LISA Simpson Take 2

The original LISA prototype was a success. She was such a success that I have no plans on taking her apart to reuse the components. (My original Simpson design has long since been scavenged.) As far as print quality, the only problem is from screw wobble artifacts. I have already posted an updated design in my gitHub that uses screws that are 2.5X stiffer which will allow the acceleration to be increased and should remove the wobble. The original LISA had some slight backlash artifacts (barely visible) but nothing that I would call a deal breaker. For now I have no plans on addressing the backlash. (Possible solutions including buying a anti-backlash nut or spring loading two nuts against each other. Both solutions add costs and according to my decision making process, removing the backlash is not worth it.) Other than the backlash and the wobble, there were only two remaining "problems".

1) She took up a 500mm diameter area for a 225mm diameter printing area.
2) She didn't fully utilize the full length of her screws.

Thanks to uncle_bob trying to squeeze a mega LISA through a doorway. We figured out a more compact arrangement for the support columns.

To better use the screws length I am dropping the bed down. This has an added benefit of letting me physically level the bed on the fly and will let me access the electronics without more intensive dissassembly.

So here is the rendering of my proposed redesign. Missing a few parts but this should get the idea across.


She now can fit in a 400mm circle and has about 35mm added to the print height.

It is going to be a while until I build my second LISA. Do you guys/gals have any comments or suggestions?

Looks cool!

1) When you womp the big circle out of the middle plate, you reduce it's stiffness a bit. Do you think it still will be good enough to keep the screws all in line?

2) You can print the part the rides on the arm so the pivot point is a bit higher. That *may* give you more run length on the drive screw. Unless you intend to drive the top into the upper plate that would reduce the value of dropping the plate. Are you driving into the upper plate?

3) You could terminate either end of the build plate adjuster into a T nut (they work well in plywood). That would eliminate one of the adjuster nuts and it's clearance issues. Issues? They normally come in coarse thread so that could be an issue.

4) You can drop the point the arms hit the effector. That will (net) raise the bottom of the hot end relative to everything else. At some point the bottom of the arms hit the printed object. Put another way - bottom of level arms = tip of hot end = optimum. If you do this does dropping the build plate still pay off?

5) If you go with a hot end that has a fan on it (or other "junk") that will increase the diameter of the hole in the mid plate. How big a hot end diameter have you planned for?

Edited 1 time(s). Last edit at 01/16/2014 12:27PM by uncle_bob.

@uncle_bob:

1) I suspect it will be stiff enough. I am considering doubling up the plate as a backup plan. However, the top and bottom plates will probably give me the rigidity I need and the mid plate only needs to register the screws in the xy direction.

2) The shoulders do get zeroed against the top plate. (I am probably close to my limit of squeezing out more build volume in the z direction. The drop plate was more for leveling and just had the side benefit of adding to the volume.)

3) Good idea. I also am considering the making the nut holder for the top reach over the side to hold it in place. That way I only have to turn the bottom one. However, those wingnuts don't cause any clearance issues. :-) You would think that they do but I check and double checked.

4) I am at my near optimum. (I use the whole length of the screws.) The only way I can get more range at this point is to bring the screws closer together and redesign the hub so that the hot end can come within 25mm of the screws. This will shorten the arms which in turn will give us more z height. I did a quick run through on that design direction and if I change the screw spacing from 300mm to 250mm it will shorten the arms to 150mm. It will also add 50mm in the z direction of the build column and will reduce the footprint to a 350mm circle. To give you a feel for the size, it will just barely cover a 8.5"x11" piece of paper.

So, I will change the wingnuts on the top to something that just holds the bolt and hooks onto the side. I will also experiment with decreasing the screw spacing. I haven't done so yet because I have been reusing GUS Simpson's hub but it might be time to diverge.

Note: I also looked at changing my columns from 2" PVC (2.375" in real life) to 1.5" PVC (1.9" in real life) and I only shaved off 30mm from my footprint. To me I don't think it is worth that change.

I have sourced some of the aluminum tube that I suggested on the other design. That stuff is *heavy*. It's also a bit exciting to cut. The PVC is probably a better way to go for a generic design.

I played with shrinking the printer in the way I mentioned above. It was quickly evident that having the dropped bed would not work with the smaller footprint. The design changes quickly started stacking up for what amounted to a 50mm decrease in the footprint. I am not one to worry about the footprint excessively so I have currently ditched any ideas of compressing the design. Bottom line: I can make her smaller with the same print volume but it isn't pretty. I vote for elegance.

@uncle_bob: Yeah, it seems like PVC doesn't get much love but it is so cheap, very easy to source, and very easy to work with. I figure that people can substitute anything they want for the columns using the tensioning technique that I use. All they have to do is print adapters that go in the end of the columns to center the column on the tension rod.

I think that moving the supports in takes care of most of the footprint issues. If you pack it in to tight, getting the print out of the machine will be a hassle. A max height / max diameter cylinder will be interesting enough to remove with a sunken print surface. Adding another tight fit isn't going to be very user friendly.

------------

So how long are the lead screws on this version? They look like 3' in the picture, but that's not a good way to figure things out.

--------

With the adjuster screws on the print plate where you have them, the weight of the plate hits pretty close to the weakest part of the mid plate. With a heavy plate (or some heavy filament ) that might give you a bit of bend.

Edited 2 time(s). Last edit at 01/16/2014 09:35PM by uncle_bob.

@uncle_bob: I bought 24" screws. This gets me around 250mm of print height.

Ignoring the ability to level the bed, doesn't the cut out lowered bed trade better
screw utilization for build complication? Personally, I'll gladly pay a little extra
for the screws to avoid this complication.

As for bed levelling, would it be possible to add a servo actuated bed probe,
have the firmware auto probe the offsets and compensate in firmware (let's
assume the bed is perfectly flat)?

I find simplicity to be one of the attactions of LISA and wouldn't compromize there.

I've done the auto bed leveling stuff on other printers. It does work, but you need to do it right. The probe and mount need to be more rigid that you might at first think. Getting a top mounted probe past the arms is a bit exciting in terms of getting the problem past the arms. Mounting the probe on an arm is a clearance issue with arms bumping each other. If you have the arms low to give you maximum print height, there's not much room to mount it below the arms.

None of that says you can't do it. The most likely solution is to mount it low and give up the build height.

-----

Since the gcode is all pre-corrected for the geometry, the standard bed leveling firmware will not compensate a LISA. You would need to use the individual probe codes to collect data and feed that back into the Python pre-processor. The bed level rotations would need to be moved to that code and re-written in Python. You would then run the reprocessed code for the print. Again, not anything you can't do. It's just not very user friendly.

Edited 3 time(s). Last edit at 01/17/2014 07:26AM by uncle_bob.

Thanks Uncle.

I'm well aware of the software flow, but you probably (?) also don't need to autolevel for every print.
Besides, I consider the current pre-processor flow temporary;
with a 32-bit controller you can move all the math to the firmware, eventually.

BTW, I should have said that I'd rather trade build height for a simpler construction.

If you are going to the effort of putting the leveling probe on the head, you might as well use it to its fullest. Plywood does change as it dries out. Variable warp with time is not at all an unusual thing. The biggest issue for probing is space in the middle of the printer. Here's a possible solution:

1) Right now all the arms meet at a single vertical pivot point directly over the hot end. That makes for simple(r) math. It also means we have a hollow shaft in the middle of the printer.

2) You *could* replace the single vertical center pivot with three pivots (one for each arm). Mounting them around the edge of a round disk would give you room for this and that in the middle of the printer. (Could be a probe setup, a compact extruder (yea!), a rotary tool, ...). The hot end would mount in the center of the disk.

3) The added pivot points mess up the math in the pre-processor. You have one more angle to figure. Since that's code on the PC, I don't see it as an issue. You would have one more variable into the pre-processor (vertical offset maybe) that would be zero for the normal LISA. It will tax a Mega a bit more. Fast ARM’s with FPU’s better get here soon!!!! Fast stepping and lots of math do not sound like a good combo on the Mega.

4) The three hubs would use the same number of bearings as the center hub. They could easily be 608's with 8mm blots in them. Hollow bolts are no longer a constraint. Cost to build is same / less with this approach.

5) The center disk would be < 4” in diameter on a normal sized LISA. How much smaller depends on clearance issues. Even at 3” you have a lot of room for stuff. On a scaled up printer you (obviously) would have even more room.

6) Print area / print height would not be directly impacted by the change. (At least that’s my guess). The arm “reach” at the full extension points is not changed. That assumes the distance to the inner horizontal pivot in the arm is not changed.

7) A practical implementation might lose you a bit of build height (maybe an inch) simply to get everything to clear. A lot depends on the distance from the vertical pivot to the inner horizontal pivot.

8) I believe that accuracy would not be impacted by this change. The disk could be pretty heavy duty.

9) Ideally I’d put the pivot arms *under* the disk. That way you don’t lose build height. That puts some constraints on the design in terms of bumping. I’d have to do a bit more work on this to see how severe those issues are.

10) We already have one printer design and one set of math that does not fold into Marlin. This would add a second set of math to that pile. Having that math variations grow faster than the number of printers built may not be a good thing.

Yes I know, it would be a lot easier to work out what the heck I’m talking about if I did a sketch ….


(Truth in lending --- I'm mostly thinking about doing this to put a compact extruder and probe on a scaled up version. The post here is simply a sneaky attempt to get the alternate math folded into the main core of the project. Also to get somebody other than me to do the math…).

@uncle_bob: Unfortunately the arrangement you propose doesn't produce constrained motion. The central hub will be able to rotate and raise given three constant steppers rotations. (Technically you are constrained unless you are in the center but I wouldn't trust it to not produce nasty results.) There are options but you will have to break the symmetry. http://youtu.be/N5CgIq2YWps I have a few symmetrical solutions but none of them are simple or really worth sharing. I would just go with bigger bearings or lazy susans if you need an area to work with in the middle.

@tth: I wouldn't call the dropped bed more complex. It adds 3 springs and an extra plate. (Maybe $1-$2 in cost and a small amount of fabrication time.) However, it allows for easier access to the electronics without removing the tensions rods. It also lets me level without having to induce missed steps or using software leveling.

I am staying away from limit switches and autoleveling probes for a few reasons. The main reason is that the firmware doesn't exist. Even an act like homing would require a special routine because a carriage position is a function of all three stepper rotations and you don't know the initial conditions when you are homing. This is not a trivial math problem. Even my less than rigid prototype was able to print pretty level so with only a minute adjustment to a screw every few prints I should be able to keep her printing level.

My main reason for avoiding adding switches is that it will require more time to install/maintain/debug them than I will save. LISA returns back to the top after a print so I don't have to do much to home her manually and I usually stick around for the first layer of a print so it is no big deal to tweak the level as the skirt is being printed. I made it through several runs that require nothing besides print removal.

@uncle_bob: I suspect that putting the inverse kinematics into firmware will get easier and easier. I am avoiding writing firmware for a few reasons. Eventually, I will be addressing the firmware. I would expect a solution in the future that only requires that you specify the inverse kinematics and the rest will be taken care of. I think Marlin is almost there but you have to be mindful of hardware restrictions.

Also, in regards to the drop plate connecting at the weakest point, I can move it next to the column. However, besides looking bad if it droops and warps, it won't matter as long as I don't pick up any bad vibration during a print. Easy change.

Warp:

My concern is mainly the warp being humidity / time dependent and messing up your bed level.

Three pivot:

Darn that physics stuff .... I had been looking at it only at the edges. The center does indeed create an issue. I'm wondering if a heavy center would seek the lowest point . I don't think I would want to depend on that. To much chance of it turning in to a very cool looking rotary pendulum.

Bed level:

I would not try the bed level code without limit switches on the printer ..... You *could* do it, but there would be a lot to it. .... actually I think there are some initial states that would crash before any leveling code could do it's thing. Limit switches would be part of it.

Edited 3 time(s). Last edit at 01/17/2014 07:36PM by uncle_bob.

Here's an alternative to sinking the build plate. I'm not claiming it's a *better* alternative in all cases...

Make up a C shaped structure to elevate the bearing that holds the lead screw above the mid plate. Have the open part of the C towards the build plate. There is a flat plate on the top of the C shape that the bearing mounts under. Think of it as a pipe supporting the bearing mounting plate with one side cut out of the pipe. The pipe supports the plate on three sides so it's not going anywhere. The side towards the build is fully open so you have no (major) clearance issue in that direction. There are a *lot* of ways you could build the structure. You would need a bit more than just the pipe to make it work (like some thread rod ...).

If the bearing is below the bottom of the hub arms, you lose no print area at all when the hub approaches the lead screw. You *could* set it higher and only loose the radius of the bearing past the lead screw.

If you want the arms to go level, that sets a max height on the bearing. Your nuts have to get down on the lead screw.

The motors would mount below the mid plate. If (for some odd reason) they don't have 6" shafts on them, you would need a shaft extension.

On the plus side:

This gets you back to a solid mid plate. The build plate would be up where you can see it and poke at it. Your hot end could have all sorts of junk hanging all over it with no major restrictions. Pulling full height parts out of the printer would not be an issue. Adjusting bearing location with it above the mid plate would be a bit easier.

On the minus side:

You have three gizmos to make up. I'm not sure that a printed part would be strong enough, but it might be. Certainly an aluminum plate (or three) and three big thread rods would make an adequate structure for each support. You have one more thing to get wrong / out of level / drift / mess up in the design.

Alternate:

You could make the pipe solid all the way around. That would not let you have a hot end with several inches of stuff hanging off of it. It would give you support on four sides. Moving the bearing around for alignment might be a bit more complex. You are pulling up on the pipe (pre load on the bearing) so having it whole probably keeps out one possible way for it to tip.

Truth in lending:

Yes, this is a way to get back to a design that fits an 18.6” build plate through a specific width door.
.

Edited 1 time(s). Last edit at 01/21/2014 12:54PM by uncle_bob.