Before I got into CNC I’d never considered the concept of Backlash. And yet, 100 years ago folks were writing books about the subject. Like this one. Sometimes you see little adjustment screws on machines (like router fences etc) and the adjustment screw includes a little spring. Turns out that this is just another example of an anti-backlash device.
Backlash is just the ‘play’ between two or more mechanical devices that need to transfer energy between themselves. Backlash is present in gears, rack and pinion systems and between nuts and screw threads. Consider two gears that inter-mesh. Spaces exist between the teeth due to the manufacturing tolerances of the gears. As you start rotating one of the gear wheels it will turn independently of its slave until its teeth have cleared the space and come into contact with the teeth of the slave gear. At that point any additional rotation on the drive gear will be transmitted as rotation to the slave. As you continue to turn the drive gear it successfully transmits rotational energy to the slave because the ‘slack’ in the system has been taken up.
Now consider what happens if we reverse the direction of the drive gear. As the gear rotates in the opposite direction the pressure between the teeth is relieved and rotation of the slave halts. The master continues to rotate but there is no movement of the slave until the opposite edges of the drive gear’s teeth come into contact with the teeth of the slave gear. As you continue to rotate in this direction the slave gear begins to move and will continue to move in unison with the drive gear.
For many mechanical systems this backlash is not important. In a drill, for example, the fact that a user may have to wait a few milliseconds after the motor starts turning for all the play in the gears to be taken up and the chuck to move, is inconsequential. In accurate positioning systems, however, this backlash will lead to errors. Consider that you may be trying to drive a gear an exact number of degrees. A stepper motor or servo will accurately turn the first gear the required number of degrees but backlash between that gear and its neighbour may mean that only some portion of that rotation may be transmitted. Reversal of direction will certainly cause the final output gear to lose some rotation. The bottom line is that you might turn the drive gear 5 degrees and that might result in a 5 degree rotation of the slave gear (let’s assume they’re the same size!) if the slack in the system has already been taken up. Alternatively, you might rotate the gear 5 degrees and for 3 of those degrees the teeth of the gear could be moving in the slack before they engage the teeth of the slave , hence the slave only moves 2 degrees! That’s the problem with backlash – it has a maximum value but it can be anything from 0 to that maximum value.
Anti backlash systems seek to correct for this backlash, thereby improving the overall accuracy of the final system.
Backlash in threaded rod and drive nut systems
Backlash in a threaded rod and drive nut combination is fundamentally the same. In contrast to a geared system, where rotation of one gear translates to rotation of another gear, with a threaded rod, rotation of the rod results in linear movement of the drive nut. The backlash in such a system can be seen in Figure 1.
Click on the image and you can see a cross-section of a threaded rod (purple) with a nut cross-section. Notice that there’s a gap between the thread of the nut and the thread of the rod. That’s the backlash. Figure 2 illustrates that by application of a constant load (force) on one side of the drive nut, the backlash can be removed. With the constant load in place, pressing the nut thread against the rod thread, the two threads stay in constant contact irrespective of which direction the threaded rod is rotated. No more backlash.
There are many ways to achieve this tensioning, one of which is illustrated in Figure 3. Two nuts are tensioned against one another. This can be achieved by actually tightening two nuts so that their faces press against each other. This approach can be used but the problem is that it introduces quite a bit of friction and there’s a tendency for the nuts to lockup if the threaded rod is not completely uniform (which it won’t be). A compromise approach is to create the tension between the two nuts using a spring. The other advantage of the spring solution is that it is able to adjust for wear of the threads. The spring needs to be selected so that it is strong enough to overcome any resistance encountered during movement of whatever it is that the drive nut is moving and weak enough that the friction caused by the loading is not a problem for the motor driving the threaded rod.
This is not the only way to overcome backlash, but it is the one that I chose for my second CNC design. Other methods include the use of a split-nut approach which I may try at a later stage.
My Anti-Backlash design
In my first CNC build I did not go for mechanical backlash compensation, instead choosing to compensate via software. This gave pretty reasonable results but not quite good enough for my intended application. For the second CNC build I decided to design some form of anti-backlash compensation. I looked at several approaches:
- Split Delrin nut design: A nice solution and relatively simple to construct but i saw two immediate problems with this one. First, i didn’t have any Delrin and secondly the solution does not compensate for wear.
- Split Delrin nut with spring for wear compensation: I like this solution because it provides the simplicity of the split nut but adds the wear compensation. Unfortunately i figured that i would need a lathe to make this part, which i didn’t have.
- Ballscrew system: Professional threaded rod system that addresses backlash by creating an incredibly close toleranced mating of nut and leadscrew using ball bearings. As you can imagine, the phrase ‘incredibly close toleranced’ can be substituted for the phrase ‘expensive’. This approach really produces some of the best accuracy but it’s just too expensive for my tastes. A possible solution here might be to scour eBay for potential second-hand bargains.
- Two nuts preloaded with springs: Machined out of Aluminium but with brass inserts for the threads (brass has lower friction and better wear characteristics than Aluminium).
Here are the 3D models for the design:
The rectangular sections are made out of solid Aluminium bar which is 20mm x 20mm square and 50mm in length. Although Aluminium can be tapped to accommodate my M8 Stainless Steel threaded rod, my fear was that it’s just not strong enough and would begin to wear rapidly. After a little research i discovered that Brass seems to be a common material that is used for drive nuts because of its excellent low friction properties. I considered making both A and B parts out of Brass – but that would have been pretty heavy! I had drawn up a very elegant design made completely out of brass, but it required a lathe which i had no access to. So my rather intricate compromise was to have brass inserts which would be threaded to M8. These round inserts would sit in a 16mm cylindrical hole cut into the Aluminium and would be secured to the Aluminium by 4x M4 grub screws; 2 on the top and 1 on each side of the aluminium blocks.
The assembly would be repeated to create two identical 50mm long, 20mm square Aluminium bars which had 16mm round M8 tapped brass inserts about their centre. These two ‘drive nuts’, labelled A and B in the diagrams would be held together by two M4 shafts. The ends of the M4 shafts which fit into the A piece would be tapped so that they would be secured to part A and not movable. The other ends would simply slide into the M4 holes on part B and so part B would be free to slide closer and further away from part A. Finally, springs would be fitted through the M4 shafts and these would act to push the two parts away from each other, hence providing the necessary tension required to push both sets of threads into contact with the stainless steel threaded rod threads (not shown in the diagram). The outcome should be much reduced backlash.
Initial tests with a basic Aluminium model proved successful in that construction was exhibit significantly less ‘play’ when compared to a similarly dimensioned simple nut. Specific backlash measurements were not made at this point.
…to be continued…