by Pomona ). Cable assemblies are pretty expensive, same for connectors.
Majority seems to be manufactured by
http://www.pasternack.com/bnc-triax-adapters-category.aspx
http://www.amphenolrf.com/connectors/triax.html

[TiN]

Don't we all love plesa's posts, which always share bits of good details (at least for me, lol).

My CNMC triax cable is graphite based as well (inner core of middle shield to core wire).

And DD-52 is indeed Noritake VFD made custom for KI.

[Smith]

I really hate the yellow triax cable too. The keithley stuff is great.

I have heared  a lot of people about the double sided tape on the single row units. I thought it was quite easy to remove.

[plesa]

Don't we all love plesa's posts, which always share bits of good details (at least for me, lol).

My CNMC triax cable is graphite based as well (inner core of middle shield to core wire).

And DD-52 is indeed Noritake VFD made custom for KI.

LOOL, I really appreciate Prof. Keithley bow
Your CNMC looks interesting. Does it have SPC shielding? Keithley cables shield is difficult to solder.
Your CNMC is made by http://quirkwire.com/ by the way.
Few types are also made by Habia.

I really hate the yellow triax cable too. The keithley stuff is great.
I have heared  a lot of people about the double sided tape on the single row units. I thought it was quite easy to remove.

Belden 9220  is not for low current measurement, but Pomana used it anyway for triax adaptor, which is weird.
It is much more easier if you heat up the PCB e.g with hot air gun.

[plesa]

Another 6517A electrometer repaired. It did not fall down from rack, but it has been damaged by overvoltage on input.
Which resulted in indicated overrange on all current and voltage ranges. On input was present 60V ( same on preamplifier output).
First of all I expected that input protection has been damaged
LS313 dual NPN transistor made by Micross was OK.
https://www.micross.com/pdf/LSM_LS313_TO-78.pdf
But it was ok including the input operational amplifier LMC6001
http://www.ti.com/lit/ds/symlink/lmc6001.pdf
All discrete transistors in shielded electrometer area inspected (2x 2N3904, 2x2N3906, 3x Supertex VN0550 and 3x Supertex VP0550) and Q402 (2N3904) and Q403 (2N3906) were damaged.

After replacing two NPN/PNP transistors unit is working again!!

Really helpful was schematic available in Keithley 617 ( pages 113-120 and  schematic 161)
http://www.univie.ac.at/photovoltaik/umwelt/ws2015/K617_original.pdf

Difference between 617 and 6512/6517 are in discrete FET input stage compare to low input bias current OpAmps used in recent versions.
Thats pity that Keithley do not provide schematic but Keithley gear seems to be more repair friendly ( only few custom made parts, but poor official assistance with repair and support).
I hope this will help to restore more Keithley electrometers

Be carefull and avoid contamination of PCB traces and resistors by fingerprints. PCB is quite sensitive to pads delamination. PCB needs some cleaning with IPA with toothblush and now it is going to be baked for 1h@50°C.
Firmware upgraded to B04.

[TiN]

 
Firmware B04? I have only C-versions on my collection. Perhaps upload for addition?

[plesa]

Firmware B04? I have only C-versions on my collection. Perhaps upload for addition?

It was typo, it has been upgraded to C04.
Today during calibration found that ranges 2nA, 200pA and 20pA are off by 0-75%, upper ranges are ok within calibration limits..
+2nA  1.5nA
-2nA  -1.4nA
+200pA  47pA
-200pA  -200pA
+20pA  10pA
-20pA  -15.7pA
Voltage measurement is spot on.
Needs to think about what is root cause, if some DG411, LMC6001 or something else.

Update - It was faulty cable

[Smith]

I had recently replaced my display, and last week I cleaned the input on the connector and PCB after it had excesive noise and offset after being stored for a while. I cleaned in with IPA and let it dry in a dry cabinet. After that I had let it run for about 6 houres to stabilize the device and calibrate the offset.

After about 5 houres the display turned off. It was smelling like hot electronics. All voltages seem fine, except the VFD voltages, and the device reacted to the buttons. So it had to be the VFD module, or the PSU. Replaced the VFD with the old one, and the display worked. The two transistors driving the transformer where toast (BC868). It seemed like the 60V VFD drive voltage was overloading the PSU. I quickly reverse engineered the VFD module PSU, and found some VFD drivers, caps and resistors on the 60V rail. Then I unmounted parts until I found the problem. .Turned out C912, a 100V 2,2uF cap turned bad after less than 8 houres of use! It had a DC resistance of 450Ohm. I replaced it with the same cap of the old VFD board, and everything booted right up.

[TiN]

Interesting catch.
That trace from Q901 going to xfrmr looks unhappy too.

[Smith]

The trace is discoulored indeed, but there was no real damage. It's running for a few hours again to see if everything is fine, and to check the input. All seems fine, and the unit stays a lot cooler now. The input seems more stable too.

[plesa]

Someone interested in repair not working 6517A from ebay?

http://www.ebay.com/itm/KEITHLEY-6517A-ELECTROMETER-HIGH-RESISTANCE-METER-/111994284036?

current bid is $350

[Smith]

It's getting more expensive. These are nice machines, but the service manual is quite limited. Looking at this unit is seems to boot fine, it just shows overflow on voltage setting. Current will probably do the same. My guess is a dirty connector or PCB. With such an input impedance the smallest leak will generate an overflow.

[plesa]

It's getting more expensive. These are nice machines, but the service manual is quite limited. Looking at this unit is seems to boot fine, it just shows overflow on voltage setting. Current will probably do the same. My guess is a dirty connector or PCB. With such an input impedance the smallest leak will generate an overflow.

Yes, the unit seems to be in good shape just upgrade firmware and repair input stage But actual price is too high.
I will check preamplifier output, This can help identify problem in input stage or in ADC.
With high probability it is related to input stage- check input protection (LS313 ) and opamp LMC6001.
If both are OK, I will check the 2N3904/2N3906 transistors in input stage.

[Smith]

This may come in handy: I recently had troubles with a 6517B (rental unit this time). This seems to be the same problem I have heard a lot of complaints about.  I was measuring cable leakage, until at one time the AUTO setting gave me the default not auto-ranging right. If I remember the 20nA range and below. The current displayed seemed right, but it was not very stable. It was reading about 2nA on the 200nA range. But all ranges below gave me an overload. Shortly I was afraid I fried the input as I just had a short on the 1kV output.

I connected my calibration resistor (with good coax connections) and all ranges where working OK again. I took out the MM, and connected it to the rear pre-amp output. DC ranges where good, both with the cable, and the resistor. The cable measurements were not very stable though. I thought maybe it's AC related. On most ranges using the resistor it was less than 5VAC. With the cable it was just above that in some ranges. Then I switched to the 20nA and 2nA range, I had an overload warning on the Fluke 87V! Eventually it settled to about 180-200VAC. So AC was the culprit of the overload messages.

[PREDATOR]

Hello!

I'm new here and I see a lot of skilled people, maybe someone could help me with a very strange problem that I have with Keithley 6517A.
First of all, I'm not an electronics engineer, so my knowledge in this area is limited.
I'm using this equipment to measure samples with very high electrical resistance (>100 GOhm). A few months ago I had to measure a sample with very dynamic resistance (ranging from hundreds of GOhm to several KOhm), so I was using autorange function.
I saw that when range changes from 200 pA to 2 nA, the displayed value is 10 times smaller (sometimes even negative) than it should be, so i decided to do a manual masurement. I sourced +10 V from the internal V-Source and measured the current through a 85 GOhm fixed resistor using Keithley low noise triax cable with Meter connect ON (it internally connects V-Source LOW to Ammeter LOW). The results were:

200 pA range -> +117 pA
2 nA range -> negative or 10 times lower current value (11.7 pA)
20 nA range -> +117pA
200 nA range -> +117 pA

Then, using the 85 GOhm resistor, I inserted one end of the resistor in V-Source High, and the other one (aprox. 10 cm long) in input High. Sourcing +10 V again, it measured:

200 pA range -> +117 pA
2 nA range -> +117 pA
20 nA range -> +117pA
200 nA range -> +117 pA

I extended the part that goes into the input High by 10 cm (20 cm total length) and measured again, using the same settings:

200 pA range -> +117 pA
2 nA range -> negative or 10 times lower current value (11.7 pA)
20 nA range -> +117pA
200 nA range -> +117 pA

It seems that if the input High connector is longer than 1x cm (no matter how good is it screened), the ammeter doesn't measure correctly only on 2 nA range, the other ranges work great. This problem doesn't always show up, sometimes the electrometer works fine, other times it doesn't.
I tried resetting the instrument using default bench settings, using digital filter, different integration periods (measurement speed), nothing seems to work. Any help would be much aprreciated.
I'm sorry for the bad english, I'm not a native english speaker, I hope I made myself understood.

Best regards,
Ovidiu

[Kleinstein]

Form the schematics shown so far, the positive input terminal works like the input of an trans-impedance amplifier, when measuring current. If there are no countermeasures, such an TIA might start oscillating, if there is to much capacitance on the input.

Usually there are internal caps parallel to the feedback resistors to make it less / not sensitive to input capacitance - so maybe there is a broken contact on the capacitor used in the 2 nA range.

[Smith]

Well, hearing you are not an electronics engineer: How about your measurement enviroment? Any heavy equipment / strong lights within several meters of your measurement? The 6517 is sensitive to noise, specially in the lower levels (read my previous posts).  Try to shut down some lights/computers/power supplies or try your measurement a few meters away.

[PREDATOR]

Hey guys! I would like to thank you for the answers.
This instrument is driving me crazy. It refused to work in the past few days in our lab on 2 nA range, so I decided to isolate it from the other equipments :1 DMM, 4 PCs, 7 lcd monitors, one  bench power supply, a fridge, 2 air conditioners, 2 lab ovens, 2 printers, one network switch, as Smith suggested.
I removed it from the lab and I plugged it in a power socket located many meters away from any electronic equipment and it worked like a charm. I returned to the lab, plugged it back in and worked again.           
I guess it must be a bad solder joint...

[PREDATOR]

UPDATE: I used the unit for current measurements, by-passing the 2 nA range and it worked like a charm until yesterday, when the V-Source refused to turn on.
If I set the voltage to any value and press the operate button, the operating light doesn't turn on and the output voltage is - x.xxx mV. 
I also tried to send the command through GPIB, then interrogate the instrument and it's saying that the V-Source is turned on. 
Well done Keithley, well done! 

[voltampere]

Nope, I had contact with Keithley, and C03 is their final FW. C05 didn't work and was never meant to be released as far as they know. The device still has some problems, but it doesn't look like they will ever be solved

I am about to repair a Keithley 6517A that is running on firmware C05.
I am not absolutely sure that is was delivered with this version but i guess so.

The thread is here https://www.eevblog.com/forum/repair/repair-of-a-keithley-6517a-drifting-v-offset-and-i-bias-offset/

I would be happy if you 6517 ( and 617 ) experts could take a look at my problem.

Thanks

Joerg