bson, thanks!
OK, time to fess up.
Context: Just for fun I'm playing with and modding K2001's ADC. To learn about finer points of ADCing and attempt to reduce Keithley's noise in the process.
K2001 is very conveniently designed for messing with it.
Among many other very cool things, I've stumbled upon the following:
When converting analog signal into digital form, this ADC uses two counters. One for counting the difference in the number of up and down slopes during dual slope portion of the cycle, another is simply a clock count for the duration of the final slope.
And, apparently, due to some major screw-up, designers come one bit short for the dual slope counter if integration time is too long (e.g. 10 NPLC). 16bits are not enough!
Thus, this counter goes through integer overflow!
On my device (A08, single ROM), in 60Hz country, overflow happens around +- 7.5 to 8V at ADC, which translates into +- 15V at the front panel (12.5V in 50Hz zone?).
When I realized that, my first instinct (for some reason) was try to fool my Keithley by switching overflowed voltages so very fast that the micro wouldn't be able to catch the wrap around.
And it WORKED!
Overflow, basically, maps two different voltages to the same value on ADC's output. For example, looking at the raw readings (on my K2001) +13.7V and -17.3V produce the same integer value!
Why the same trick doesn't work on later models? I do not know.
Several theories:
1) They fixed the hardware - the big chip sitting on the ADC board. That's expensive. Unlikely.
2) They (hardware designers) added an RC filter that slows the transition enough for the microcontroller to see through the ruse.
3) They added like 0.1NPLC peek to check which ballpark they're landing at.
4) You guys suck. And your power supplies suck too.
Long leads and small, high ESR caps might not be able to fully recharge Keithley's internal circuits in under 160ms required to trick the micro. I don't recommend switching at the multimeter end - might tarnish the contacts.
So, here you have it. Small mystery solved. This is FUN!