Re: Pulse Train Record

[Kurt Goetze <goetze@aps.anl.gov>]

Hi,

I have done something similar to what Bob suggests, in order to have
a BI record (set to I/O Intr) trigger something else after a delay.  With
sysClkRateSet(2000) set in my st.cmd file, I was able to set delays
with a resolution of 1msec.  Jitter was surprisingly good, a few % of a 
msec,
iirc.  The setup was VxWorks / MVME5100, in a crate with lots of other 
stuff.
The binary input was an IP-Unidig input.

I experimented with higher clock rates but 2000 seemed to be the cutoff
for high resolution coupled with reasonable jitter, on my setup.

One thing I wanted to point out is that any old (in my case SNL) code using
vxWorks taskDelays will be affected by this.  Obviously your delays will 
get
much shorter with a faster clock rate!  I found this out the hard way.  
Another
reason taskDelay() calls should be converted to epicsThreadSleep().

Kurt Goetze

Dalesio, Leo wrote:
> If you have the OS tick rate at 1 K Hz,
> You could set the BO's HIGH field to something .001 - 1 millisecond
> set the SCAN rate to .01 Hz
>
> It is unclear what the jitter would be on the scan task - it should be under 100 usecs if you are on vxWorks or RTEMS.
> Does jitter matter or just the pulse rate?
> How wide of a pulse is required?
>
> It would be interesting to hear your results.
> Bob
>
>
> -----Original Message-----
> From: tech-talk-bounces@aps.anl.gov on behalf of David Dudley
> Sent: Wed 6/25/2008 9:54 PM
> To: tech-talk@aps.anl.gov
> Subject: Pulse Train Record
>  
> Is it possible to use a bo for a pulse train output?  I have an
> application where I need to pulse a chemical metering pump which
> requires a pulse train to control it at between 0-100 pulses per second,
> and was wondering if I could use a simple digital output bit to do it,
> or if I need some device support as well.
>
> David Dudley
> City of Corpus Christi
>
>
>