Reply to post: Your test isn't working!

Your test isn't working!

I may have posted this before but it seemed appropriate.

Many years ago (early 90s) I was a test engineer for a company that made 'smart' payphones; smart in the sense that they were actually small microcontrollers with support electronics that operated as a payphone*.

I had designed some of the actual electronics on the newer variant, designed the test interfaces / measurement circuitry** and wrote the various tests*** for production units.

The actual build of the PCB assemblies was contracted out (not at all uncommon even then for a company that made perhaps 1000 units per month at the time).

One fine day, the operations manager (who was a total <redacted>) came to my desk and very loudly proclaimed that my test was not working as all the latest build was failing a particular test in the sequence although he had not noted which step it was.

I grabbed a known good unit (in its anti-static bag, natch) and set off for the 100+ mile drive to the build house.

When I got there and observed the test, the unit was failing a check for the escrow relay firing circuit. The escrow relay is a hunk of metal that takes (or refunds as appropriate), the money in the hopper. My test instructed the microcontroller to charge up the circuit, which was then measured (it should be 120V) and this step was failing.

I put my known good unit on the tester and it passed without issues, so I knew the test was fine****. I duly grabbed a few failing cards and returned to the office.

An 'associate' engineer (who tended to look down his nose at test engineers) put one on his bench setup and ran through the escrow relay operation and declared it to be fully functional. I had noticed that the operation seemed sluggish (and his 'test' was hardly scientific) so I took the unit and rigged up the same circuit on a bench that the test equipment used (a resistor!) and probed the firing circuit with an oscilloscope.

When I operated it, the circuit charged up to about 80V and promptly collapsed to about 40V. The energy was stored on a capacitor after charging in preparation for firing.

When I had someone desolder the capacitor that was supposed to hold the charge, it had clearly delaminated so the breakdown voltage of the part was not the specified 250V but more like 80V. This showed me that the temperature in the build process was way too high in that zone of the board.

Replaced the cap and bingo, everything worked.

The operations manager had brought the CEO and VP of engineering, with the associate engineer in tow, round to watch my tests (in the hope of seeing me fail, probably) so when they saw that it was not the test that had failed (and had prevented defective kit going out of the door) the CEO congratulated me for a well crafted test. The operations manager and the associate engineer looked as if they wished the earth would swallow them up.

I got on the phone to the build house and explained the problem (they were quite decent about it) and they fixed the bad batch and the process.

The joys of schadenfreude...

* This was in the USA where it was legal for business to hang their own payphone on a wall (subject to varying regulations) where the local telco had not put one and (hopefully) profit.

** Bespoke electronics in a dedicated box.

*** Using EEPROMs in the game slot of a Tandy Colour computer! 6809E assembly code. The test box and target was controlled by some ports we had added to the game slot interface card.

**** On as previous occasion, engineering had changed a circuit (added a diode) without telling me and the test was failing as it had not taken that into account. Morons.