[Bdi4emc-help] Quickstep driver

Sat Feb 18 20:52:16 CET 2006

Paul> Using a more realistic period of 7.5uSec.

I got a similar report from Jeff.

Well, I'm glad y'all can run at a such a ridiculous speed, but you really 
shouldn't need to.  10us is really fast!

I'm thinking about why you might want to do that...  it occurs to me that 
freqmod is limited to generating pulses at even multiples of the PERIOD 
frequency (call that f), and the pulse frequency is kept constant throughout 
an entire servo cycle time.  This means that when moving quickly the step 
pulse frequency is alternating between f/2 and /f3 each servo cycle -- a 50% 
change in velocity that can be really hard for the motors to follow.  You 
may not lose steps, but I bet it makes a lot of  1KHz noise, and it 
certainly reduces effective torque.

Quickstep does things differently.   At each period, it either pulses or 
not, depending on which choice brings winding currents closest to their 
ideal phase.  At 10us PERIOD, if your driver electronics do x10 
microstepping, and you're travelling at 4000 full steps/s (fast!) then the 
full step phase in the windings jitters w.r.t. the shaft position by at most 
+/-7% or so -- not enough to significantly affect torque to the shaft.  If 
your electronics only do x4 microstepping, this error can get to 15%, which 
is still just fine, and the 3% error introduced by 10us time quantization is 
insignificant compared to the 12% introduced by the microstep resolution.

Unless you really want to go a lot faster than 4000 full steps/s, or you 
have really fine microstepping resolution, you'll be happy with a 10us 
PERIOD.  That Celeron could probably use the extra cycles for something more 
productive.

--
Matt