dsp-based ESCs ?

[savuporo]

hey, i have seen great threads on these forums of people building ESCs around 8-bit micros.
has anyone tried a hand at dsp-based solutions yet ? there are a bunch of reference designs from Microchip, Freescale and ST on quite cheap chips, basically able to do full FOC control.
I have a ST7MC-based refkit on the table, with 1Kw and 3KW reference power boards, currently im lacking a motor to start experimenting. any recommendations ?
( note, ST7MC isnt a DSP per se, doesnt have many MIPS, but its a dedicated BLDC chip ) Freescale has DSP56F805 and relatives, Microchip has dsPIC30F.

I havent settled on the micro/dsp yet, im intending to get the refkit or two working first, but after that i'd like to go up in power into ten-kilowatt range.

[vintage1]

I am sure it would be an interesting project, but DSP's are optimised for things that aren't really required in an ESC.

[savuporo]

well, it depends on quite a lot of things. on power level and on what characteristics do you want from your motor.
Im trying to get sensorless, and as few bits of extra hardware in power stage as possible, to ultimately yield a low-cost system design. All that with a great starting torque and a good power delivery over wide range of RPMs. Ultimately in a 10-20KW power range.
That basically calls for a FOC algorithms, which are best suited for DSP.

[Comatose]

When you're in the 20kW region, the cost of your FETs and your PCBs is going to absolutely swamp any savings you can get by going sensorless. Further, great starting torque and sensorless don't go well together. You can be as clever as you want with algorithms or rotor saliency or whatnot, but the the whopping buck three hall sensors would have cost, you'll end up with a much more robust system.

not that FOC is a bad thing, but it isn't a magic bullet.

[savuporo]

im not dead set on sensorless. i want to break my teeth on smaller power range first, to get an idea of what works better in that application.
So far i have the theorethical side somewhat covered, and from practical side i have some experience with different motor types in different applications. i have worked with industrial controls quite a lot, never designed a power stage myself, however.

You are probably right, the cost of hall sensors or even low-res optical encoder wouldnt add anything, except for making the mechanical design a wee bit more complex. Its probably worth it, though.