Re: freqmod - step pulse timing

["Joel Jacobs" <jj-at-netexp.net>]

----- Original Message -----
From: "Fred Proctor" <proctor-at-cme.nist.gov>
To: "Multiple recipients of list" <emc-at-nist.gov>
Sent: Tuesday, March 21, 2000 6:04 PM
Subject: Re: freqmod - step pulse timing

Hi Fred,
Thanks for the info on the step pulse timing.  I fixed my controller program
so the step interuppt latency in only 3-4us max and it no longer misses any
pulses.

Joel


> As you discovered the EMC motion controller assumes a step on the rising
> edge of the pulse. Also, if the direction bit changes, the step bit
> change is deferred for one cycle so they don't change at the same time.
>
> The choice of a step on the rising or falling edge of the output pulse
> should be .ini-file-configurable, but it's not. I will change this so
> it's configurable on a per-axis basis. Also, the clocking is inhibited
> when the motion controller is computing zero speed, which may happen in
> mid-pulse. So, you may see your output left high half the time when
> moves are over. I will change this so that the cycle always completes.
>
> The output should be a 50% duty cycle, to within one task cycle time,
> e.g., you'll either have:
>
> 1111000011110000... (4 task periods up, 4 down, 4 up, 4 down, effective
> period 8 X task cycle time)
>
> or
>
> 11100001110000... (3 up, 4 down, 3 up, 4 down, effective period 7 X task
> cycle time)
>
> but never
>
> 110000110000... (2 up, 4 down, 2 up, 4 down, effective period 6 X task
> cycle time)
>
> instead, it would be
>
> 111000111000... (3 up, 3 down, 3 up, 3 down, effective period 6 X task
> cycle time)
>
> --Fred
>