Presumably, it can measure down to 0.1mV as I get one decimal place measurements. What I don’t understand is the flipping of the toggle switch. Shouldn’t I see some difference there? What’s the point otherwise?
Don’t know if you’ve seen Ian Fritz’s article on the subject, it’s here.
The point of the right switch is that flipping it exchanges the roles of the two 100k resistors (or, actually, in the version of the circuit here, one 100k resistor plus one side of the trimmer versus the other 100k plus the other side of the trimmer). If the two resistances are mismatched then flipping the switch will change the measurement.
Conversely — if the measurement doesn’t change (except the sign) — that means the resistances are exactly matched. Or exactly enough not to change the measurement at the level of precision of the multimeter anyway. Put it another way, in terms of Fritz’s formulas, g is equal to zero. Which is not a bad thing, in fact it’s a great thing. Dave Jones says in the linked discussion:
You adjust the trimpot by switching the meter probes back and forth, turning the trimpot until you get the same reading with just the +/- sign changing.
So the behavior you see is the goal.
When I did transistor matching I didn’t do the swapping. In fact I think I probably skimmed over that part of Fritz’s explanation. (I was paying more attention to Kassutronics’s description, which doesn’t seem to mention it.) I don’t think it’s critical for our purposes. Fritz was using his method to match transistors at the 50 µV level. Per Kassutronics, 2 mV is probably good enough for synth building.
1 Like
This is very helpful! I went through Fritz’s document but as the circuit I built was a simplified version, I was not sure of what was applicable and what not. I thought I was supposed to flip the toggle switch and take two different measurements, subtract the numbers, and note down the measurements to find the right pairs. However, what I found was that the measurements were identical (including the sign) no matter what transistor pair I put in the ZIF socket. I thought it was highly unlikely that all the random pairs had no differences.
So if I understand this a little better:
- the switch is about calibrating the circuit.
- I should toggle the switch and adjust the trimmer so I get identical with opposite signs measurements with whatever pair I have in the ZIF socket.
- Once I have that calibrated change one of the transistors and take measurements (without flipping the switch) and note whichever two transistors give me the same measurement.
So far I get identical measurements when I flip the switch, but always with the same sign.
Well, as I understand it, that’s wrong. The switch is supposed to change which multimeter lead connects to which resistor, that is, it should be equivalent to exchanging multimeter leads. When you do that (if you’re measuring voltages) the sign will flip.
1 Like
Swapping the multimeter leads does result to identical measurements of opposite sign, flipping the toggle doesn’t. I checked continuity in the toggle switch (and reflowed the solder just in case) and everything is fine. Still, no difference in the signs.
Besides that, I am still not sure why Ian Fritz asks us to:
(5) Calculate half the difference of the two readings. That’s your mismatch.
If the difference between the two readings is always zero (identical readings with opposite signs, then half of zero will be zero and all the pairs will be assumed to be perfectly matched, which does not make sense.
That certainly means something is wrong with the circuit; there’s no way you’d get the same sign if it were correct. If you test continuity between the points where the multimeter would be connected and the transistor emitters (or the resistors), with a correct circuit you’d see each multimeter point connects to one transistor (and one resistor) with the switch one way and the other transistor (and other resistor) with the switch the other way. That can’t be what your circuit is doing, or the sign would flip, because it’s exactly the same thing as reversing your multimeter leads.
In Fritz’s original circuit there was no trimmer (and no switch). If the resistors were mismatched you’d get different (magnitude) readings, and the difference in magnitude would tell you how to apply a correction to your readings to take that mismatch into account. That difference would have to do not with how well the transistors are matched, but with how well the resistors are matched.
For example: You put your reference transistor in one side and another transistor in the other side. With the multimeter leads one way you measure 2.35 mV; with them the other way you measure -1.87 mV. They differ in magnitude because the resistors are mismatched. The actual transistor mismatch is (2.35+1.87)/2 = 2.11 mV.
You could do the same with all the other transistors you’re measuring. But there’s an easier way, using the two readings from the first transistor. If you calculate what Fritz calls g you get g = (2.35-1.87)/2 = 0.24. Now you can put a different transistor in (along with the reference transistor) and maybe you measure 2.13 mV with the multimeter leads the first way around. Now you don’t need to reverse the leads, you can just do 2.13-0.24 = 1.89 mV and that’s the actual transistor mismatch for that pair.
But if you use the circuit with the trimmer, it’s even easier. You can adjust it with the first pair of transistors so you get the same (magnitude) readings, and having done so (if you leave the trimmer alone from then on), g = 0 and no such correction needs to be applied.
1 Like
I really appreciate your taking the time to explain this! If I understand this correctly, the toggle switch should be only used for estimating the resistor mismatch g or, in the presence of a trimmer, calibrating the circuit to make g = 0. If what I just wrote is correct, there should not be any flipping of the toggle switch during the transistor measurement. It’s an one off thing, and probably a waste in the case of the trimmer circuit as one will only need to swap the multimeter probes a couple of times to get g =~ 0.
Of course the above conclusion assumes that the toggle switch is implemented correctly in the circuit, which I am not sure at all that it is the case. I will have another closer look out of curiosity mostly, and to warn others of the problem, as I lean towards leaving this issue behind to proceed with some actual transistor matching (I have some 900 pieces to go through).
[Revised]
Enough for a lot of modules.
I see this response in another forum:
The switch toggles the bias currents to the two transistors under test, not the meter connection to the two transistors. If trimmer T1 has been adjusted correctly, the current through resistors R1 and R2 will be identical so switching them will have no effect on the Vbe difference between the two transistors so it is correct if you don’t see a sign change when flipping the switch.
Looking more closely I see this is correct and my earlier thought was wrong. It’s not switching the meter probes, it’s switching the resistors. So once the trimmer is set correctly then, yes, you should see the same value with the same sign. Sorry if I caused confusion.
1 Like
I use transistors quite a lot for different circuits so I thought I should take out the matched pairs before I get my hands in this particular stash!
Alright! Here’s what I’ve done.
-
I adjusted the trimpot as best as I could by ignoring the toggle switch while I had a pair of transistors in the socket.
-
I then put in a reference transistor and went through all the transistors noting the measurement (the range of measurements I got was 0 to 2mV in a multimeter that measures down to 0.1mV)
-
Each time the measurement took a minute or so to settle because I touched the transistors with my figers. To avoid confirmation bias (i.e. removing a transistor once the transistor gave a familiar measurement, that is, a match), I measured them again, this time, using tongles.
-
I sorted the 12, or so, best pairs according to the double checked measurements.
-
I am now testing the transistors in each “matched” transistor pair against each other. I get differences around 0.8-1.2mV.
My question is this. Is this latest measurement the actual measured (mis)match in each pair? Not zero (due to measurement errors and some actual mismatch) but, likely, good enough?