The Spark Gap

18th December 2014

I don’t often post about other podcasts that I like or that I listen to because I don’t think it’s that interesting a subject. However, every now and then I come across a podcast that I find amazing that I’m really drawn into.

I came across this when doing a vanity search on my own podcast. Thanks to the new Twitter advanced search I found a reference to The Engineering Commons, Pragmatic and something called "The Spark Gap." The show is hosted by Karl Bowers and Corey Lange who are both Electrical and Electrial/Computer Engineers respectively, much like I am.

Love the name too. When I think of a Spark Gap I think of a radio transmitter and I think of Hertz and the early radio experiments of the late 1800s. Already a regular listener to TEC, I gave TSG a shot, and fell in love with it almost instantly.

I started on Episode 21 about PCB layout: a subject near and dear to my heart. I considered doing an episode of Pragmatic about it but I decided it was too niche and decided not to. Thankfully they have, so now I don’t have to…

Their discussion on decoupling capacitor placement had me fist-pumping the air in heated agreement. Not a discussion I’ve had in years and one that annoyed the crap out of me when I did layout in my early career.

One aspect of capacitor use that they didn’t reference that I came across when debugging one of the designs I worked on, was the use of the capacitor as a current tank. The idea is that certain ICs can have high current draw requirements and that draw spikes particularly at high frequencies. The issue is that power propagating across the power plane might not "travel" quickly enough (due to inherent inductance/capacitance between the tracks) to supply power to that supply pin causing the voltage to drop at the input of the IC momentarily. This can then affect the entire ICs performance, and in the cases I was troubleshooting, lead to data corruption on one of the data buses.

We initially used Tanatalums as current tanks due to their low inductance but eventually switched to low inductance ceramics due to reliability issues Tantalums have especially at high temperatures. That, and the fact that their failure mechanism was typcially quite impressive and destructive whereas the newer chip-ceramics were quite stable albeit physically larger. One of the things we tried hard to do in the dependability team was to drive Tantalums out of the boards.

Another issue they discussed was component layout: Single vs Double Sided and so on. At Nortel we restricted most boards to chip caps, resistors and inductors on the backside of the boards as we found the heavier QFPs moved around on the second pass through the reflow oven. That said, epoxy dots would usually keep the bigger components in place on the backside of the board it was still frowned upon as a general practice.

I digress: What they didn’t talk about was the component physical alignment. Cooling profiles and heat loading for ambient air surrounding the components is also a factor for vertically mounted PCBs and we found that with naturally cooled (not forced fan or ACU) had better long term reliability if the components allowed channelled airflow, similar in concept to tall buildings in the middle of a big city, such that air could flow easily between them and rise unimpeded.

There’s no doubt that this podcast is really in-depth, with the double-episode about PCBs delving into Net naming, power and ground planes, vias and stacks more (get it? Uh…never mind…) with so much really good PCB layout advice that those two episodes alone are a must listen for those that want to improve their PCB layout skills.

That said I realise it’s a niche podcast that may well not appeal to many people. For me though: I’m working my way through now from Episode 1. To Corey and Karl: Great stuff guys!