propeller corrections was Re: stepper-servo system

I made a couple of mistakes... Corrections in <CAPS>
Note that "future"<G> propeller chips will execute instructions in 2
or 1 clock

>In CCED, "ballendo" <ballendo@...> wrote:

Hello,

Propellor has a 20Mhz crystal--or internal osc that is x4 for 80mhz,

Most instructions are CURRENTLY FOUR clocks, for 20MIPS.
(PER 32BIT COG)
<end of corrected text>

> Keep in mind that this is a PARALLEL processor, with 8 "cogs"

running

> at this rate! Memory can be an issue, as was brought up originally,
> but IMO this chip is "smokin"!
>
> (Seems a bit expensive, and the SPIN programming language is still
> getting sorted out; but this one's MADE for our kinda stuff IMO.)
>
> The SPIN topology is aimed at reusable modules of code that you'll
> simply string together... (Some of the promised modules look pretty
> interesting)
>
> It was covered for three months running in recent Nuts and Volts

mags.

>
> Hope this helps,
>
> Ballendo
>
>
> --- In CAD_CAM_EDM_DRO@yahoogroups.com, Thomas J Powderly <tomp-
> tag@> wrote:
> >
> > Graham,
> >
> >  I'm interested in your EDM work.
> >
> >  What is the instruction cycle time on the propeller?
> >  My work with 1656/7 pics as edm osc's were choked by slow

clocks

> ( 20mHz ) and clumsy loops.
> >  
> >  How many cycles to 'normal' instructions?
> >  This determines the max oscillator freq.
> >  But you may want 2 cpu's clocking each other ( one for on, one
> for off )
> >
> >  Or you may be looking at a dynamic offtime based on gap
> condition..
> >  Something like Iso-energetic pulses where the on time is

preceded

> by a process dependant voltage
> > ( the sequence is: wait for the gap to allow a high-ish voltage
> (say 90 to 120 V=)
> >   wait for the high-ish voltage to drop into the discharge zone

(

> say 35 to 28V=)
> >   then begin decrementing the 'on-time'
> >   then begin decrementing the off time,
> >   repeat 4ever )
> >
> > Now for your 'stepper in a servo loop',
> > I think you mean to control a stepper motor by the process, not a
> velocity of position loop.
> > I dont know of public algorithms, but there has been work done.
> > A hardware system I've used had 2 thresholds, which evaluated
> average gap voltage into 3 classes,
> > open, cutting, bad. ( A window comparator )
> >
> > Open conditions allowed forward pulses ( changed state of
> amplifier's DIR signal, and enable the Step input )
> >  The 'velocity' was proportional to the difference between a
> desired target voltage ( say 30V )
> >  and the actual average gap voltage ( say 28V ) , this gives a

+2

> error.
> >  The proportion is determined by (Err / Range) * Gain
> >
> > Good conditions were a null point in the control system.
> > The Step line is disabled ( or just not en-abled).
> > No motion is desired when the cut has no error.
> >  Just wait a few pulses and it'll shift off this nullpunkt :-)
> >
> > Bad conditions changed the sense of the amplifier's DIR line.
> >  A separate gain can be used for 'running away'
> >
> > I wrote up an Eagle circuit for the EMC group
> > http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl/emcinfo.pl?
> ElectricalDischargeMachining
> > it's a gap sensor with gap-error out, with 0 to 5 V signal (where
> 2.5 is null, 0 is runaway & 5 is charge! )
> > So it's all analog, you'd have to add the window comaprator to

get

> the 3 state output ( fwd,wait,bwd).
> > and some oscillator that output STEP pulses porportional to the
> ERROR size would be needed.
> >
> > I used a higher gain on runaway than for the infeed ( safer )
> > You may even add another rate device for a 'jump'
> >
> > keep in touch with your progress.
> >
> > oh, and the guys at Heidenhain suggested a low pass filter at

500Hz,

> > they said just chuck anything over that!
> >
> > If you use a dual trace scope with 1 on the sensor output
> > and the other on the process, you can evaluate the circuit on a
> working system.
> > At this time the circuit is merely evaluating, not commanding any
> motors.
> > It's leeching off the working system.
> >
> > By hooking the steppers up with no electrode, but driven by a
> running system,
> > and you previously tested sense circuit...
> > you can evaluate the motor response ( I like to stick a junk
> screwdriver into the gap
> > at this point to see how quick the system can react! Very effect
> and very practical.)
> > These motor's are just 'pupetting' ,just reacting.
> >
> > Last to close the loop, attach the electrode to you system, move
> the power leads
> > over to your tool and try cutting. :-)
> > Leave the motor to leadscrew connection exposed ( belt or coupler)
> > It's motion is more informative than a scope!
> > Nervous but progressing is good. Sluggish is bad. Buzzing is bad.
> >
> > regards
> > TomP
> > tAG
> >
>