Diffusion in diodes

That has been discussed here before. The Spice diode model is inadequate (aka sucks :-)

I've worked on and off trying to make such a model, but have always hit snags :-( ...Jim Thompson

Here are some 1N914 turn-on waveforms. The diode is along a 50 ohm line, line-to-ground, midway between the 50 ohm pulse generator and the 50 ohm scope.

Fuggly. So I guess my PD model will probably have to have a lumped RC transmission line in it someplace.

Some PDs speed up amazingly in the span of a few volts right up near breakdown. Seems like you can deplete the epi and even most of the contact region.

Cheers

Phil Hobbs

We have noticed continuous speedup in silicon PDs as voltage increases, well past the point where the capacitance quits changing significantly.

We had some VCSEL laser diodes that seemed to have an inductor/PIN sort of effect. A current step, applied cold, would make a huge voltage overshoot. If the pulse were reapplied within maybe 50 us, there was no second overshoot. Wait a few milliseconds, and it comes back. These were telecom parts, good for a gigabit per second, but they had to be kept biased. We work in pulse time domain, so we couldn't use them.

Yup. You keep on depleting the N+ epi and P substrate. They slow down quadratically as the undepleted thickness increases. Each dz of depleted material looks like a sheet of charge, which produces a constant E field, so the voltage goes up quadratically with depletion depth, i.e. the depth goes like sqrt(V).

Those were those Optek ones, iirc. I didn't realize there was a big voltage overshoot--I thought the light just came on slowly. They must have doped the substrate like crazy.

Cheers

Phil Hobbs

Sure, greater electric field will sweep the charges through the bulk side a bit faster.

Jon

Yeah, SPICE doesn't do forward recovery.

Maybe a current-dependent inductor, to complement the built-in charge-dependent capacitor? Starts large, becomes small. The energy stored needs to be fairly small, and out of sync with the control variable (i.e., maybe it's charge controlled as well, not current controlled), so it looks like loss and doesn't show up during reverse recovery.

On a related subject, do you think IGBTs experience modulated turn-on as well? I'd suspect turn-on from cutoff is slower than turn-on from "simmer", and that Vce jumps anomalously during a positive Ic step of sufficient magnitude (say 10x more current). Another way to put it is, even with Vge "on", the collector isn't really "on" until some current arrives to inject charge.

(No particular importance, the thought simply occurred to me in relation to IGBTs in pulsed operation. A current step waveform would be relevant for something like, an IGBT driving a saturating magnetic compressor device.)

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

Coincidentally, I've been thinking about improving my PD model to include carrier diffusion effects. Diffusion slows down PIN diodes by a factor of diameter**2, which gets ugly except in very small units.

I started out by changing TT in an ordinary 1N914 model, and immediately ran into something I didn't understand.

I expect Debye shielding (free carriers) plus slow diffusion to make the forward voltage overshoot, but in SPICE it doesn't. It does the opposite--current overshoots and voltage lags. What am I missing?

Thanks

Phil Hobbs

=============== Version 4 SHEET 1 880 680 WIRE 208 96 128 96 WIRE 336 96 288 96 WIRE 128 160 128 96 WIRE 336 160 336 96 WIRE 128 272 128 240 WIRE 336 272 336 224 WIRE 336 272 128 272 FLAG 128 272 0 SYMBOL voltage 128 144 R0 WINDOW 3 283 129 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value PULSE(0 1 10n 100p 100p 5n) SYMBOL diode 320 160 R0 SYMATTR InstName D1 SYMATTR Value Ddiff SYMBOL res 304 80 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 100 TEXT 8 312 Left 2 !.model Ddiff ako:1N914 TT={transit} CJO=1f TEXT 88 40 Left 2 !.step dec param transit 0.1p 10n 1 TEXT 384 96 Left 2 !.tran 100n TEXT 440 152 Left 2 !.model Dideal AKO:1N914 CJO=1f

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

hobbs at electrooptical dot net

Is the E-field any bigger? (I'll have to look.) I thought the field was sorta constant, just a larger depletion region. Of course that's for the "physicist" brick wall doping, "real" doping profiles may be different.

George H.

The dominant effect is reducing the diffusion delay time.

Cheers

Phil Hobbs