Got two Keythley 191 units, one with AC option is working very nice. The second, without AC option is misbehaving.
First the display was flashing erratic.
Open the burn smelly unit and found the R159 (10Ω) burn out, it connects the rectifier bridge CR102 negative output to the C114 and 7915 VR102 regulator via power on switch.
Replaced resistor and found shorted C113 tantalum 10uF cap at the output of 7915, replaced the 3 tantalum caps (C110, C113 and C116) from +15, -15 and +5V, just in case, by 10uF 25V electrolytics.
Power on the unit all voltages are very good now, ripple on all voltages below 25mV.
Now display is steady, but measurements are crazy.
Used a 20Ω resistor for measuring ohms, scale 200Ω goes into overflow, it should not. The same 20Ω resistor shows 0.089 on 2kΩ scale, what is very wrong.
VDC scale is crazy.
Measuring 1.5Vdc on input, scale 200mV goes correctly overflow, scale 2V shows 1.01V, 20V scale and up shows 0.00000
Measuring 5Vdc on input, scale 2V goes overflow correctly, but 20V, 200 and 1000V scale shows 0.00000.
The only gray wire connected to Ohms switch, goes to center of transistors and white tubular cap, is scaling voltage correctly, the 5V on input, 4.66V on that wre, when switching scales, change to 0.466, 0.0466, etc, means the attenuator is working nicely.
Can't understand why it shows 0.00000 with 5V input at 20V scale, but overflows on 2V scale.
Exchanged the processor board between the working and not working units, not the problem.
Exchanged the display board, same thing.
I wonder if the switches are not bad contacting.
As far as I understand from the schematic, the ADC doesn't play any different role when manutally switching scale for voltages,
The ADC selects Vin (for 200mV), Vref, Vsig (for >200mV) and Vzero, as in page 5-3 of the tech manual, the display is 2*(Vsig-Vzero/Vref-Vzero), and that doesn't depend on the input voltage since the scale switches are manually operated to keep Vsig max 200mV. So, the ADC doesn't even know what scale it is operating, considering Vsig is no higher than 200mV, or overflow will show. So, the problem of showing 0.00000 for 20V and up, has nothing to do with ADC or processor, it must be something around the attenuator switches. But the gray wire shows attenuation correctly.
One strange visual observation, from time to time, a sequence of 500ms display on, very fast (20ms?) display zero and repeat, like if the processor is trying to indicate something wrong. This is not all the time, this not happen when the processor is showing something different from zero.
Ohms scale seems to understand short circuit at input, shows less than 1Ω with leads shorted, but as said before, 20Ω is shown as 0.08 on 2k Scale, overflows on 200Ω scale.
The same lower ohms measurement (20Ω as 8Ω on 2kΩ scale) is observed with measuring 1.5V on 2V scale (1.01V), may be ADC Vref is out with other problems.
I think I will need to power-on both units side by side and scope (Tek 2235) trace everything.
I don't believe the shorted tantalum cap on 7915 and burn resistor as consequence would cause any other problem for the op-amps. May be I could swap the 4 socketed LF351 between the good and bad units, just for diagnostic.
======================
Okay, scoped and measured all the ADC diagnoses based on the tech manual, Table 6-4, page 6-8.
Items 1-10 are okay.
Item 11, C112 (the integrator cap) 0 volts to ±15 mV, as "integrator input" is not exactly like that, it is not ground, there is a digital 2V signal there.
Item 12, C112 (the other side of the cap), it is said "see waveform per figure 5-7". That also is not correct.
Figure 5-7 shows an integration voltage, most analog, ramps, etc, see figure 191-1 attached.
What it shows is square waves, and it never reaches zero or negative voltage, figure 191-3 attached.
The negative voltage is important to trigger the single slope comparator U07 output pin 7, figure 191-2 attached.
The input pin 2 of U07, or Q109e, never goes below ground (see 191-3), so its output never goes down.
This is explained at paragraph 5-18 on page 5-6... "The single-slope comparator output is also positive, and it will not switch until the integrator output crosses zero..." Indeed it never does. And that it seems to be the failure here.
Any scale of VDC does not produce an image as 191-1 on the integrator capacitor, not even close, it changes a bit according to the selected scale, but it is always like 191-3.
Now trying to figure out why the integration over the cap is not linear ramps, but only digital.
I removed the C112 integrator cap, measured 3.58nF, pretty close to the 3.6nF specified.
Temporarily solder another cap close value to see if integration works, no game.
R137F and G measured okay with ohmsmeter.
V-F pulses at U104a-2 output shows activity, so the U106+U116 counter/divider is working.
Q109 seems to be working.
Single Slope trigger is the last to check and could be the culprit.
U108-6 has positive pulses from the processor.
U108-2 has different pulses.
Measuring the transistors inside U108 (CA3086) using Diode function of a multimeter, found Base-Emitter (U108-6 to U108-7) as 2.4V both polarities, it should be 0.7V in way direction. Crude diagnostic defined open NPN U108C, U108-6 being positive must show the same polarity on emitter 108-7, instead, emitter is way negative by the -15 via R137E. Culprid.
Temporarily soldered a 2N4401 NPN TO92 bipolar transistor over the pins U108, 6(B), 7(E), 8(C), and badabim badabum, working.
Now lets go calibrate all pots for Ω and V.
Never give up.