Keithley 2440 Sourcemeter voltage read back fluctuation

[analogRF]

Finally I got my first Keithley sourcemeter (already had a bulky and long Advantest R6243). It is a 2440 5A version which was booting fine without any errors but there was nothing in the output.
So it turns out the +/-15VF rails were dead which was easily fixed and the unit works fine now albeit a bit out of spec in lower ranges but spot on in higher ranges.
But there is one problem that I cannot figure where and how to troubleshoot. Although the I/V source is rock solid but the readback specially in voltage has this weird fluctuations on the last few digits when in low ranges which gradually disappears as i move to higher ranges with less decimal digits. For example if I am measuring 1.000000 the last 4 digits fluctuate in what I believe is some kind of pattern. It appears they increase over several triggers and then suddenly drop to almost zero and this pattern repeats. At least that's how it feels. In higher ranges with less decimal digits the fluctuation is less and finally disappears for example when I measure 40.000 V

Current measurement seems much better although again the last 2-3 digits seem too unstable for me.

Any idea where and what I should be looking at? Supply voltages have no noise or spike.  What are the components involved with the A/D?
Manual says U610 Gate array controls the ADC but for the life of me I cannot see U610 anywhere. I see U600 - U609 and then U650....

attached is a picture of the analog board
 

EDIT: is there any schematics from Keithley multimeters that can help troubleshooting the "multimeter" function of this sourcemeter?

[analogRF]

it appears there is an error in the manual and U610 is in fact U609 (Altera FPGA) on the board. That is the only gate array I can see.
there are also some stuff that are wrong on the PCB silk screen such as +/-85V and +/-42V which are in fact is +/-58V and +/-23V, respectively.

But still the question remains which components make up the multi-slope ADC? Does anybody know? Is it similar to some Keithley DMMs with available schematics perhaps?

[julian1]

The altera max and circuitry underneath it (I think a hct175 flip/flop, and sd5400 fet switch and comparator) is the adc. It has to send the serialized counts back through the optocouplers on the rhs of the board. So perhaps it is worthwhile to inspect the soldering joints/supplies around the cpld/optocouplers. Although if the error is range dependent that seems a bit odd.

How did you fix the +-15V supplies? With so many rails voltages being wrong, maybe that's a better place to look. I *think* that +24V is the main high current supply, from which everything else gets derived. It comes from the bottom board, but there should be a test point on the top board somewhere.