Screw comp on master/slave axis

[ekbiker]

Hi, I plan to use AcornSix with gantry style Y axis with software defined master/slave. Is there separate screw comp for Y1 and Y2. And let's say that Y1 is master and I run linear encoders, do I need an encoder on BOTH the master and slave? Or can I calibrate the slave Y2 screw comp using an encoder temporarily mounted, and then just run the encoder on the master Y1?

[cnckeith]

short answer is no to different screw comp for a paired axis . this feature is complicated and not implimented as of yet and is fairly far down the to do list.
what is your application? i don't see many gantry machines that are required to hold tenths.

[ekbiker]

No specific application, I'm just trying to build the most accurate DIY gantry machine possible under $30,000. I'm using rolled C7 ball screws but with linear encoders to compensate. Ground C5 screws cost 4 times as much in the lengths I need.

What would you suggest I do?

What I could do is this: map Y1 and Y2 separately and write down the numbers. Analyze the map and find where the pitch errors are out of sync and average the difference. Since if the errors are in sync, then there is no problem with a single comp table.

If I get really lucky, the error on the two screws are all in sync.

Another way is to run parts in the middle of the gantry for more accuracy, since the out of sync error differences will be somewhat minimized.

Another question - while running with encoder feedback, is the comp table constantly updating? Or is the comp table static and the encoders are used instead?

Just throwing this out there: If you let me look at the source code I can probably figure it out. Since your software doesn't work without the board anyway.

[suntravel]

C7 ballcrews are cheap, I would try to find a good matching pair and run them with two nuts, one spring loaded.
Scale on one side and use the volumetric compensation.

Uwe

[cnckeith]

C7 ballcrews are cheap, I would try to find a good matching pair and run them with two nuts, one spring loaded.
Scale on one side and use the volumetric compensation.

Uwe

exactly!

[ekbiker]

I'll try. But I don't have the luxury of picking which screws I like. I get what they send me, that's it. Any pitch error mismatch out-of-sync will be magnified on the side that's not running the scale. But will be somewhat okay on the side with the scale. Large parts will have trouble. Doesn't Volumetric comp still depends on the scale mounted to only one side? I guess I haven't seen VC done and might not fully understand what it does. And it's an Ultimate license feature. I don't really want to get that and find that it doesn't do what I want.

[IPGregory]

C5 (rolled) Screws can be had for not much more than C7s.

[ekbiker]

Uwe, I watched your youtube video on Volumetric Compensation. It will not work for a gantry machines because the Y scale isn't directly coupled to the spindle. And since Y1 and Y2 screws have different pitch errors, this means the perpendicularity to the X axis varies as it travels down the Y axis. VC works by having a perpendicular reference, but in this case of the moving gantry there is no perpendicular reference.

For this to work, it must have independent screw mapping for Y1 and Y2; unfortunately it is so far down on the to do list.

[suntravel]

That is why I suggest to find a matching pair.
C7 is +-0.05mm / 300mm

Rolled HiWin T7 is max 0.052mm / 300mm
They are 3 times the price of cheapest no name C7, but I have measured some 16x5 900mm and they had < 0.025mm deviation /900 mm

So If you have a 30000$ budget, spend a bigger part of it for two good Y ballscrews
Straightness is also something to consider for good precision and speed.

Uwe

[ekbiker]

I'm buying directly from HiWin, so there's good chance the two Y screws are from the same lot. Which means they should be close (fingers crossed). I'm using the R1 rotating nut in 20/20 2650 mm length. Buying ground screws and scales will blow my budget.

Ideally, there should be individual screw comp for paired axis (Centroid Support, pretty please). I'm also a software engineer, so if Centroid will let me have a look at the source code, I can probably figure out how to implement it.