Double-decker build

Yes that’s correct in what I am aiming for and what I think will happen. The tilt of the Z axis should be extremely low due to the second gantry but the horizontal deflection of the gantry rails will be essentially the same as a regular MPCNC and I’m hoping will be the new weak point.

Im thinking stickout should have minimal effect on stiffness but it does have some. At zero stickout it has effectively the same stiffness as the lower gantry. At say 1 foot stickout the leverage causes lateral forces at the tool to be multiplied into larger forces at the lower gantry, lets say 1.5 times to keep the numbers simple. With 1.5x the force, the lower gantry then deflects 1.5x the distance and again because of leverage this 1.5 at the lower gantry translates into 1.5*1.5=2.25 times the distance at the tool.

This leverage factor is influenced by the height of the second gantry and is the reason for making it very tall.

So if the goal is a large z axis but still stiff wouldn’t a lowrider style rig be a better option? Or aluminum extrusions?

As I understand it, Lowrider is not much better at remaining stiff over a large Z range, although they say it is stiffest at the bottom of its range, while MPCNC is stiffest at the top of its range. As for aluminum extrusions, according to my measurements the largest source of deflection comes from twist/tilt of the Z axis away from vertical. I am not seeing how aluminum extrusions would be used to improve this, but I am willing to entertain any suggestions.

Here is the Z axis and motors installed:[attachment file=109873]

I have not yet designed the mechanism to secure the belts in place. I guess that’s next, and wiring, and we’ll be ready to make a crown.

1 Like

So, will you wire the X motors on the bottom in series and the X motors on the top in parallel with each bottom motor? Or do the whole set of four in series? Or all four in parallel?

All four in series is the plan. I got a 24V power supply and a board that supports it, so in theory each motor’s torque/speed should be just as good as running two steppers in series at 12V.

Some of the people on this forum are actual psychos.

I love it!

2 Likes

Some?!..Most! Wouldn’t have it any other way!

3 Likes

Okay, it’s moving and running gcode! Sorry for the crappy video (unedited).

Wires are still a mess obviously. That will need to get get cleaned up.

I had always planned diagonal braces or solid panels to keep the whole machine from racking. These are not in place yet and when jogging, the whole machine wobbles. Once the braces are installed, it should be much better.

I’ve got the electronics on a little board that is hinged underneath, so it folds up out of the way where metal chips can’t fall on it, but it also folds out for easy access. Right now it’s a MKS Gen L and 24V power supply and a Raspberry Pi with OctoPrint. To set up the servo for the tool changer I’ll add a buck converter, and then there’s the LCD and joystick which will have to hang off somewhere.

I’ve got two power strips, one is always on for the raspberry pi and maybe other things in the future, and the other power strip is connected to the E-stop button, so the 24V PSU that drives the steppers and any power tools will be hooked up there. This way I can cut power to the steppers and router without resetting the Raspberry Pi or the Arduino.

4 Likes

Very cool. The space under the xy pipes is clear, if you need to add anti racking pieces. Although a piece of ply on the whole side would do it.

I like your controller area. Easy maintenance and no frills, but also clean enough to keep sanity.

Just imagine. Now you can use really thick pieces of paper :slight_smile:

2 Likes

Getting really close now.

  • The wooden cross braces on the upper part make a night-and-day difference in wobble, as you might expect.
  • I've also got a tape measure and sleeve around the X axis wires.
  • Also installed the joystick into the table.
  • Also got a long cable for the Raspberry Pi camera so I can see the workspace with OctoPrint.
  • Also got the blinky thing (MAX7219) installed, had to lengthen the wires for it to be visible when the board is stowed
I'm not sure what I'm going to do about the LCD. It works but the wires are too short for it to be convenient when the board is stowed. But on the other hand, with OctoPrint, I don't really need LCD access. But I still might want to print from SD, so maybe I will mount it somewhere even if it's not easily visible. Replacing the wires seems like a lot of trouble.

Next I think I will be trying to measure stiffness in the different modes of deflection!

[attachment file=110807]

2 Likes

There are longer lcd wires. Ryan sells them too. It would be a quick swap.

1 Like

Oh cool. Good idea!

I haven’t got quantitative measurements yet, but it’s looking like I still have work to do. To my eye, just pushing with my hand, it is not as stiff as I had hoped.

To my eye, it looks like the two worst modes of deflection are legs bending and belts stretching. The leg bending caught me by surprise because I was expecting that constraining the upper part of the leg would keep it vertical, and the conduit resistance to bending would be plenty to keep the side rails from moving horizontally. This is in contrast to the standard MPCNC where the resistance to racking would be mostly dependent on the strength of the corners and feet, not the bending of the metal.

[attachment file=111430]

The remedy for this would be giant gussets or extra wood panels that fit under the side rails. I will still measure the deflection and only attempt this remedy if it is truly a major contributor.

If I grab the side rail and push on one of the rollers with my thumb, I can see the belt stretch. I am wondering if perhaps I have crap belts. My other machine has some deflection in the belts but it was in the zip ties and I cannot perceive any change in length in the belt at all. I’m going to see if I have extra, or I’ll remove the belts if I need to, and measure the amount of stretch of the belts themselves. If I do have crap belts like I suspect, then I’ll buy some good name brand ones and that should solve that.

Meanwhile I need to upgrade my tools to get some real quantitative stiffness measurements. I think I can use the same weight hangars I used before, but I have a little 3d printed stand for my dial indicator, which is too light and too small to confidently hold the dial indicator where I need it. Straightforward enough to build a larger, heavier one.

Also, I’m a little bit puzzled that my motor drivers (A4988) are getting quite hot. I had to turn them down well below 0.7 V to keep them from going into thermal shutdown. All three are behaving the same, and I even swapped one out with a spare and it did the same thing. With current turned down, the holding torque seems low, yet the drivers are still so hot I can’t keep my finger on them for long. One problem at a time, I’m deferring this until I get the stiffness better under control. Maybe in the end it just takes heatsinks and a fan.

leg_bending.jpg

It would be interesting to test this on belts outside of the machine and zip ties. There have been lengthy discussions about this on the forums and it’s hard to tell what is belt, zip tie, or just worry that is the root cause.

You have a 24V PSU and all 4 motors are wired in series? You could wire up another driver to the same socket. It definitely seems like it should be doing twice the work, so get twice as hot, but I’m not sure where the heat is coming from. I could see the double motor resistance or double the supply voltage being the culprit, but for opposite reasons :).

1 Like

These belts I have, at 48 inches long stretch by about 4mm when under 5.5 kg of tension. I am not sure if this is typical. I’ll buy some name-brand belts and see if they are any different.

My test setup loops around one bearing and basically applies axial load to one of the side rails, which should have insignificant deflection relative to the 4mm I am measuring. By looping back to the anchor point I don’t have to try to watch both ends of the belt for deflection.

[attachment file=111497]

[attachment file=111498]

For some reason it didn’t occur to me that belt length is one of the contributors to decreasing machine stiffness as the build gets larger. I guess this is why the LR2 has 10mm belts. It’s obvious when you think about it, but belts were not one of the things I was considering.

You can also try doubling them up. Don’t mesh the teeth or anything, just put another belt behind the first, teeth pointing the same way.

That is more than I would have expected. I would assume that is linear so even 12" would have 1mm of stretch. Hmmm.

Yup, exactly why the LR went to 10mm.

4mm is a lot more than I expect though. Another point would be The quick and dirty we have for aluminum is 1.9kg of cutting force and at worst that is distributed to at least 2 belts. So maybe at what point does the belt show stretch? I believe we are only supposed to tension them to 3-5lbs.

Today I measured another belt, left over from my first machine, and the stretch was the same, to within what I could discern with my eye, so I would say 10% difference or less. Maybe they ultimately come from the same factory in China and there are a thousand resellers. Seems plausible but I don’t actually know. These are the two products I ordered

It seems unlikely that I was unlucky, getting a particularly bad set of belts twice. I also have these name brand belts on their way, which I will measure when I get them:

They are pricey but I trust E3D not to cheap out with an “equivalent”.

For the heck of it, I also measured some Dyneema fishing line I have, with 10lb max strength. It has very low stretch compared to other string of equivalent strength (similar to aramid they say), and my string measured roughly 10 mm stretch under 2kg of tension, at the same 48" or 1.2m length. This puts it at about 7x more stretchy than the GT2 belt. Of course the ultimate strength and the construction is different, but hypothetically if I cast my own rubber containing 7 strands of 10lb Dyneema fishing line, I would end up with about the same overall stiffness. (I have no intuition about cable or belt stiffness and this is me trying to get a feel for it.)

Finally got to cutting. This is Ryan’s crown gcode.

I measured the stiffness at approximately 0.01 inch of deflection per kg, and the best part is that it is not hugely dependent on the length of stickout, maybe about 50% worse at maximum extension compared to movement of the gantry rails.

6 Likes

That thing is a beast.