My main bench multimeter is an HP3466A that I found at a hamfest a couple years ago. Mind you it’s older than I am, but it looked to be in good functional order, and $30 seemed like a fair price for a 4.5 digit bench meter, so I picked it up.
After a while though, I noticed that its DC voltage readings seemed to be low when probing circuits containing larger resistances; anything around 100K ohms caused a noticeable discrepancy in the expected voltage reading, and it worsened with larger resistances.
Since the discrepancy appeared to be dependent on the resistance of the circuit being probed, I immediately suspected a problem with the multimeter’s input impedance. Consider the standard voltage divider, where R1 is the impedance of the circuit being probed and R2 is the impedance of the multimeter:
The HP3466A is rated at 10M ohm input across all voltage ranges, so an R1 impedance of 100K ohms shouldn’t cause any noticeable discrepancy in the measured voltage; it was clear that the input impedance of the multimeter had to, for some reason, be lower than its rated value.
However, using a second multimeter, the HP3466A’s input impedance measured 10M ohms, just as the manual specified. Something wasn’t adding up (in hindsight it’s worth noting that this measurement was made with the HP3466A powered off).
After some poking and prodding to make sure the physical range select switches were functional, I pulled the whole meter apart to start examining the PCB. That didn’t get me very far, but the HP3466A manual kindly included a well-annotated schematic:
Note that there are two paths to the input amplifier: the “normal” upper path, and the lower path which passes through the switch labelled K100. As the schematic helpfully notes, this switch is only on when either the 20mV or 200mV ranges are selected.
Further examination of the schematic showed that K100 was a normally-off relay, driven by transistor Q100; when sufficient current is applied to Q100’s base, Q100 is driven into saturation, allowing current to flow through K100’s inductor, thus closing the switch:
This was an important clue, as further testing revealed that the input impedance was in fact 10M ohms in the 20mV and 200mV ranges, but not in any other DC voltage range. I initially suspected that the relay was bad and possibly stuck in the on position, but a quick test showed that it was in fact functional (in hindsight this should have been obvious, as a bad relay would likely be stuck in it’s normally-off position).
The next thing to check then was Q100. Since the relay should only be closed when either the 20mV or 200mV ranges are selected, we’d expect Q100’s collector to be low (Q100 in saturation) when one of those two ranges are selected and high (Q100 in cutoff) when any other range is selected.
However, measuring Q100’s collector voltage showed that it remained in saturation regardless of the range selection, keeping K100 always on! A quick voltage check on the base of Q100 verified that the circuit was attempting to drive Q100 into cutoff when voltage ranges above 200mV were selected, but the transistor was just never going into cutoff. This suggested that Q100 was bad, and would explain the input impedance discrepancy for all ranges above 200mV: with K100 always on, there were two parallel paths from the probes to the input amplifier, resulting in a much lower input impedance.
The HP3466A manual listed Q100 as a general purpose NPN transistor; replacing it with a 2N2222 did the trick, and the HP3466A has been a staple of my bench ever since: