Search This Blog

Sunday, November 27, 2011

Worst Case Circuit Analysis Without a Computer

Paul Rako / Analog Editor over at EDN posted a story about the Linear Tech – Switching Regulator Universal Compensator Adjuster { http://tinyurl.com/c59rqbm }

Years ago (early 90's perhaps) I remember seeing Jim Williams had a decade Resistor and Capacitor box and called it a Worst Case Circuit Simulator – or something like that. And my memory is foggy on this point, I think it was Jim, but it may well have been someone else (Let me know if you remember this better).

I too had used my very old Heathkit Resistor and Capacitor Substitution boxes for this sort of thing, but that gave me the idea to....... make it right.

So I stuck them together and put a “Professional” label on it and called it the “Hagtronics -Monte Carlo Circuit Analysis System”



1960's Heathkit Technology at it's finest - And still useful today, I might add!

It has served me well over the years and has been the cause of much instability for the most sensitive loops, because frankly it is too big and picks up too much garbage what with it's long leads and all, but it has helped stabilize many of the less finicky loops that I have run across.

I also use it on the output of active circuits to get a gauge of how stable to loading they are.


The joke around design circles then was that the "Loop Design" procedure was something like this,

1) Find any setting where the loop didn't oscillate.

2) Vary R and C to find both the upper and lower bounds of stable oscillation.

3) Pick values exactly in between and put them on your "Production Ready" schematic.

Sounds funny, but you would probably be surprised, perhaps even shocked at how many power supplies in the world have been compensated this way.

First - because in the 60's to 70's switching power supplies loop dynamics were not well understood.

Second - Even when we understand the loop dynamics we are not given all the internal gains of many of the regulator IC's to really make a complete model for real analysis anyway.

Third - Many of today's switching regulators operate in Current Feedback Mode. Current Feedback Mode is rather interesting because it removes a pole from the regulator state equations and turns a nominally second order loop into a first order loop, making compensation much easier and less likely to go bad down the road.


So there you have it – Computer Optimization the way Jim would have wanted it! As long as you understand basically what you are doing to the loop, this isn't a bad way to start out compensating a loop or even to help worst case test it.

Thanks Paul, for reminding me of this.

Steve Hageman
AnalogHome