Dave rant: Robustness of modern ICs? Static? Over voltage? Under?

[george graves]

Robustness of modern ICs vs the old school stuff - that I don't even know about..

Lets call it "Dr. Davelove or: How I Learned to Stop Worrying about static, and Love the arduino!!!!"

(kidding about the arduino stuff - just a google juice thing)

[SeanB]

I had a computer where the 5v rail had a little oops, and applied the unregulated rail to the output. The overvoltage crowbar as well turned out to be faulty and non functional.

Look at DVM during testing and it shows 14.xx Volts. Check range, then quickly turn the power off.

Changed every part in the 5V regulator aside from the capacitors. Transistors, diodes, voltage reference diode, resistors and the 2 power transistors on the big case warmer heatsink along with all 3 parts in the 5V crowbar that did not work. Turn it on and 5V rail is back, and within limits, but the fault is still there. Fault finding time, so simply change each card starting from the back to do a quick isolation ( having a bright orange unit marked Maintenance really helps here for known working stuff) and after 4 cards it works. Luckily a card that rarely fails, so plenty of spare ones to put in and close it up and test then send out to the waiting plane.

Later look at the old one to see why, and one flatpack is looking somewhat discoloured on top. PM5403 hex inverter with open collector outputs, ironically a 30V rated part on the outputs, had gotten really hot and melted it's lid off, and then went open circuit on all 6 sections. Funny enough this is used to implement the 5 second delay, and simply drives a relay. It has an enable that is wire or'd from the power supply and will only start timing if the supplies are all over 3V. 14V and it died, at well over the 7V abs max for TTL. The rest of the devices surived unharmed, including EPROMS and some NMOS DAC's.

[rob77]

i was repairing several 8bit computers (local ZX spectrum clones) failed after someone swapped the 5V and 12V rails. it had a 5pin DIN connecter for power and when the cable broke, people were tending to repair themselves - of course soldering the the 5V and 12V wires in reverse
the only parts failing were the DRAM chips, everything else survived - Z80, eprom, ULA, all the TTL logic.. so back in time the chips were much more robust than nowadays.

[T3sl4co1l]

"Much more robust"

In the best case, I would think.  The prevalence of NMOS LSIs probably had better tolerance than most (I don't know if they ever avalanche, or if the resistor loads just sit there and cook, like they're intended to do anyway), but most other things (DRAMs, TTL) have hard avalanche limits, that are process dependent.  With variable yields, I can only imagine the breakdown voltages varied quite a lot as well.

In the worst case, a 5V family might avalanche at 7V.  Which isn't any better than today.

And anyway, today's parts are going to be better, in that they are more consistent.  I would expect 3.3V logic chips to avalanche pretty consistently at whatever voltage they do.

Tim

[SeanB]

Most IIL logic ( you only really find it these days as a clock driver) is incredibly hot running at the best of times, most seem to be quite happy running hot enough to discolour the board, and do so for decades.

[coppice]

In general ICs are a lot more robust against damage than they used to be. Despite the fine tolerances most ICs are harder to damage with static than they used to be. You don't have the awful self destruct quirks of many old families - like early MOS, where if the -5V wasn't present before the +5V and +12V the chip melted.

Operational stability is another matter. Many fine geometry parts, especially consumer stuff targeting the 2 layer board markets, can be much easier to crash than their older cousins. The industry is going through a major learning phase on this right now.