A close-up pic of that Garmin servo label would be interesting. It should mention the TSO standard achieved
C198.
The actual component must cost $200 to build and hold in stock, even with the usual aviation mark up I can’t get my head around the suggested price for this OH component!
How about a cool 30k for a new KI256 😬
Have just received an eye watering quote for a simple stabilator pitch Warrior servo (OH, not even new component). Any suggestions for a supplier either UK or Stateside?
The actual component must cost $200 to build and hold in stock, even with the usual aviation mark up I can’t get my head around the suggested price for this OH component!
Yes these -55C rated CPUs are interesting and to comply with the TSO (?) requirements (assuming they are relevant to GA; I don’t actually know what the environmental spec is for GA servos – does anyone know?) one would use one of those. Then one could do it all in software.
A close-up pic of that Garmin servo label would be interesting. It should mention the TSO standard achieved.
Historically, my interest in this was in the development of an “STC modification” for the King servos to fix the burnouts, and the other issues like the tachometer failures which are absolutely guaranteed after x hours.
Peter wrote:
This makes it very hard to do a brushless design because none of the control chips (either for 3-phase brushless motors, or for stepper motors) are rated down to -55C.
Not only that automotive AEC Grad 0 chips are generally tested down to -65C (>500 cycles), there are also significant number of off the shelf microcontrollers available that are rated down to -55 or even -65 (e.g. the TI HT -line).
Hint if you can’t find them: Typically they are marketed as high temperature versions as outside of aviation temperatures below -45C are quite rare and these Chips have an extended temp range that qualifies them for very low and very high temps
Here is a bit of PR for the Garmin servos which use brushless motors.
They must have known of the King servo issues which were massive (on the KFC225, at least) by around year 2000 but the issue was well suppressed until they stopped making it some years later.
So by the time they did the GFC700 they knew they had to avoid using brush motors. Especially the crappy Globe motors which most autopilot servos use, presumably because they are specced down to -55C.
They seem to be using 3 phase brushless motors (there is, practically speaking, no such thing as a “DC brushless motor” which the article mentions) and some electronics.
I wonder if they use shaft feedback or just run them open loop? The geardown ratio is massive (of the order of 10k:1) so the motor doesn’t feel the control surface forces at all.
You were making 80k/year and you tested every one at -40C and +85C?
Due alone to the time it takes to change the temperature by that much, the chamber must have been huge.
We have produced Automotive electronics and 100% full range temperature tested them, for -40 to +85. This was for a Japanese car company and volume was in the region of 80,000pcs/year.The test kit was not unusual or particularly expensive.
The same kit could do -55C
because they obviously don't test at -40C and +85C
Huh? Not every chip is necessarily tested at other than room temperature, but the design is certainly evaluated at these temperatures.
The problem at low temperature is hold violation, and no, you don't need interesting logic design practices to run into this problem if you don't care.
Processors will be tested like any other complex logic i.e. with test vectors which are designed to exercise every part of the chip. I used to do this on ASIC designs.
Obviously you can't test a 1 million gate chip with every logic combination.
I don't know what chip makers do (because they obviously don't test at -40C and +85C) but they probably run the vectors with a certain speed margin. In lower temps, chips get faster so you are "allright" testing at +20C - so long as you didn't engage in certain "interesting" logic design practices which make stuff like self modifying code look entirely respectable 
You have a big job doing that with a processor because any tiny part of it could stop working at a given temp outside the data sheet spec.
Sorry to break the news, but no sufficiently modern processor (or other device) has 100% test coverage at speed.
If a design breaks at -55C, it was marginal at -40C already, the difference being only 7%.
Most problems at low temperature are mechanical, for example different dimension changes of different packaging materials and the PCB, etc.
