Gear Reduction

Hey all,

Has anyone ever looked into a gear reduction for X & Y on the LR3? Just wondering if it would make more precise cuts through harder materials with a deeper pass depth or with larger diameter tools?

What would be the optimal ratio?
What is the max speeds that can be achieved using reduction?
Can the belt system handle the torque?
Will the cnc be more prone to deflection? (I’m not sure what it’s capabilities are)

I’d love to hear your input and pick someone’s brain.

A number of people use 1-start and 2-start lead screws on their LR2 machines rather than the recommended 4-start lead screws. The same should work for the LR3. These alternate lead screws give a 2X or 4X increase in torque for the Z axis, but you pay with a lower max Z feedrate.

Any gearing system will have backlash, introducing potential imprecision, and would also result in a lower max feedrate. There are other ways (besides gears) to increase torque if needed, but Ryan’s speed test indicates the stock machine is more capable than many believed.


The Z axis is plenty torqy for me. I was just wondering about X and Y. Regarding backlash wouldn’t the belt system introduce the same issue with any slack or stretch?

Topic is “Gear Reduction”, but wondering if you saw these cool belt mod experiments to help make more precise cuts? Not sure what reliable performance improvements they observed.

Have seen posts by folks accidentally using pulleys on X and Y axis with different teeth than the recommended 10mm GT2 16 Tooth Pulley mentioned in the LR3 docs. Not seen anyone intentionally change gearing for more torque? Seen Nema 23’s mentioned, but these relatively few folks are usually adventuring far away from what most do. No idea if these folks change and/or add more belts, or different mechanism.

wrt zero backlash gears… Recently stumbled onto harmonic drive strain wave gear for zero backlash? Neat, but not sure if/how you’d use in a LR.

Disclosure: Am relatively new to all of this stuff and am just using stock parts from V1E shop’s LR3 kit. Still don’t know what my weakest part(s) are that are worth optimizing, still learning…


I’m just a hobbyist with this kind of thing with most of my “knowledge” gained by reading forum topics and web pages. Someone with an engineering education will have a much firmer knowledge about these things. Here are some stream of conciousness thoughts:

  • The backlash from belts is primarily due to the movement of the teeth of the pulley in the grooves of the belt, not any sort of slack or stretching.

  • Assuming the machine still uses belts, gearing will add backlash. That is, backlash is compounded by the two forms of gears. There is backlash compensation in Marlin, so you might be able to tune the backlash out of the cutting.

  • The faster you spin the stepper motor, the less torque it has. Assume a 5:1 gear ratio. To move at the same feedrate, you will need to spin the motor 5X faster. You will gain torque from the gears, but, depending on your target feedrate, lose torque to spinning the stepper faster.

  • Various control boards have limits to the number of pulses they can generate per second. Higher gear ratios will mean reaching the limit of the board sooner. You might be able to compensate by reducing the microstepping, but then you are trading the accuracy that perhaps you hoped to get by the gearing for faster feedrates.

  • Geared steppers are expensive…$40 vs $15 or less for non-geared steppers, and the stepper holding torque might not be as high as the common, high torque steppers.

  • At some point, more torque will lead to significantly more deflection.


If you added a 2:1 ratio:

  • To move the cutter at the same speed, the motor rpms would have to move twice as fast.
  • Stepper motors lose torque with RPMs. So they would be more likely to skip.
  • But the same amount of motor torque would be twice as much at the tool.
  • Except for losses in the extra parts.
  • There would be more backlash. The pulleys/belts have a little, and you would be tripling it to make the 2:1. If you used herringbone printed gears or something, it might be even worse.
  • There would be more parts. More to assemble, more parts to fail, more parts to tune or align.
  • Friction in the gears won’t be reduced by the 1:2 ratio. The reduced torque from the motors going faster would have to fight that friction. This is something we see in the Z leadscrews. Maybe it would not be an issue on X,Y.

Most people are having trouble going faster, or dealing with flex, not ultimately with torque. Those problems are solved witg CAM techniques.

In my mind, I don’t think the juice is worth the squeeze. But I only have my mind because I have not done it and I would be happy to see it. I would be interested in seeing how you did it, and any measurements showing increased force at the tool before skipping steps, and any changes in the top speed.


Thanks Jeff, always appreciate your input as well as the others contributing.

I’m a lifelong tinkerer and love experimenting. I actually bought another set of steppers to test with.

I’m just thinking about trading top end feed rates for more aggressive cut depths, but I’m afraid the deflection will overcome the rigidity of the cnc since I’m running a 50"x100" work area.

As far as CAM I use Vectric Aspire since I get it for free as a dealer.

My reason for developing a set GT2 rails is based on observations of my full size LR2 that those with desktop LR’s might not see.

With my gantry at the mid-point of the Y-axis (10’ span) the belt tension is the nominally same on either side of the gantry. I have adjusters at both ends that I can dial in the tensions. However, as the gantry moves along the Y-axis the belt tension on the shorter length increases while it decreases on the longer length. So each time the motion reverses direction the delta of the tensions between the 2 lengths begins it’s shift in the opposite direction as well.

My rail is not a length of belt because the profile was not designed to mesh together that way. I printed PLA segments with a profile based on the GT2 specifications.

Tensioners are not needed for the long span of the axis but rather in the short span between the pulley and the rail. The motion is now consistent across the length of the axis.