I have turned my burly into something else

Hi all. I haven’t been on these forums for some time. No real reason, just the ebb and flow of life and hobbies.

Last time I was on, I was having some major accuracy issues that only got worse. I didn’t always have these issues, and I suspect that they may have even be fixable…but I always knew I built my burly larger than advisable (24x48) so I got this crazy idea in my head to beef up the machine.

I went back and forth between building a primo, trying to strengthen the burly, or doing something completely different. The path I chose is something of a hybrid solution.

For z, I went with a “radial” design, rather than a cantilever. I designed a “sleeve” for the dw660 that allow the unit to slide on 8mm rails and bearing. This is kind of a 3d printer inspired solution.

For x, I went with import clone HGR20 linear slides with a 5mm ball screw. This is the biggest departure. I had to employ a 1:3.75 gear ratio between the stepper and screw in order to preserve torque and this jog speed, as well as go to a single nema 23 motor. This is mounted to a large plywood plate. If this machine pans out, I may cut a rev2 of this plate, and possibly opt for stiffer material. Still, the rails are BEEFY, and lend quite a bit of stiffness here.

For y, I stuck with the mpcnc rollers. I moved the old x rollers to y, where I now employ 4, at the corners of the big plate. I like the torsional stability this provides. Downside is, square is built in…no adjustment. I think I got close, and will improve either by iteration, or skew adjustment in marlin (never messed with it before)

I’m almost at the point of testing. I need to re route and tension belts, address cable management, lower the legs a tad, and add more supports to the log rails (the gantry IS HEAVY)

I hope it all works out. The burly and this community was a hell of a gateway drug. Also, I don’t mean this as an indictment of the mpcnc, I’ve made a ton of cool shit with it! Kinda just taking the next step. If it doesn’t work, I kept all the old pieces intact, or I’ll just build a primo or LR3!

(The table is very messy, might be hard to see what’s going on here)


I am generally a fan of ‘unintended uses’ and this I would say qualifies.

How are you driving the NEMA23 motor and what are its specs? I was under the impression that the usual DRV8825 (or similar) couldn’t drive the current needed to get full performance out of NEMA23 motors, but I haven’t really researched it and I would love to be wrong.

Yeah, the screw drive was the big snag in this project. The screw came packaged with the HGR20 slides and when I clicked “buy” I didnt even consider the value of the screw’s lead. At 5mm, it is over 6x “slower” than a 16t pulley, which has a 32mm lead.

At first, I tried to drive the screw directly with one of the NEMA 17s. Obviously this didnt work. The screw needs to turn at around 500 RPM to achieve jog speeds. I thought at first that the 17s just didnt have the torque, so imagine my surprise when the 23 couldnt do it either. At some point in troubleshooting I then realized that neither motor would turn that fast while NOT connected to anything. They lose timing. The gear ratio solved that.

Long story short, the NEMA 23 I bought is on the smaller side, and calls for 2.8A (compared to 1.5 for the 17s). Its rated at 269 oz in vs 76 for the 17s. I am driving it with the mini Rambo. This board can do 5A total for the motors across all channels. The default current is 1A for each axis. So, I can run the 23 at 1.5A or 2A safely, and still have enough current to feed the other drives.

I am not getting the FULL power out of the 23, but even at less than full power, its much more torquey than the 17s, so it should do the job with a big factor of safety. In testing, it barely gets warm. Overkill, maybe, but safer than trying to feed 1.5A to the 17. The NEMA 23 was a $30 motor. Not too bad, especially if you only need one.


I think (though not sure) the issue with trying to drive a stepper too fast comes from a high microstepping multiplier. IIRC I was able to get a stepper with a planetary gear reducer to run faster by going from 32x to 4x on my driver configuration (though I was changing all sorts of things at the time). You would loose rotational position resolution but not torque capacity. When you run a speed increasing belt you loose both.

If you have adequate torque from the nema 23 your config is a non issue, but I figured I throw it out there for someone to confirm or shoot down.

1 Like

Interesting experience, thanks for sharing.

At first, I thought microstepping was my problem as as well, so I kicked the multiplier down to 4, then 2, then 1. For me, it made no difference. As soon as I changed the steps/mm to suit the given microstep setting, the motor would again lock up during jogs.

It occured to me that rpm is rpm, regardless of your microstep settings (as long as the CONTROLLER has enough frequency to keep up with the pulse per second required). Observing the same behavior at both 1x and 16x microstepping proved the issue wasn’t controller frequency.

Based on the reading I have done, the issue is likely the power required to run the motor that fast, both the current AND the voltage. (More V = more torque). There are also some EMF and inductance aspects to it that I don’t really understand either, but the consensus of the suggestions I saw on Google was to up the power if you could. This is where Jaime’s comment comes in. I’m likely not getting FULL power out of the nema 23 with the mini Rambo, but I’m getting enough extra to overcome the extra torque required to increase screw speed via pulley ratio, while keeping the motor RPMs down.

I also noticed that dropping the acceleration helped to am extent. Timing was being lost early in move, as the pulses were likely outpacing the inertial forces resisting movement when starting from a rest. Unfortunately, the effect of lowering acceleration alone never got me near 500 rpm, even when the accel values became so low as to be unusable.

I think this is where the power came into play. More power allowed for more acceleration via overcoming resting inertia.

Just my $.02. still not at accuracy testing, but I think the belt reduction will work. The belts have a constant pitch, so the ratio should be consistent with no slippage. I wouldn’t expect the pulleys to have an adverse effect on accuracy. Backlash in the lead screw/nut on the other hand…

A better way to think of it is more V = more speed, and more A = more torque.

What voltage are you running for the motor supply? If you are running 12V then its an easy upgrade to 24 and you should get higher top speeds. Or if you are happy with the speed, you can change the pulley and get more torque.

1 Like

12v. I am happy with the pulley and the speed (now, after trial error and learning). Was trying to avoid buying a new power supply. I had the pulleys lying around and the mounts are printed (close to free).

Of course if I bought the new psu maybe I would have needed the bigger motor. Oh well, done is done.

Had another weird problem tonight. My mini-rambo won’t save my e-steps and current limit settings to eeprom. M500 command results in “eeprom datasize error”. I tried to reset with M502, but no dice.

On the bright side, principal assembly is finished and I managed to clean up. Now I need to manage the cables. Need more cable chain too.

If testing works out I am going to increase the size of the long axis so I can cut a full project panel.

Also confirmed that backlash on the ball screw is about .004", but it appears to be consistent.

1 Like

Looks like I have to update my firmware. Haven’t done so since I bought the board. I’ve never needed to save any settings before. Found another thread about it on these wonderful forums!

I’m interested to see how this turns out. 0.004" backlash ain’t that bad. Wouldn’t it be cool if Marlin or GRBL could use DRO feedback to get rid of backlash? If they could, I’d go for an all ball screw rig in a heartbeat.

I had similar results as you when I was playing with lower microstepping on my primo (drv8266 @12V 1.4Arms) after changing the z from an 8mm lead to a 2mm lead (to lift a much heavier spindle). Going from 1/32 to 1/4 didn’t seem to make any difference in the max rapid speeds I could get. I swapped in 2209’s in standalone spread-cycle 1/16 mode 1.4Arms and they are slightly better. I am upgrading to 24V to get my rapids back though.

Something like this @24V (48V even better if you can afford it) would work much better than 8255’s for your nema23:

The cost is not in the spirit of MPCNC I think. That is stepping in to ‘big boy’ cnc territory, where costs go up quickly in general. However it’s a possible solution if you run into problems with torque.


1 Like

I’ll be investigating 2 Marlin features I’ve not used before. M425 is backlash compensation. Supposedly Marlin will apply a factor of your choosing to each move in a particular axis each time there is a direction change.

Also, M852 bed skew compensation. This will apply a correction factor when parsing g-code to account for axes out of square.

I may try and apply the backlash comp right away, but will likely wait on bed skew. It will probably be easiest to determine the degree to which my machine is NOT square by cutting a large test piece and measuring it, rather than measuring the machine itself. The solution will likely be iterative. Cut measure comp, cut measure comp, etc

1 Like

I have not yet found the perfect application for the Odrive (https://odriverobotics.com/) I bought (or at least not perfect enough to raise its priority level), but in principle it should be able to take step/dir inputs and drive a big BLDC motor with position feedback. It can’t drive a bipolar stepper, but big BLDC motors are much, much cheaper than equivalent high power steppers.

For proper control you would need software to take both the rotor position feedback and the DRO position feedback, but I’m sure this has been done before. You should be able to find someone’s footsteps to follow. Which is not to say that it’s easy, but at least you won’t be all by yourself in territory that nobody has explored before.

1 Like

The firmware upgrade worked, I can save settings to eeprom now.

In the process, I did realize that the current limit for x and y on the mini Rambo are LINKED. I will be feeding the nema 17s the same as the 23. I hope that doesn’t come back to bite me.

I get no skipping on jogs at 1.5A, which the 17s ARE rated for, but I hate bumping up against the limits.

So, maybe someone can help me out here.

I finally got around to doing some test cuts and I observed some strange behavior. On the first test cuts I had HUGE issues with skipped steps on the x axis (screw driven, gear reduced, nema 23 motor).

Only happened while cutting, not during jog or travel. The first thing I tried was to try a 24v PSU. This actually made it WORSE. I started to have problems jogging as well. For some reason, I decided to try turning the max current DOWN with m907 x1000 (it was set at 1500ma).

And that fixed it. My test cut went good with no skips. I don’t know if its FIXED, I don’t know if it will still be a problem if I push the tool harder, and I have no clue how LESS current helped. Can anyone educate me? Haha.

Driver got too hot and shut down briefly maybe, resulting in the missed steps? I think certain drivers can do that. Not sure if it applies to your case.

1 Like

The TMC drivers will reduce their max current when they overheat too. So it might be they were switching between 1500 and 500, until you made them 1000 and they just stayed at that level.

I agree that it is probably overheating on the drivers. They often print (to the serial console) when they overheat. You can also see the current setting and OT (over temp) warnings with M122

I read that this could happen, but why only x? The way the mini Rambo works is that m907 sets max current for BOTH x and y to be the same (you can only set max current for x/y, z, and e, you can’t set for x and y independently).

The nema 17s on y were getting pretty toasty, but the nema 23 on x wasn’t even warm. I wonder why only x would shut down if they were getting the same current. Maybe the x is drawing more due to more torque? X is bumping up against the max where y isn’t?

Based on @jeffeb3’s reply, the explanation could be that the NEMA 17s handle the reduced current fine without missing steps, while it ends up being too low for the NEMA 23s. I suppose it would just depend on the torque characteristics of the specific motors in question. That or the 23s end up using slightly more power (higher voltage to reach the current or such) and thus get hotter, or maybe even something like the position of the drivers on the controller and the airflow around them being different.

And you may be aware already, but it’s the driver’s that get to hot and turn off/reduce the current, not the motors (there’s no sensor on them for the driver to get feedback about that).

1 Like

A mini Rambo has a4982 drivers with digipots, not TMC. So M122 won’t do you any good. Allegro chips are rated to 2A… rule of thumb that’s good for ~1A design. Based on my aging experience with the similar a4988, I’d keep it 1A or less, or upgrade to a more capable driver. I never drove my allegros to fault condition. So I’m not sure how a4982 handle OT… reduced current, a latching limit, hiccup mode?

Is your fan cooling good and even? If not, SirNate0’s suggestion is a good/easy one to follow up.

Looking at 2 common stepper specs on amazon, the 17 is 1.4ohm/3.0mH, the 23 is 0.9ohm/2.5mH. This is generally true for motors… larger will be lower R and L. All else equal, the larger motor needs less volts while on (due to lower resistance), and has a shorter rise time so can reach higher speeds before it runs out of volts (due to lower inductance). All of these factors point to the smaller motor being the first to stall in your setup (stall via OT due to more power required, inadequate volts, or inadequate torque). The physics tells us that something else is at play, like driver OT or something.

It seemed to work when you reduced current… which rounds up the story I think. Hopefully 1A is enough to keep it going well. If not, it may be time for an upgrade.

I have finally gotten around to cutting the 1st real job since the rebuild. In honesty, I’m not sure if I have made anything better…but at least I have a working machine again.

The good:
-With the current limit back to 1000, I had no issues with missed steps.
-The outside edges looked excellent

The bad:
-Chatter on V-bit, as shown below
-The parallel edges around the text portion appeared to be .008-.010 inches too big. I took a .010" full depth finish pass, followed by a .000 “spring pass”

Toward the end of my troubles with the Burly, I was getting anywhere from +.015-.030 on ODs, so I guess its improved. More testing is needed, and I also need to check all the usual suspects like ACTUAL tool diameter, loose grubs, loose clamp bolts, etc.

Also, my theory on the chatter is that the spindle is “bouncing” due to the flexible coupling between the z-stepper and the lead screw. Not sure why this would be any worse than the Burly that used the same coupling, but either way I am going to make a solid shaft adapter and try again.

1 Like