What's new
What's new

Renishaw probe; best way of finding center of a square tilted at 45 degrees?

dandrummerman21

Stainless
Joined
Feb 5, 2008
Location
MI, USA
Running some 2" square shafts on a horizontal mill. I need to probe the ends of these shafts to find center of the square on each end. they are hanging out of my fixture about 8" on both ends and the shafts are 42" long, just coldrolled stock. I've run them before, and know they aren't straight enough to assume they will be in the same location within a few thou every part.

However, when I ran them before, i was holding 1pc at a time with toeclamps against an angle plate. Probing was easy because the square was not rotated and I just found center in x and y. This time, I made some V fixtures so I can run several at a time, but since they are in v blocks, they are tilted 45 degrees.


On to my question, I'm conflicted which is the best way to probe these. What I want is to probe it on all 4 flats and find the center, similar to finding the center of a boss. Machine has mid 2000's version of inspection+ installed, no gui or anything.

So, aside from writing my own program (which I'm not sure I'm ready for), I can think of a few options that I'm not crazy about.

First option, use O9812 (WEB) to find center of x and y, somewhere just off of center so I am not hitting the corner, but rather some section of the flat. Of course, the machine will approach from one axis only, instead of normal to the surfaces. I know this can kind of work, but i feel approaching normal to each surface would be more accurate.

Second option, use O9822 (ANGLE WEB). I am not sure if this really does what I want. I guess you would run it once at 45 degrees, reposition to the new location it sets, and then once at -45 degrees to kind of tweak it in. Would/should I run each one twice to creep into a more accurate location? Or take the average of the two? Does this really work for this kind of purpose?

Third option, use O9823 (3 POINT BOSS). I really wish this was a 4 point boss program instead of only 3. I was thinking of using this twice and taking an average of the 2 locations.


I didn't see any custom ones on the internet that would accomplish what I want simply, but I was hoping that I just didn't look for the correct terms.

What would you guys do in this situation?
 
If the angle is exact 45 deg, the X coordinate of the center would lie at the mid point of the two X coordinates when the surfaces are touched from both directions (+X and -X).
Likewise, Y coordinate of the center can be found out.
 
I've read this 3 times and still can't figure out how the square is oriented in the machine. A picture or drawing would be very helpful.

EDIT: Like this? If yes, then hitting the part where the probe stylii are drawn will work fine, but it'll be at the mercy of the quality of the stock surface and the accuracy of your 45º fixturing. But really, that applies to if it were at 90º too.

Untitled.jpg
 
Last edited:
Sorry, i almost illustrated it but thought i could explain it away.

I'm on a horizontal doing the ends of the square shafts. Xy center of the square.

Pardon my terrible phone illustration.
 

Attachments

  • Screenshot_20240106_130139_Chrome.jpg
    Screenshot_20240106_130139_Chrome.jpg
    52 KB · Views: 24
Ok, so really the same theory applies. You can hit two angled surfaces, as long as they are really 45º, and the control will still find the center. I've done this same thing, but on machined surfaces, and it worked really well. The probe will deflect off the angled surface, but it will deflect the same amount on both sides.

I imagine, since you're using mill finish bar, that the location tolerance isn't all that tight?
 
Running some 2" square shafts on a horizontal mill. I need to probe the ends of these shafts to find center of the square on each end. they are hanging out of my fixture about 8" on both ends and the shafts are 42" long, just coldrolled stock. I've run them before, and know they aren't straight enough to assume they will be in the same location within a few thou every part.

However, when I ran them before, i was holding 1pc at a time with toeclamps against an angle plate. Probing was easy because the square was not rotated and I just found center in x and y. This time, I made some V fixtures so I can run several at a time, but since they are in v blocks, they are tilted 45 degrees.


On to my question, I'm conflicted which is the best way to probe these. What I want is to probe it on all 4 flats and find the center, similar to finding the center of a boss. Machine has mid 2000's version of inspection+ installed, no gui or anything.

So, aside from writing my own program (which I'm not sure I'm ready for), I can think of a few options that I'm not crazy about.

First option, use O9812 (WEB) to find center of x and y, somewhere just off of center so I am not hitting the corner, but rather some section of the flat. Of course, the machine will approach from one axis only, instead of normal to the surfaces. I know this can kind of work, but i feel approaching normal to each surface would be more accurate.

Second option, use O9822 (ANGLE WEB). I am not sure if this really does what I want. I guess you would run it once at 45 degrees, reposition to the new location it sets, and then once at -45 degrees to kind of tweak it in. Would/should I run each one twice to creep into a more accurate location? Or take the average of the two? Does this really work for this kind of purpose?

Third option, use O9823 (3 POINT BOSS). I really wish this was a 4 point boss program instead of only 3. I was thinking of using this twice and taking an average of the 2 locations.


I didn't see any custom ones on the internet that would accomplish what I want simply, but I was hoping that I just didn't look for the correct terms.

What would you guys do in this situation?
1. You mentioned, that the probing process was satisfactory when you measured the part with bar surfaces oriented parallel to XY machine axes. This means that the square was "squared" good enough for your demands.
2. If so, it is still "squared" good enough when the part is rotated by 45 degrees. Using 2 O9812 routines (one in X and one in Y) will give you the results of the same quality as you got while the part was aligned with machine axes.
3. In any case all "vector" routines induce a lot of inherent errors. I recommend, whenever possible, to use the "standard" routines with measurement moves parallel to machine axes.

Stefan
Cogito Ergo Sum
 
3. In any case all "vector" routines induce a lot of inherent errors. I recommend, whenever possible, to use the "standard" routines with measurement moves parallel to machine axes.

I've been curious about this - are the Vector routines bringing anything to the table vis-a-vis the ball hitting an angled surface? Or are they just giving you the ability to move the probe at an angle?

Is it a big deal that the stylus is hitting this 45 degree surface as far as accuracy goes?
 
Since the probe mechanism is 3 points on the older probes, it seems to me that there will be some small error when probing an angled surface when the probe is typically calibrated touching flats or tangent points on an arc.

Newer probes using strain gauge technology would minimize an already small error as they do not require as much deflection to register a touch.
 
Simplistic but if your V locations are precisely in the centre of whatever they are made from and the location is down into the V the centre of the stock will be the centre of the V in one axis. This is the benefit of using V blocks as you automatically have repeatability on one axis. But you still have one axis to probe. Having only done a little probing I wouldn't lose any sleep probing at 45.
 
Is it a big deal that the stylus is hitting this 45 degree surface as far as accuracy goes?
It is if you wish to find the position of certain point on surface. In such case, in order to know which point on probe's sphere is touching the surface, you must to move normally to the surface. Standard (tripod) probes are calibrated in 12 discrete position every 30 degrees, and the data of the angle closest to the surface nominal angle is used for calculations.

Stefan
Cogito Ergo Sum
 
Since the strike is symmetric from both sides, any error caused by angled approach would cancel out.
As mentioned already in the past, vast majority of the probes (tripod based) are not symmetric.
The pretravel (distance the probe moves from the physical contact of it's sphere with the measured surface to the moment when trigger is generated) is not the same on both sides. The difference is small, but they definitely do not cancel each other.

Stefan
Cogito Ergo Sum
 
In the 2-touch method, where the second strike is at a low feedrate, the difference (error) may not be significant.
Moreover, this difference will be there even in 90-deg strikes.
Here, the point is, can 45-deg strike at all be used?
If not, what can be done?
 
Has anyone thought of using Vee blocks? Hold one to the left side, probe it, set Zero, hold to the right, probe it, divide by two. Resulting number is also the distance from the top when it is probed.
Did I miss something?

Ed.
 
It is if you wish to find the position of certain point on surface. In such case, in order to know which point on probe's sphere is touching the surface, you must to move normally to the surface. Standard (tripod) probes are calibrated in 12 discrete position every 30 degrees, and the data of the angle closest to the surface nominal angle is used for calculations.

Stefan
Cogito Ergo Sum

Can you M19 R180 the spindle so the probe is hitting from the same quadrant, just turned around? Would that comp out the error?
 
I sketched this in cad and it looks like you are onto something with the O9822 angled web cycle.

It looks like you run it at 45 degrees and 135 degrees, record the x and y error for each cycle then depending on renishaw's convention for defining error you do one of the following:

If the X & Y error is measured to the center of the web along the probing path,
add the x errors to generate a total x shift, and add y errors to generate the total y shift. see if this diagram helps.

1704667692902.png

if X and Y error define the intercepts for the web centerline, you would average the y errors and x errors to determine total fixture offset difference as you mentioned.
1704669113750.png
if it's worth your time, you might have to run some tests to see which convention the macro's use. In both cases, you need to retain the positive/negative sense of the error in your calcs, not absolute value. Also, any twist of the bar stock will induce another error that this method won't compensate for. It should compensate for the stock being a rectangle.

The renishaw manual states that you need to perform a vector calibration before using angled web measurement. I'm sure Probe and Sinha understand that part better than I do, but I would guess renishaw has a way to reasonably compensate at 45 degrees vs the 30 and 60 degree calibration points. Like other's mentioned, your stock surfaces are likely a bigger source of error than the probe calibration.

good luck, let us know how it goes.
 
Thanks for all the replies. A few things:

Probe is older, an MP10, so it's a "tripod" type.

Probe is calibrated with vector cycle. I believe it's every 30 degrees.

Part is sticking out 8" from the V and stock is not (very) straight. If I had to guess, I would say location might vary .010" part to part, maybe more. I'm milling a round boss on each end that needs to be within .020" TIR to the square on both ends so would be pushing my tolerance limit without probing.

I believe regular 9812 cycles off center so i dont hit the corners should be accurate enough. But was wondering how much more or less accurate other methods were.

I feel like angled web should be more accurate but I would have to physically do some macro math to average them (or take the errors and apply them).

While I feel like angled web should be more accurate, I know from experience that 3 point bore is less accurate than a regular 4 point bore cycle, and multiple iterations are required to get it to be more accurate. So I was guessing the same would be true for the angled web macro.



I have the fixture made but haven't programmed it yet, but plan on getting it running today. I may try both probing methods and see what I get.
 
Thanks for all the replies. A few things:

Probe is older, an MP10, so it's a "tripod" type.

Probe is calibrated with vector cycle. I believe it's every 30 degrees.

Part is sticking out 8" from the V and stock is not (very) straight. If I had to guess, I would say location might vary .010" part to part, maybe more. I'm milling a round boss on each end that needs to be within .020" TIR to the square on both ends so would be pushing my tolerance limit without probing.

I believe regular 9812 cycles off center so i dont hit the corners should be accurate enough. But was wondering how much more or less accurate other methods were.

I feel like angled web should be more accurate but I would have to physically do some macro math to average them (or take the errors and apply them).

While I feel like angled web should be more accurate, I know from experience that 3 point bore is less accurate than a regular 4 point bore cycle, and multiple iterations are required to get it to be more accurate. So I was guessing the same would be true for the angled web macro.



I have the fixture made but haven't programmed it yet, but plan on getting it running today. I may try both probing methods and see what I get.
why not use taller v blocks to control the deflection of the bars better?
 
why not use taller v blocks to control the deflection of the bars better?
Part is 42" long and pallet is 26" square. V-blocks are custom machined to hold 2 bars 6" apart.

Similar parts, same customer, also come in 2.5" and 3" square, with similar lengths.

Aside from buying a HUGE piece of material to cradle them in, or welding up something to extend another 4" each side (milling 3.5" deep on both sides), this was a quick solution.
 








 
Back
Top