[Stefan Heinzmann]

Hi all,

I'm wondering what the reason may be that there seem to be no
transistors incorporating a means of sensing the die temperature (i.e. a
diode). Wouldn't there be a lot of advantages to this, given the
difficulty of mounting a sensor device in close physical proximity to a
conventional power transistor? A lot of circuits (bias temperature
compensation, overtemperature protection) rely on such a sensor device.

I know that there are MOSFETs with integrated protection circuitry, but
that's not useful for bias compensation, is it? In many cases such
MOSFETs seem to be built for switching applications anyway.

Well, if you can integrate complete protection circuitry with a
transistor, integrating just a diode would be a piece of cake, wouldn't it?

 

[Jim Thompson]

Hi all,

I'm wondering what the reason may be that there seem to be no
transistors incorporating a means of sensing the die temperature (i.e. a
diode). Wouldn't there be a lot of advantages to this, given the
difficulty of mounting a sensor device in close physical proximity to a
conventional power transistor? A lot of circuits (bias temperature
compensation, overtemperature protection) rely on such a sensor device.

I know that there are MOSFETs with integrated protection circuitry, but
that's not useful for bias compensation, is it? In many cases such
MOSFETs seem to be built for switching applications anyway.

Well, if you can integrate complete protection circuitry with a
transistor, integrating just a diode would be a piece of cake, wouldn't it?

It's done on microchips all the time.

If you want to know how hot a discrete transistor is, examine its B-E
voltage at a fixed low current in between power pulses.

...Jim Thompson

 

[Stefan Heinzmann]

It's done on microchips all the time.

Yes, I know. That's exactly why I think it should be possible on power
discretes.

If you want to know how hot a discrete transistor is, examine its B-E
voltage at a fixed low current in between power pulses.

And if you'd want to do it continuously, i.e. for bias compensation and
overload protection in a linear amplifier?

 

[Jim Thompson]

Yes, I know. That's exactly why I think it should be possible on power
discretes.


And if you'd want to do it continuously, i.e. for bias compensation and
overload protection in a linear amplifier?

It's quite easily do-able with external circuitry, but I can't
presently divulge how... I'm considering if it's worthy of patent
application expense. Makes it a piece-a-cake to produce stable,
low-distortion A-B bias.

National used to make a "protected power transistor", the LM195 that
had thermal sensing built-in. (It was just a mask-set variation on
the LM117/217/317 regulator chip.)

...Jim Thompson

 

[Stefan Heinzmann]

It's quite easily do-able with external circuitry, but I can't
presently divulge how... I'm considering if it's worthy of patent
application expense. Makes it a piece-a-cake to produce stable,
low-distortion A-B bias.

National used to make a "protected power transistor", the LM195 that
had thermal sensing built-in. (It was just a mask-set variation on
the LM117/217/317 regulator chip.)

Yes, I remember that. That solves the protection problem only. ST has a
number of protected MOSFETs that supposedly can also be used in linear
mode, but I have seen nothing that would allow me to use the measured
temperature in external circuitry. It is as if you either get nothing at
all or a complete all singing and dancing automatic protection system.
So I guess I still wonder why there seems to be no market in between.

Maybe there's no market any more after you publish your cunning circuit.

 

[Paul Burridge]

Hi all,

I'm wondering what the reason may be that there seem to be no
transistors incorporating a means of sensing the die temperature (i.e. a
diode). Wouldn't there be a lot of advantages to this, given the
difficulty of mounting a sensor device in close physical proximity to a
conventional power transistor? A lot of circuits (bias temperature
compensation, overtemperature protection) rely on such a sensor device.

I know that there are MOSFETs with integrated protection circuitry, but
that's not useful for bias compensation, is it? In many cases such
MOSFETs seem to be built for switching applications anyway.

Well, if you can integrate complete protection circuitry with a
transistor, integrating just a diode would be a piece of cake, wouldn't it?

I should imagine so. It's such a good idea I can't imagine why it's
not been done before - at least for power BJTs commonly used in amp
outputs. I've never had much luck with external compensating diodes.
There's too much of a thermal and time lag no matter how closely
they're coupled to the power device from what I've experienced. The
TO220 package is pretty bad in this respect, I've noticed. :-(

 

[Jim Thompson]

I should imagine so. It's such a good idea I can't imagine why it's
not been done before - at least for power BJTs commonly used in amp
outputs. I've never had much luck with external compensating diodes.
There's too much of a thermal and time lag no matter how closely
they're coupled to the power device from what I've experienced. The
TO220 package is pretty bad in this respect, I've noticed. :-(

And would you know what to do with them if there were diodes in the
same package? I doubt it ;-)

...Jim Thompson

 

[Winfield Hill]

Jim Thompson wrote...

National used to make a "protected power transistor", the LM195 that
had thermal sensing built-in. (It was just a mask-set variation on
the LM117/217/317 regulator chip.)

There are many close simularities between the two schematics, but
there are many differences as well, re-routed wires, added parts,
changed part values. Do these differences represent refinements
to the older LM317 design?

Thanks,
- Win

whill_at_picovolt-dot-com

 

[John Fields]

And would you know what to do with them if there were diodes in the
same package? I doubt it ;-)

 

[Jim Thompson]

Jim Thompson wrote...

There are many close simularities between the two schematics, but
there are many differences as well, re-routed wires, added parts,
changed part values. Do these differences represent refinements
to the older LM317 design?

Thanks,
- Win

whill_at_picovolt-dot-com

I was retained by ICE (Integrated Circuit Engineering) around 1980 to
analyse the LM195/LM117. At that time the circuits were virtually
identical.

I have a very faded copy that I will attempt to restore by scanning
enhancement. If I succeed I'll post it to my website.

...Jim Thompson

 

[Jim Thompson]

Naaah! I just find Burridge terribly tiring... talks a great tune,
but can't find his own asshole using both his hands ;-)

...Jim Thompson

 

[John Fields]

Naaah! I just find Burridge terribly tiring... talks a great tune,
but can't find his own asshole using both his hands ;-)

---
Maybe, but he's getting better, don't you think?

At least he had the chutzpah to put his ignorance on the line and ask
for help with his field strength meter PCB, no?

And then, when he took the advice he got and came back with the report
that it worked "out of the box" and thanked us all for our help, I
thought that was pretty cool...

 

[Paul Burridge]

And would you know what to do with them if there were diodes in the
same package? I doubt it ;-)

That part's easy. But if they *did* start making transistors with 5
leads that might prove a bit more confusing.

 

[John Fields]

That part's easy. But if they *did* start making transistors with 5
leads that might prove a bit more confusing.

 

[Paul Burridge]

Assuming you could somehow allay the confusion by reading the data sheet
and identifing the diode's pinouts how would you _do_ it?

See AoE p.93.

 

[Jim Thompson]

See AoE p.93.

Bwahahahahahahahahahahahahahaha ;-)

...Jim Thompson

 

[Paul Burridge]

Bwahahahahahahahahahahahahahaha ;-)

Too simplistic for you? I assumed John just wanted some indication
that I was familiar with the principle behind using diodes for thermal
compensation.

 

[Jim Thompson]

Too simplistic for you? I assumed John just wanted some indication
that I was familiar with the principle behind using diodes for thermal
compensation.

Except the scheme doesn't work very well. It has a fundamental flaw.

...Jim Thompson

 

[Ralph & Diane Barone]

---
Blah, blah, blah.

Assuming you could somehow allay the confusion by reading the data sheet
and identifing the diode's pinouts how would you _do_ it?

I missed the start of this thread, but Sanken makes transistors with built
in temperature sensing diodes in a 5 pin package. Check out page 2 of
http://www.allegromicro.com/skncatlg/semicond/pwrxtrs.pdf

 

[Paul Burridge]

Except the scheme doesn't work very well. It has a fundamental flaw.

Which is?