# backlash & feedback, is there a strain gauge cure?

30 Jun 2011 20:12 #11034 by andypugh
1:1 wrote:

Or have I fluffed my basics ?

Possibly, I think you can consider the X and Y velocities and accelerations as independent vector components.

01 Jul 2011 23:53 #11048 by 1:1
They are...

But in coordinated motion aren't we well, coordinating these two independent vectors into one motion? Just hitting the end point of a motion path at the right end time (i.e. area under the graph is correct) is not good enough, it has to be within an acceptable error at all points in time...

Backlash means something that should have an acceleration is undertaking zero velocity change so the other coordinated axis will have to account for that or you'll end up a zigzag in the coordinated motion path. Its either a zigzag or the other axis has to also stop and wait for the other axis to roll through its slop until they can run again, which begs the question of how to maintain constant feedrates.

I'm not saying there isn't a solution, I'm just wondering how its done - the control side of it more than the mechanical that is.

02 Jul 2011 00:36 #11049 by andypugh
1:1 wrote:

Backlash means something that should have an acceleration is undertaking zero velocity change

True, but by definition, at the point that the backlash is taken up the instantaneous velocity is zero. (as it is a reversal).

The problem is not coordinated motion, actually. The problem is cutter forces. The pathological case is miling a slot with blocks of material alternately on the left and right of the cutter. The speed is constant, but the cutting force keeps reversing.

02 Jul 2011 01:02 #11050 by 1:1
hmmmm,

ok cuter forces - something for me to ponder - but why isn't coordinated motion a problem ? off topic or actually not a real problem ?

02 Jul 2011 01:38 #11055 by andypugh
1:1 wrote:

why isn't coordinated motion a problem ? off topic or actually not a real problem ?

If all you are trying to do is motion, then it is just about possible to compensate. Because during a reversal of direction the axis speed is zero, there is plenty of available motor speed to run to the other side of the backlash. Also, the axis has just stopped, so there is time, too.

But, that is only true if all you want to do it move. Add cutting forces and it is all a lot more difficult. Clever G-code can help. After all, Manual machines always have lots of backlash, and you can think your way round it.

02 Jul 2011 01:57 #11056 by 1:1
I'm talking about coordinated motion between two axes - like a circle ...

The velocity profile (and accel and position) of a circle plotted in two orthagonal dimensions is two sine waves out of phase with each other by 90deg

If there is any lag or stopping in one wave (axis) the other must compensate 'immediately' (or within acceptable limits) so as to keep the resultant profile of the two axes combined (i.e. the circle) on path

While one axis hits the time it needs to change direction - i.e. backlash - the other one at its fastest velocity ... but to maintain the circle profile it needs to stop and wait for the other backlashing axis to catch up

max velocity does not equate to zero velocity - a contradiction in requests to the axis

I'm starting to repeat myself - hinting to me that one or both of us is either wrong, or not understanding the other (I'll note that although most cry the opposite that the internet is hardly a ideal place for communication )

02 Jul 2011 10:31 #11064 by andypugh
1:1 wrote:

While one axis hits the time it needs to change direction - i.e. backlash - the other one at its fastest velocity ... but to maintain the circle profile it needs to stop and wait for the other backlashing axis to catch up

The axis which is changing direction is passing through zero velocity. The effect of backlash with compensation is that the length of time that it is stationary is a little bit longer than it should be. If you imagine machining a circle then the effect will be to put an actual flat on an arc segment that is nearly flat anyway.

I think we are arguing different points. You are saying that backlash compensation is not good enough, I am saying that the mechanics makes backlash compensation easier than it might first appear. And that it is not good enough. And that in any case cutter forces make the whole question moot.

02 Jul 2011 19:53 #11079 by 1:1
I'm using argument as a method of understanding/learning - something I've learned my personality type does often - hope no offence taken :

I'll certainly look into the cutter forces complication, its something I'd like to get my head around. Got any reading online you could link to ?

02 Jul 2011 22:54 #11080 by andypugh
No references, but google for "climb milling" (which will presumably mention the alternative too)

03 Jul 2011 12:50 #11096 by BigJohnT
1:1 wrote:

I'm using argument as a method of understanding/learning - something I've learned my personality type does often - hope no offence taken :

I'll certainly look into the cutter forces complication, its something I'd like to get my head around. Got any reading online you could link to ?

Like Andy suggests just google a while... the fact remains that you can't program out slop in a CNC machine no matter how clever you are. Any backlash will show up in the final part as flats and chatter marks. Even drilling will not be precise as the drill bit (of any size big enough) will pull the part around within the limits of the backlash and produce a hole that is not round. Even vibrations of the spindle might move the part around while drilling depending on the machine.

Some clever people have used spring loaded double nuts on acme threads to reduce backlash to the point where the machine can have acceptable backlash for CNC operation.

John