https://en.wikipedia.org/wiki/File:Toronto_Propane_Explosion.jpg

This post is a bit technical, but it summarizes my (mostly failed) experiences with power electronics over the past 5-10 years and my recent progress towards getting power electronics into our knowledge base.

I’ve been generally interested in electronics since I was a child, and more specifically interested in it since I met my wife. (Praew got a PhD in electrical engineering.)

At work, we have built many of our own electronic projects to our own specifications that most other companies would need to buy off the shelf (at a high price, and without the features that we would want). This was done with the help of our very talented engineering team, two of whom were Praew’s former students. We built systems to automate the office light and ceiling fans that also integrate the air conditioning on/off logic. We have simple circuits that turn on and off the cooling fans in the racks that we use for many of our workstations around the office based on whether any of the computers inside the rack are turned on.

We even went so far as to design and build our own biometric entry systems for keeping the doors locked during the day, but allowing authorized employees to come and go easily.

And all of this is in addition to all the amazing things that Praew has been doing in her lab at work. (Humanoid soccer playing robots, presentation robots, agricultural robots, pipe and boiler inspection robots, robots that teach children about money, and more!)

But the one thing that I’ve always wanted to be able to do but has continued to elude me for years is to design and build our own motor drive circuits. This first became obvious when we converted a Nissan NV Wingroad from petrol to an electric vehicle (EV) about 7 years ago. In order to hook up an electric air conditioning unit, we tried to build our own switching power supplies to adjust the 110V-140V of the primary batteries to the lower voltages used by the peripheral systems. We even tried to build our own inverters to drive AC motors from the DC sources.

These components when purchased off-the-shelf are very expensive (maybe $1000 per unit), and even these commercial ones have blown out on us more than once. But our own attempts to build them have always failed within a matter of minutes or even seconds probably due to the high transients we get on our switching circuits.

The same thing has repeated itself many times with high power robot motor drive circuits, where we have always ended up needing to buy commercial products, and even these have failed on occasion.

These problems have remained a “black box mystery” to me ever since. Now it is an issue again for the motor drive circuit needed for our water circulation pumps.

But I think I might finally have a strategy to get a handle on this.

My theory goes like this…
Electronic circuits that seem to function correctly for a time and then fail are likely failing for one of 3 culprits:

  1. Voltages transients
  2. Current transients
  3. Power dissipation causing parts to overheat

Each electrical part in the circuit has a voltage rating, a current rating, and a maximum power dissipation that can be read off of datasheets and the like.

Voltage transients are observable on an oscilloscope, but the other two require doing indirect measurements and then calculating. And when you consider measuring all of these things on multiple parts all at once, with the correct isolation for measurements that aren’t referenced to the same ground. This requires a large amount of expensive hardware that we simply don’t have enough of.

In comes the electronic circuit software simulator.

When we were struggling with the power electronics prototypes we were building for the EV, we bought this book about snubber circuits. It was very helpful and did result in some slight improvents. Snubber circuits can help to filter out transients and greatly reduce voltage spikes, but they come with a catch. Their very property that allows them to reduce voltage spikes and the resulting high frequency ripples can create current spikes and even increase overall heat build-up in the switching components. (That is in the case when you got them right. When done wrong, they can make all 3 worse.) Thus fixing one problem threatens to cause another.

I learned from that book that building snubbers isn’t all about “adding a good thing” to the circuit, but more of a trade-off in order to find a sweet spot where voltage, current, and power dissipation (heat) all sit within acceptable ranges.

And that was where I first encountered SPICE (Simulated Program With Integrated Circuit Emphasis). By simulating the switches (such as MOSFETs or IGBTs) with various snubbers, you can easily experiment with and observe all properties associated with your circuit while simultaneously measuring voltage, current, and power across each component. When one of those parameters goes outside of the acceptable range, you don’t have to deal with a blown out part that needs a half hour of work to replace (or days of delay sourcing replacements!). Instead you simply observe the problem and run experiments to address the issue. In the case of snubbers, the author wrote about measuring the voltage, current, and power dissipation across the switches as they transitioned from open to close or vice-versa (with switching times as low as a few nanoseconds), and let theory guide you in trying different snubber values to alter the behavior of the circuit until the results are improved.

But at that time SPICE was a mystery to me, and there were other more pressing matters that prevented me from exploring it as fully as I would like.

Until last night…

(to be continued)