Thread: Is My Rotary Table a POS?

I am mostly a lathe guy lately, but do some setups on VMCs. A co-worker is on vacation for the next week, and dropped a 4 axis mill job in my lap.

The workpiece is a weldment. He made a fixture that holds it pretty well.

The first problem I ran into is that the rotary table is not 100% accurate. I skinned the top of the fixture at A0, then flipped to A180 and skinned that. This created two surfaces that were out of parallel by .012" over 8". I measured the out of parallel with a mic, and confirmed it by indicating the top of my fixture (0 - 0 at the ends) and the bottom reads 0 -12 (one end to the other). I can rotate the table at will, and as long as I move A+ to my desired angle, it repeats very well.

The error calculates to .09 degrees. So, I re-skin the fixture at 179.91°. Lo and behold, I have made it parallel to the 0° surface.

I'm not sure what is up with the rotary error, but I extrapolate from this that I should adjust my A90 to A89.95 (to compensate for the error). But, no.

Long story short, I come up with A0, A90, A179.91, and A269.96 are the numbers I need to use to make 90° angles. Once I had skinned my fixture at these numbers, I could rotate to any of them (in a A+ direction), and I could indicate the fixture (top, bottom, front, or back) and see less than .0005 over 8".

So, my rotary table seems to be accurate to ±.1°, with a repeatability of at least 10x that. I'd expect the errors to be linear, but that doesn't seem to be the case, for whatever reason.

Am I expecting too much accuracy from the rotary? Is this SOP for a rotary?

EDIT: I forgot to mention, the mill is a Milltronics VM20 w/ Cent7 control, and the rotary is a Troyke.

Take a look at this link:
http://www.troyke.com/adjustments.html

Also, try to always approach from the same direction.
M11 ; unlock table
G00 A-5.0
G00 A-0.0
M10 ; lock table

....

M11 ; unlock table
G00 A85.0
G00 A90.0
M10 ; lock table

....

M11 ; unlock table
G00 A175.0
G00 A180.0
M10 ; lock table

How old is the table?

.09 degrees = 5.4 arc minutes. On Troyke's website, they mention that some of their tables with scale feedback are accurate to +/- 2 arc seconds.

Maybe your table has seen better days...

Rotary tables have compensation tables just like ballscrews. The laser calibration guy checks and comps out any error on our FANUC based machines once a year.

Possible other issues being encoder loose or dirty, wormscrew backed off from gear, loose worm thrustbearings, etc.

I am moving to each angle from the same direction.

Repeatability is very good.

I am pretty sure that this rotary has no scale feedback.

I guess I just created my own compensation table for these 4 angles. The angles in between are still in question, but I only need the 4 (0, 90, 180, 270) for this job.

It's very possible that if the table is older, there is more wear on half the gear. This is much more probable if the same programmer has been there a long time as everyone has a tendency to turn the indexer the same direction all the time when doing 4th work and you do a lot of 0/90/180 work and very little 270 stuff.

Originally Posted by Tonytn36

It's very possible that if the table is older, there is more wear on half the gear. This is much more probable if the same programmer has been there a long time as everyone has a tendency to turn the indexer the same direction all the time when doing 4th work and you do a lot of 0/90/180 work and very little 270 stuff.

I'm afraid that you are probably right. Wear or damage to the worm seems likely. I think it's about 8 years old. It doesn't get used daily (is out of the machine most of the time). Though, when it does get used, it may get treated a little roughly. By that, I mean that guys will try to obtain low cycle times first, and think about wear/tear second.

I think I'll be fine with the job I am setting it up for. All the close tolerance stuff is done at A0, the work on the other 3 sides is less critical (even if I hadn't found the error, it would have been fine).

When the job is done, I'll see if I can find what is wrong.

I pretty sure that it's a DL-6.5-B (I'll have to doublecheck tomorrow).

Troyke catalog shows the current DL-6.5-B to have an accuracy of 60", so I am over 5x that with my error.

Time for repair/rebuild.

Just out of curiosity.... You say that when you spin the table 180 degrees it does not quite go to 180, if you tell it to spin 360 degrees does it repeat the same error or does it accumulate error with the amount of rotation? It may be the settings for a full 360 rotation are off or as has been stated previously the unit needs repair or has damage inside it somehow. If it turns quietly and smoothly I would probably look at the settings first. Also how much measurable backlash are you seeing? Is it possible that it is just backlash you are experiencing? Again approaching from the same direction every time will account for most of that. Peace

Pete

Originally Posted by Pete from TN

Just out of curiosity.... You say that when you spin the table 180 degrees it does not quite go to 180, if you tell it to spin 360 degrees does it repeat the same error or does it accumulate error with the amount of rotation? It may be the settings for a full 360 rotation are off or as has been stated previously the unit needs repair or has damage inside it somehow. If it turns quietly and smoothly I would probably look at the settings first. Also how much measurable backlash are you seeing? Is it possible that it is just backlash you are experiencing? Again approaching from the same direction every time will account for most of that. Peace

Pete

Actually, it overshoots at 180. It overshoots 270 (by about half as much). It is right on at 0 and 90.

I never tried going to 360, to see if it was the same as 0. Good idea.

I don't think it is a backlash problem. I approach each angle from the same direction, and the same distance (10°) in G0. I tried slower rates in G1 (like F10.), but got the same reading.

Edit to add: It was late Friday when I was looking at this, I didn't see much backlash with my indicator (usually .002° to .006°), but I am unsure if there is any backlash comp being done by the control. I will have a better look at the rotary on Monday. I am sure I have a manual for it, so I will probably be able to give the thing a good mechanical look-over.

Originally Posted by Tonytn36

It's very possible that if the table is older, there is more wear on half the gear. This is much more probable if the same programmer has been there a long time as everyone has a tendency to turn the indexer the same direction all the time when doing 4th work and you do a lot of 0/90/180 work and very little 270 stuff.

Good point. Wear in one spot is a strong possibility.

We had add on Tsudakoma tilting rotary tables on our smaller VMCs. The gear to tilt the table 0 to 90° would get worn out due to the fixture weight. Maint would disassemble the table and index the worm gear 120° and again 240° to get more life out of them before they replaced them.

Make sure that the table locks are disengaged when checking for backlash.
You may have to disconnect them to make this happen depending on how the PLC side is set up.
You also need to check lash at all four positions (assuming the worm ratio breaks even at 90 degrees)

Rotary tables rarely overshoot unless the deccel rates are very high and the inertia on the platen is very large.
First guess is that zero is at a "lagged" worm/wheel position, this makes some other positions seem high.
This is a sign of a worn/damaged worm gear but most tables will put the worm at the same position at 90 so this makes little sense here.
Usually you see this at weird angles like 32.1 degrees and your errors are much higher than normal.

Your corrections are bit puzzling to be but I don't know the worm ratio or table size.
Does position zero repeat? How about after 10 full revs in one direction checking it each time?
First test I always do is 50 revs in one direction stopping every full turn and then checking if zero is still zero.

First you would worry about encoder errors or a coupling slipping.
Then try to see if the worm wheel was worn. (cyclical errors that follow the worms revolutions)
Next worm gear bearings (somewhat random errors) followed by worm wheel concentricity to the rotation center problems (long term errors that map to one rev).

There is also the possibility the the locks are not generating enough force but this usually results in a "bowed" cut if the ends of part are spaced equally on both sides of center.
Of course a knowledge of any major machine crashes would make "bent" gear teeth a possibility.

Bob

I didn't get to look into much today. Except to find that there is a fair amount of backlash at some angles, and not so much at others.

It looks like the worm to me.

Don't know specifics about that table, but most have a hardened steel worm-screw, and a bronze worm-wheel. If that table is overstressed at some point in its rotation, like from a crash, the worm-wheel can be distorted. This can show as backlash slop if the wheel threads have been spread apart, or other threads could be jammed together, causing a tight spot. Either way, the error should be repeatable as long as the table indexing directing remains the same, to remove backlash as an issue.

Some controls allow setting pitch-error comp for a table every few degrees, which can improve a condition like this. I don't know this control to say if it could or could not have that feature. I would guess that the worm-wheel has sustained damage.