Re: [CAD_CAM_EDM_DRO]glass scales vs rotary encoders

>
> Backlash is the problem with scales. There is a recent post in the emc
> archive that talks about problems that one fellow (Dave Engvall) has/had
> with servo tuning and lash between his motors and the glass scales.
> Perhaps he will elaborate on his servo experience for us here.
>

MOST OF THIS IS THE EMC POST:
Hi everyone,
When I automated my knee mill (Cincinnti contourmaster....think BP ) I used
acu-rite 0.0002" glass scales on X, Y and W. It came with a lashup on Z but
50 thou backlash doesn't lend itself to good tolerances.

Jon speculated some time ago that with 0.004 backlash and glass scales I
might have some trouble tuning. It turns out they are tunable but the
results might not be the best.

The other day I converted the X axis to an 250 cpr encoder on the ball
screw. This was too good a chance to pass up so I hooked up the glass scale
to an Acu-rite II DRO and proceeded to compare results.
Using G0 (60 ipm) moves it tracks rather well considering the backlash.
G0 X 0 followed by G0 X n produced agreements of usually less than 0.001 and
often 0.0002. Slowing things down to my usual operating speeds of 10-15 ipm
the dynamics change and disagreements got larger but never more than 0.002.

The real change is in tuning: I needed a D of 6 to 7 with the glass scale to
get reasonable following which also produced funny noises in the servo motor
as it tried to follow the commanded position....and dithered plus or minus a
thou or so at audible frequencies producing a solid band when (commanded -
actual) was plotted with gnuplot.

With the encoder I was able to reduce the D to 1.5 and get much quieter
operation and occasional wandering of one or two encoder counts. The next
step here is to change to a higher resolution encoder and see if it gives
better control.

Taking this one step further I may try putting the encoder on the servo
shaft. [1]

In retrospect: glass scales might be a good choice if one has tight ball
screws or is working in a position where static control is good enough ie G0
to a position with G61 enabled and then drill a hole or in CMM applications
but probably not the best choice in dynamic situations with older equipment.

Dave


1:
Just did that: moved the 250 cpr encoder to the Y servo motor. Have not
changed the tuning yet. Can't do everything at once.
Added 0.004 backlash to ini and with some coaching from Ray it took off.
The plans are to put a 1000 cpr encoder on the X ball screw and then go thru
the tuning again.
I tune by setting P with the ZN procedure (Check with Les Watts). Then crank
in some FF1, then FF2, then D and maybe some I. After each change I log a
move then look at actual position - commanded. Ray, bless his heart, is
working on documenting this for everyone. Plotting is done with gnuplot.

Sorry to make this so long but hope it helps not confuses. <grin>