2009/9/30 Steve Blackmore <[email protected]>:

> I watched your video, but couldn't determine how much lead in you were
> giving the thread?

I am trying to cut a 4mm lead at 200rpm. After a suggestion from
alex_joni on IRC last night I intend trying a bunch of stuff tonight
with the motors turned off so that I can try much higher velocity and
accel limits without having to worry about if the hardware is capable.

> My acceleration values are high, BUT I still always start 5mm off the
> end. Delay on 1st pass for spindle to get to speed,

I cut (another) thread on my component last night with a 2" lead-in
and it nearly always managed to almost synch before entering the cut.
That "nearly always" is the killer.

> I think you have two problems, noise and not enough lead in.

I am intending to make a better encoder. I think my problem is an
unfortunate combination of things one of which is too-infrequent
position updates,

<Wild guessing starts here>
Looking at the source code there is what looks like a very simple
proportional-only controller (the coders might not even have
recognised it as such) and that clamps both accell and velocity to
zero in the case of an overshoot. I think I might have enough dither
in my encolder position to produce false overshoots, and then the
system is not recovering from that dead-stop. I am not sure that
clamping accell to zero is a good idea, in fact.
http://git.linuxcnc.org/gitweb?p=emc2.git;a=blob;f=src/emc/kinematics/tp.c;h=033198b6b0401d58edfff961aa01cb15ca97c03d;hb=HEAD#l896
Line 569 zeros newvel, then if newvel is zero line 578 sets newaccel
to zero. I would need to look more closely at what actually happens in
the code but can the system achieve zero velocity from some velocity
with zero accel?
I am still stumped as to how a system with a cycle time of 50uS and no
"memory" can produce behaviour which is cyclical in the 100mS range.
</wild guessing>

-- 
atp

------------------------------------------------------------------------------
Come build with us! The BlackBerry&reg; Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay 
ahead of the curve. Join us from November 9&#45;12, 2009. Register now&#33;
http://p.sf.net/sfu/devconf
_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to