Re: DIY probing

Hi Chris,

What mathematical model would you recommend? DXF OK?

Should not be too hard to make a probe, I even started to collect likely
scrap parts. I'm more caught up in the software side of the task right
now. And I NEED to get my grid data collected.

Your router sounds impressive. I've heard of cutting long linear cams
to correct the leadscrew errors.

Yeah, more fun to get something and start experimenting!

Alan KM6VV

ccs@... wrote:

>
> > Then my first guess (how many guesses do I get?) would be that you'd
> > want to generate points as you went around a supposed axis, and then
> > drop down and take another contour. The points would first have to be
> > adjusted for the probe's dia, and then a tool path generated for the
> > OTHER SIDE (outside) of the horn bell. That's tricky!
>
> Probing strategy would probably depend on the performance of the
> various axis... ie, I'll build the machine first and then see what
> seems the most efficient way to run it.
>
> To correct the size, what I would first do is make a smooth
> mathematical model of the cleaned up probe data, ignoring the probe
> ball radius. Then I'd calulate the normal vector to this surface.
> You do this by calculating two perpendicular tangents - the normal is
> mutually perpendicular to these. The tangents are essentially
> derivatives, that is to say slopes or rates of data change in axial
> and radial directions. Once you have the normal vector at every
> point, you just move out on it by the ball radius to find the true
> bell shape.
>
> One can also do a planer offset, if the offset amount on each 'layer'
> is corrected for the contact angle.
>
> I'm likely to actually machine the mandrel on a lathe conversion that
> is yet to be done. The conversion of my 9" lathe went very well, and
> I expect the 13" one will work too. I don't think the CNC bridgport I
> have access to has quite enough "Y" travel to mill this shape,
> although I have certainly thought about it since it's in running
> condition.
>
> I'll put in appropriate cutter compensation for whatever machine I use
> the same way I take out probe compensation.
>
> I have a very old Starrett "last word". Although I can use it to
> "tram" > a line or surface, I don't know how you'd use them TOGETHER,
> to check > out the probe. I did write a little Gcode to just move the
> probe tip up > to a rail and then back away again, recording the
> "location" of the rail > contacts. That seemed to be very consistent,
> seldom varying more then a > .0001, as I recall.
>
> That does sound pretty good. I was thinking you could use the dti to
> verify that the axis was at the same point every time the probe
> tripped rather than trust the CNC control, but it sounds like you are
> getting results that are, as you say, more accurate than any 15"x15"
> gantry I'm likely to build.
>
> Now I just have to decide if I want to buy a MAXNC probe for a little
> of $200 or build my own...
>
> I'm thinking 1"x10tpi acme rod for the gantry screws, and using them
> for rails as well since I don't think it will flex much over 15" or
> so. Double screw on the bottom axis with timing belt linkage between
> the two. Upper axis might get linear bearing rod in addition to the
> single screw. Z axis probably initially sherline but eventually
> something with about 13" travel to probe in to where the bell starts
> to bend. I may hang some weights on pulleys for backlash
> compensation... I don't expect this machine to move fast.
>
> Just occured to me that by probing accurate shapes of some sort I
> could build a correction table for the screws.
>
> I may just have to jury-rig something to probe mouthpieces on the
> sherline tomorrow to get a feel for everything.
>
> Chris