Re: [CAD_CAM_EDM_DRO] Digitizer algorithms (mine worked)

Graham Stabler wrote:

>
> After spending 9 hours digitizing a 45mm*13mm area (containing a
> plastic propeller blade) with my Proxxon and home made touch probe I
> have started to think about optimising the digitizing a little (tht
> was over 13,000 digitized points.
>
> Turbo CNC simply pecks a bed of nails from a Z datum height.
>
> Enter my fast scan algorithm, I say mine but I have probably re-
> invented it.
>

Hi Graham:

I had posted a similar algorithm to yahoo Turbocnc Tuesday May 14 2002
which makes the probe do 'double duty' for maximum efficiency:

It takes a point not only 'on contact', but moves while on contact,
and records another point 'on non-contact'!

Here is an extract (but unfortunately has not yet made it into TurboCNC
source code yet):

-----------------------------------------

For digitizing, I think we all agree the obvious, but straight forward
and slow method is the 'upside-down bed-of-nails' approach, where all
X&Y movement happens at the upper Z clearance plane, and at each XY grid
point the Z comes down until it contacts the surface, the controller stops
the motion, records the XYZ point, and retracts up to the clearance plane,
and continues along the search grid.

This UP & DOWN to the clearance Z plane is obviously a great waste of time.

Here is an alternate 'stair-step' algorithm:

While following the XY grid along a row-major or column-major
path (lets assume incrementally along X):

Loop1 (probe currently not in contact):
- increment X, if that results in contact (running into the side of
something) move up until its not, record XYZ point, go to Loop1

- but if no contact, move down until contact, record XYZ point, goto Loop2

Loop2 (probe currently in contact):
- increment X
-- if the probe is still in contact, move up until its not, then record XYZ, goto Loop1
-- if the probe has lost contact, move down until it does, then record XYZ, goto Loop2

Thats how I see it (you'll need a probe which is compliant in both
X,Y,and Z in order not to damage it). But this should result in
a much faster scan.

------------------------------------------

However while reading your post, I had another idea: in order to miss
going down to the table, preface the matrix probing with an 'object outline'
probing. This could be done very rapidly by 'feeling' all the way around the
object at a constant Z (known to the human to be everywhere below the surface
of interest). Then the XY matrix probing never has to go beyond the bounds of
the object. This would even work nicely in a "flower" type object, such as a
multi-blade ducted fanblade (ie rotor) with a solidity < 1.0

Doug Fortune
http://www.cncKITS.com























.