PING: John Larkin

Or tell you that the part is just peachy, when reality and a nose will tell you otherwise.

Failure is a great learning tool, too. A hot wire across the 9V battery, mentioned above somewhere, will teach the mush-head, fast.

That's silly. The data sheets and the models are often wrong or incomplete. NO simulator will predict failures accurately, and most will miss entire failure modes. Some parts can be run reliably at 2x or even 4x datasheet/model limits.

They call it simulation because it's not real.

Well, I don't just hit them with hammers.

Do you ship products that uses devices outside the datasheet guaranteed limits? Those that you deem "run reliably at 2x or even 4x datasheet/model limits" ??

...Jim Thompson

You do *what*??

I doubt that he ships his product with over-rated devices inside. If he did, the first fire would cause an insurance company to trash his ass... big-time. ...Jim Thompson

One DUV laser controller that I designed controls the exposure of about 1/3 of the ICs made in the world. Thousands have been built since 2002. The field failure rate has been exactly zero.

I'd tested an IRF740 to somewhere around 140% of the datasheet rated power before it failed (at voltage, too).

Though that's more a matter of range than rating. They give max RthJC, not typ. The die was huge, unsurprising RthJC was much lower than max.

They're slowly removing DC SOA from old parts, too. This exact part didn't have a DC curve in its datasheet, even though the original (IR) did.

Well, a bulk semiconductor physics simulator would certainly be able to model failure, down to the puncture of gate oxide and melting of the damaged area.

Actually, I wonder if they do temperature and conductivity; that's kinda sorta multiphysics, but it's just another bulk parameter so probably. Or easy enough to add.

Tim

IR tends to be wildly optimistic about their mosfet current and power ratings. I'm impressed that you got to 140%.

When power dissipation is important, we test mosfets to failure

and either overkill on power capability and heatsinking, or realtime compute Tj and shut down if that gets scary.

Except that we don't have the data or the compute power to do that. I expect that the semi makers don't either.

I don't blindly trust data sheets. Even typicals are way off some times. A data sheet switching time for a mosfet might be 10 ns typ, but it can get under 1 ns if handled right.

We use a lot of RF parts in pulse applications, where we generally have no useful data.

It could make up stuff like that, maybe. Detailed prediction is another issue because fault initiation is stochastic.

Cheers

Phil Hobbs

You would have to initialize the semiconductor matrix with a reasonable doping fluctuation (including crystal defects and impurities as well as intentional dopants). Also, you'd have to modify your geometry so your junction walls aren't ideal planes, but rough and lumpy things... :)

Tim

er

not

n't

aged

As a kid I found a big disparity between specs & real life ability sometime s existed. I don't make a habit of trading on that, but I would expect if y ou took the time to test the parts properly to determine that they are capa ble that should suffice as a defence. However others will surely be more fa miliar with US law.

And of course there's more than 1 way to protect against fire. A monitored tr can diss more P than an unmonitored one.

NT

No design should catch fire because a semiconductor failed.

imes existed. I don't make a habit of trading on that, but I would expect i f you took the time to test the parts properly to determine that they are c apable that should suffice as a defence. However others will surely be more familiar with US law.

ed tr can diss more P than an unmonitored one.

That's not the point, you need more than one barrier to fire, as you know p erfectly well.

IRL electronics do sometimes catch fire when semis fail, despite designing things not to, so you also need to apply due care to semis surviving.

NT

Not mine.

This thread motivated me to pull _Analog Circuit Design_ (Williams) down from the shelf to read that chapter for the very first time. About the time that Gilbert hit four transistors it became obvious to me that my transistor design skills had become rusty over the years. So the next book to come down from the shelf was the third edition of AoE.

My plan is to read Chapter 2 of AoE before returning to "Where Do The Little Circuits Come From?" My restudy is now at 2.2.2 Switching circuit examples. The subject matter is mostly review, which makes it an easy read for me. The only thing that slows me down is my desire to actually build the examples and tinker with them.

Thank you,

Keep the faith baby!

Cheers

Phil Hobbs

I'm puzzled. I've got Jim Williams' The Art and Science of Analog Circuit Design, and there is no such chapter in it. I've googled "Where Do The Little Circuits Come From?" and the only hit is this thread.

NT

He did two volumes. Gilbert's chapter is in the 1991 book.

There's at least two Williams' books with similar names.

_Analog Circuit Design - Art, Science, and Personalities_ has an old Type 547 Oscilloscope on the cover. That's the book that contains "Little Circuits." ISBN 0750696400.

_The Art and Science of Analog Circuit Design_ has a bird on the cover. ISBN 0750695056.

Thank you,