RocketMonkey
Active Member
I was going to say submerge the whole rig...
Matching Vbe of power transistor
- Thread starter Alan0354
- Start date
I thought of that also, that's why I was looking for turkey pan!!!
I just did the easy way, I use the fixture, have everything ready to probe and write, Just turn it on, reference to the emitter of one transistor and measure the difference voltage of the other emitters. Turn off the power, sort according to the voltage, redo again. After a while, there is consistency. Regroup, verify and bag them in groups. Took a few hours, but I sorted 5 sets of 9s. Best is 1mV, worst is 2.8mV spread in the group.
So I still have the PNP to look forward to later or tomorrow.
It's just too much hassle to dunk in water and what make people think it's any better? My last sort using counting by heart result only 3mV spread, doing it in room temp, no warm up. This should be a least better as I am measuring differential voltage referencing to one transistor. If transistor heat up during measurement, there is temperature compensation by the reference transistor.
Actually when you buy bulk, most are very close. Only like 4 that are way out, most conform to within 6 to 7mV.
I since modified the way I do it. I lower the collector voltage to 0.6V only. I experiment from 20V down to 0.6V, it really does not make a difference in matching. So I lower to 0.6V to minimize heat. Now it is a lot more consistent and I can leave them on longer and do measurement.
merlynski
Curmudgeon Electronicist
Rod Elliot of ESP has a very well done article and test bed circuit for power transistor matching, as well as excellent information on Audio electronics, on his web site.
Link to tester: http://sound.whsites.net/transistor-matching.htm
Link to tester: http://sound.whsites.net/transistor-matching.htm
I think my setup is more complete, I mainly test for quiescent current which is 100mA per device, that's the biggest heat generator. My setup test 60 device at a time.
The problem is not about measuring the Vbe, the hard part is when you heat the transistors, the Vbe keep changing. YOu cannot mount the transistors onto a heatsink ( you have 60 of them and they have to be swap easily!!!). Also, For the current you have to be careful the socket contact resistance. The way the article using switches and all will introduce error as we are matching to 1mV. I don't even measure on the fixture, I measure directly at the emitter of each transistor.
I literally do a quick read, then swap transistor to group them. Retest, swap until I get a set of 11 at close match. Let it rest to cool everything down. Then retest to verify. It's the tempco that is the main problem, not the test fixture design.
The problem is not about measuring the Vbe, the hard part is when you heat the transistors, the Vbe keep changing. YOu cannot mount the transistors onto a heatsink ( you have 60 of them and they have to be swap easily!!!). Also, For the current you have to be careful the socket contact resistance. The way the article using switches and all will introduce error as we are matching to 1mV. I don't even measure on the fixture, I measure directly at the emitter of each transistor.
I literally do a quick read, then swap transistor to group them. Retest, swap until I get a set of 11 at close match. Let it rest to cool everything down. Then retest to verify. It's the tempco that is the main problem, not the test fixture design.
IMO, you've gone above and beyond the call of duty with matching. FWIW, curve tracers have a low duty cycle to avoid burning things up. You could probably do the same thing, pulsing the base so the temperature doesn't change much. Check two at a time doing a differential measurement between them.
I did the differential matching, that was my first attempt, it really did not work out last time. The problem I discovered is not the method of measuring, it's the inability of keeping the temperature of them consistent and track without actually mount them on the heat sink. Remember we were joking about putting them in water? It is not practical to mount on heatsink to do matching as I have too many of them!!!! I need to match sets of 9 or more.
I finally experiment with lowering the collector voltage. I power the transistor up, then change the collector voltage from 0.6V to 20V and found the Vbe stay constant to within 1mV. So the last match was doing in collector voltage at 0.6V. We'll see whether I can get better than 2mV between the 9 transistors this time. My best so far was to 3mV matching.
I finally experiment with lowering the collector voltage. I power the transistor up, then change the collector voltage from 0.6V to 20V and found the Vbe stay constant to within 1mV. So the last match was doing in collector voltage at 0.6V. We'll see whether I can get better than 2mV between the 9 transistors this time. My best so far was to 3mV matching.
merlynski
Curmudgeon Electronicist
I actually meant to point you to Rod's article discussing transistor testing where he covers many of the same subjects you are dealing with in your process, rather than the test fixture itself. I was thinking along the same lines that ConradH was in Post 26. I have read (and appreciated) many of your posts and your engineering background, and am NOT criticizing, just providing a link to related material.
Sincerely, merlyn
BobbyN
Well-Known Member
I hope someone didn't already mention this and I missed it so just in case...
Try this simple circuit. First use a DVOM or??? to match the HFE, then use this circuit. Take the closest matching HFE's and use this circuit tester. Once power is applied leave it alone for a few minutes and/or until the voltages and temperatures stabilize. You'll see what I mean... I use it and it appears to work very well.
http://www.dragonflyalley.com/synth/images/TransistorMatching/ianFritz-transmat0011_144.pdf
Bob
Try this simple circuit. First use a DVOM or??? to match the HFE, then use this circuit. Take the closest matching HFE's and use this circuit tester. Once power is applied leave it alone for a few minutes and/or until the voltages and temperatures stabilize. You'll see what I mean... I use it and it appears to work very well.
http://www.dragonflyalley.com/synth/images/TransistorMatching/ianFritz-transmat0011_144.pdf
Bob
I put in the set that I matched onto the board, it is disappointing, 4mV variation. The problem is not only you match on the bench, you also have to worry how even the temperature distribution on the heatsink. I have to have the board mounted towards one end of the heatsink, the side of the heatsink where the transistors are close to the edge have lower Vbe because it's hotter than the other side. I can definitely feel the side of the heatsink where transistors are close to the edge hotter because the heat has no where to go at the edge of the heatsink, Just by one or two degrees different is enough to introduce 4mV of mismatch.
You almost have to pick the transistors with a few mV less Vbe to put to the end of the board as the last transistor is always a little cooler. Also, you have to put transistor with a few mV more Vbe at placed that is going to heat up more.
Matching Vbe is a lot more complicated than people imagine. It's not about the circuit of the tester, using winestone bridge, differential voltage measuring all sound good, but it's not the problem of measuring, it's everything together.
You almost have to pick the transistors with a few mV less Vbe to put to the end of the board as the last transistor is always a little cooler. Also, you have to put transistor with a few mV more Vbe at placed that is going to heat up more.
Matching Vbe is a lot more complicated than people imagine. It's not about the circuit of the tester, using winestone bridge, differential voltage measuring all sound good, but it's not the problem of measuring, it's everything together.
I am just anal, If I get into it, I try to go all the way. I guess I failed. Ha ha, I am retired, I have time!!!
It is more forgiving in my case using 0.27ohm emitter resistors. Doug Self advocates using 0.1ohm emitter resistors and run about 250mA per output pairs. That becomes very critical if you have 4mV difference in Vbe between adjacent transistors. On the side note, do you know Bob Cordell and Doug Self got into quite a hot debate in running such high current and using such low value emitter resistors?
It is more forgiving in my case using 0.27ohm emitter resistors. Doug Self advocates using 0.1ohm emitter resistors and run about 250mA per output pairs. That becomes very critical if you have 4mV difference in Vbe between adjacent transistors. On the side note, do you know Bob Cordell and Doug Self got into quite a hot debate in running such high current and using such low value emitter resistors?
Just an idea...
Mount a batch of say 6 to 10 transistor on a large heat sink. Hook up all transistor collectors to one end of a 12V supply. Each transistor has its emitter hooked up to other end of power supply through its own 1R 3W 1% . Then connect each base to the output of individual opamps. One input of the opamps is connected to each transistor's emitter while the other input is connected to a common reference voltage source . This in terms make each transistor's emitter current programmable to say 200mA each. To measure the Vbe just turn on the 12V supply and measure each transistor. Since the current through each transistor is the same( or within 2% due to tolerance of the emitter resistors) and the power dissipation of each transistor will be the same, the heat up rate of each transistor will be about the same. Of cause it is based on the assumptions that the DUTs are properly mounted on the heat sink and the heat transfer from the dies to the heat sink are pretty close. Sorry for my English and I did not attach a schematic, however I guess you guy can figure it out
Mount a batch of say 6 to 10 transistor on a large heat sink. Hook up all transistor collectors to one end of a 12V supply. Each transistor has its emitter hooked up to other end of power supply through its own 1R 3W 1% . Then connect each base to the output of individual opamps. One input of the opamps is connected to each transistor's emitter while the other input is connected to a common reference voltage source . This in terms make each transistor's emitter current programmable to say 200mA each. To measure the Vbe just turn on the 12V supply and measure each transistor. Since the current through each transistor is the same( or within 2% due to tolerance of the emitter resistors) and the power dissipation of each transistor will be the same, the heat up rate of each transistor will be about the same. Of cause it is based on the assumptions that the DUTs are properly mounted on the heat sink and the heat transfer from the dies to the heat sink are pretty close. Sorry for my English and I did not attach a schematic, however I guess you guy can figure it out
I feel that you may be over-thinking this a bit. When you look at commercial power amps using paralleled output devices they are invariably mounted on the same heatsink, with their respective emitter resistors mounted in freespace (raised up above the circuit board) to provide natural cooling. The driver transistor for each PNP/NPN side is located centrally on the same heatsink to provide overall temperature stabilisation. This seems to work very well with both class AB and class A amplifiers (the main difference being bias levels).
With the high bias values you are proposing to use, even with variations in effective gain balance between the output devices, as none of them will ever approach their low bias cutoff point, they will all still be operating within their linear operating band and the slight gain differences should have little effect on overall performance. As has been mentioned before, with careful heatsink mounting, Individual variances in junction temperature should balance out through combined heating across the heatsink.
With the high bias values you are proposing to use, even with variations in effective gain balance between the output devices, as none of them will ever approach their low bias cutoff point, they will all still be operating within their linear operating band and the slight gain differences should have little effect on overall performance. As has been mentioned before, with careful heatsink mounting, Individual variances in junction temperature should balance out through combined heating across the heatsink.
