RE: DSPRSSTC?

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "Steve Conner" <steve.conner@xxxxxxxxxxx>

>we had a DSP failure in the audio system...
>Equipment loss was approximatly $12,000.

Wow, that makes the loss of a $40 set of IGBTs look like peanuts. I bet the
DSP box makers wish they had put a watchdog timer in that mutes all the
audio outputs ;)


>If a DSP is to be used, make sure
>there is a protective shutdown system in case there is a "freakout".

Yes, this is (or should be) standard practice in any case where something is
controlled by a uC or DSP, in such a way that a crash could cause material
damage. One of our products has a DSP synthesizing waveforms to drive a
laser diode that costs about $1000, so we used a backup current limiting
circuit to keep things under control if the DSP went nuts and asked for a
crazy level of current. Which it did on a daily basis during development.

In SSTC systems with a uC I prefer to keep it out of the very lowest level
where the actual drive waveforms are generated. For the job I worked on with
Finn Hammer, I used 555 timers and logic gates to do the IGBT drive and
protection circuits, and a PIC to generate the special sequence of bangs
that makes the output look and sound like a lightning flash.

I designed it so that no matter how nuts the PIC went it could not damage
any of the power electronics. The worst it could do is generate a stupidly
high break rate that would blow a fuse in the charging circuit.

Having said that I know a few people, such as Jimmy Hynes, Alan Sharp, and
Jason Judd, who use a uC for everything including generation of the gate
drive waveforms. But I think it should at least be set up so a uC crash
can't cause cross-conduction. I believe their systems all meet that
requirement. Transformer gate drive is great from that perspective because
cross-conduction just isn't possible by definition.


Steve C.