This week I thought I’d diverge a bit from the Energy Efficiency theme and share some funny engineering experiences I’ve had over the years. I’ve been a practicing engineer for over 30 years and have been involved reviewing projects and circuit designs a good part of that period. I’ve collected a list of several of my favorite oversights (the names have been withheld to protect the guilty). Hope you enjoy these as I did when I found them!
One of the funniest (and probably most embarrassing) "faux pas" I’ve seen was the attempt to provide high voltage isolation for a power supply. In a design review I had recommended a 0.100" clearance between the primary and secondary side. About two months later the engineer called me and was concerned about something in the design that was causing them to fail their HV isolation test. He emailed me the layout and in about 10 seconds it became apparent what had happened. The designer passed my recommendation to the layout technician who instructed their auto-router to provide the 0.100" clearance between the primary and secondary side... the auto-router did just that - funny how machines do exactly what you tell them. The software provided exactly 0.100" clearance between the PCB lands on the primary and secondary sides of the transformer but left the default clearance of 0.010" on all the others. It takes about 70kV to jump 1" in dry air... it would only take around 700V to get across the barrier the auto-router provided thus the failing board. Luckily, it was an easy fix and the board passed without issue... the moral of the story? Don’t think your auto-router knows more than you do...
Stability Is Your Friend
One day I get a call from an aerospace engineer that was concerned that a batch of voltage regulators were not working properly in their circuit. The design had been working fine for years, but lately they were experiencing failures due to an apparent oscillation in the output voltage of our regulator... this was a very "mission critical" military application that was using a linear voltage regulator to provide a clean voltage in a projectile. I asked the engineer to describe the design. He went on to describe the input power stage and the large heat-sink mounted on the side of the projectile where the regulator was mounted. The regulator’s leads were then wired to the PCB where they provided the regulated voltage - about 16 inches away. The first question I asked was where the output capacitor was located - it was on the PCB 16 inches from the regulator! I choked for a moment realizing that it was amazing these things were working at all - real kudos to the National Semiconductor engineer that designed the regulator. Effectively what the aerospace engineer had done was place parasitic inductors (the wire) in series with the capacitor shifting the stability point of the regulator right to the edge... most of them worked at the temperature where these "projectiles" were deployed... the new batch of regulators had higher gains and thus were oscillating. I gently told the guy on the phone the bad news and referred him to the data sheet where it clearly stated how close the output cap had to be to the regulator to guarantee stability. This raised the next question, "if they are working now, will they continue to work?" The answer was - maybe. It depends on so many factors including process aging, temperature, the type of wire used and gauge. The engineer on the phone suddenly inhaled, thanked me for my help and hung up... I assume that the problem was resolved. I didn’t sleep well for about a year after that.
The Linear Boost Converter - Not!
OK, this one really made me think... where did we go wrong in writing the data sheet. I received this email from an "engineer" that was wondering why his circuit was not working properly. I replied and asked for that section of the schematic so I could review the design. Within the hour an email with an attachment showed up. I opened the PDF and had to stare for a minute... this could not be right. They were using a linear regulator where the input voltage was lower than the programmed (resistor divider value) output voltage. They were supplying 5V to the regulator and expecting 12V at the output! This is fine for one of our boost simple switchers, but not going to work for a linear regulator. OK, now I wondered how I was going to respond to this... so I sent a copy of the data sheet pointing out the "drop-out" voltage or loss component of the regulator along with a boost power supply application note and introduced this person to the world of switching regulators. I doubt that would happen today since integrated switching regulators are so common and most likely taught in university programs.
Any similarities to the above problems are coincidental and not intended to make you feel bad if you made the same mistake... we’ve all made mistakes - the real question is "did you learn from them?" Engineering history is filled with stories of failures or bad decisions ("let’s use the most reactive chemical in the world to float a dirigible" kind of thing). But the main thing is that we learn and improve our skills as engineers - I certainly have tried and continue to do so every day. Hope you enjoyed these tails from the past. Till next time...