Matching Vbe of power transistor

[Alan0354]

I have been trying to match output transistors for Vbe as I use 9 pairs in the OPS stage. It is much harder than it sounds as tempco of Vbe is -2mV/deg C. You handle it, it start changing. You breath on it, it starts changing. People suggested immerse in water to do it, it's way too messy. So this is a way I come up with as see what you guys think.

See attached immage, I use the proto plug in type board, I just ordered 60 of the 51ohm 1% 2W 100ppm/deg C resistor. I am planning to put in all 60 of the power transistors in like the circuit shown. I am drawing 100mA through each transistor like in my amp. I am going to let it sit for say 1 hour to let the temperature settle at higher temperature. I then measure the voltage different between the emitter pin of two different transistors and look for the difference. I just keep going and note down the difference in voltage between two emitters.

Notice I stress to measure on the leg of the transistor as I don't trust the interface between the leg and the plug in prototype board, there might be a drop across the contact and it's not the same for every transistors. I think I can trust the base contact of the prototype board as the base current is only 1mA, it's not going to have much voltage drop across the contact even thought it's not a very good contact.

The good point is all the transistor are in the exact environment, so even the room temp change a degree or two during measuring, all transistors subject to the change at the same time and it will null out the difference.

This might sound tedious, but if you think of the alternative, dumping in water and measure one by one?!!, you have to use a thermometer to measure the temperature of the water, stabilize the temperature every time, then the drying and cleaning. The water is going to heat up gradually when you keep doing it, all it takes is 1deg change and you ruin your measurement.

Let me know what you think. I just ordered 60 resistors and a 5V 10A supply( $13 only free shipping, very cheap).

Notice I put 0.1u ceramic cap in the circuit, I think the transistor might oscillate without them. I have been doing matching, so far I have no luck at all.

 

[BinaryMike]

Are you planning to eliminate emitter ballast resistors in your output stage?

 

[Alan0354]

Are you planning to eliminate emitter ballast resistors in your output stage?

Oh no!!! I am using 0.22ohm emitter resistor and run 100mA to get Oliver's condition. I am running 9 stages, so I want to be careful so no one transistor carry a big load.

 

[ConradH]

So, didn't like my water suggestion.

I think the goal is load sharing under high currents, less so at idle. I've got a Crown note on how they measured theirs, and Vce=4V and Ic=3A. Or is the amp class A and that 100 mA each is the operating current?

 

[Alan0354]

It's not a class A amp, I want it to be high bias class AB, 9 stages gives me about 1A( well I crank it up a little), that gives me 8W of class A with 4ohm speaker or 16W with 8ohm speaker. I have no intention to run more than 40V rail, so each transistor is going to be about 1A max current only.

From my understanding, Vbe relation should be close regardless of the current as long as it's the same current and as long as you don't push it to beta drooping region. I would not come close to the drooping level with 9 stages and rail under 50V.

It's too inconvenient to use water, also you can only do one or two at a time. You keep sticking the hands in the water and the fact the the transistor and resistors generate heat that will slowly warm up the water will create error in reading. My point is run all 60 of them at the same time, give it time to settle and I am looking for difference in Vbe, not the absolute Vbe. The Vbe might change over time, but if they are in the same environment and power the same way, they track, so I can take my time to do the measurement. I can even repeat the measurement over and over to confirm the result.

Another thought is maybe I should raise the collector voltage and make the transistor heat up above the room temperature, with more heat generated, effect of the room is going to have less effect as the heat source is going to be more dominant inside the transistor. Say if I raise the collector voltage to 10V, I would dissipate 1W/transistor, without heatsink, even the TO264 can get quite toasty. Or even 15V and make it hot. That should eliminate the room temperature changes.

Just thinking out loud. I really want to run 100 to 120mA per stage. I spent $350 for a new chassis that can dissipate more power than the one I have, I want to run it with 0.22ohm resistor and meet the Oliver's condition described in Cordell's book. I just sunk another $40 on 60 of the 2W 1% 100ppm/deg C resistors and a 5V 10A supply all to find a match set.

 

[Alan0354]

I spent some time building this as I expect to use this more than twice in the future, took me a few hours to get to this point:





I am still waiting for the resistors and still need to wire up the rest. I decided to put all 60 in the same row, equal spacing ( well almost, there is a slight difference between two proto board) so they all heat up the same way, affect by the next door neighbor the same way except the two end piece.

I am going to leave it on for say 1hr. to let the temperature stablize and then measure the difference between the voltages of emitters ( measure differential voltage). This should be very close to how it is in real amplifier.

 

[Alan0354]

I finished building the tester, but I don't think it will work. I have my OPS( output section) pcb built, test to a certain point and power up to measure the Vbe. I did two test:

1) Without mounting on the heatsink, I adjust the rail to about +/-7.5V to lower dissipation and adjust to about 1A total current. I recorded the Vbe of both stone cold and ran for 15minutes, I see big difference in Vbe particular when warm, it ranging from 35mV to 18mV across the 9 emitter resistors. It goes into current hogging as the one that is slightly lower Vbe will take on more current, in turn getting a little hotter, which lower the Vbe and take on even more current. The difference is Vbe when cold is not that big.

2) I put it onto the big heatsink, cranked the rail to +/-30V and draw 1.1A. I let it sat for 30mins until it got quite hot( still touchable, can handle with hands, but I would not put my hands on indefinitely). Sadly, the current dropped to 0.9A which is like 20% drop when hot. I don't like it, it's not good enough. But back to the point, I measured all 9 Vbe, they were closely match. From 19.5mV to 21.5mV. Only about 2mV difference. I don't think I can do better than this.

Long story short, the test has to be done with heatsink to buffer the temperature. Or else, the one with lower Vbe get hotter and hock more current and further lower the Vbe. There is no correlation between Vbe measurement between with and without heatsink.

Let me know how you feel. I think I just wasted two days of time and money buying the extra proto board and 5V 10A power supply. Conard might be right after all !!! Only way without the heatsink is dip in water, on top, you have to stir the water to keep the temperature even. That's too much trouble. Amy other idea?

 

[gslikker]

So as I understand your last test is having the actual amp circuit total circuit bias go down 20% between cold and a reasonable temperature when "hot" ?

It would give me a safe feeling, as a prevention from thermal runaway, so actually your results are quite good?

 

[Alan0354]

So as I understand your last test is having the actual amp circuit total circuit bias go down 20% between cold and a reasonable temperature when "hot" ?

It would give me a safe feeling, as a prevention from thermal runaway, so actually your results are quite good?

Thanks, My first pcb run stay within 10%. Adjusting R16 will improve it. The new version is a completely different layout, I was afraid of that will happen. I made the mistake putting this R16 on the bottom side and is a SMD, I already removed the board off the heatsink to change it. I am going to put a trim pot on the top side to adjust.

 

[elnaldo]

Which circuit is so sensitive to this differences? Is it a DIY Amp? I don't think a factory does that accurate matching. I think you should stablish a tolerance and stay between those +- points.

 

[Alan0354]

Which circuit is so sensitive to this differences? Is it a DIY Amp? I don't think a factory does that accurate matching. I think you should stablish a tolerance and stay between those +- points.

Yes, it is my own design and build with my own layout pcb and all, from scratch.

You mean the amps on the market are not this accurate? I have no idea, never look at one before. You mean they drift even more? I am hitting the wall right now, I don't know what is supposed to be good enough. I thought I need to keep it at least down to within 10% or better.

It's the Vbe drift with temperature of the power transistors. I put in a lot of effort to compensate. This is my schematic of the OPS pcb. You see I use opposite polarity transistor on the 3EF output stage. I bolted the PNP predriver and the NPN driver transistors together ( the bottom side is NPN to PNP driver) so the drift cancel out. Then I bolt the bias spreader Q15 on top one of the big power transistor to compensate. So it's a two stage compensation. The pre-driver and the driver at bolted on a separate heat sink so it's not affected by the heat from the big transistors.

I had to experiment with different transistors to match last time. But this is a different layout, I might have to do it again to improve it.

 

[elnaldo]

I"m sure that comercial amps are not that accurate. You need to stablish a tolerance and move within that range.

Conrad H mentioned a crown paper, I think that paper specifies the different ranges to clasify the transistors.

I can't tell about your circuit, I'm not an expert . But I know that a good design can handle the components tolerances.

 

[Alan0354]

I"m sure that comercial amps are not that accurate. You need to stablish a tolerance and move within that range.

Conrad H mentioned a crown paper, I think that paper specifies the different ranges to clasify the transistors.

I can't tell about your circuit, I'm not an expert . But I know that a good design can handle the components tolerances.

I remember I saw Conrad referred to some crown paper, I do not remember where. I mainly want to know what is the % in the name brand amps.

 

[elnaldo]

I remember I saw Conrad referred to some crown paper, I do not remember where. I mainly want to know what is the % in the name brand amps.

It's like a 20%

 

[Alan0354]

I tried the setup, it really did not work that well so far. I let it warmed up for 30mins, measured and wrote down the differential voltage using one transistor as reference. I then grouped them and chose out 2 sets of 9 transistors that matched to within 3mV. I packed it away in separate bags. But out of curiosity, I put them back and tried again, it's different!!!

I already soldered in a set from the former way of matching, that is turn it on and measure right away. The set matched within 3mV in the real circuit with heat sink. Maybe that's a better way doing it cold after all.

 

[elnaldo]

I think 3mV is not that relevant for this application. I'd just install the matched sets you have now, and enjoy the amp or spend the time tweaking other things.

 

[RocketMonkey]

Mostly just to satisfy my curiosity I'd like to know more about how you perform the water test? Are you testing the components wet or just synchronizing the temperature with water and then testing? I might have an idea.

 

[Alan0354]

Mostly just to satisfy my curiosity I'd like to know more about how you perform the water test? Are you testing the components wet or just synchronizing the temperature with water and then testing? I might have an idea.

I don't do water test at all. I try to avoid thinking about it so far. But I am really out of ideas. On this run, I have an unforeseen problem. The stupid emitter resistor (50ohm) get so hot, I think it affect the transistors. The sickening thing is they all plugged onto the protoboard, contact is not consistent between transistors and the heat transfer is not consistent between them, all it takes is one or two degree variation, it'll destroy the measurement.

I would like to hear your idea.

 

[Alan0354]

I think 3mV is not that relevant for this application. I'd just install the matched sets you have now, and enjoy the amp or spend the time tweaking other things.

All else fails, that's what I am going to do. The first set I built on the pcb passed all the tests and matched with flying color. They match within 3mV on real heatsink, real environment.

I just want to exhaust all possibility first for now. I really want to match to 3mV or better because I want to lower the emitter resistors further which make matching that much more critical.

 

[Alan0354]

$hit!!! I find myself looking at turkey roasting pan on Amazon today!!!! It's all your fault Conrad!!!!

I am running out of ideas, taking a swim might be the way!!!

I shouldn't make it on a plank of wood, I can see it keep trying to pop up when I try to measure, stupid wood floats!!

AND I thought I was so damn smart to put everything in a long piece for easy measurement, now I have to find a way to cut in to fit in the turkey pan!!!

If anyone has better idea, please, I am all ears.