BJT switch temperature compensation

[likephysics]

Is there a way to minimize the temperature dependence of VBE.
I am trying to design a BJT switch, that will turn on at a particular voltage(43v).
ckt attached.

The base sees 0.65v when Vcc is 43v or more (resistor divider bias). This turns on the BJT.
It should be off when Vcc is 42V or less.

I tried adding Re, but that didn't help much during simulation.

 

[jsgruszynski]

The dominate temperature effect is BJT drift is from Vbe variation (with temperature) which has an exponential effect on current. For this reason the standard method of dealing with it involves using matched pairs of BJTs with the right kind of circuit design that can use the matching of Vbe values in current mirror pairs to achieve compensation.

See diff-pair matching layout techniques and band gap references for examples of doing this kind of matched pair for temperature compensation.

Whether this makes sense for power devices is probably dubious because you normally only worry about temperature drift for linear operation and not for cut-off/saturation mode switching. But if you are trying to operate linear these may gives you some ideas.

 

[likephysics]

I already looked into diff pair matching and current source matching. They are different compared to the switching application.

Does Mosfet gate threshold also change with temperature?
Couldn't find any info in datasheets- irf540 etc.

 

[uart]

If all you want is a simple circuit to turn on at about 42 to 43 volts then try something like my attachment. I've left it an open collector output.

 

[likephysics]

If all you want is a simple circuit to turn on at about 42 to 43 volts then try something like my attachment. I've left it an open collector output.

wouldn't Vbe of NPN still change with temperature?

The emitter of PNP is at 33V. So the 47K and 4.7K, divide 33V, correct.

 

[uart]

wouldn't Vbe of NPN still change with temperature?

Yes it will, however the switching point is now primarily determined by the zener voltage. This circuit will switch more sharply and at a more consistent voltage level than the original.

Another advantage is that the two transistor solution provides an output of the correct phase to apply a small amount of positive feedback (not shown) if you want to make it switch even more sharply.

The emitter of PNP is at 33V. So the 47K and 4.7K, divide 33V, correct.

Yes it gives about 3 volts with a Thevenin resistance of about 4.3k, however these values are not critical and have little to no effect on the switching threshold.

 

[likephysics]

Another advantage is that the two transistor solution provides an output of the correct phase to apply a small amount of positive feedback (not shown) if you want to make it switch even more sharply.

How do I add positive feedback?

 

[uart]

It depends a bit on what type of pull-up you want to use on the open-collector output. Eg, pull up to +5 supply, to +12 supply or perhaps pull up all the way to the main supply (42V or whatever it is).

Say we assume a pull up to +5V at the open collector output. One example would be to split R2 (33k) into two parts, say 27k and then 5k to ground. Then apply a fairly high resistance feedback resistor of around 150k from the open collector output back to the junction of the 27k and the 5k resistors.