Measuring Fundamental Machine Errors - Mill

macardoso

H-M Supporter - Silver Member
H-M Supporter - Silver Member
Joined
Mar 26, 2018
Messages
2,724
Hi All,
I have an odd love of metrology after working a few summers in a metrology lab while I was in school. I have a small shop at home like many of you and I run a CNC mill (G0704) and a lathe (Enco 12x36).

I'd like to start an open discussion about what are the fundimental errors in a machine tool (in particular a mill), how to measure them, what instruments are required, and maybe a brief discussion on how to correct them.

Here are the logical sections I have come up with. I'll try to make a new post for each as I have time. I really value all your input. My hope is that I can fully measure my machine, and maybe use it as a starting point for future work.

  1. Spindle Errors (Runout, Taper form errors, angular misalignment, etc.)
  2. Z Axis Errors (Tram, Column twist, Column roll, Head wedge, Nodding, screw lead, etc.)
  3. Y Axis Errors (Saddle wedge, Saddle twist, Saddle roll, Saddle sag, screw lead, etc.)
  4. X Axis Errors (Table wedge, Table twist, Table roll, Table sag, screw lead, etc.)
  5. Axis to Axis errors (XY Perpendicularity, XZ Perpendicularity, and YZ Perpendicularity)
Ryan (691175002) on CNCZone had a discussion about this, but I'd like to dive deeper and learn as much as I can for all of you.

-Mike
 
Ok... Section 1:

Spindle Errors

Best I can see, there are 3 fundamental spindle errors:
  • Runout - non-concentricity between the axis of rotation and the axis of the tool
  • Angular Misalignment - varying runout at different distances from the spindle. Tool axis and rotation axis not parallel
  • Taper form error - out of round taper, curved taper, or taper cut at wrong angle
The hardest part I see is measuring the actual machine error and not the error of the collet, tool holder, or test bar. Ideally you would have a "master" test bar which is known to be true and rigid. The method of testing using one of these is described in Dr. Georg Schlesinger's "Testing Machine Tools" (7th edition). Unfortunately, very few hobbyists will have access to a precision standard such as this.

In order to measure runout, and have that measurement isolated from the errors as best we can, I'm thinking of using a high precision collet with a gauge pin inserted. We would use a .0001" test indicator to measure the TIR against the side of the gauge pin. The measurement should be made as near to the spindle nose as possible to minimize the impact of angular misalignment. By using a high quality collet and test bar (gauge pin), we also minimize the error due to the test components. We cannot isolate any taper form errors in this way as the collet will ride on the taper and transmit those errors through.

As a thought, maybe the taper form error isn't a fundamental error, as it can be deconstructed into both runout and angular misalignment? But if this is the case, how would one know if they need to fix the taper angle or the spindle bearings?

To measure Angular misalignment, we repeat the runout test at varying distances from the spindle. by noting the increase in TIR over a 6 to 12" span from the spindle, we can calculate what amount of error is due to runout, and what is due to the non-parallelism between the axis of the gauge pin and the axis of rotation. This is great for vertical mills, however one would need to compensate for the droop in the test bar due to gravity on a horizontal mill or lathe. If we use the mill's Z axis to move the indicator, we introduce the column errors to our measurement. How else could we measure this?

Finally we look at the taper form errors. I would say this is easiest when the bare spindle is removed from the machine and tested between V blocks on a surface plate, but we can do it on the machine. To measure the taper angle, we run an indicator in an out of the spindle taper (without moving the spindle) and note the indicator measurement over the distance moved. Again this introduces column errors. The measurement should be repeated for several indexes of the spindle to get an understanding of how the spindle form looks around the circumference.

Please share thoughts and ideas!
 
Here are some badly drawn pictures of what I'm talking about...
Runout.jpgAngular Misalignment.jpg
Taper Angle.jpg
Taper Form Error 1.jpg

Taper Form Error 2.jpg
 
Tolerances in the collet and gage pin will add up, will be additive to the spindle itself, and determining the actual spindle runout from the results will be confusing at best. The first place to start is with a careful visual inspection of the spindle taper and the cylindrical seat, carefully cleaning, correcting and dressing any swarf, cosmoline, burs and dings, then testing the taper and seat directly with a tenths indicator. Testing the taper and cylindrical portion at multiple heights and making a map of the results of the taper and cylinder testing combined, all referenced to specific clock positions, would check for 3D tilt and runout from the spindle axis without introducing confusion from intermediate tooling. Whenever possible, measure what you want to qualify, not a surrogate.
 
There is a way to accurately measure radial and angular spindle runout. It requires being able to rotate the R8 test bar in the spindle socket. If the R8 key is removed you're set. Otherwise, it would be necessary to be able to rotate the R8 test bar 180º. This could be done by cutting an additional keyway in the R8 test bar or determining the position of the keying pin in the socket and cutting an annular clearance groove at that position.

It works like this. The test bar consists of an R8 tool holder with a machinable bar. The test bar is inserted into the spindle socket and the mill is used as a vertical lathe. A lathe tool is set up in the vise or other suitable fixture and the test bar is turned down to a slightly smaller diameter. You now have a test surface that is perfectly concentric with the mill spindle axis. Rotate the R8 test bar 180º and measure the TIR. The spindle TIR will be the measured value.

Both radial and angular runout can be measured in this way, providing the turned section is long enough. I would suggest something like a 100 mm turned section with the first measurement made no more than 10mm from the spindle face and the second made at the far end of the bar. Parsing the measurements onto the respective radial and angular components can be done by knowing the distances from the spindle face for each measurement.

To do it properly would require adding a degree scale to the test bar as the angular runout and radial runout aren't necessarily in the same plane. This will provide direction as well as magnitude of the two runout components and will allow determination of where in the spindle geometry the angular runout occurs. This adds complexity and isn't really necessary unless correcting the spindle is planned. For most users, being able to assess the TIR would be sufficient.
 
I really appreciate your thoughts! I agree on measuring right on the taper but couldn't wrap my hear around mapping that directly to the actual values of runout and angular runout. Indexing the spindle seems to be a nice way to go about it.

RJSakowski, I like that method! it is very similar to the method you might use to true a lathe. I see that this method COULD impart error due to the fit between the spindle taper form and the R8 collet taper form.

What kind of accuracy would you guys consider acceptable for a bridgeport class machine as far as the spindle radial and angular runout? What about the perpendicularity of the nose of the spindle to the axis of rotation?
 
What kind of accuracy would you guys consider acceptable for a bridgeport class machine as far as the spindle radial and angular runout? What about the perpendicularity of the nose of the spindle to the axis of rotation?
.0001" or better (preferred) to .0002" maximum, for total combined runout. More runout will still cut metal, but not nearly as well or as accurately.
 
My Tormach 770 has a maximum specified runout of .02mm at the spindle and .03mm at a distance of 100mm from the spindle. The machine actually measured at .01mm (.0004") and .02mm (.0008") respectively.

Normally, the spindle face isn't used in machining (setting the face as the reference surface for tool offsets being an exception) so the perpendicularity to the spindle axis isn't an issue Tormach also uses the spindle face as a registration surface for its TTS system and has a technical bulletin regarding checking an correcting that surface prior to converting a non Tormach mill to use the TTS system. My forty year old mill/drill had no measurable variation around its circumference.
 
I
RJSakowski, I like that method! it is very similar to the method you might use to true a lathe. I see that this method COULD impart error due to the fit between the spindle taper form and the R8 collet taper form.
You are correct in that an issue with the form of the R8 test bar could influence the alignment. Ordinarily, I would expect that any imperfections would rotate with the bar. That is you had zero runout as turned and the spindle taper were correct, when the test bar was rotated 180º, you would still have zero runout.There is a slim possibility that imperfections in each would cancel each other out in one alignment and add in the other. For instance, a burr on one fitting into a divot on another but I expect that is unlikely.

Tormach has a one piece test bar which is machined to be concentric and which they use to measure runout. For $250, you can buy one from AliExpress. A bit much for what would be once in a blue moon use. I had considered making my own but was concerned about getting the taper accurate enough that it could be considered a standard. This is a problem with hobby based metrology.
 
Last edited:
This is a problem with hobby based metrology.

For sure! I will have to measure the spindle on my machine this weekend to get a feel for how it looks.

I will write up the next section soon; this is a great conversation.
 
Back
Top