Re: Fwd: emc rtlinux & irq

[Jon Elson <elson-at-pico-systems.com>]

Ray Henry wrote:

>When i use this system with a low rotation of encoder (500 - 600 rpm ) its
> all work very well , but why while the
>enkcapture routine can read very fast rotation of encoder emc interface
> have problem to display the step?
>  
>
If you are detecting encoder position entirely with software, it is easy 
for the
sampling interval to miss encoder counts.  If you sample at 1000 per second,
then if the encoder is moving faster than 500 count/second, you may start
to miss counts.

>Make I the rigth thing when i use shared_memory to exacnge the step
> encoder position?
>  
>
You probably have to, at some level.

>It's possible make a prg in rtlinux to generate pulse for pwm to pilot
> directli the h-bridge motor driver?
>  
>
I think this is totally impossible entirely in software, unless the PWM 
frequency is less
than 100 Hz.  You need fine resolution of pulse width, and at the same 
time, a fairly
high frequency.  This requires timing resolution in the microsecond 
range.  Some kind
of hardware is needed to generate these varying-width pulses.

>My personal goal is to make a motion control with a low number of
> componets that is possible.   And i think that this can be useful!
>  
>
I have done essentially this with a one-chip solution.  it is a 30,000 
gate FPGA.
A slight change would make it produce PWM waveforms.  But, without feedback
of instantaneous motor current, the performance would be sluggish. 
 Motor inductance
causes a substantial lag between the change in PWM duty cycle and change 
in motor
current, which is what produces torque.  if you don't actually measure 
motor current,
then you have to drastically reduce loop bandwidth to prevent instability.

Jon