Lathe first retrofit

[cyclestart]

I just remembered, that backlash on the rotary axis may not be as import if that axis is just rotating in one direction. There may be a way to home the rotary axis in the same direction for subsequet operations.

That should work. If the spindle has movement while the drive system is stationary (free play) that's another problem and more difficult to solve.

Even so it would be interesting to know how to measure backlash on the rotary axis?

If you have a 4 jaw chuck clamp a bar across the face of the chuck and measure the movement. It's like a large indicator needle, a bit of math will translate that to degrees.
If it's a 3 jaw or some other system do whatever works.

EDIT/ Don't hit the bed with the bar. This could happen if you home the machine with the bar in place.

[ZeroBacklash]

Ok, so I just did some backlash tests on the machine and am getting some ridiculous numbers
aronud 2mm for the X axis and 1.5mm for the Z axis.
So I don't know if there is any general method or gcode to run the backlash test.
But I used the steps per mm dialog in Mach and tood the machine to move 5 mm. Of course I didn't continue with the motor tuning step (I cancelled it when it asked me how much did the machine move)
Then I told it to move -5mm. And used the difference to get the backlash number.
There are 3 levels of gears in the system, for eg. on X axis I have the bed moving on a rack and pinion, then then I have another set of gears that couple the X to the gear on the motor shaft and then the motor itself is internal gearing. All these seem to be adding to considerable backlash.
Is this machine usable if I use backlash compensation in Mach?

[Zorbit]

Do you have a dial indicator ?

[ZeroBacklash]

Do you have a dial indicator ?

I have a dial indicator and I did the measurements with that.

[hfjbuis]

Ok, so I just did some backlash tests on the machine and am getting some ridiculous numbers
aronud 2mm for the X axis and 1.5mm for the Z axis.

On my small lathe the backlash for X = 0.25 mm and for Z = 0.83 mm. The numbers you have found seems OK due to the fact you have some addition gearing.

There is no difference in the influence of backlash to accuracy wether you turn conventional or CNC. On a conventional lathe you can turn as accurate as on CNC. Despite backlash and no ball screws, i can "easily" (CNC) turn within 0.01 mm.

[ZeroBacklash]

On my small lathe the backlash for X = 0.25 mm and for Z = 0.83 mm. The numbers you have found seems OK due to the fact you have some addition gearing.

There is no difference in the influence of backlash to accuracy wether you turn conventional or CNC. On a conventional lathe you can turn as accurate as on CNC. Despite backlash and no ball screws, i can "easily" (CNC) turn within 0.01 mm.

Thanks for sharing your experience and I'm amazed to learn that you can turen within 10um with the said backlash?
So how does this workout, because as it should be the machine would lose count of the steps when the axis is changing direction and although the controller is sending steps the axis is not moving at all while it is eating up the backlash.
So kindly share what additional steps you are taking to get those kind of accuracies?

[cyclestart]

WARNING: turn down the volume if clicking on the vid

That kind of profiling won't love backlash. There are software solutions to backlash but eliminating it in the machine is best.

[hfjbuis]

So kindly share what additional steps you are taking to get those kind of accuracies?

• During turning the bed is protected to avoid chips coming between the bed and the carriage
• On the small lathe the carriage is pulled down on the bed (no play) by 2 preloaded ball bearings mounted on a 3 mm steel bar that functions as a spring. On the large lathe, the weight of the carriage itself is large enough to keep the carriage on the bed (no play) during turning.
• During CNC turning the top slide gib is tightened to avoid play.
• The final pass is always 0.02 mm when using insert tools or 0.01 mm when using HSS tools.

I do CNC turning the same way I do conventional turning. That means backlash is eliminated during the turning pass by always moving to tool towards the workpiece. That way the tool determines the cutting depth, not the workpiece.

For example turning a outside taper, the thicker part is at the chuck side (Z-).

Wrong procedure:
Start at the thinner side (end of the bar) and turn in the direction of the spindle (Z-). To position the tool (cutting depth) you move in the direction of X-. During turning de tool is gradually moved in the direction of X+. The change in direction causes backlash. The workpiece has to push the tool in position. The results are depending on the backlash, the drag of the cross slide and the material you are turning etc.

Right procedure:
Start at the thicker side (spindle) and turn in the direction of the end of the bar (Z+). To position the tool (cutting depth) you move in the direction of X-. During turning de tool is gradually moved in the direction of X-. There is no change in direction so there is no backlash. The tool determines the cutting depth.


If the lathe was equipped with backlash free ball crews and driven directly by the stepper, then there would be "no" backlash and it wouldn't matter how you turned the taper. However, if you have trapezoid spindles and geared steppers like me, you have to CNC turn the "conventional way" to get good results.

[ZeroBacklash]

• During turning the bed is protected to avoid chips coming between the bed and the carriage
• On the small lathe the carriage is pulled down on the bed (no play) by 2 preloaded ball bearings mounted on a 3 mm steel bar that functions as a spring. On the large lathe, the weight of the carriage itself is large enough to keep the carriage on the bed (no play) during turning.
• During CNC turning the top slide gib is tightened to avoid play.
• The final pass is always 0.02 mm when using insert tools or 0.01 mm when using HSS tools.

I do CNC turning the same way I do conventional turning. That means backlash is eliminated during the turning pass by always moving to tool towards the workpiece. That way the tool determines the cutting depth, not the workpiece.

For example turning a outside taper, the thicker part is at the chuck side (Z-).

Wrong procedure:
Start at the thinner side (end of the bar) and turn in the direction of the spindle (Z-). To position the tool (cutting depth) you move in the direction of X-. During turning de tool is gradually moved in the direction of X+. The change in direction causes backlash. The workpiece has to push the tool in position. The results are depending on the backlash, the drag of the cross slide and the material you are turning etc.

Right procedure:
Start at the thicker side (spindle) and turn in the direction of the end of the bar (Z+). To position the tool (cutting depth) you move in the direction of X-. During turning de tool is gradually moved in the direction of X-. There is no change in direction so there is no backlash. The tool determines the cutting depth.


If the lathe was equipped with backlash free ball crews and driven directly by the stepper, then there would be "no" backlash and it wouldn't matter how you turned the taper. However, if you have trapezoid spindles and geared steppers like me, you have to CNC turn the "conventional way" to get good results.

Those are some very good suggestions.
Lets say that we are turning a 10mm rod down to 8mm.
So we will be doing it is multiple passes of Z axis and X axis right. On each pass X will move deeper into the workpiece by 0.2mm or so?
So lets say we start at the chuck end of the job give a feed on X for 0.2 and do a cutting move towards the end of the rod.
Now once we have reached the end of the rod, we would have to return Z back to the start position (towards the chuck) right and hence a change in direction? So how would we eliminate backlash here?
I agree that for turning there is no change of direction in X and that's why the turned piece would be accurate in diameter. However the Z end, of the workpiece at the chuck would have a fillet and not a right angled step. I guess we cannot eliminate that right and we would need an additional step to lathe the fillet down to a rgiht angle.

[hfjbuis]

[QUOTE=ZeroBacklash;2225632]Those are some very good suggestions.
Now once we have reached the end of the rod, we would have to return Z back to the start position (towards the chuck) right and hence a change in direction? So how would we eliminate backlash here?

On a manual lathe, at the end of the cut, you would retract X beyond the starting point, move Z back beyond the starting point, move Z to the starting point, move X to the starting point and then start cutting.
You could do the same during CNC cutting. Alternatively you could move X back to the starting point + the backlash X, move Z back to the starting point + backlash Z, move X to the starting point and start cutting.

[ZeroBacklash]

[QUOTE=hfjbuis;2225650]

Those are some very good suggestions.
Now once we have reached the end of the rod, we would have to return Z back to the start position (towards the chuck) right and hence a change in direction? So how would we eliminate backlash here?

On a manual lathe, at the end of the cut, you would retract X beyond the starting point, move Z back beyond the starting point, move Z to the starting point, move X to the starting point and then start cutting.
You could do the same during CNC cutting. Alternatively you could move X back to the starting point + the backlash X, move Z back to the starting point + backlash Z, move X to the starting point and start cutting.

So would'nt this need some kind of absolute sensor where the machine will return after each pass. That way the motors no longer affected by lost "steps" and blindly return to the absolute 0 position after a cutting pass.
Somthing like what you suggested.

Set a absolute 0 for them machine on the surface of the workpiece where you want to start cutting. We may need some kind of sensor / limit switch for this.
so lets say if the chuck face is Z 0. And I want to start cutting from Z10. So I will position the sensor at Z10 and maybe another sensor for the X?
What kind of sensors would be accurate enought, I guess the limit switches levers will have sufficient play to defeat this method. May hall effect sensor or something on those lines. And they would be needed to be wired to the inputs of the CNC controller.

[hfjbuis]

So would'nt this need some kind of absolute sensor where the machine will return after each pass. That way the motors no longer affected by lost "steps" and blindly return to the absolute 0 position after a cutting pass.

The controller knows the position based on the steps send to the stepper. If you overload the stepper, the part you make is out of spec regardless if you correct the position afterwards by some kind of closed loop system.

What kind of sensors would be accurate enought

If your controller can handle the input signals, you can use a glass ruler for absolute positioning. There is also a cheaper less accurate variant the kind like Igaging sells.

I don't use limit switches and I have never read how accurate they really are. But considering the technic used in a limit switch (mechanical, optical, hal, etc) , any DRO is more accurate.

[ZeroBacklash]

The controller knows the position based on the steps send to the stepper. If you overload the stepper, the part you make is out of spec regardless if you correct the position afterwards by some kind of closed loop system.

Actually I was worried of the lost steps due to backlash. Where in the controller is sending steps to the driver but the bed is not moving until the backlash is eaten up. And and the controller doesn't know about those lost steps since it thinks it has achieved its target position.

[hfjbuis]

The controller has to put the stepper at the requested position. It is the task of the controlling software (CAM) to compensate for backlash by adding the backlash to the requested position at the appropriate time depending on the direction of move and the direction of the force to the tool.

[cyclestart]

It's a good thing backlash doesn't accumulate. If a manual machining strategy is used all will be good. My feeling is it places some limitations on contouring but that's not typical lathe work.
btw: CAM generally means software that creates a part program, the way hfjbuis used the term isn't wrong but it's less common.

[skrubol]

Well, for a lathe, it could be done in CAM for simple stuff (no curves that need to reverse direction.) Just make sure your retracts all go at least the amount of backlash beyond where they need to go. Machine controllers (Mach, LinuxCNC, etc.) usually have backlash compensation in them as well, and best to use it if you need it.

[hfjbuis]

My feeling is it places some limitations on contouring but that's not typical lathe work

You are right, backlash is killing for turning contouring. I can turn a sphere despite the lack of ball screws. For this to work, the cutting depth must be so low that the drag of the carriage is greater than the cutting force. If you do turning contouring often, the easy way is to fit ball screws.

[ZeroBacklash]

Well, for a lathe, it could be done in CAM for simple stuff (no curves that need to reverse direction.) Just make sure your retracts all go at least the amount of backlash beyond where they need to go. Machine controllers (Mach, LinuxCNC, etc.) usually have backlash compensation in them as well, and best to use it if you need it.

So how does this work out? In this technique better than the one where we
1. retract to beyond an absolute 0 set position on the machine (Defined by a sensor)
2. Move to the absolute 0 set position.
3. Start the next turning pass from the 0 position. (without any change in direction)