Design problem

Did not build these yet..the "response" was from Spice. Like i said elsewhere, i have used zeners to 185C and found leakage well under 20uA and have had no complaints at 204C at 60uA.

Now _THAT_ is very spiffy! What makes the high temp possible is the CMOS process, which i believe is a version of the Dielectric Isolation process; killing most of the leakages. All of the bipolar references have glitches in the I-V curves in various temperature regions, typically above 160C; some start exhibiting an "N" I-V curve where the peak current can grow so bad that the unit if on, could never turn on if power was interrupted when at high temp. That means, the actual circuit design of this beast (if it behaves) is rather different to say the least.

text -

Thanks!

Well, Spice says so, but i refuse to believe it as my experience suggests useful operation to a minimum of 185C and in certain cases to

204C; see my comments in other places.

collar

d

nt (Q3)

ly lower

ming to

o e

etc.

e

At what breakdown voltage? "Zeners" with breakdown voltage below about

5V actually breakdown by the Zener mechanism, while higher voltage - including the 6.8V part referred to in your original post - depend on avalanche multiplication. I've got no idea whether this makes any difference to the high temperature behaviour - except that the avalanche breakdown voltaage rises with temperature, while the Zener mechanism gives a breakdown voltage that falls with rising temperature

- but it seems unlikely that the two sorts of "zener" diodes will behave identically in all otehr respects.

-- Bill Sloman, Nijmegen

llar

(Q3)

lower

ng to

e"

ge

You haven't specified the breakdown voltage of the part that gave this

- remarkably low - leakage current.

You've seen a diode with less than 20uA leakage current at 185C. You'd expect roughly 2^16 - 65,536 - less leakage current at room temperature, or less than 0.3nA which is low.

Looking at a specific part

we can say that if you'd used a BZV85-C6V8 you'd have been promised less than 2uA at 4V at room temperature, which equates to less than

130mA at 185C. Higher voltage parts offer lower leakage currents - anything above 15V is guaranteed less than 50nA, which equates to 3.3mA at 185C, and the 50nA probably has more to do with the limitations of their test gear than the properties of the diodes, so the 70V parts may well do quite a bit better, maybe even down near your 20uA at 185C.

-- Bill Sloman, Nijmegen

Problem with FETs is that their leakage starts getting bad above

185C, so the usability goes down. STM used to make one that was decent at 185C and 500V, but now one has to test every part of a particular one in what is left.. Not very good idea to test in "reliability" that way..

As everyone should know, those leakage limits are from high speed testing and exist due to that limitation...high speed testing. The limits are orders of magnitude higher than reality in most cases.

That would be very good ! That's the way it should be of course, I would think.

I will have to watch for that then next time I buy a bunch of replacements. Almost forgot about those plastic versions ! Now I sorta remember them.

Hopefully changing the emmiter resistors to more suitable values is all that would be necessary, if at all necessary, to better match transistors in the circuit.

boB

The 50nA maximum leakage for BZV85s that break down above 15V is pretty clearly a reflection of the limitations of the test gear.

For the lower voltage parts, the tolerance on the leakage current is going to be in the same ball-park as the tolerance on the forward current at the knee - which is closer to a three-to-one range.

The current through the "zener" diode at any voltage is just some multiple of the low voltage leakage current and you can't get away with orders of magnitude variation on what it essentially your primary production parameter.

The manufacturers have to control chip area and doping levels fairly closely to get decent yields of diodes that fit their specifications, and an orders of magnitude uncertainty in leakage current isn't an option.

--
Bill Sloman, Nijmegen

Seems if i parallel transistors, then the lower current in each one results in lower gain.. ..So there is a conflict between low current (==power dissipated) and gain (== more current).

That's engineering. You change the design to favor not cooking the output transistor, but you may lose a little gain to do so. I want the thing to run longer, so I use a bigger battery, but now it weighs more. It produces heat so I include a fan, but now it's bulkier. etc.

Can you give some detail as to what you have in mind? It may not be as problematic as you may think.

Ed

** Power transistors like the MJ15024 are often paralleled to increase dissipation and max current ability - up to six devices in a group is not unusual. The current gain for a matched, parallel group is the SAME as for one device.

For equal sharing, devices need to be of the same type and brand plus ideally from the same batch. Surprisingly, matching Hfe values does not give equal sharing - matching the Vbe of each device at some Ic like 1 amp does.

Along with the use of emitter ballast resistors of say 0.47 ohms and NO resistors in the base feeds.

... Phil

1970s and

the

wire

=20

2N3773s and

in=20

2N3055G.

several=20

=20

lead=20

same=20

areas=20

earlier=20

not=20

I don't follow that one. As the collector currents go down so do the = base currents. The aggregates remain about the same. Try it any which way = you like. The gain reductions are small to completely avoidable.

and=20

70s and 1/4 the

wire

=A02N3773s and

n 055G. n

as

several

ead

ame

n areas

ier

the

ot

ase

y you

d

i_C =3D I_s*exp(v_BE/V_T)

This is a non-linear curve. Transconductance (gm) is the small signal slope at a bias point on the curve. More current steepens the slope, aka higher gm. It is why "multipliers" like the MC1496 (gilbert cells) work.

=A01970s and

about 1/4 the

Al wire

legacy

=A02N3773s and

ones in

=A02N3055G.

doubt.

1990's on

lost

good as

though several

pack

=A0Because lead

exact same

termination areas

earlier

=A0as the

may not

one

the base

way you

and

And in some real transistors other effects result in modestly changing (

and

several

Beta nonlinearity can be quite a bit worse than that.

Cheers

Phil Hobbs