LISA Simpson

My guess is that the Big LISA will come in over $1K by the time I'm done. If it's $2K, that's not great, but it won't stop the program. If I get it done in 6 months and two passes, that's fantastic.

As far as a "normal" printer goes - somebody needs to get the Marlin boys working on the firmware. Yes I could do that, no I'm not raising my hand to do it. For right now the Python code looks like an adequate way to go. I'd rather get things up and running with Python and then mess with firmware at some later date. The whole end stop issue also needs to be checked out if I indeed need it (I'm about 99% sure I will).

Edited 1 time(s). Last edit at 12/27/2013 11:32AM by uncle_bob.

@CTF: The RSMax has a community behind it and a track record so IMHO it would be a better 1st printer. That said I will help you however I can if you go the LISA route.

@Uncle Bob
If the budget gets over $1K for a 2X, well, that is expected and not an issue.
The firmware is indeed what may steer me towards purchasing an RSMax for now, although I have the feeling that I will regret that decision (vs. the Lisa) almost immediately...

@Nicholas,
Thank you for the info.
I build lots of things, so I am not too concerned about that, I just would prefer to purchase the best I can with my money.
If you think that the RSMax will be equivalent in print quality to the Lisa, then I may go this route for now.
On the other hand, I have that fresh TechShop membership that is only waiting to be used.
I love to build things, as long as I know that the end results will be worth it.
I would certainly be very frustrated if I had lots of issues that I could not resolve because my knowledge is not good enough in that domain. (Firmware programming for example).
Once I get a little bit more feedback, I will decide which route to go.

I appreciate your support.

As far as I know, the number of LISA's that have actually done a print can be counted on one hand, without using all the fingers. Some of us are heading off into crazy derivative builds without a whole lot of analysis (build it and see what happens....). I certainly would not recommend anybody take *anything* I've posted as a firm basis for a real printer until I get things up and running. I'd very much recommend sticking very close to the design that Nicholas has documented for anything other than an experimental build.

Ok, if I'm going with 3/4" 0.750 OD thread rod (Roton Hi Lead) is:

[www.vxb.com]

A rational bearing to use?

The price certainly is right.....

@uncle_bob: Looks good. I would verify that the screw is .75- and the bearing is .75+. I would assume that is the normal way this would be but with the weird 19.05mm bearing I would double check.

Quote
uncle_bob
I certainly would not recommend anybody take *anything* I've posted as a firm basis for a real printer until I get things up and running.

Well, thank you for being our beta tester.
I look forward to your findings before I take any decision then.

Quote
nicholas.seward
@uncle_bob: Looks good. I would verify that the screw is .75- and the bearing is .75+. I would assume that is the normal way this would be but with the weird 19.05mm bearing I would double check.

Buying one to check is a good idea.

Of course this means I'd have to also get some thread rod. Actually mounting one on the other is a pretty sure way to check them. Four feet or five ....guess it's time to decide.

-----

There is of course the approach of *asking* if it will fit a 3/4" shaft. (far less fun). I just contacted them. The answer I got back is that it's an SAE bearing and it will fit a 3/4 shaft.

Edited 1 time(s). Last edit at 12/31/2013 09:42AM by uncle_bob.

@uncle_bob,

Bearings:
Typically you want a press fit between the shaft and bearing.

The Roton lead screw will be about 0.0035" undersized.
The Dia 0.7500" bore of the bearing will be about -0.0005" undersized.

Bearing ID: 0.7500 - 0.0005" = 0.7495"
Lead screw Dia: 0.7500 - 0.0035 = 0.7465"
Total clearance: 0.7495" - 0.7465" = >0.0030"

0.0030" is really loose, but I'm guessing that the RPM's are really slow, so it's not going to shake it's self apart.
If it's really sloppy loose, I'm probably going to take a chisel to the threads of the lead screw to increase the O.D.
That decision depends on how it feels to me after assembly.

Arm material:
I have decided to go with cedar for the arms with plastic printed ends.
Super cheap $2.98 each (1.5"x1.5"x48") x 3 = total cost $8.94.

Cedar kiln dried over 100C becomes stronger.
It can be kiln dried up to 260C, so it's safe to use in a heated enclosure.

If it improves the arms, I'm going to seal the cedar with acrylic garage floor sealer.
I'm not sure about the out-gassing, so I'm holding back on my decision.

If any one else is going to explore this, note that "ideally" you want to have at least (6 rings/inch), knot free.

In my case I'd rather have it fit and then shim it than not have it fit. Exactly how I'd shim it and whether I'd get exotic about it is TBD.

The bearing needs to go over a modest (inch or two) length of the rod. If it's a real tight fit that's going to be "interesting". If you ever want to pull this beast apart, getting a gear puller down in there could also be "interesting". I can see a situation where you decide to put new nuts on the shafts and need to pull it all apart to do so. That said, I'll probably have the bottom end of the shafts on pretty tight and plan on pulling the top's to get at things.

The right way to do it is to turn down the ends of the rod below 0.58" and fit the bearing to than surface. You probably could mill a flat on the drive end as well. That all sounds like more excitement than I want to get into. Casting a spacer / plug with one of the Cerosafe like alloys or epoxy is about as exotic as I'd want to get.

Quote
uncle_bob
Did some indirect digging on Marlin and the Wantai stepper drivers. It appears that the current version of Marlin puts out 3 us step pulses. The controller should be happy with 1.5 us, so there's some margin.

I'm curious where you got that 3us figure... Did you calculate that from the code or actually measure it with a scope?

My measurements showed about 1.88us:
[www.youtube.com]

I'd also like to know where you found the documentation on the minimum step pulse width for the DQ542MA; I can't seem to find it on any of the available documentation from the mfgr. If it's a software driver (based on Atmel) then I would be amazed if it can actually accept a 1.5us pulse. Very curious to see if this really works. Software based drivers typically need closer to ms (e.g. a few hundred us) to respond. I certainly hope I'm wrong about that.

The Polulu drivers require 1us and are very rare in accepting that short a pulse. The THB6064AH chip based driver I have been using needs 2.3us and we get past that with a pulse stretcher to make it work with Marlin, so I'm guessing you will need to use what you have built into your adapter PCB to get your drivers working.

Of course, pulse stretching does limit maximum frequency, as you already noted, but not enough to be a serious problem if it stays in the us range.

---

[www.linistepper.com] Open source stepper motor drivers.

1.5 us to 2 us is what people seem to get with the older Marlin code. The newer code has a bit more in-between the turn on and turn off. People seem to be getting about 2X the pulse width with the newer code. If it turns out to be an issue, dropping another 1us delay in the Marlin code is a real simple fix.

The minimum pulse width for the new version of the DQ542 came out of the Wanti doc's. The 5us comes from the docs on a re-labeled version of the same controller. Those doc's also have the same minimum pulse width info.

Thanks for the heads up in your previous post that got me digging into all of that.

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

Just to link things up for any poor soul trying to follow my twists and turns in building a franken-LISA, there is a thread on Concept Forge with the details.

[forum.conceptforge.org]

Edited 1 time(s). Last edit at 01/01/2014 12:39PM by uncle_bob.

Here is my first good print! I think most of you will agree this is pretty close to the gold standard.

Videos

Slow Print!
[youtu.be]

Upside down print.
[youtu.be]

I did a whole laundry list of tweaks. I designed and tested a better filament drive. I shortened my bowden tube. I decreased the current for the steppers. I added a spool holder. I am going to see if I can get calibration going tonight to enable full bed prints. It doesn't sound like much now that I list it but as lazy as I am it took a ton of Twilight Zone episodes to get it done.

Edited 3 time(s). Last edit at 01/01/2014 06:16PM by nicholas.seward.

The links to the videos are backwards ....

Other than that - VERY impressive!!!!

----

How much did things change when you inverted the printer?

Fixed the links!

When I went from normal to on its side the print had a 1mm shift. Upside down vs right side up are about the same so if you can quickly/gentry invert it during the infill you won't see much but a slightly larger layer in the middle. The 3/8" screws are springy enough to explain most of shift. The plastic and the joints should contribute to this a little. All in all I am impressed. It will be fun to compare this to when I have the 7/16" screws installed.

Well, that's a lot better than what I get when I invert the Z axis on my i3 (which then falls to the floor more or less).

How fast were you printing on each print?

About the only other thing I can see to check in this vein is twist. Not quite sure how, but it would validate the PVC pipe / thread rod part of the design.

(yes I am sitting here looking at tube for my Franken-LISA).

Edited 1 time(s). Last edit at 01/01/2014 06:37PM by uncle_bob.

@uncle_bob: It is hard to judge but I can put weight on top platform and wiggle it by 1-2mm if I really try. If I had to guess I would say that the wood is flexing and not the pipe. If you use solid 4" PVC or metal with 3/4" flatstock I wouldn't expect any problems.

I was using slic3r defaults. 30mm/s for outside perimeters and 60mm/s for everything else. BTW, I ramped it up to triple that for a while. Of course acceleration didn't allow it to truly triple but the machine was able to hang in there. The outside surface quality suffered but not as much as I would expect. (I would say that all the surface issues came from the lag from the bowden tube.) Bottomline: go as fast as you want but slow down to 30mm/s for the outside perimeter.

This is a very nice first print!
Would you say that this is comparable to a Rostock, or better than a Rostock?
I do not really have a comparison between both at this time.
The inverted idea is very neat, I did not realize the advantage about doing this until I realized the overhangs were now easy to make.

@CTF: I also have never used a Rostock so I can't really compare. I would guess that this is just a bit slower.

Additionally, orientation has no effect on the overhangs. The angle of the nozzle is all that really matters.

Very impressive. Congrats.

@Nicholas,
I was more interested in a final print quality rather than a speed comparison.
The print you made with the Lisa looks very good, and I was wondering if the Rostock would be better or worse than that.

@CTF

Print quality is a function of fine tuning the printer. Given about 5Kg of filament and three months you can probably optimize everything on a pair of printers for a given set of prints. Until you do that with a consistent set of expectations, it's going to be very tough to decide if one or the other is "better". Best guess if you did do all that work on any rational pair of printers - they both do a pretty good job.

-----------

Design question:

The standard LISA mounts the motors to the bottom plate (I think ... if not simply ignore this ):

If you mount the steppers to the middle plate, then mounting them is a bit more complex. That goes for mounting the couplers as well as the motors themselves. In a large printer, the plate may be a double thickness. That would make some of that work a bit easier. It also would mean that the motor's torque goes directly to the mid plate. There would be no thrust / twist on the bottom plate at all. One could then go with an abbreviated bottom plate.

Am I missing something here?

---------

Back to measuring the width of the doorways. Building a printer that can't move out of the room it's in seems to be a bit of an issue.....

@uncle_bob: I don't follow. The motors are attached to the bottom. You might have to draw me a picture. (I have consided mounting the steppers on top of the build plate a few times since the height of a stepper/coupler stack is about the same length as the unused thread. However, this solution is fraught with a few engineering gotchas so I keep discarding it. If I really wanted anouther 3+ inches for free I would keep everything as it is but cut an access hole in the mid plate and move the build platform to the bottom. This doesn't make sense for a desktop solution but it might for a 1.5X+ LISA.)

@CTF: LISA's print quality is as good as I have ever gotten/saw. I would put her up against most any printer. Getting good prints out of most printers is kind of like continuous arm wrestling. Sometime you make improvements and some time you slip up. It always takes effort. :-)

@Everyone:

115x115x115 print
&
190x190x190 print


First layer


Bridging (71mm bridge at the top and then each one is 10mm shorter)

Lots of good news. All the distances are so well defined that LISA holds to a plane like a champ. I have done no tweaking to keep LISA from doing the tradition Rostock cupping problem. I was able to use the whole print bed with only a couple of judiciously induced skipped steps. I am very happy with all the results and would use this over my Printrbot for most things.

Some not as good news. I tried taking pictures but there are some very subtle artifacts from screw wobble and for some strange reason there is a hint of a backlash artifact from one screw. Again, I tried to take pictures of these flaws but they are beyond the ability of my camera phone to capture. Most people wouldn't notice unless I pointed it out to them and made them run their fingers over it. The artifacts are some inconsequential that I am considering leaving LISA-0 with the 3/8" screws (It prints great every time so I think it will get inserted into my student's bot farm.) and building another one with the 7/16". I want to try a few more things so maybe another build is the way to go.

I think I am going to add to very stiff bed leveling springs to make in flight layer adjustment a snap. There is no need for a leveling probe.

Edited 1 time(s). Last edit at 01/02/2014 11:58AM by nicholas.seward.

@uncle_bob - I have been thinking about polycarbonate tubes for the arm lengths instead of aluminum. Higher temperature than PVC with less sagging and good strength, light weight. I think you could use polycarbonate with a moderately heated chamber?

[www.mcmaster.com]

Or let's go in on an X-Winder and make a few dozen carbon fiber arm tubes.. [www.xwinder.com]

Edited 2 time(s). Last edit at 01/02/2014 12:06PM by jason.fisher.

@Nicholas

I *think* (and I may well be wrong) that the bolts that fasten the motors down go through the plywood that is at the very bottom of the assembly. Put another way, they fasten below the motors.

What I'm suggesting is having the motors stay below the build plate, but bring the mounting screws up rather than down. They would fasten to the "build plate" plywood. With double thick wood, countersinking the bolt heads would be fairly easy.

@Jason

As you may gather, I've gone through a lot of gyrations on the arms. I finally decided that drilling a pair of holes in a solid bar was simple enough that I'd go that way. Aluminum is pretty stiff, and it's also fairly cheap. If you are doing PLA, I'd bet that polycarbonate would be fine no matter what you did for heating. The other printed parts in the system will have trouble before it does.

Edited 1 time(s). Last edit at 01/02/2014 12:45PM by uncle_bob.

Scale up question:

The central axis of the machine is a hollow plastic shaft that's printed as part of hub_bottom. The shaft appears to be 15mm in diameter. I'm *guessing* that scaling up that shaft diameter would be a good idea on a 2X larger printer. 25 mm and 30 mm seem like pretty good candidates for a scale up from the standpoint of bearings to replace the 15mm 6702's.

So now the questions:

Is the shaft dimensioned based on the need to mount the hot end on the bottom or is it critical? (Do I need to scale it or not?)

If I need to scale it, is 1.5X (25mm) good enough or is 2X (30mm) the way to go?

Yes, I could go to something completely different. I'd prefer to go in as few directions at once as possible. The more I can retain from the original the better. At this point scaling is ok. Machining a different design from a solid block of tool steel - not so much.

@uncle_bob: My constraints when design were to have a hot end mount in the bottom, have a 4mm hole for a PTFE liner up the middle, and have the outside dimensions at 25mm. I bolt with or threaded rod with a 4mm hole or a 2mm polished one would do the trick. I figured that would not be a hardware store find so I settled for a plastic shaft. I ran through all the bearings that would give me a reasonable sized shaft but wouldn't bust through the 25mm outside dimensions. The list was quite small.

6701
6702
6703
6800
6801
6802
6900
6901

I then went through them and saw how much they cost and how well the worked with the design.

So to answer your question, 608 bearings could be good for a 2X design if you can find magical bolts with longitudinal through holes. If you wanted to go massive and just use sheets of plywood for the hub you could even use lazy susans. If I was to build a 2X I would us the 6805s (25x37x7) that I have left over from THOR Simpson.

I looked at :

Hollow studs:

[www.mcmaster.com]

(not very precise)

For a lot of money you can get precise (but no threads):

[www.mcmaster.com]

There are also a lot of smooth tubes in various other levels of precision.

To your point, none of it looked very promising. The plastic shaft seems to be a pretty good alternative.

I'm may need to re-do the hub parts to get them to mate up with my aluminum bar arms. Once I need to re-draw them, doing this or that while I'm at it isn't that big a deal.

Thanks!

Dual 608's:

Right now the arms use dual 608's at the pivots. That's going to spread some of the forces out over more plastic. It also will constrain the shaft through the 608's quite well.

An alternative would be to mate the shaft tightly with one 608 and use a material like aluminum to mount the outside of the bearing. The aluminum isn't going to let the bearing go anywhere. The tightened down shaft won't go anywhere inside the bearing. The bearing isn't perfect trying to do this (single row of balls).

How much do you loose going to one bearing in this case?

I'm not really after a 0.00327" sort of answer. "A lot", "not much" and "don't do that" are all a lot better than what I have now. My concern is mainly making the arms more massive than they need to be. Cost is not really to crazy either way, but it's lower for the simpler arm.

The short answer is you can't get away with it. 1X LISA couldn't and it gets worse the bigger you go.

The medium length answer is that we need two side loaded to remove all play and get perfect rotational motion where the joints add nothing to the play. It will be hard enough to make sure the members can minimize the rest of the play. I am 100% sure that it would be a hot mess if I went with 1 bearing on LISA with all other thing left the same. We are talking about long arms here. My arms are 200mm long so if you had just 0.1degree play this would be 1/3 of mm play.