Re: [CAD_CAM_EDM_DRO] Servo amplifier boards.

johndroc@... wrote:

> Yes, I meant shield - brain fart.
>
> Is there a trick to getting the amplifiers to work? I have turned and
>
> adjusted until my fingers hurt! I can get the amp adjusted so that it
>
> doesn't "drift," but it won't do anything else, either. When I try to
> jog, I
> get the Following error. At one point, I had the Y axis kind of
> working, but
> that was a few days ago. I've managed to zing the table back and
> forth a few
> times (and even smack it pretty hard), but there is no method to my
> madness.
> Is there a general guideline to follow (eg. adjust signal to a certain
> point
> then tach and then comp) or anything I should be looking for? Also,
> do I
> need to have the PID values setup for this? Thanks.

Well, it is pretty hard to diagnose the problems from a distance, and
these amps are not quite like mine, but should be similar enough.

First, set reasonable numbers for the EMC .ini file's parameters.
Something like
P=100, I=0, D=0, FF1=0, FF2 = 0, FF3 = 0, and set the following error to

about 1.000 inch for now. Do you have the units set up right so the
readout
is correct when you move manually? What do you have the max velocity
set to? This needs to be set right so the max velocity produces 100%
range on the STG's velocity output. This is in units of in/sec, so 120
IPM
equals 2 IPS. There is a default and a max, the max is what counts for
the
calculations, the default is what sets the jog rate slider on the
screen.
One other thing is the accel parameter,
which is in user units/sec^2, and the default is too high. (I broke
belts
on my CNC due to that.) Try 2.0 for both max and default acceleration.
These appear in the Traj section of the .ini file, not in each axis'
section.

Now, for the amp, turn the tach pot to the middle range, turn the sig
pot to about
10% up from zero, CL about 20% up from zero, and CMP to the middle
of its range. Power the servo amp up, and make sure it is stable. Set
the
bal pot to null out the drift to .0001"/sec or so. Don't work hard on
this,
as it will need to be readjusted after changes to the tach or sig pots.
Now, make a slow jog, and observe the movement. If there is 'bouncing'
when it starts or stops, decrease tach and sig about the same
percentage.
If there is no bouncing, but the start and stop are sluggish rather than

fairly 'snappy', turn up tach and sig about the same percentage,
carefully.
When you get too much tach gain, the servo will oscillate, possibly
violently.

Now is the time to get the logging and graphing features working. You
can download xgraph from the NIST FTP directory. The logging
dialog box will output files to xgraph, and may properly format output
so you can see actual position vs. commanded position, or actual
velocity vs. commanded velocity for one axis. This is extremely
useful for tuning the servo. You may have to use the Xemc GUI, as
some versions of the TkEMC don't handle all the logging modes.

For the first part of the tuning, you need the velocity (actual vs.
commanded)
graphs. You will see the ramp up, coast, and ramp down sections
of a jog move when you get the log file to include enough samples
to hold all this. You can do a move in about 1 - 2 seconds and hold
all that in 3 - 4000 log samples. You may want to skip 1 or 2 samples
between logged ones to make file size tolerable.

When you get the log data graphed, you will probably see the ramps,
but in the coast part, the actual velocity slowly approaches the
commanded.
You need to increase sig to try to get the actual to follow closer to
commanded. When you get it right, the lines should be pretty close
to each other. You may need to turn the CMP pot up or down
to get the best results. If it starts vibrating, turn CMP up (or if it
is way up, turn it down), or turn SIG down. If you have to turn
sig down to prevent oscillation, you also need to turn TACH down
proportionally, to keep the ratio between these two inputs the same.
To correct the velocity tracking as described above, if you need to
turn sig up, but it will cause oscillation, then turn tach down to get
the same result.

When you get these going well, so the trapezoid-shaped graph tracks
well at a variety of jog speeds, then you switch to the position
graphs, and observe the performance at the discontinuities between
the velocity ramp and the constant speed part. The servo system
will always have some trouble tracking this change of acceleration.
This is where FF1 comes in. Set FF1 to some small value, like 1.0.
Adjust in small amounts, like adding 0.5 at a time, and regraph a
jog at fairly high speed (like 60 IPM). When the graphs get VERY
close, you are DONE! And, you may have a system that outperforms
a number of good, commercial servo motion systems. I have gotten
my system to track within .003" or so, for all motions under 100 IPM,
including full-scale acceleration. It generally tracks within .001"
at constant velocity.

Now, of course, this takes a while, and you may not be able to get
such good results right away.

Good luck, and feel free to ask questions,

Jon