Re: re:Automated "Manual CNC"

Brian Collier wrote: >earlier ballendo replies below
-new ballendo replies

> Not sure what you're hoping to gain here. You'll go 90percent of
>the way to cnc, but without getting the benefits?!

Aren't I only missing the speed of part creation?
The end result should be the same and the effort producing the part
is still being handled by the machine. Sure it's not as elegant as
CNC, but wouldn't it get the job done accurately?Actually, wouldn't
it be better than open loop and ball screws with respect to accuracy?


-Let me be sure I'm on the same page here. You want to output a file
(from CAD) which consists of only lines .001 long(or some other
arbitrary value).

-This part seems EASY, but due to CAD/CAM conventions(see WAY
below),may be the hardest part to achieve.

-Then you want to drive your standard axis powerfeed motors slowly
enough so that accel/decel is not an issue, and using your linear
scales as feedback devices, stop at each point in your file. If I've
got this part right, let's refer to the next block of conversation
below.

> It would appear the motivation is to avoid retrofit of screws and
> motors. Screws you don't hafta change(just be willing to accept

their

> limitations, which you've already done with your proposal), motors
> are easy and cheap.

The only limitation I have accepted is of speed, unless there is
someting I am missing. Accuracy is right on due to measuring actual
table movement and not lead screw rotation. What software supports
closed loop feedback from the table and not just open loop step
movements? I could spring for motors and drivers and build an
interface for my scales to the driver, if I could find software that
could use closed loop feedback from the table.

-Actually NO, you've accepted much less. What you've described is
very similar to the early point-to -point (PTP) CNC machines. Back
then, the issue was lack of computing POWER. Let's use an example to
show you what you're dealing with: A 3inch square,with one corner cut
off at the 2 inch mark on the affected edges. Let's also assume that
we're talking tool movement here, so the real part is 1/2 tool
diameter smaller. OK? Here goes:

-We begin driving the x axis powerfeed, reading the scales until we
have moved 3 inches.Stop x, start y,again reading scales til we have
moved 3 inches.Stop y,change the direction of x,and begin moving x
whilst reading the scales, thistime looking for a 2 inch movement.
Now we Stop x (while we change the direction of y;if we can change
the powerfeed motor direction while moving another axis we don't have
to stop, just add y in). Now to cut our 45 degree corner we start
both x and y (which MUST be calibrated to move at the same rate, I'll
have more to say re: this later),reading the scales until we reach
the starting coordinates for our final 2 inch line. This will be a y
only move, and the direction is the same as we are already traveling.
So, there we have it, a simple part cut(somewhat, see below) as you
described.

-First let's talk about that 45. IF both motors aren't moving EXACTLY
the same speed, it WON'T BE 45! If your response to this is that
we'll be moving in smaller steps, then you've got REAL problems! You
will then be moving a powerfeed motor in units of .001, start, stop,
start, stop, etc, reading encoders at each(or some arbitrary) point.
This is what STEP motors are made for, and what sevo SYSTEMS deal
with!

-You have now to deal with the dwells and jaggies of this type of
cutting.(and using powerfeed axis drives which are NOT designed for
this type of movement, these parts will be Uhhhggleee!)

-Now what if that simple square w/ chamfer is rotated 5-30 degrees
from the axis travel direction?

-Now you HAVE TO cut using this stop,go method and won't get any
advantage of inertial smoothing since we're moving so slowly!

-As I mentioned before, this WAS the state of CNC (well,NC really) in
the 50's and 60's. You could move PTP, and cut lines which were
parallel to axis directions. And also again, this was due more to the
state of computers at the time, since steppers and servos were
already well known.

-There are three ways to cut arbitrary(any shape) tool paths.

-1.FIXED rate motors,as above,lines parallel to axes,45's IF
calibrated for speed,everything else is cut in stair steps.

IMO,This is the BEST you can hope for with your presently described
system.

-2.VARIABLE rate(speed) step motors which can be altered in speed
accurately, and on-the-fly. These will do all the above moves,AS WELL
as any arbitrary shape without the EFFECTS of the stairsteps
(actually still used), since the tool diameter in relation to step
size combines with the inertial smoothing due to the somewhat
constant movement.

-3.SERVO systems where the motor speed is again controlled on-the-
fly, but which move smoothly through points as they "seek" to reduce
the error between where they are, according to the scales (or
encoders), and where they SHOULD be,according to the desired position
coming from the control.

-Have I convinced you yet? :-)

> Have I convinced you yet? :-)

I still need a little more convincing. . . ;)

BTW, what is the easiest way to output XYZ coordinates in the correct
sequence using tool compensation with a CAD package? I thought this
would be a job for a CNC program like MasterCAM. What programs can do
this?

-This was answered well in some previous posts. Keep in mind that
VECTOR is several hundred dollars, and you don't NEED it if you use
regular CNC.(no offense, Fred, I like VECTOR, and the critical word
above is NEED)


Will the XYZ coordinates be also separated by a programmable spacing
as well as being in the right sequence (i.e. along the toolpath)?

- Not sure from Freds' answer(earlier post) if it will give you the
.001 length lines you mentioned.Probably not. It is somewhat harder
(or at least unconventional in CAD/CAM) mathematically to follow an
arbitrary curvy line and generate points a certain distance
apart,than it is to generate points within some error(deviation) from
the desired path. It WILL give you destination points to achieve in
xyz, as will ANY CAD pkg supporting HPGL(they all do) or DXF(most all
do).

<repeated from above to tie this all together, I hope :0>
Actually, wouldn't it be better than open loop and ball screws with
respect to accuracy?

No again. parts made with Open loop (with A/B ballscrews, and
microstepping drives) are indistinguishable from the BEST servo
systems, so its not the "open-ness" of the loop that matters!

Your leadscrews and nuts will most likely have(actually it's not too
hard to say WILL HAVE)backlash which will become apparent at EVERY
direction change(even with backlash compensation and slow movement!).
The cutting forces and/or inertia/momentum of the machine will get ya!

So, as I said before, I believe this is a valuable thought-experiment
and you may still wish to pursue it, but if your goal is to make
chips, I think you are giving up much more than you will gain.

Ballendo (of the tired fingers)

P.S.
There are a few s/w cnc controls which support closed loop operation.
Search previous posts for more info.
1.EMC (NIST)
2.DeskNCrt (Carken)
3.Microkinetics millmaster/lathemaster/Instep library
4.MaxNC-Servo (really a closed loop stepper)
5.SimpleStep (JRKerr??)
6.PCNC (Bestsoft)requires expensive motion card