Ajax CNC encoder mounting positions

[Piobaire]

Hi there!
Is there anyone that have tried and have experiance of both motor fitted encoders and ball screw fitted encoders on a Ajax system? I now have linear scales on my milling machine and an Anilam controller, but replacing it.

1. How much backlash and belt tension error are the belt drive producing and could it be compensated at all?

2. The screw pitch error could be compensated, but how about the pitch error while temperature changes on the ball screw?

Screw mounted encoders seems that at least one error source are eliminated, but one still exists. Motor mounted have two error sources, the questions is if it have practical influence of the workpiece??
What experiances have you got, my fiends?

Best regards from Sweden,
Piobaire

[NC Cams]

I just went through a major tune-up of a Bridgeport EZTRAK mill and as it pertains to this subject, here's what I found:

Mechanical lash is the real culprit that has to be dealt with. In our case, we had dwell flats of 0.001 to 0.003 at the direction change points at 12, 3, 6 and 9 o'clock when milling cirular and eliptical parts.

This was traced to an interactive relation of the following:
a. Ballscrew backlash
b. Belt drive looseness
c. Clearance/insufficient axial stiffness of the ball screw support bearings
d. Looseness in the table gibs/box ways/guides.
e. Inconsistent lube film from way oiler.
f. tach feedback from the servos.

We first took the rolled thread OEM ball screw and had it reworked with slightly oversize balls to eliminate backlash.

We then shimmed the OEM ball screw bearings for more axial preload. When we had the bearings shimmed enough to eliminate slop, we couldn't turn the infeed handles.

So, we took out the OEM abec 1 ball screw bearings and installed TRUE abec7, high preload, high contact angle machine tool grade ball screw bearings.

Force to turn the handles is a slight bit higher than OEM unshimmed bearings but there is NO slop whatsoever. Pricey bearings but deadly accurate.

It took quite a bit of tuning of the gib adjustement to get the tables "tight" but not binding. We now can get 0.0005" movement when we hit the left and right arrow 0.0005" feed buttons. This took backlash comp tuning of the CNC to make happen.

Accuracy prior to the tach tune-up??

How about SHAPE accuracies of withing 5-10 parts per million and only a 0.0001" profile error the seems to be confined to the table direction changes. This is compared to a CNC ground precision baseline sample. All shape creation is done via point to point milling cuts at 1/4 deg intervals of a circular shaped form

All this with 1995/1996 vintage 250 count OEM encoders mounted on the motors that drive with a 2:1 belt to a 5 rev/inch ball screw.

The machine software and servo tuning and encoder position remained the same throughout - the only time spent was on eliminating mechanical slop.

The tach senders are currently being serviced due to other issues and have not been returned yet.

Point is this: higher encoder resolution and/or mounting position will NOT fix sloppy iron. Don't even think that electronic comp will fix it either.

If you put a optical feedback on the table via linear encoders, you MIGHT be able to get around sloppy iron as you'll theoretically "know" where the table is. But even so, I don't think you'll be able to totally compensate for slop in the mechanical drives.

Besides, if the table starts to move when you cut due to slop, you can bet the motors will try to compensate and that can lead to some real bizarre interaction between the iron, cutting tools and controller.

In short, I don't think it is WHERE you mount the encoders as much as HOW WELL YOU ELIMINATE SLOP so that things go/stay where you tell them to.