sot jfet

Alas, a jfet would be ideal in a closed-loop current source: low Cout, zero base-current error, unlikely to oscillate. But the pinouts are wrong, so I can't drop one into the existing boards.

John

Indeed, and nicely put. If you look closely enough, there is an array of L's and C's inside every transistor, along with gain. Hence the old ROT "R's in 2 of 3 leads"

Cheers Terry

Nor will I. Thanks.

AABA, [background] I once had a nasty problem with an IGBT gate drive circuit. When we designed the first version, we didnt test it on the largest devices (none available at the time) and so unsurprisingly it didnt work right (although the 8-of-20 smaller products all worked fine). Being a gate-drive, the symptoms were highly destructive. We re-designed the circuit, and re-spun our smt pcbs. As usual a draughting error crept in (later we modified our check/approve system to catch these types of mistakes), and we had the +ve and -ve inputs of an LM393M reversed.

[interesting part] We re-worked the SOIC8 parts, and began soak-testing drives. Our MTBF increased, but some still blew up, about half an hour into soak testing. Damn. Much analysis, simulation and measurement took place, and we were convinced the circuit *should* work. Eventually some deft work with a heatgun led to a reproducible result - heating the SOIC8 would turn the IGBT on. We couldnt confirm the exact cause (too hard to disassemble the SOIC8) but it appeared that bending the legs up to attach wires had damaged the internal bond wire connection. Enough heat made things move far enough apart that the contact was broken, and KaBoom!

We quickly re-spun the pcb with the correct pinout, and the circuit behaved beautifully. We had one unit on soak test for years.....

That taught me never to bend leads on ANY smt parts that go to customers. If I have to vandalise an smt part thusly, I get a new one and bend the leads very carefully, clamping the lead before it enters the housing.

[hilarious part] In a fit of inspired stupidity, or CEO purchased 1,500 pcbs of the new, completely untested (although heavily SPICEd) circuit, despite our loud, written protests that it *probably wont work* (ah, the joys of smt prototyping). He actually went to the purchasing dept and made them change the order - we had ordered 10. Of course they didnt work, and the pcb's had to be scrapped. Next management meeting, R&D were criticised by said CEO for wasting $10,000 on pcb's.

Cheers Terry

Alas, our R&D manager was a complete sissy, but several of us were quite vocal. About a month later, the R&D manager showed everyone at lunch a scan of his unborn child, and remarked that you could see the spine. Some card muttered "well its obviously not yours then", resulting in much hilarity. He was not amused.

Cheers Terry