Belt Stretch and Stepper Holding Measured

is this new motor a complete drop in as-is, or was there more work to replace?
(wondering if I should click on amazon’s next day delivery :wink: )

It’s very straightforward to replace the X motor. It has the same mounting holes, but is about twice as long.

You’ll need an 8mm, GT2, 20-tooth pulley (the stock servo is 6.35mm or 1/4"):

And if you opt for one of the 10mm-wide pulleys, like I did, might as well grab a 10mm steel belt:

The 10mm belt fits the existing idler pulleys and belt clamps fine (little bit of squeezing on the belt clamps to get the belt threaded).

The stepper does not include an appropriate connector, it is just bare wires. I snipped the connector off the stock stepper and connected it to the new stepper with four 2-way lever nuts.

I’ve only run with the new stepper for maybe half an hour total. No issues so far… should it turn out to have some unseen problem, I won’t have wasted much money or time.

I went ahead and measured deflection at a few other places, to be sure it’s the belts stretching and not something else.

Deflection at the HDZ router clamp was identical to deflection where I measured it on the router, so no flex there. The X-to-Y plates + Y V-wheels account for 3.3% of the deflection, and the Y extrusions + Y-to-base-plate brackets account for 1.7%. So, nothing significant relative to the belts.

Any ideas for how to stiffen the belts, short of switching to ball screws? Is this large amount of stretch not really a problem in practice and I’m making a big deal out of nothing? It seems like a good explanation for the visible deflection I see in the cutter when making heavy cuts.

1 Like

Are you using V-Wheels on the X and Y axes?

Yes. Planning to add linear rails in the future, but I don’t think that’d help with the belt stretch I’m seeing. If I push on the HDZ along the X axis, the X stepper pulley doesn’t move, but the X belt and X V-wheels do move (and the HDZ along with them).

I suppose I could try even wider belts, which I suspect would give a proportional improvement, but not solve the problem. Maybe glueing the belts to the extrusion and creating a makeshift rack and pinion? :slight_smile:

Check out this interesting discussion from a while ago about doing exactly that.

I wonder if this “worst case” belt stretch is relevant though, since most (all?) people doing high-precision parts will use a separate (and light) finishing pass to get to final dimensions, and very little force = very little stretch then?.

Some people seem to achieve 0.0002" tolerances on parts with a Shapeoko with regular belts.

This being said, these comments comes from the guy who spent an unreasonable amount of time mapping the belt stretch on his machine, to finally realize that it’s way simpler to just run a job, measure offsets to target dimensions, adjust stock to leave/offsets in the design, rerun, done. It was a fun journey though.


Thank you for the great links. I was brainstorming some ideas last night and this link I found in there

already explored the ideas I was thinking of:

  1. Double belts, bottom belt glued to the extrusion, top belt meshes against it
  2. Similar to 1, except the top belt is just a “tank tread” looped around some pulleys
  3. Glue a belt onto a raised portion of the extrusion and press the servo pulley against it, forming a simple rack and pinion

Interestingly, the author of that post started with double 15mm belts, ran that way for a while, then found it was too much trouble to maintain and switched to a single 9mm belt. I might do a quick and dirty version of #1 this weekend to see if it helps any with my manual stretch test.

I’m not so worried about tolerances of the finished part. I think you are right that the SO3 is actually pretty good in that department, with some extra care… and conservative settings + extra finishing passes.

For the current piece I’m working on, those extra passes and conservative cuts are costly, though… turning 10+ hours of machining into 20+. It wouldn’t be a big deal if it was 20 minutes vs 40 :slight_smile:

So, I’ve been cutting pretty aggressively - 5mm DOC, 1.25mm WOC, 150 +/- 30 IPM on a .25" single flute. The Makita seems to have enough power, and it’s not pushing the theoretical limits of the SO3’s steppers. In practice, I’m seeing some bad signs:

  1. Carbide end mills are snapping easily, and don’t seem to maintain a constant engagement
  2. I can see the end mill (actually, the HDZ itself!) visibly deflect in XY when entering cuts (in the range of .5 to 2mm I’m guessing)
  3. I occasionally lose a step here or there (hopefully solved by my new X stepper)

The stretch I’m seeing with an aggressive cut is easily enough to exceed the recommended max chipload for the end mill, too, which could be problematic for subsequent passes.

I’m about to upgrade to a spindle that should (in theory) give me 3-4x the MRR, within the SO3’s force limits. However, the belt stretch has me worried that I won’t be able to utilize anywhere near that much force.


What router speed? Are you sure that it’s belt stretching and not other deflections (V-Wheels, etc.) like I suspect? Minimizing torque on the router mount, gantry, and side extrusions would help minimize its impact on the belt stretch measurements.

1 Like

30,000 rpm.

I was worried it might be something other than the belts, too. I repeated my tests, this time measured at different locations:

X deflection (mm) per 1 lb of force:
.028 @ router body
.030 @ HDZ router clamp (I suspect it is, within the margin of error, identical to the router body)
.00198 @ left XY plate, Z position near mid-extrusion (between the upper/lower V-wheels)
.00072 @ left Y extrusion
.0015 @ right XY plate
.00046 @ right Y extrusion

(Keep in mind, this is me manually pulling on a luggage scale while leaning over to look at the indicator… the numbers are not perfect)

The V-wheels, extrusions, end plates, HDZ, router clamp, etc. seem to account for a measurable amount of deflection that I’d want to correct for eventually, but less than 10% of what I’m seeing at the HDZ and router.

This is with my 10mm steel belts, tightened as much as I could (probably pre-stretched a bit). The stock belts are 2-4x the stretch, depending on tightness.

What got me thinking it was the belts in the first place was that I could see the X V-wheels rotate and the X belt tighten / loosen on either side, but the X servo pulley was not rotating and the portion of the belt engaged with the pulley was also not moving. So, the HDZ was moving, the belt and V-wheels were moving, but it wasn’t because of the stepper giving way or the surrounding SO3 frame bending. I didn’t precisely measure it at the time, but to the naked eye, the amount of belt stretch and V-wheel rotation I saw was in the same order of magnitude as what I ended up measuring with my experiment.

1 Like

So pulling along the Y-Axis on an endmill close to the waste-board causes essentially the same measured deflection of the endmill tip as pulling on the center of the gantry?

No, I didn’t measure at the end mill tip - I measured at the router body just under the router clamp - that’s the first number. The second number is taken slightly higher, at the carriage’s router holder.

In all cases, the force was applied at the bottom of the router body (below the carriage center, but above the collet and end mill).

Force and measurement are along the X axis.

I’m trying to measure X axis deflection/stretch that is not attributed to the end mill itself flexing.

Perhaps to your point, the total deflection would be even larger if I applied the force to (and measured at) the end mill tip.

To clarify, I’m always applying the force at the same location. The different numbers are measurements from my indicator being placed at other locations while I repeat this test.

A 1/4 endmill will have negligible flex. Machine deflection at the endmill cutting edge is what really matters.

1 Like

I can measure at the end mill… but surely that will be even larger than the .5mm+ deflection I see at the router body?

Edit: deflection at the end mill cutting edge is what ultimately matters, but I already know that I have an excessive amount of deflection there, much more than I can attribute to end mill flex or runout. I’m trying to locate the source of this, hence the measurements at the different locations, which seem to validate my guess that it is the belts. Yes, the force while cutting, which creates the deflection, will originate at the cutter itself… but for the purpose of answering “are the belts the weak link”, there shouldn’t be much of a difference between pulling on the bottom of the router body vs 1.5" lower on the cutter.

That’s my point, I suspect you’ll see a lot more deflection where the router bit would normally contact the workpiece because of the resulting force increase on the V-wheels The distance from the cutting edge of the endmill to the V-wheels (the length of the lever arm) determines the magnitude of the increase. Your 0.5mm is a lot less than what @luc.onthego and I measured and documented in the previously linked thread. But, we didn’t have HDZs.

OOPs! I just noticed that you’re only been testing the X-axis! So the forces have been causing radial loads on the V-wheels rather than the far more challenging side loads that Y-axis forces would produce. Try the Y axis?

This is about cutting forces as well - slow your feed rate, and there will be less runout and deflection. The combination of the two consumes cuttters faster than anything. You should be able to hold ~3-4 thousandths without much effort - as you have shown, the trick is to keep cutting forces down. I’ve noticed the steel belts also stretch quite a bit less when tensioning than the glass belts. The kevlar belts I find to be ok, but I’ve broken a couple, and haven’t broken any of the steel ones.

Also keep in mind that holding torque from the steppers (as well as stepping torque) is directly related to available current from the drivers. The stock drivers are adequate for the stock motors. If you want to go much bigger and upgraded driver will be needed to get all the toque the stepper is rated for (which is a lot of work to do - you need to go to an external driver, which really necessitates a different controller board) . It’ll still work, and you’ll still get more from an upgraded driver without it, but you’ll get more with a properly sized driver.

Yes, but I’m intentionally testing forces parallel to the X axis, because I’m trying to measure X belt stretch. (I’m sure the Y belts stretch, too, but I’m focusing on X for now).

I’m not as concerned about the V-wheels flexing. I’m planning to switch to linear rails soon, which should remove that problem. But, I am very concerned about the belts stretching under load. I had been assuming the steel belts would be more than adequate for cutting forces of 20lbs or less, and that there was no significant value in switching to some other drive system (ball screws). But, the stretch in that range is much larger than I thought.

I did re-measure with the force applied to the cutter tip, as you suggested. That didn’t significantly change the results. I did notice that there is a bit of rotational deflection, likely due to the V-wheels, but it’s not as large as the belt stretch (and I’m less concerned as the linear rails would likely remove this).

@mikep yeah, a controller and driver upgrade is likely in my future. I’ve seen your posts on that, which have helped a lot. Also, yes, I could certainly hit that level of precision if I backed off on how aggressive my toolpaths are, but… my concern is that I’m upgrading to a more powerful spindle in the hopes of reducing overall job time, which will necessitate more cutting force in order to use the spindle’s full power. I feel like I need to get the major sources of deflection under control a bit, or I’ll just make a mess of my cutters and workpiece with that extra power.

Also, @gmack, I did read through the other thread and the deflection you saw pulling in Y (about 1.5mm or 1/16”?) is in line with the belt stretch measurements I saw for the stock belts…

One thing I didn’t see in your thread (or may have missed) - how did you conclude it was the X-axis V-wheels at fault, and not the Y-axis belts stretching?

I did a quick test of using two belts, with the bottom belt adhered to the extrusion (actually, sitting on a 1.5mm shim to make things tight) and meshing into the top belt as the carriage rolls over. It reduced X deflection in my test by 30%, but it also noticeably increased “rolling resistance” of the carriage by about the same amount when manually sliding it around… so I’m not convinced this did anything other than just scale the required force. Similar to over-tightening the V-wheels.


Yup, at 10 lbf. But based on how it “felt” and experience/intuition (which may have been misguided) I didn’t consider drive belt stretch. Unfortunately Gates seemingly doesn’t make modulus of elasticity data for their GT2 (or newer GT3) belts readily available. But, its easy enough to measure to enable simple belt comparisons. Just clamp some nearly fully open calipers to the belt and measure the length increase as a function of force (with weights?). What size Shapeoko do you have?