Transistor selection for minimal thermal drift?

Thread Starter

B Scott

Joined Jul 23, 2013
9
I am using a current mirror to control current in a number of locations on a circuit board. I have tested various types of transistors and found that the variation between transistors from the same lot is minimal, so that is not a concern. What is a concern is the change in current with temperature, as there may be differences in temperature on different locations across the board. I've found that the variation in this regard from one transistor type to another is quite significant. I need to know what transistor characteristics to look for for a transistor with minimal amount of thermal drift (in an NPN current mirror application).

Thanks,
BS
 

#12

Joined Nov 30, 2010
18,224
I'm afraid that minimizing thermal changes is always done by keeping the temperature of more than one transistor equal to each other. The best I can do is point you to a chip.
 

Attachments

crutschow

Joined Mar 14, 2008
34,285
All transistors have nearly the same drift with temperature so the trick is to keep them both at the same temperature. Using a dual transistor in one package is a good approach. Otherwise attach them both to the same small heatsink with thermal grease.
 

Thread Starter

B Scott

Joined Jul 23, 2013
9
We're not talking about a couple of transistors, but 30 of them, and they are spread across 18 inches worth of circuit board. Keeping all transistors together was my first thought, but that is not possible. That is why I am focused on minimizing the problem through the use of transistors which exhibit the least thermal drift. Of the transistors I have tried, I have seen about a 3:1 range of drift between various transistor types. With identical part numbers that are binned according to gain (BC847A2 vs. BC847C2), the higher gain parts exhibit drift of about 1/2 that of the lower gain version.
 

Ron H

Joined Apr 14, 2005
7,063
BS, post a representative portion of your circuit.
From a stability standpoint, putting each transistor in its own feedback loop is the best option.
Failing that, a resistor in series with each emitter will partially swamp the thermal differences.
 
Last edited:

WBahn

Joined Mar 31, 2012
29,979
What is your mirror topology?

How are you dealing with the combined effect of the base currents from 30 transistors resulting in the mirrored current being considerably less than the programming current?

Are you using ballast resistors?
 

Thread Starter

B Scott

Joined Jul 23, 2013
9

I can't tell whether the attachment worked... will assume so.

Ron, can you explain further the mechanism by which the emitter resistor helps the thermal drift? I did do one test with various emitter resistance values (100, 1k, 10k ohm - same base input voltage), and saw the same drift with all.
 

Ron H

Joined Apr 14, 2005
7,063

I can't tell whether the attachment worked... will assume so.

Ron, can you explain further the mechanism by which the emitter resistor helps the thermal drift? I did do one test with various emitter resistance values (100, 1k, 10k ohm - same base input voltage), and saw the same drift with all.
Your attachment didn't work. You can test it prior to posting by clicking on the preview button.
 

kubeek

Joined Sep 20, 2005
5,794
Don´t you just need to use the transistors in pairs? You should be able to cascade current sink mirrors and current source mirrors to distribute the curent to many stages.
 

Thread Starter

B Scott

Joined Jul 23, 2013
9
Well, the attachment didn't work. How can I attach a file from my computer? Just checked the FAQ and find nothing on "attachments".
 

WBahn

Joined Mar 31, 2012
29,979
Your attachment didn't work (at least from what I am seeing).

In a basic current mirror, the transistor temperature is a function of a number of things, but the driver is the amount of power dissipated in the transistor which, in most instances, is dominated by the collector current and the collector-emitter voltage. Since the programming transistor is diode-connected, it will have a collect-emitter voltage of about 0.7V. But the mirror can have many volts for Vce and hence be dissipating many times the power, resulting in a higher temperature compared to the programming transistor even if all else is held equal. This is why you want to thermally couple the two (or however many) so that the heat in the mirror can heat up the programming transistor to about the same temperature. But this is not practical in your case.

At a constant collector current, the Vbe of an NPN transistor drops about 2mV/K. Around room temperature, the collector current doubles for an increase in Vbe of about 18mV. Thus, if the junction of your mirror transistor is just 9°C higher than the programming transistor, you will have about twice the current in it. Just a 1°C difference will give you a mismatch pushing 10%.

But if there are emitter-degeneration resistors (also known as ballast resistors), then as the voltage dropped across the resistors increases compared to Vbe, the current is dictated increasingly by the resistors and less by the transistor Ic/Vbe characteristic. While the resistance is also temperature dependent, it is not very much (say 50ppm/°C for metal film resistors and 200ppm/°C for carbon) and the effect is linear, as opposed to the exponential effect of the change in Vbe on collector current.

If you know the current you will be running at and the temperature coefficient of the resistors you are using, then you can size the ballast resistors to compensate for the change in Vbe, provided you can adequately thermally couple the mirror transistor to its ballast resistor, which should be practical in most instances. Of course, the resulting resistor value may or may not be practical.

But if you can drop several volts across the ballast resistor, then you can virtually eliminate the influence of the temperature dependence of Vbe. In fact, you can make a mirror that uses a small-signal transistor and a power transistor and end up with very good stability.
 

WBahn

Joined Mar 31, 2012
29,979
Thanks Kubeek - here it is
I assume the bottom of R7 is at ground?

You might want to put a resistor in the collector path of Q3 and size it so that the Vce of Q3 is approximately what you expect the Vce of the mirror transistors to be.

470Ω ballast resistors should work pretty well. What current levels are you typically running?

Could you provide your data showing that you have a temperature mismatch problem and not just a component mismatch problem?
 

Thread Starter

B Scott

Joined Jul 23, 2013
9
WBahn - I'll have to read your first post later when I have time to digest what you're saying.

Yes, bottom of R7 is ground.

"Could you provide your data showing that you have a temperature mismatch problem..."

If I understand your question correctly: I don't have a temperature mismatch problem, I have a potential temperature differential problem where the transistors in different locations may see enough temperature difference from one side of the board to the other to cause a current mis-match problem. So I am trying to minimize the variation between "mirror" transistors with temperature.

Sorry Ron, Vref ranges from .03 to .6 volts.

BS
 

WBahn

Joined Mar 31, 2012
29,979
WBahn - I'll have to read your first post later when I have time to digest what you're saying.

Yes, bottom of R7 is ground.

"Could you provide your data showing that you have a temperature mismatch problem..."

If I understand your question correctly: I don't have a temperature mismatch problem, I have a potential temperature differential problem where the transistors in different locations may see enough temperature difference from one side of the board to the other to cause a current mis-match problem. So I am trying to minimize the variation between "mirror" transistors with temperature.

Sorry Ron, Vref ranges from .03 to .6 volts.

BS
You've confused me totally. Here you are saying you are dealing with a potential problem by in an earlier post you stated, "I did do one test with various emitter resistance values (100, 1k, 10k ohm - same base input voltage), and saw the same drift with all."

What drift did you see? Same drift as what? As the other resistor values or compared to no resistor (basic current mirror)? What is your data? How much current mismatch can you tolerate.

Temperature differentials across the board ARE a temperature mismatches. It can be characterized two ways. The currents in the mirrors are not the current that is being programmed, or the currents in the mirrors are not the same as each other. Which is more important? One way to get at this is to ask yourself which is worse - that the current, on average, in the mirror is within 10% of the programmed current but that there is only a 1% range between min and max, or that the current, on average, is within 1% of the programmed current but there is a 10% range between min and max.
 

Ron H

Joined Apr 14, 2005
7,063
Why don't you tell us what your circuit is supposed to do. Sometimes, when we get the big picture, someone will come up with another way of doing the same thing, but without the problems.
 

Thread Starter

B Scott

Joined Jul 23, 2013
9
What drift did you see? Same drift as what? As the other resistor values or compared to no resistor (basic current mirror)?
With a set base voltage, I looked at the emitter voltage change with temperature. With the 3 different emitter load resistors listed, that change with temperature was very much the same, one to the other. Or in other words, the emitter voltage drift due to temperature did not change across that range of emitter load resistors.
 
Top