LowRider 4 questions - belts / tubes / other

LowRider CNC Advice
1

It’s been 6 years since I built an MPCNC. The first year was spent testing, and after that, it ended up serving as furniture to store all kinds of junk - like all flat surfaces eventually do in a workshop :).

That was my first attempt, and it had its quirks - partly because of the components I chose, and partly due to the original design itself.

1) Rigidity / Holding Strength

I used 6mm belts, a 12V power supply, and stepper drivers rated for a maximum of 1.5A, even though the stepper motors were rated for 2A. That obviously limited performance.

2) Usability

The setup used a RAMPS board (Marlin) and a typical 3D printer-style screen/controller. I had to juggle SD cards and run back to the PC for every small change - super annoying.
Another usability issue: the tubes running all over the machine always got in the way.

3) Dust Collection

I never found a dust collection solution that fully met my needs. I even tried designing my own, but the design of the core made integration difficult. Doing it the way I wanted would have required reprinting the entire core, and working with STL files in Fusion is no fun.

4) Wire Management

Just… not great. But that is a combination of mostly may laziness and a bit of the design

Despite the limitations, I spent around $150 back then - including the main board, drivers, stepper motors, and even the router (a Makita knockoff that still works great).
So I’d say it was worth it just for the learning experience.

Now, I’ve finally decided to tackle those issues :slight_smile:
There’s a project I’ve been procrastinating on long enough, and having a CNC would greatly simplify it.

First attempts to fix the issues

  1. I ordered a 24V power supply and plan to switch to TB6600 drivers (unless there’s something better in that price range). I also replaced the RAMPS board with a TinyBee, and upgraded to 15mm belts.
  2. I removed the screen and Marlin entirely, and added a Raspberry Pi running Klipper with Mainsail. I installed an extra Wi-Fi adapter on the Pi-one connects to the local network, the other works as a hotspot/AP. If I’m not in range of the local network, I can just connect directly to the Pi with my phone.
    I loved this setup, especially Klipper’s ability to reconfigure things on the fly.
    That said, Klipper is still missing a few features - it’s possible I could fix or work around some of them.

I was planning to print the Primo parts to improve rigidity, but…
After seeing Doug Joseph’s beta tests of the LowRider, I decided to scrap that plan and go with the LowRider instead.
Yes, I’m easily distracted :).

But considering I’d need to reprint most of the Primo parts anyway, why not try something that solves other issues too - dust collection, wire management, tubes getting in the way, machine size, portability, etc.

Questions Before I Start

  1. Belts - I already have 15mm belts. What would I need to modify to use them? Or might they fit without any changes?
  2. Tubes - The most expensive part of my original MPCNC was the stainless 25×2mm tubing (25.2–25.4mm outer diameter). It’s very rigid and still in great condition—it would be a shame to toss them, especially with today’s metal prices. Can they be reused with the LowRider, or would that require too many modifications?
  3. Asymmetric X Axis - I mostly understand the benefits of this approach - though I admit it ticks a few of my OCD boxes :).
    One thing that concerns me: If I understand correctly, the LowRider rides on top of the right tube, which guides it along the Y axis. But there’s nothing really holding it down in the Z direction except for gravity and the belts - right?
    That’s fine when using down-cut spiral bits, but what about straight bits, carving bits, or up-cut bits? Couldn’t that cause lifting or instability?
  4. Workflow - As mentioned earlier, I really liked the Klipper + Mainsail workflow. But with all the changes, I plan to try GRBL.
    Is it possible to achieve a workflow similar to Klipper/Mainsail with GRBL? Specifically:
  • Configuration without recompiling the firmware?
  • Uploading files from any device, without juggling SD cards?

What Sells Me on the LowRider

  • The ability to easily disassemble the machine for storage.
  • Simple adjustment of Y-axis length when needed.
  • Better wire management.
  • A more practical dust collection setup - I’ll probably tweak it a bit, but it already looks far more manageable, and it won’t take up valuable workspace.
  • And of course, the performance shown in the beta videos.
2

LR4 is designed to use pulleys and belts that are 10mm wide. You can’t use 15mm wide belts.

Cannot be used for LR4. It needs 29.5mm, 30mm, or 32mm OD tubing.

Both sides of the LR4 Y axis ride on something and are located by gravity and the tension of the Y belts. The rail side rides on (but is not captive by) the Y rail, and the other side rides on the table. Yes, it is asymmetric. It’s a good feature of the machine, preventing overconstraint of the Y axis and making for a much less finicky build. No special alignment of parallel rails needed.

The router is what does all the work, if your feeds and speeds are proper for the bit you are cutting with this isn’t a problem. The trim routers we use are meant to be hand held, which should inform the nature of the forces needed to cut with them.

If it does, you’re cutting a material using a process unsuitable for the LR4. People regularly mill aluminum with LR4s. If you need better than that, LR4 isn’t the machine for you.

Certainly doable, but… Why?
FluidNC controllers like Jackpot are under $70.
A small minority of users are running Klipper on the LR4 systems.

FluidNC does this, as does Klipper.

FluidNC and Klipper can do this.

MPCNC is a great machine if built smaller, and it shines in particular if you need lots of Z travel.

LR4 is as you’ve assessed- more capable for many uses and in particular for larger sizes, dust collection, and ease of removal/transport.

2 Likes
3

Yes, I’m aware of that and have looked through the instructions. My question was more about the modding aspect - how hard would it be to modify the design? Since there are only STL files available, I would need to recreate a few models (in the worst case, almost all of them) - at least the ones for the tubes.

Working with STL files in Fusion is no fun. Been there, done that with the MPCNC - I had to recreate the central gantry, and I’d really like to avoid doing that again.

As for the belts, maybe editing the STL files directly to add extra clearance would be enough?

Is there any noticeable difference between 30mm and 32mm tubes in the rigidity aspect? Will probably go with 2mm. I’ll be able to source precise tubes - not necessarily cheep… But here, it’s either this or stainless… The stainless is cheaper but it is with an inner seam and i’m not sure how precise it is.

When mentioning GRBL, I was speaking more generically and had in mind all the forks out there - FluidNC is on the list to try. From what I’ve read so far, it’s basically GRBL on steroids, and TinyBee seems to be compatible (ESP32).

I’ll be keeping the Klipper configuration in case I want to go back. I managed to get it almost working (homing and manually moving the axes), but there were some out-of-range issues when trying to run certain G-code files.

There was a post-processor for Fusion, but it was incompatible with the latest version, and I didn’t have time to dig deeper into that. With some effort, I’m confident I could get it working, but for now, I’ll try the new board and firmware(s).

Since I plan to keep the Raspberry Pi, flashing new firmware over SSH will be easy.

If anyone has some suggestions for good stepper drives that would handle 2A motos - still haven’t decided, would be appreciated :slight_smile:

4

2amps motors, wow i can smell the melted plastic.

Tmc2209 can handle the 2A of course you need to provide sufficient cooling to them. A lr doesn’t need 2A running motors.

1 Like
5

Yup, you can mod yourself whatever you want. Just a quick advice:
-get your lowrider working STOCK, then mod wat you want. Please understand everything first, and move forward with the basic knowledge.
-please: follow the models license rules

6

Ryan does not share the LR4 CAD or STEP files, so you’d be doing a lot of work to modify them. You’d need to recreate all the models that have tubes, and be certain you’re aligning everything properly because you would not be working from parametric CAD.

This is a good time to ask… Why?
It’s your machine, and I understand how fun it is to design and modify machines- but we should be clear here. We encourage users to build the V1 machines “Stock” first before embarking on wild modifications.

The LR4 is already defined by the spindles (Trim routers) we use, and the machine overall is already way better than the spindle is. Thinking that you’re going to massively improve the machine with wider belts, larger tubing, higher current steppers just isn’t going to yield the returns that you think it will. Even if you massively enlarge the spindle, you’ve now set about on a path where to re-optimize things it would not be or look like a LR at all.

Here again, yes you could subtract space to “fit” the pulleys, but as you carve away space be careful you don’t cut into internal features or remove the load paths in the parts. It bears repeating that the 10mm belts and pulleys are already way more capable of handling the loads than the machine needs. You’re doing all this work to avoid buying $60 worth of belt. Is that worth it?

No, the tubes as spec’d are already better than needed. Pay attention to wall thickness. Folks tend to find thin wall stuff, and that’s the major place you would hurt your machine rigidity.

FluidNC started out as a GRBL fork, but has diverged and added capabilities to the point you shouldn’t think of it as a GRBL variant. You should think of FluidNC as a next-gen firmware that maintains some similarity to the GRBL command set.

TinyBee is capable of runnign FluidNC, and isn’t terrible to start with. That said, the board is fairly limited compared to Jackpot or one of Bart’s newer controller designs. Pay attention to those limitations and study up on this if you use TinyBee.

Klipper is completely capable, definitely keep this option for yourself.

People usually use FluidNC from the WebUI over WiFi.
In theory you might be able to flash FluidNC from the Raspberry Pi OS as it’s just a Linux variant- but I’ve only flashed FluidNC from x86 Linux, MacOS, or Windows.

If you run Klipper then the Pi will be central to your system. For FluidNC, the Pi doesn’t add much value, frankly- unless you plan to use it for a G code sender. Even there, most of us would recommend you stick to running FluidNC from the Web UI.

The TMC2209s that most of us run top out at 1.2A. You’re back in territory where we need to ask you why. The .59Nm steppers that we run on the LR4 are again more than capable. The only debate there is with the X axis, and there’s a bit of a chance this is one spot where some improvement might be had for edge cases.

You aren’t going to get those 2A drivers on Jackpot, or TinyBee, or most other boards. For that you’d be better off with external drivers and a board like Bart’s external driver board (Presently the Boxer is the latest variant of that). With those boards you’re also open to experimentation with the latest batch of closed loop steppers. If you’re really pushing your machine, particularly in X, this might be interesting to work with.

If you do decide to try and use 2A current with the LR4, you likely will need to upgrade to a higher temp material like PET-CF for all of the parts with steppers, as 2A will run hot in the stepper.

Most folks here are using Estlcam for post-processing. There are forum posts for experimental Fusion360 post proxessors- so that might be something you’d explore. I don’t use them so cannot comment further. Hopefully others who do will jump in.

1 Like
7

Why are you wanting to modify the design before even trying it like it is? There are SEVERAL of us that have running LR4’s exactly how it is designed by Ryan and have zero issues with the functionality. My suggestion is to build it how its meant to be built and try it before you say the design needs to be changed.

8

It isn’t actually difficult. I have already done one mod will test later.

10

I modded the z steppers holes to accomodate some larger steppers (the ones Phillip shared with the magnetic encoders) thats the mod im going after.

1 Like
11

That is definitely the best approach - assemble everything first, and then think about how to eventually modify it.

Either way, I won’t be touching the tubing, because there would be a lot to modify, and a lot could go wrong with that. The belts, on the other hand, are still debatable. First, I’ll take a closer look at the files and see if it’s worth the effort.

The main reason is to salvage as many parts as possible from the old machine. Not going well so far :slight_smile: . I don’t have space for two machines, which is why the ability to disassemble is appealing.

At least one of the steppers is having trouble holding torque and is losing steps. It might be a stepper driver issue. It’s been a long time since I last tuned them, but I believe I did it fairly evenly.

One more question.
The T8 lead screws with brass nuts usually have a lot of backlash (at least the cheaper ones). Won’t that cause issues? Wouldn’t it be better to go with a nylon nut? Or a combination of both.

//Edit:
One more.
Those are large prints. Has anyone printed the parts with a 0.6 mm nozzle instead of a 0.4 mm without any issues?

12

You have the weight of the gantry and router always pushing down on them. This works as a very effective anti-backlash mechanism.

Not in this application.

People use .6 all the time for these. The key thing is not to go crazy with perimeters. There are special ‘hints’ in the files which force the slicers to lay out internal features correctly. If you go beyond the recommended number of perimeters, you risk screwing up the slicing of your part.

I’ve even printed a core with a .8 nozzle on my old Taz 5, but that’s a lost cause as it is flow-limited by the E3D V6 hotend on that toolhead. .6 and .8 were the same exact speed for me, and the .8 starts getting fiddly.

The lesson to learn there is to understand the maximum flow of your printer and stay within its’ capabilities.

Modern drivers like the TMC2209s let you set each driver’s hold and drive current over UART, so you have fine-grained control over this directly from your config file. Don’t try to re-use your old drivers; that’s a recipe for trouble- particularly if you are getting a new controller anyway. Use modern drivers on your control board.

This is anther limitation of the TinyBee, to fit the minimal IO use, these have funky/inoperative UART control over the drivers. Jackpot and all of Bart’s boards are radically better in this respect.

1 Like
13

I used 0.6 for everything, it all printed nicely

1 Like
14

I use .6mm for all of mine as well. Pretty sure @vicious1 does now also

15

I’ll chime in as a klipper cnc user with full endstop and probe functionality on the LR4. Works fine. The camera isn’t auto rendering timelapse video of each cut, but could. I’m using an old trigorilla board and a mini E3V3 controller because the trigorilla blew a couple channels because of an unforturnate event and the e3v3 was available. Klipper can split the motor functions over multiple boards as long as both axis motors are on the same board. the majority of the folks are using the jackpot board now, but if you have old printer parts like me, putting those to work with klipper makes a lot of junk box parts useful again.

1 Like
16

Yes, I’m using an old Prusa MK3 - since it works well, I didn’t upgrade to the S/+ variant. A 0.8 mm nozzle wouldn’t help much in this case either. It has good presets, and if I want anything accurate, I stick with them. If I don’t care as much about precision, I have a 0.4 mm nozzle preset that beats the print times of the default 0.6 mm “fastest” preset. :slight_smile:
Since multiple people said that 0.6 mm is fine, I’ll go with that. Anything between 0.3 mm and 0.4 mm layer height should be fine, right?

@CesarH, @Jonathjon Thx

Can you elaborate? What I found is that there can be issues when using both a display over UART and stepper drivers. I don’t want any displays, so if I’m not mistaken, it should be fine. I may actually go with the TMC2209.

Yes, I also have that working fine. The only problem I’ve run into is finding a working post-processor for Fusion that’s compatible with Klipper. Granted, I haven’t spent much time searching yet - I might try Estlcam.
Are you using the official Klipper or Naikymen’s fork? Any macros?

2 Likes
17

I run a .3 layer height. 3 top/bottom and 3 walls. Cubic for infill at the recommended settings.

18

TinyBee does not use a UART to talk to the TMC2209s. That sucks.
In your FluidNC motor configs you’ll likely find a section that looks like this:

stepstick:
        step_pin: I2SO.1
        direction_pin: I2SO.2
        disable_pin: I2SO.0

This means FluidNC is using the TMC2209s in legacy fallback mode, so you’re back to monkeying around with the pots on the board or getting whatever the default is in the driver chip.

That’s not a thing in FluidNC. There is one ESP-32 UART that is used to communicate with FluidNC (e.g. for gcode senders, pendants, or FluidTerm), and a different ESP-32 UART that is used by FluidNC to communicate with the TMC 2209 drivers.

That 2nd UART is the thing that isn’t set up in the TinyBee.

I believe you’re mistaken.

Then you should go with one of Bart’s boards, a Jackpot, or some other FluidNC board that hooks up an ESP-32 UART to the TMC2209s.

19

I’ve been hanging out on the FluidNC Discord long enough to see the support burden they face for boards like the Tinybee where the manufacturer does not support the project and they spend a lot of time dealing with it. For that reason alone, I would not use it.

I will note that they do have details about getting drivers to work with UART via board modifications.

http://wiki.fluidnc.com/en/hardware/3rd-party/MKS_TinyBee#modifications-to-allow-uart-control-of-drivers

20

At some point I think MKS made some kind of contribution to FluidNC for TinyBee, so at least that’s better now as far as I can tell.

This gives me a little pause:

All of this said, the time to modify a cheap board to reach the basic functionality of Bart’s 6 driver boards or a V1 Jackpot doesn’t seem to make the cheap board worth it…

1 Like
21

It is closest to marlin. I use a modified marlin post processor in estlcam for gcode generation. Klipper doesn’t like the leading zero in gcode commands like G01 vs G1. You can deal with this in a postprocess like estlecam by defining that pretty easy. Others have made a macro for G01 to call G1 and pass the arguements. I do have a couple simple macros and a couple settings Ive posted about here.