I'm having a hard time with this circuit. I can't get either LED to light.
I am using a REF102 to supply 10 volts to power the circuit. I had to use the recommended PNP transistor in the datasheet to provide enough current. I'm getting exactly 5 volt drop across each of the 390 ohm resistors. I used a 156k ohm resistor for the base connection.
Any ideas?
Did you use R4; which value?
Try disconnecting the collector of one of the DUT. One of the LED should light.
Otherwise, it has to be a silly problem of wrong polarity, wrong pinout or wrong connection.
There is so little that can go wrong in such a simple circuit, that should make troubleshooting an affair of 60 sec. at most
This is weird. I removed the jumper between the LEDs and the transistors' collectors. I inserted my ammeter in its place and there was no current flowing even when there was current flowing elsewhere in the circuit.
At the suggestion of a friend, I replaced one of the 390 ohm resistors with one of different value (in this case 1k ohms) and voila! The red LED lit up.
The only thing I can figure is that if the two transistors match exactly and the 390 ohm resistors match exactly, what is created is a perfect wheatstone bridge so that no current flows across the circuit.
Now I'm going to put the original 390 ohm resistor back in the circuit and try a NPN transistor which I know does not match.
Aha! When I put the 390 ohm resistor back and inserted a different, poorly matched NPN transistor the red LED lit up. I then remeasured the current between the LEDs and transistor's collectors and it read about 150 micro amps.
At the suggestion of a friend, I replaced one of the 390 ohm resistors with one of different value (in this case 1k ohms) and voila! The red LED lit up.
The only thing I can figure is that if the two transistors match exactly and the 390 ohm resistors match exactly, what is created is a perfect wheatstone bridge so that no current flows across the circuit.
Now I'm going to put the original 390 ohm resistor back in the circuit and try a NPN transistor which I know does not match.
Aha! When I put the 390 ohm resistor back and inserted a different, poorly matched NPN transistor the red LED lit up. I then remeasured the current between the LEDs and transistor's collectors and it read about 150 micro amps.
Last edited:
So....after testing a handful of transistors I found a pair that is off by about .22 %.
Oh, I gave up on the LED and substituted a resistor to actually measure the current.
What do people generally think is close enough to be called a match? Is 1 % close enough?
Oh, I gave up on the LED and substituted a resistor to actually measure the current.
What do people generally think is close enough to be called a match? Is 1 % close enough?
transistors in parallel benefit from a close match of Vbe if they are to pass the same current when driven by the same voltage.
621mV as a drive signal to a pair of undegenerated transistors that measure as different Vbe (say 621mVbe & 623mVbe) will pass a different current when in circuit.
Adding degeneration resistors reduces the difference, but does not eliminate it.
621mV as a drive signal to a pair of undegenerated transistors that measure as different Vbe (say 621mVbe & 623mVbe) will pass a different current when in circuit.
Adding degeneration resistors reduces the difference, but does not eliminate it.
After using a DCA55 / DCA75 combination and a spreadsheet, I think I have come close to matching some transistors.
Just to see how close I was with that method, I decided to try the method in the attached document.
I matched the 100k resistors within 1 ohm using an HP 34401A. Even so, when I switched transistors per the recommended method, the difference was about .7 mV. This was way more than can be accounted for with the small difference in emitter resistors, in my opinion.
Any ideas?
Just to see how close I was with that method, I decided to try the method in the attached document.
I matched the 100k resistors within 1 ohm using an HP 34401A. Even so, when I switched transistors per the recommended method, the difference was about .7 mV. This was way more than can be accounted for with the small difference in emitter resistors, in my opinion.
Any ideas?
Fritz jig puts the resistors on the wrong side of the DUT &REF. This arrangements very effectively hides the differences between DU & Ref.
I thought I expressed this view at the time IanFritz posted his jig test.
The emitter resistors must be very small. Even zero ohms is acceptable.
The voltage drops should be measured across collector/load resistors.
I thought I expressed this view at the time IanFritz posted his jig test.
The emitter resistors must be very small. Even zero ohms is acceptable.
The voltage drops should be measured across collector/load resistors.
I tend to use 1k0 ±0.1% as the collector loads.
That lets you measure upto 10mA if you don't let the resistors get too warm. Use 100r for higher currents and 10r for power devices.
Gives a resolution down to 0.1µA using the 199.9mVdc scale of a DMM voltmeter.
That lets you measure upto 10mA if you don't let the resistors get too warm. Use 100r for higher currents and 10r for power devices.
Gives a resolution down to 0.1µA using the 199.9mVdc scale of a DMM voltmeter.
I'm thinking that I need to maintain precise control over the ambient temperature when measuring many of them.
The way I do this kind of test is to insert the whole bunch into a breadboard, then put a cover (leaving the emitters or drain or whatever you are going to test accessible from outside), let things rest and stabilize for 15 minutes, and then proceed with the tests, noting the "address" (breadboards are indexed) in front of the measured parameter.
After watching the transistor being tested drift up and down over time, I decided to tape a thermocouple to one and see if the temperature was changing. Sure enough, for no good reason that I can tell, the transistor itself is changing temperature both up and down over time. If it were self heating, I would think it would run away but it eventually cools back down again. It performs this cycle continuously....
If you use an LTP style jig with the two devices thermally coupled, you fairly effectively remove the temperature variation from the comparison of DUT to REF.
Comparison is one of the most useful measurements you can achieve.
Comparison is one of the most useful measurements you can achieve.
Does LTP mean Long Tail Pair?
I've thought about making some kind of jig to accomplish that feat but am not sure about how to actually accomplish it.
I've thought about making some kind of jig to accomplish that feat but am not sure about how to actually accomplish it.
I started with a 8pin DIP socket.
But clamping the two To92 together with the wide gap between the two rows of pins in the socket was bending the leadout a lot. Some while layer I bought some SIL sockets. They too are 0.1" pin pitch.
Cut off 3 sockets from the strip, cut off a second row of three sockets.
Glue these together to form a 2row by 3 receptacles grid. Instant glue works OK.
You have six pins out of the bottom side.
point to point wire these to form an LTP test jig.
This jig can be wired emitter to emitter and take this to 0V (-ve of a lab supply)
Wire the two bases together. This is where you input a voltage of ~600mV to get the LTP to conduct.
Attach a 1k0 0.1% resistor to each collector.
Connect the two resistors. This can be fed by two methods.
Either a direct wire from the +ve of the lab supply, or
to the tail of a CCS that can be preset to the current you want to measure DUT and REF at. Then the top of the CCS goes to +ve of lab supply. You can add another resistor here if you want a bit of protection in case the lab supply is set too high for the current rating of the devices. Another 1% 1k0 would drop 10V if your devices drew 10mA (5mA each if balanced but max of 10mA to one if badly unbalanced).
You can plug in a To92 DUT and REF facing towards each other and clamp them together with a wooden clothes peg.
But clamping the two To92 together with the wide gap between the two rows of pins in the socket was bending the leadout a lot. Some while layer I bought some SIL sockets. They too are 0.1" pin pitch.
Cut off 3 sockets from the strip, cut off a second row of three sockets.
Glue these together to form a 2row by 3 receptacles grid. Instant glue works OK.
You have six pins out of the bottom side.
point to point wire these to form an LTP test jig.
This jig can be wired emitter to emitter and take this to 0V (-ve of a lab supply)
Wire the two bases together. This is where you input a voltage of ~600mV to get the LTP to conduct.
Attach a 1k0 0.1% resistor to each collector.
Connect the two resistors. This can be fed by two methods.
Either a direct wire from the +ve of the lab supply, or
to the tail of a CCS that can be preset to the current you want to measure DUT and REF at. Then the top of the CCS goes to +ve of lab supply. You can add another resistor here if you want a bit of protection in case the lab supply is set too high for the current rating of the devices. Another 1% 1k0 would drop 10V if your devices drew 10mA (5mA each if balanced but max of 10mA to one if badly unbalanced).
You can plug in a To92 DUT and REF facing towards each other and clamp them together with a wooden clothes peg.
Last edited:
I am still having the problem of the transistors not settling down even with their "heads" clamped together. Even if the power supply I am using did not have good regulation, both transistors see the same voltage. The collector resistors are within 0.007% of each other and the bases are seeing a little over 700mv. The collector currents are about 3 ma.
Any ideas?
Any ideas?
- Status
- Not open for further replies.