I linear rails-ed my Maslow

[RandomDave]

TLDR: It’s a lot stiffer than than the current Maslow, at a trade-off some extra weight (~2kg, but we’ll see if that can be optimised), and reduced Z travel (I still get ~25mm of bit depth which is more than I need). Plus component cost + printing time of course.

I’ve been working on this for a couple of months and sitting on it for about a month because i’ve not been able to test it yet (because of a bereavement), but i’m just going to put it out there and hopefully i’ll get time to test it soon. I’ll have to post all the details in little chunks when I get 30 mins here and there (same reason), so bear with me. I should be able to do a build thread at some point with photos I took from the final build.

The problem I was trying to solve was that I was getting noticeable flex when doing X axis aligned cuts with a 1/4 bit (which is what I want to use to avoid bit deflection). As there was only Z height support on 2 of the 4 columns, I could flex it x-axis aligned. That is actually what triggered the Maslow 4x4 over-arching project i’ll discuss in another thread, but that quickly side-tracked into doing this conversion kit stepping stone.

Also it has a handle now, which is nice:

I’m working on a few designs, so to keep track of versions etc, i’ve stuck it in a fork on github:

Maslow4_hardware/Maslow_4x4_projects/Linear_rail_2_rails_2_z_4_columns at main ·...

(You can just download a zip of all the files).

It’s STLs because I started throwing things together in TinkerCad and at some point if it works well i’ll recreate it properly in a proper CAD package.

The README also has a component list that i’ll stick at the bottom of this thread.

Here’s some random photos from when I built the latest version (1.0):

There’s a bunch of design decisions i’ll go into more depth at some point, but off the top of my head:

• The top and bottom clamps / plates are 20mm rather than 11mm because it adds a huge amount of stiffness.
  • The extra 9mm goes up and down from the area that clamps onto the router - that bit is still 11mm
  • The clamp is zigzag and 2 bolts to make it stiffer, it’s also sized quite carefully, to the point it should be tight when done up, but tape to pad might be needed depending on printer tolerance.
  • The top and bottom can also be different because you’re printing them.
• Using 350mm 2040 for the uprights as well as the cross piece is deliberate - although it looks tall, the room lets you remove the router vertically to remove the arms without disassembling the sides, which are the bits that you need to fettle into place the most.

Non-printed components:

• 3 x 350mm 2040 aluminium profile.

• 2 x 350mm MGN12 rails (300mm also works but limits maintenance Z travel).

• 4 x MGN12H Carriage (2 per rail).

• 50 x M3 30mm socket head cap screws (includes spares).

• 20 x M3 10mm socket head cap screws (includes spares).

• 100 x M3 Locknut (includes spares).

• 10-30 x M3 washers (10 needed for router clamps, other per user preference).

• M3 balljoint allen key (at least 100mm, the longer the better).

• 60 x M5 12mm button head cap screws (includes spares).

• 60 x M5 T Nuts (includes spares).

More when I get a bit more free time, but feel free to ask questions and I should be able to reply…

[RandomDave]

Things i’ve already remembered I didn’t put:

• I’ve tried to stick with the sizing and bolt patterns of the original parts, though i’m not sure there’s much value of mix-and-matching.
• I made sure the button-pusher still works with this.
• I’ve included a set of shims to help construction as it all needs to sit well to not bind the rails.
• I’d say it’s…a little bit more tricky to put together than the OG Maslow 4, but mostly it just requires more patience as you need to tighten and loosen and generally fettle stuff to get everything sat correctly and tightened snug but not overtightened because linear rails are finicky.
• I stuck in a printable nut-holder because some of the holes are deep, and you really need a decent ball-end allen key to drive some of the screws.

[dlang]

heh, I was just talking to people a couple hours ago about trying to print new uprights that are wider to give more stiffness to the machine.

[dlang]

so, here is a 3d printed post support design, you embed 5 nylock nuts at the bottom and 4 at the top (top is optional, it gives you a place to put a handle or brace across the top)

the bottom plate will use 40 degrees, which should keep it just inside the belts, the top stays within 30 degrees.

I haven’t tried printing these yet (will see about getting that queued up tomorrow)
onshape cad is at Onshape

stiffer post supports 3dprint.3mf (151.5 KB)

[RandomDave]

@dlang

so, here is a 3d printed post support design, you embed 5 nylock nuts at the bottom and 4 at the top (top is optional, it gives you a place to put a handle or brace across the top)

the bottom plate will use 40 degrees, which should keep it just inside the belts, the top stays within 30 degrees.

I haven’t tried printing these yet (will see about getting that queued up tomorrow)
onshape cad is at Onshape

Interesting - printing a single piece like that certainly reduces complexity. I’d tend to think the cross brace is pretty important.

Purely based on poking things with large sticks (ie levers), I found to reduce the flex you need three things:

• Stiff clamps / plates and columns - to form a ‘cage’ around the router that is very stiff. (helps with both X-axis and Y-axis flex)
• A stiff square with the sled as one edge (the sled can flex) for the cage to travel on (helps with X-axis flex).
• Wider load distribution of the sides attachment to the sled (helps with XY-flex/twisting).

I do suspect there could be a less beefy set of conversion components that gets you 90/95% of the additional stiffness, possibly utilising more of the original components such as the bars and linear bearings rather than linear rails though - which would probably be along the lines of what you’ve done + stiffer plates. Possibly utilising aluminium profile still as it’s very stiff for the weight

[dlang]

Dave wrote:

apologies for mixing threads here.

Interesting - printing a single piece like that certainly reduces complexity. I’d tend to think the cross brace is pretty important.

I have been routinely picking up the maslow by grabbing the tower support, it
seems to be very stiff in one direction, much less so in the other. I’m leaving
a provision for a handle/cross brace in case it helps (either in stiffness, in
picking it up, or as a place to mount electronics )

adding that stiffness was a significant part of the reason for this.

the other reason is that I’m using a non-standard spindle, and instead of two
big power coreds (router and 24v to the electronics), I’m planning to have a
high power 48v power supply attached to one post, and then on the other post I
will have the motor driver and a 48->24v converter to power the
electronics/motors with only one power cord going to the sled

the wider crossmembers give me better options to attach them

Purely based on poking things with large sticks (ie levers), I found to reduce the flex you need three things:

• Stiff clamps / plates and columns - to form a ‘cage’ around the router that is very stiff. (helps with both X-axis and Y-axis flex)
• A stiff square with the sled as one edge (the sled can flex) for the cage to travel on (helps with X-axis flex).
• Wider load distribution of the sides attachment to the sled (helps with XY-flex/twisting).

my solution to this is to eliminate the cage and the thin components holding it
in place and instead of a column anchord all the way around to the sled (the red
tube in the cad I sent you)

I do suspect there could be a less beefy set of conversion components that
gets you 90/95% of the additional stiffness, possibly utilising more of the
original components such as the bars and linear bearings rather than linear
rails though - which would probably be along the lines of what you’ve done +
stiffer plates. Possibly utilising aluminium profile still as it’s very stiff
for the weight

my only concern with using 2020 type rails is attaching to them solidly so the
joints do not flex (especially on just one end) solid plastic may be as good, or
at least very close

but you have working hardware, I don’t yet

David Lang

[RandomDave]

@dlang

I have been routinely picking up the maslow by grabbing the tower support, it
seems to be very stiff in one direction, much less so in the other. I’m leaving
a provision for a handle/cross brace in case it helps (either in stiffness, in
picking it up, or as a place to mount electronics )

adding that stiffness was a significant part of the reason for this.

the other reason is that I’m using a non-standard spindle, and instead of two
big power coreds (router and 24v to the electronics), I’m planning to have a
high power 48v power supply attached to one post, and then on the other post I
will have the motor driver and a 48->24v converter to power the
electronics/motors with only one power cord going to the sled

the wider crossmembers give me better options to attach them

Ah that’s very interesting - eliminating the power brick and cord would be a very nice thing to do!

Am I right in thinking/remembering you are using one of the 500W, quite narrow diameter spindles?

@dlang

my solution to this is to eliminate the cage and the thin components holding it
in place and instead of a column anchord all the way around to the sled (the red
tube in the cad I sent you)

Yeah, I’m very interested to see how that works out - it should be very stiff when you get it to work!

@dlang

my only concern with using 2020 type rails is attaching to them solidly so the
joints do not flex (especially on just one end) solid plastic may be as good, or
at least very close

I think 2040 rather than 2020 Makes a large difference to how strong a joint you can get - with the corner brackets or the upright attachments I am using, I have no real concerns - if the sled didn’t flex i’m not sure a cross piece would be needed, though i’d probably still have one just because its an easy win.

[bar]

This is super cool and an excellent write up!

[dlang]

Dave wrote:

Am I right in thinking/remembering you are using one of the 500W, quite narrow diameter spindles?

700w bldc, most of the body is 55mm

David Lang

[RandomDave]

It lives!

Bit of a Hello World cut, I got it to calibrate, but it still needs dialing in i’m sure.

I also threw the brushless franken-router from Interesting find - brushless Franken-router setup in there, with ribbon cable extensions to relocate the board…because changing 3 major things at once never causes any issues

This is 12mm ply, cut in 2 x 6.5mm passes (need to work out if I didn’t have it properly z-zerod as it’s not gone all the way through everywhere), at 800mm/s feed rate and 400mm/s plunge rate, using a 6.35mm (1/4") bit.

It was solid as a rock.

It probably is over-engineered and suffering for it slightly, the plunge rate might need tweaking (I was running it at quite a low speed and it bogged a bit in the plunge), but i’m pretty happy with that.

Now to try my luck with some full depth cutting…

[bar]

This looks awesome!

How would you say the noise compares with the brushless spindle?

[RandomDave]

@bar

How would you say the noise compares with the brushless spindle?

Hmmmmm, it’s quite difficult because both are not too loud with no load, but are then load enough when cutting to need ear defenders (especially in the shed).

I do think it’s a bit quieter, especially at slow rpm. I think the cutting noise of the bit becomes the prevailing noise, whereas I think the brushed dewalt is still a lot of spindle noise under load.

I’ll have to download a dB app on my phone or something and actually compare it.

I was also (this time) looking at whether I could get single pass cuts, and how low I could run the spindle speeds while doing that (with interesting results i’ll talk about in a sec) - it might be more of a difference if i’m running shallow passes with the router turned right down, which I also need to investigate.

[RandomDave]

So. Ahem.

Yep, I played ‘how slow can I run the router at 13mm depth’ and the result was the router (I think) not having enough torque, the bit starting to slip in the collet, and then getting pulled down (because it’s an upcut bit) in a positive feedback loop.

I actually was getting some cutting at quite low rpms:

But I think I was pushing it too far. I do wonder if it was more an issue that the collet is a bit crap and that was the failure point, but it did feel like I was pushing the router harder than it could take when I turned the speed down a lot.

That was actually the second attempt at a full depth cut (third cut overall) though, the first one I messed my CAM settings up and it still did 2 passes. But what I noticed at the end was the two Z screws weren’t matched any more (might be able to tell on this, though its tricky):

I’m not entirely sure when this happened though - I have a feeling it might have been on the original cut because I had the router low when it plunged and it struggled a bit - I wonder if one of the steppers struggled to push it down and skipped steps.

I realigned it and whacked the current to the steppers up to 1.2A each. I’m not sure how far out of spec i’m getting for the steppers or the board with that @bar? I’m happy enough that it’s my own fault if it goes wrong, but what do you think I can get away with? 1.2A? 1.5A?

Overall though, I did a final cut where I kept the router running reasonably fast. And it cut 13mm with a 6.35mm (1/4") bit in one pass at 600mm/s:

A bit of tearout / fuzz, with a bit of sanding it seems like quite a clean cut though:

The motivation for this is seeing if I can get to the point of using a compression bit and doing one pass cuts. I’m not sure this cheap brushless router can do that, it might be asking a bit much, and a couple of passes might be a reasonable limit.

One point is, the little PSU I got has some 7-segment displays. If I believe it, the max draw I saw on it’s little PSU was about 15-16A (@ 20V):

Which is only 300W. A bit of googling suggests dewalt batteris can do 20A reasonably sustained, 40A for bursts, though I don’t really know what their brushless trim router would actually draw, and whether the limitation is the PSU or the router (I suspect router). That’s another investigation for another day though, I was happy with the final cut, though i’d likely back off a little for a longer cut, maybe 500mm/s for a single pass in 12mm, but probably just 2 passes.

[bar]

I’m pretty sure the stepper drivers are rated for 1.5 amps so you should be totally good. (although you can never really trust the data sheets on that stuff)

I bumped mine up to 1.3A for a while and I didn’t have any issues with them

[RandomDave]

Awesome!

What’s the headroom on the boards - is that 1.5A, or (theoretically) can the boards go higher than that?

[bar]

I looked it up and they’re actually rated for 2A continuous and up to 2.8A peak…that being said some larger heatsinks might be in order to really push those numbers. I’ve found the current ratings in the datatsheets are usually under unrealistically ideal circumstances

[RandomDave]

Yeah, that makes sense good to know for when I try some slightly larger motors!

[RandomDave]

I tidied up the wiring a bit, added some handles:

and did a few more experiments.

I was able to run a 1/4 compression bit full depth on 12mm at ~400 mm/min. The cut itself was clean, but it felt like there enough force needed that it shifted position / settled when raising/lowering the bit.

Interestingly, I saw 20A draw from the PSU, so running at 400W.
It did feel like that is pushing things too far to get good accuracy though.

I went the other ways and did 3 passes using a 1/8 bit (upcut for now) at 4.5mm per pass and 1000mm/min - that gave me pretty nice clean results:

There is definitely something funky going on with the Z axis though - it’s consistently ending up lower than it should be at the end of a cut (in the region 3/4/5mm too low).

I assume it’s missing steps when being raised, but I need to verify that.

The normal suspects would be skipped steps if too much force is needed or skipped steps due to interference I guess. I pulled the motors and between them they’re lifting ~5kg - as it’s on an enclosed frame i have some thoughts on bringing the force needed down. It could easily be the dust extraction or the power cable for the router / the router itself, so i’ll try running a ghost cut and see what thing it points me towards!

[dlang]

Dave wrote:

There is definitely something funky going on with the Z axis though - it’s
consistently ending up lower than it should be at the end of a cut (in the
region 3/4/5mm too low).

I assume it’s missing steps when being raised, but I need to verify that.

It could also be that with an upcut bit, the force from the bit is pulling down
hard enough to move the Z (simple force may not be enough, but force combined
with vibration could be)

as a simple test, try using a straight cut or downcut bit and see what happens

David Lang

[RandomDave]

@dlang

It could also be that with an upcut bit, the force from the bit is pulling down
hard enough to move the Z (simple force may not be enough, but force combined
with vibration could be)

as a simple test, try using a straight cut or downcut bit and see what happens

Ah, i’d not considered that! Especially if there’s already lots of vertical weight, the hold current / strength of the steppers might be not quite enough. I think alleviating some of the weight would help in that case, i’ll pick up a straight cut too if I don’t get anywhere else.