[denvermatt]

I'm using Mach 4 and a PMDX SmartBOB. I finally got my Z-Axis running correctly, and I was in the process of calibrating it when I discovered that no matter what value I put in the backlash compensation field it didn't make any difference in my backlash.

After doing some research I discovered that Mach 4 doesn't really do any compensation for backlash at all, and have instead thrown this down to the plugins to handle. The field in the motor controller section of the configuration tab just collects the information, but does nothing with it.

PMDX lists this as a "Known Issue" (PMDX-SmartBOB-USB Mach4 Plug-In Release Notes) in their release notes, but have not produced a fix for it despite having it in their known issues list for over 2 years.

This is extremely frustrating. I really don't want to go backwards to Mach 3, and I've put a lot of money into the PMDX Board.

Question: does anyone have a recommendation on how to get around this? The backlash is more than I'm willing to accept.

[ger21]

1) fix the machine.
2) switch to a motion controller that supports backlash comp.
3) ditch Mach4 and move to UCCNC or LinuxCNC

[louieatienza]

I'm using Mach 4 and a PMDX SmartBOB. I finally got my Z-Axis running correctly, and I was in the process of calibrating it when I discovered that no matter what value I put in the backlash compensation field it didn't make any difference in my backlash.

After doing some research I discovered that Mach 4 doesn't really do any compensation for backlash at all, and have instead thrown this down to the plugins to handle. The field in the motor controller section of the configuration tab just collects the information, but does nothing with it.

PMDX lists this as a "Known Issue" (PMDX-SmartBOB-USB Mach4 Plug-In Release Notes) in their release notes, but have not produced a fix for it despite having it in their known issues list for over 2 years.

This is extremely frustrating. I really don't want to go backwards to Mach 3, and I've put a lot of money into the PMDX Board.

Question: does anyone have a recommendation on how to get around this? The backlash is more than I'm willing to accept.

Fix the backlash. Way better than any software compensation, and you'll have better parts and finish....

[RCaffin]

1) fix the machine.

How very strange. I wonder WHY Mach4 does not support backlash correction?

Could it be that backlash correction NEVER works satisfactorily anyhow?
Could it be because trying to run a decent CNC machine is simply incompatible with the whole concept of backlash correction?
Yup.

Cheers
Roger

[ger21]

How very strange. I wonder WHY Mach4 does not support backlash correction?

They do, but it needs to be handled by the motion controller, as it can't be buffered.
Apparently, PMDX does not see it as a priority.

[RCaffin]

Hi Gerry

Thanks for the correction, and yes, I do understand.
I seem to remember that Gary of Warp9 had the same opinion (or lack of urgency) regarding backlash handling in the ESS.
And even in Mach3, the backlash correction only covered some operations, not all.

I wonder whether one could point out the analogy with trying to steer a car down a busy city road with a loose steering box?

Cheers
Roger

[OlfCNC]

Hi Gerry

Thanks for the correction, and yes, I do understand.
I seem to remember that Gary of Warp9 had the same opinion (or lack of urgency) regarding backlash handling in the ESS.
And even in Mach3, the backlash correction only covered some operations, not all.

I wonder whether one could point out the analogy with trying to steer a car down a busy city road with a loose steering box?

Cheers
Roger

Whom it may concern the UC300ETH handles the backlash very well. Even with Mach3 it works without stopping the axes not like how it works with LPT ports.
When I first started using it I thought the backlash is not working, because the controller adds the backlash so smoothly.

[ger21]

This is one of the main reasons that I'm not using Mach4.
Which functions you have available, and which ones actually work, vary depending on which motion controller you are using.
So, in a lot of ways, you're Mach4 experience has more to do with the motion controller vendor's support, than Artsoft.

[louieatienza]

Hi Gerry

Thanks for the correction, and yes, I do understand.
I seem to remember that Gary of Warp9 had the same opinion (or lack of urgency) regarding backlash handling in the ESS.
And even in Mach3, the backlash correction only covered some operations, not all.

I wonder whether one could point out the analogy with trying to steer a car down a busy city road with a loose steering box?

Cheers
Roger

Great analogy... you either start turning sooner, or get it fixed. Obviously one is the better solution.

[samco]

Granted in a perfect world backlash would be zero - backlash comp works though. (at least I know it does with linuxcnc)

Linuxcnc applies the backlash at every axis reversal.

This cut shows a circle cut 2 times - one with backlash comp off - and one with backlash comp on. (can you tell the difference?)

sam

[RCaffin]

Yes, there are some narrow classes of situations where you can make backlash compensation work. Cutting circles may be one of them. But that is because when you are cutting a circle you are always driving the backlash axis against the nut. Even so, I think you need light cuts and big inertia to get anything usable.

Try using an old manual mill with backlash - ACME thread is always good for this. Now try switching from conventional milinng to climb milling, and watch the whole table jump. A very simple case of this is to mill back and forth in a straight line against an edge. In one direction you are conventional milling, and all goes well. But when you reverse direction you switch into climb milling, and the table jumps. You can smash a cutter like this.

Rely on backlash correction if you want - it does not worry or concern me. But be honest and call it what it is: a poor SW hack to try to cover up a serious hardware fault.

Cheers
Roger

[samco]

This is why pretty much all commercial controls impliment backlash compinsation? I agree backlash is bad. Fix it if you can. But backlash compinsation in the control does work well when implimented correctly.. (yes you can break cutters if backlash is too much 'and' the table can be pulled around)

Sam

Yes, there are some narrow classes of situations where you can make backlash compensation work. Cutting circles may be one of them. But that is because when you are cutting a circle you are always driving the backlash axis against the nut. Even so, I think you need light cuts and big inertia to get anything usable.

Try using an old manual mill with backlash - ACME thread is always good for this. Now try switching from conventional milinng to climb milling, and watch the whole table jump. A very simple case of this is to mill back and forth in a straight line against an edge. In one direction you are conventional milling, and all goes well. But when you reverse direction you switch into climb milling, and the table jumps. You can smash a cutter like this.

Rely on backlash correction if you want - it does not worry or concern me. But be honest and call it what it is: a poor SW hack to try to cover up a serious hardware fault.

Cheers
Roger

[denvermatt]

Ok, I get it and agree that ideally I would eliminate as much backlash in the machine as possible. And believe me, I'm working on it. However I don't think you can completely eliminate backlash - some always exists otherwise the mechanism would bind up. Additionally, I'm not working with the best machine to begin with. So my options are limited.

I just wanted to warn others who may be considering going this route that although some of the SALES LITERATURE for Mach 4 says that it does backlash compensation, it DOESN'T. And that before you spend hundreds of dollars on this stuff, you should be aware that it may not work as you expect.

Thank you Gerry for recommendations on alternatives to Mach 4. I had difficulty getting LinuxCNC working on this machine and ultimately installed Windows so I could run Mach 4. It appears that UCCNC requires that I change out my BOB entirely which is undesirable. Although I will keep it under consideration if I can't get the backlash in the machine reduced to an acceptable level.

Thanks guys!

>Matt

[samco]

Actually you can. Pre-loading ball screws (dual nut with loading between them) gives you low friction zero backlash.

This is with no backlash comp.. (and a rather large machine) - that indicator reads in tenths of an inch.

However I don't think you can completely eliminate backlash - some always exists otherwise the mechanism would bind up.

>Matt

[RCaffin]

However I don't think you can completely eliminate backlash - some always exists otherwise the mechanism would bind up.
Ah well - you can and you can't.

You can:
Double opposing ball nuts and double AC bearings will give you zero backlash when properly adjusted - in the conventional practical sense. The key point here is that you must switch from sliding surfaces - which will bind or wear badly, to rolling contacts.

You can't:
Rolling contacts in the above rely on elasticity in the steel to work. Without some elasticity even ball races would not work. But having elasticity in the bearings means that when you have high enough loads there will be some movement, due at least to the elasticity. This could mean that a 50 kg load change could cause a 1 micrometre movement. Mind you, to call this 'backlash' is technically wrong: it is 'compliance'.

In practice: those with carefully engineered 'rigid' machines will happily expect repeatability below 5 microns. If you want anything better than that, you are going to need air conditioning, temperature stabilisation of the machine itself (eg water channels through the castings), some rather expensive rails (or air bearings), ball bearing leadscrews (of extreme accuracy), and seriously expensive metrology equipment (usually based on laser interference). This regime is where a factor of two improvement can mean a factor of 10 increase in the cost. Perhaps not needed for a hobby?

Cheers

[Daniel92563]

Vital systems integra board seems to support backlash compensation for mach4.

here's a nice overview
https://youtu.be/aWv4R_ftbIs

I just ordered the hobby version, I will try it out this weekend.

[louieatienza]

However I don't think you can completely eliminate backlash - some always exists otherwise the mechanism would bind up. Additionally, I'm not working with the best machine to begin with. So my options are limited.

It's pretty common in the precision ground ballscrew world to have screws that exhibit zero-backlash properties. Even without a zero-backlash nut, the amount of backlash on them is smaller by a factor of four or more, simply because of the precision grinding. The last amounts of backlash are either taken up with a double-nut similar to that of a rolled ballscrew, or a singular nut with the bearing races offset from one another, creating a preload.

Some patience and maybe some "sniping" could lead to a very nice screw for very cheap on eBay...

[samco]

Something on the linuxcnc side is $199... And almost unlimited expansion...

https://mesaus.com/index.php?route=p...&product_id=66

sam

[denvermatt]

I do have a ballscrew/gear set. The Z-axis I'm working on does not have two opposing thrust bearings, only one which is probably where the backlash is coming from.

Based on your feedback, I'll have to work on making a new mount that will allow me to have the opposing bearings.