More battery charger, voltage shunt reference.

Nice Thanks. Any guess on a max power/ current when zenering a transistor? 100mW? Is there a list/ table web page with the temperature coef. of LED's? The little I know is that it varies by color/ material. I see numbers between -3 to -5 mV/C Maybe better to measure my own numbers with the random collection of leds I have.

George H.

Of course if you use a forward biassed red LED you've got a voltage drop of about 1.6v at 20mA

Multiply this up to 13.5V and you've got the -18mV/C temperature coefficient that Fred wants.

There are tolerances all over the place, and with a string of more LEDs and the occasional regular diode you could have enough parts to swap around to get closer to whatever it is you end up actually needing.

I should have thought of this last night, but my subconscious took a while to cone up with it - I work up with the realisation that I'd missed that particualr trick

-- Bill Sloman, Sydney

Of course if you use a forward biassed red LED you've got a voltage drop of about 1.6v at 20mA

Multiply this up to 13.5V and you've got the -18mV/C temperature coefficient that Fred wants.

There are tolerances all over the place, and with a string of more LEDs and the occasional regular diode you could have enough parts to swap around to get closer to whatever it is you end up actually needing.

I should have thought of this last night, but my subconscious took a while to come up with it - I woke up with the realisation that I'd missed that particular trick.

This has already been discussed further up this thread (or one of the related threads). .

Probably because you focus on being obnoxious to the genuinely valuable contributors here; must take all your time up, leaving none for electronics. --

"Andrey Semyonovitch really was rather stupid; he attached himself to the progressive cause and 'our younger generation' from enthusiasm. He was one of the numerous and varied legion of dullards, of half-animate abortions, conceited, half-educated coxcombs who attach themselves to the idea most in fashion, only to vulgarise it and who caricature every cause they serve, however sincerely."

- Fyodor Dostoevsky

I've said a bit about it elsewhere. Cursitor Doom won't have paid enough attention to have noticed who mentioned it.

Cursitor Doom must think that he is one of "the genuinely valuable contributors here". I'm rarely rude to the ones who do provide valuable contributions - Phil Allsion is an exception, but I do say nice things about his - rather rare - useful contributions.

Cursitor Doom doesn't seem to understand anything about electronics at all. It's possible that he is too busy being a right-wing nitwit to spend any time on displaying any electronic expertise he might have, but the null hypothesis does fit the data.

If he did, he might know about the way "obvious" ideas can take a while to surface. Even John Larkin, who isn't all that creative, has mentioned this.

Yes, the Hfe drop after reverse b-e breakdown is strongly dose dependent. After prolonged breakdown and at higher currents you would clearly see a gain drop but it eventually levels off. I understand that it also recovers to varying extents with time and temperature.

Do not put those transistors back into use in preamp front ends! Reports are that zenering the base makes them much noisier.

piglet

I think anyone (mad enough) to want to pursue this idea today would be better off using a separate diode or transistor junction to TC compensate the zener instead of using the C-B junction of the same B-E breakdown transistor.

Did you notice the bit about "selected for low reverse beta"? I think that hides a big potential flaw: reverse gain could drastically collapse the reference voltage or turn the device into a very messy negative resistance oscillator.

piglet

Yes, there was a particular transistor process that NatSemi was calling for (one of theirs, naturally). An alternate nowadays would be a dual transistor (for temperature coupling), which could also be a Vbe multiplier to adjust the tempco.

In quantity, thermistors would be a good alternative (but I don't keep an assortment of those in my stock drawers).

Unless they're monolithic, dual transistors have no better coupling than separate adjacent devices. We've been through that a few times round here.

Cheers

Phil Hobbs

There's no transistor action when the base-emitter junction is avalanching, and the collector-emitter junction is forward biassed.

Reversed transistors do act as transistors - the gain isn't great but saturation voltage can be very low, and I've exploited that in one design (for a very cheap PWM multiplier, which never made it into production - it wasn't my idea but I did get it to work pretty well) - but there the base-emitter junction is necessarily biased below breakdown.

Quite what the low reverse beta might be selecting for isn't obvious to me, but people who have abetter idea of exactly what is going on inside a working transistor may be able to work it out

The chore to be accomplished here is ambient temperature sensing; we want them to match the battery.

Ideally, a battery post would be adjacent to this.

Why? At least those two diodes would be in the same package on the same bit of silicon, and presumably at the same temperature. You'd be dissipating more heat in the base-emitter junction than the collector emitter junction, but both are thin and one lies on top of the other. 44uA at about 7V is about 0.3mW, which isn't much - about 0.1C in a TO92 package.

Batteries are susceptible to self-heating - ideally they'd be in the battery, but with six cells you'd have to pick the right one.

Batteries don't self-heat when they are being trickle-charged at close to full charge, which is the only time it would matter, so it wouldn't be worth putting in too much effort to get that right.

I just had to experiment! I took a random BC549C, forward Hfe 587, reverse Hfe 14. I wired it collector to 0V, base open, emitter to +20V via a variable resistor. With a 10nF capacitor and CRO across C-E. I found it breaks into oscillation above 600uA current.

The waveform was sawtooth, just like classic relaxation oscillator. Amplitude varied with breakdown current, from approx 200mV at onset to

I think this shows that even with B-E junction in breakdown some transistor action does indeed occur. I think that is why the 1960s researchers were careful to require a reverse hfe below one.

piglet

Why wouldn't it? There's a mondo big E field there.

Another approach is to frequency-compensate it via the base. ;)

You can sometimes make diode-connected transistors (CB shorted) oscillate in forward bias--it's the world's simplest feedback amplifier.

Cheers

Phil Hobbs

There are other explanations - at least one of which make more sense to me.

When the transistors are connected either way around.

I don't think that they did, and they probably would have had lot of trouble finding any transistor that qualified.

But what is the mechanism?

Looks like you're exhausting tangential subject matter to avoid having to produce a working temperature compensated reference for the charger.

It's trivial, and I've already done it. Since you can't work out for yourself, I'll plug kjw's design int LTSpice, stick the red LED reference which I've already identified as doing what you eventually asked for, and fill in the details. I'll probably use half an LT1013 where he used a long tail pair.

That means that 2mV/C at the LED probably isn't going to be multipiied up the to the exactly -18mV/C that you think that the battery needs, but it is probably going to get close enough that it wouldn't be worth getting it more exact.

Sounds like a lot of hand waving to me.

You do seem to be the resident expert in that.