Doug Drummond wrote:

>The reason we worry about latency and related probles is that motor 
>control is
>Gizmocopter's primary flight controls.  Total motor power == Colective Stick,
>differential motor control is equivalent to the cyclic stick on a helo.
>I'm a fixed wing pilot, but even so, I know that helo pilots NEVER EVER take
>their hand off of the cyclic stick, and they don't let go of the colective 
>very
>long.
>
>So, I would second the motion for the motor control to run directly off of
>a parallel output from the computer.  DAC into a 555 circuit is one way, a
>dedicated set of programable timers is another, probably less expensive
>solution.  You really want to load all four motor control registers
>within a few microseconds of each other with an interrupt-protected code
>fragment, so that no event (such as a timer interrupt) could cause a stutter.

>But an even bigger problem is a close coupling between the attitude sensors
>and the differential throttling.  The bad news is the difficulties associated
>with attitude sensors, but I won't get into that, this message; that is 
>another
>thick book.  I would be inclined to include some sort of hard-wired loop
>between the attitude sensors and the differential throttling, so that the
>darn thing has some tendency to want to fly straight and level.

Less latency is always better, but sacrificing other things, like 
debuggability and development convenience should only be considered when 
the reduction in processing latency is of the same order as the latency of 
the physical actuators.

In, say, a gigabit ethernet switch, microseconds latencies are something 
you want to be concerned with, because the packets may be arriving and 
departing every handful of microseconds.

Rocket motors and little propellers have actuation latencies in the tens of 
MILLISECONDS, and moderately priced gyros only have 10 hz bandwidth, so 
your total end-to-end latency for responding to an impulse event is going 
to be at least 150,000 microseconds, even with a hard wired 
system..  Talking about "a few microseconds" is silly.

There are still latencies worth worrying about, like reading a NMEA GPS 
string at 4800 baud, but keep things in perspective.

John Carmack

_______________________________________________
ERPS-list mailing list
[EMAIL PROTECTED]
http://lists.erps.org/mailman/listinfo/erps-list

Reply via email to