kbob and Impulse, the Rostock MAX

[kbob]

This post contains math. If you don't like math, look away now.

I mentioned a couple of days ago that some of the measurements on my Rostock MAX don't match the values in RepetierMAX's Configuration.h. I've spent several hours measuring, remeasuring, and calculating, and I have some numbers. Here is my report.

Methodology Notes

All calculated values are given with a precision of one micron. That's obviously bogus — my best caliper's precision is 10 microns. I just used high precision to avoid additional error from calculations. I'll list the approximate error magnitudes where I can; if it's not listed, assume it's better than the listed errors. (-:

I will not go into detail about how I got each measurement. I approached some of them obliquely. (-: (Distance from unmarked center of bed to center of extrusion?)

And I did calibrate the BSAC, as mentioned earlier.

The Parameters

Here are the original constants from RepetierMAX.

Code: Select all

#define DELTA_DIAGONAL_ROD 269.0 // mm
#define END_EFFECTOR_HORIZONTAL_OFFSET 33
#define CARRIAGE_HORIZONTAL_OFFSET 35
#define PRINTER_RADIUS 198.25
#define DELTA_RADIUS (PRINTER_RADIUS-END_EFFECTOR_HORIZONTAL_OFFSET-CARRIAGE_HORIZONTAL_OFFSET+0.0)

And here's the canonical diagram. Marlin's DELTA_EFFECTOR_OFFSET is Repetier's END_EFFECTOR_HORIZONTAL_OFFSET, and Marlin's DELTA_SMOOTH_ROD_OFFSET is our PRINTER_RADIUS.

(Shamelessly stolen from somewhere on the Internet)

The Measurements

I measured DELTA_DIAGONAL_ROD at 270.029 mm, +0.104 mm, -0.335 mm. One of my arms was ~0.4 mm shorter than the others. If you discard that arm, the other five are all within 0.1 mm of one another.

END_EFFECTOR_HORIZONTAL_OFFSET came to 31.638 mm, ± about 0.1 mm.

CARRiAGE_HORIZONTAL_OFFSET is 37.970 mm ± 0.1 mm.

PRINTER_RADIUS I measured just above the bed and also just below the top. The average of all measurements came to 200.823 mm. The minimum was 200.741, and the max was 200.972, a difference of 0.231 mm. (These measurements should be accurate to about ±0.04 mm.) I have not yet used my alignment jig to square the towers up, but I will, and PRINTER_RADIUS will change slightly.

Using these numbers, DELTA_RADIUS is calculated at 131.216.

The two numbers that are actually used by Repetier firmware are DELTA_RADIUS and DELTA_DIAGONAL_ROD. DELTA_RADIUS is off by ~0.97 mm. DELTA_DIAGONAL_ROD is off by 1.03 mm.

My best calibration efforts without changing the firmware constants left me with the hot end about 0.2 mm lower in the center of the bed than at the edges. I honestly don't know whether the errors I measured would account for that -- I find it hard to reason about how changes in rod length and printer radius affect the delta warp.

Anyway, that's where I am. Suggestions welcome.

(BTW, as a software guy, I'm fascinated and horrified at how hard it is to pin any of these measurements down. Pick up the caliper, put it right back down, and get a different reading. I did what I could to find repeatable ways to measure these things. Measuring the effector was hardest.)

[kbob]

Following up on myself... Early this morning I wrote up all the facts, but didn't summarize.

Conclusion

The dimensions in the firmware are all wrong. Not drastically wrong, but they're all a little bit off.

• Delta arm length: off by 1.0 mm.
• Effector offset: off by 1.4 mm.
• Carriage offset: off by 3.0 mm.
• Printer radius: off by 2.6 mm.

That bugs me -- it indicates that SeeMeCNC was just hacking at it until it appeared to work. Either that or there are deeper reasons not to use the actual dimensions that I don't understand yet. Or maybe some sizes changed between the kickstarter generation and my May, 2013 kit. But still. How can they all be off by that much? Bad data for one or two dimensions would be understandable. Not good, but understandable.

And there is good news. The good news is that the errors all more or less cancel each other out. Delta radius, which is derived from the last three, is only 1.0 mm. That's 0.7% undersize. And the delta arm length is also undersize by 0.3%. So the overall misalignment is, uh, something smallish less than half a percent. (I am not going to do the trig right now.)

Of course, if you're trying to get a thin layer of plastic to adhere to the whole bed, you need much better accuracy than that.

[ApacheXMD]

So when you used your measured values, how did the prints turn out?

When i measured the stock arms, they were spot on at 269mm.

[kbob]

So when you used your measured values, how did the prints turn out?

When i measured the stock arms, they were spot on at 269mm.

You should ask that question in the future tense. (-:

I removed the effector and arms to measure them, and will probably not get another chance to work on the printer until next week.

I wonder if my arms are "special".

[CJGerard]

I wonder if my arms are "special".

HEHE they rode on the seemecnc short bus, we are all special in one way or another.

[doctorgonzo]

I believe the proper spelling is "speshul."

Could this issue have anything to do with heat cycles or stress causing the stock arms to "drift" in length over time?

[foshon]

I wonder if my arms are "special".

HEHE they rode on the seemecnc short bus, we are all special in one way or another.

Did they have to wear the helmet?

[lordbinky]

I wonder if my arms are "special".

HEHE they rode on the seemecnc short bus, we are all special in one way or another.

Did they have to wear the helmet?

All but one did, that's why it was ~0.4 mm shorter.

[kbob]

I have a cheap dial indicator. I got it at Horr'ble Fright last year for another project. http://www.harborfreight.com/1-inch-tra ... r-623.html

I designed this nifty little mount for it. I designed and printed it back in June, but didn't get a photo of it on the printer until tonight.

OpenSCAD FTW

Here's the real thing.

Not a render.

I made the central hole a little undersized. The three slots around it let the hole stretch to make a nice friction fit around the indicator's shaft. I'm starting to figure out how to design for ABS.

You can also see the hair scrunchies that tension the delta arms. There are three more up by the CheapSkates.

Here is the OpenSCAD source file.

Dial Indicator Mount.scad

Dial Indicator Mount - OpenSCAD

(1.65 KiB) Downloaded 822 times

My printer is back together now. I aligned the towers using the alignment jigs, I've changed the printer dimensions in the firmware, and yes, I've used this dial indicator to see how well it's aligned. I haven't had a chance to print anything yet.

More on that later; right now I just want to tell you that using a dial indicator to adjust the limit switches is about 20 times faster than sliding a piece of paper under the nozzle.

[kbob]

Did you miss me?

On July 17th, I took the delta arms off my printer. That started an odyssey of calibration, rebuilding, repairing, and upgrading that finally ended this morning at 12:15 AM when I successfully printed a calibration cube.

Here is the cube that I printed. It measures 20.00 mm across.

(Desk inlay not included.)

I typed in about four pages of text describing the ordeal in great detail. I edited it for two hours, then SeeMeCNC's web site threw the whole thing on the floor because I attached an oversized image. Bah. I'll repost parts of it someday. It's late now.

[doctorgonzo]

As a carpenter/furniture-builder guy, I dig the inlay. The print seems pretty good, too. Did you build the desk?

[kbob]

As a carpenter/furniture-builder guy, I dig the inlay. The print seems pretty good, too. Did you build the desk?

My friend Jacque built the desk, but I "helped".

More desk photos here. http://www.perfect45degree.com/gallery/

[ApacheXMD]

So your measured geometry values worked out well I presume? That hollow cube looks pretty good!

[kbob]

So your measured geometry values worked out well I presume? That hollow cube looks pretty good!

It looks like my last post about calibration was on July 18th. Between then and now, my printer has been inoperable most of the time for a bunch of reasons.

I measured the distance between each pair of towers both at the top and at the bottom. I picked the tower that was furthest away from the others as my reference. Then I squared that tower to the base. (It had been square once. But I've picked the printer up, transported it in the car, and generally manhandled it enough that it had gone out of square.)

Measuring distance between towers

Once that one was square, I locked it down and installed the triangular alignment jigs. One triangle went at the bottom and rested on the Onyx. The other triangle went at the top. It rested on some clothespin clamps, and the top assembly rested on it. The alignment jigs had to support the weight of the towers and the top piece, so I made sure they were clamped tightly enough.

Then I loosened the screws holding the other two towers (being careful that the alignment jigs prevented them moving vertically), and carefully adjusted the nuts on the threaded rod until all six triangle sides were exactly the same length, as measured using the above technique. Well, almost exactly. There is still 0.2 mm difference between the longest and shortest. Of course, re-tightening the screws changed the alignment, so there was a lot of trial and error.

Then I updated the dimensions in the firmware, installed the dial indicator and started testing. My dial indicator is not metric, so readings are in inches instead of mm. If I remember correctly, the first reading put the corners about 0.004" (0.1 mm) lower than the center. So after all this effort, I went back to tweaking PRINTER_RADIUS to find the value that works best. The best result was 0.675 mm less than the calculated value. With that done, I got the gauge to read the same number to within 0.0004 at all three corners and the center.


Here's a pro tip for bed calibrators.

The procedure in Gene's manual measures the corners 90 mm from the center. That was probably a good choice for the early adopters who had 8 inch square beds. But the Onyx is big enough to measure the actual corners. On my printer, the delta arms are vertical at 129 mm from the center. (129 ± 0.5. I wasn't real precise.) So I measured at X, Y = (0, +129), (+111.72, -64.5), and (-111.72, -64.5). Keeping the delta arms vertical improves resolution a little.


That's where I was on July 24th. Then stuff started breaking, and the printer did not run again until August 7th.

On the 7th, I set the absolute Z height. The dial indicator could only verify that the four points all had the same height, not that they were the right height for the hot end. I followed the usual procedure: heat the bed and hot end, move gradually down until a piece of paper drags, copy the calculated number into EEPROM.

Then I checked that the paper dragged at the corners. It was gratifying that the drag felt exactly the same in all four spots. I checked three times, partly to be sure, but partly to feel how nicely it repeated.

Then I printed a calibration cube. I'd switched to a new color of filament (Neon Purple from Toner), and needed to calibrate the extrusion rate. The second calibration cube is the one that measured 20.00 mm (+/- 0.02 mm) across. The walls are a tad thin on that one: 0.53-0.54 instead of 0.55. (photo)

Next, I printed the Rostock Onyx Bed Leveling Aid. It took a few tries to get a good glue layer on the board, but I eventually got a good print of it. The glass plate seems to be a little low halfway between the Y and Z posts (at 2 o'clock: you can see where the skirt came unstuck). I don't know how a glass plate can sag, nor how to compensate for it. Maybe it means I haven't finished calibrating. Or maybe the bed is cooler on that side. I don't know.

Can you ever really have too much glue stick?

Then I printed another calibration cube -- er, brick. This one was 20 by 150 mm. After the bed cooled, it measured just slightly oversize, 0.3%. (After I peeled it off the bed, it was too flimsy to measure.)

This caliper is neither substantial, alloy, nor from Birmingham.

So I'm happy. You can see that the printer is still not perfect, but it is a lot better than it was, and I bet it's better than the average working Rostock MAX.

Next: How I burnt out my RAMBo board.

[JohnStack]

Great post, trying it in a couple of weeks.

[kbob]

Great post, trying it in a couple of weeks.

Have fun. It was an odyssey.

[kbob]

About July 28th, I was heating my printer, getting ready to calibrate its height. Several minutes into the heating cycle, the LCD display went wacky. I think it turned into all white blocks, but my memory could be faulty. After I poked at it a little, I got the LCD panel to come back. I don't know how. But the USB interface was dead.

Here's the thread in the Troubleshooting forum.

I'll cut to the root cause. I had soldered the hot end thermistor leads to stranded wires and wrapped everything up in heat shrink tubing. But one strand of wire was peeking out from under the heat shrink. When the wires were bent just so, that strand would contact the crimp connector on the resistor, which was uninsulated and short the thermistor to +12V.

Thank you, John Stack, for pointing me in that direction.

The wires did not normally touch. In fact, I used the hot end for six weeks before they did. After the damage was done, all circuits measured fine because the wires were no longer touching.

I can confirm that the USB circuit on the RAMBo is definitely dead. The board does not identify itself as a USB device. The LCD panel and pushbuttons are alive, and I can control the motors and limit switches from the front panel. I don't know whether the thermistor A/D works. There are no visible signs of damage (e.g., burn marks).

It's possible, though I think it's a long shot, that the ATmega 32U2 which handles USB could be reflashed and revived. I haven't tried that, since I replaced the whole RAMBo board. It would be nice to have a spare RAMBo (or to have one to make another printer), but that's not a priority yet.

It might be fun to repair the board while teaching myself to solder SMT, since the upside is big (new RAMBo) and the downside is small (dead board remains dead).

SeeMeCNC had new RAMBos in stock. The new one came with the RostockMAX firmware preloaded and a working bootloader.

I managed to destroy my thermistor shortly after the new RAMBo arrived. Fortunately, a friend here in Eugene had soem spare thermistors in his parts bin. It is not the original part, but it is 100K and its Beta is close enough that I did not have to change the firmware.

I also took the time to get some diamond thermal paste (Antec Nano Diamond Formula 6 from a local Staples store). Now the thermistor is embedded in thermal paste, and the hole is sealed with RTV. I also had some extra PTFE sleeving, though it's bigger than the original.

So the thermistor is installed a little better than before, and the hot end works. I have not yet calibrated the new thermistor against a thermocouple.

Next: Installing motor dampers.

[kbob]

The day before my replacement RAMBo board came in, the motor dampers from astroboy907's group buy came in.

So I installed them. They work great. I'm not sure how much quieter they really are, since the printer was down for three weeks, but they _seem_ quieter.

While I had the motors out, I decided to dress the wires. I used the old paracord sheath trick. That, along with some heat shrink tubing and label tape, made decent looking motor wires. In my ever-so-humble opinion. (-:

1 use for a dead RAMBo.

I need to do something with the limit switch wires, too. And then, if I'm going to high-mount my extruder, I'll want to redo the wiring to the extruder and hot end too. In other words, I want to redo all the wiring. (-:

[geneb]

I'd love to learn more about the "paracord sheath trick" you used.

For routing wires from bottom to top, I'd recommend printing a pair of brackets that I designed (available on github.com/seemecnc/RostockMAX in the "printed parts" directory) that you use in conjunction with a 3/8" dowel rod. I'll have really good pictures of a good installation of it soon - I've got a pair of RAMBo boards with dodgy bootloaders, so I have to wait for my new programmer to arrive. (I didn't want to spend the time to dig up the STK500 I've got stashed somewhere).

g.

[kbob]

I'd love to learn more about the "paracord sheath trick" you used.

"Ah, yes. The old paracord motor wire sheath trick." — Maxwell Smart

Updating a classic.

I learned the paracord trick from Bart Dring, who runs buildlog.net. But I can't find a description, nor even any pictures, on that site now.

Parachute cord has a nylon sheath around a bundle of longitudinal fibers (also nylon, I think). You don't need the mil-spec stuff; wire dressing is not life critical. It comes in lots of colors, so you can color code your wiring, but I just used black. Many sporting goods stores sell it. I got mine on Amazon.

Cut a piece of cord about an inch longer than you need. Pull out the inner fibers and discard. Slide the sheath onto a metal tube. (This is the hard part; see below.) Using a soldering iron you don't like very much, melt-cut a ring around each end of the cord. This leaves a fused nylon end that is round, not flattened. Pull off the frayed ends and discard.

I hate you, solder tip! I coat you with gummy nylon!

Next, stuff the ends of the wires into the tube and slide the sheath off the tube and onto the wires. Slide on a piece of heat shrink tubing.

Note: these wires are too big for paracord.

Attach the connectors to the ends of the wires. Slide the sheath into place, then the heat shrink. Heat it and shrink the shrinkable, being careful not to melt the nylon. El voila!



How do you get the sheath onto a metal tube? I'll tell you how I do it, but I'm sure there's a better way waiting to be discovered.

I got a length of 5/32" OD brass tubing from the hobby shop. First I burnished and rounded the end so that it would slide in a little easier. That sort of worked, but wasn't great -- the tube edges would still catch the nylon fibers and snag.

So I made a needle end. I got some 6awg wire at Lowe's. They sell it, uninsulated, by the foot, and I only needed 6 inches or so. 6awg is just slightly bigger than the 5/32" tubing. This is the wire that grounds your house's fuse box (if your house is wired to US electrical code).

I put the wire in a drill press and used a piece of sandpaper to make a blunt point on one end. On the other end, I turned it down (still using sandpaper on a drill press), then gave it threads using a handheld die.

The pointy bit.

Is this the vertical mill? Why is the cutting tool made of paper?

If they think you're technical, go crude.

I tapped matching threads on the inside of the tube. Now I can screw the two pieces together and use the needle point to thread the paracord sheath onto the tube, then unscrew the needle, insert wires, and slide the sheath off. The threads inside the tube grab the wire insulation and hold it in place.

Someone with actual metalworking skill or even with actual metalworking tools could make a much better version of this tool.

[lordbinky]

Quick, someone post an STL of that finished tool

[JohnStack]

Holy Crap Kbob, I just read through this post again. Holy Crap!!!

You're like MacGyver. You don't have a mullet, do you?

[img]http://ia.media-imdb.com/images/M/MV5BM ... 4,317_.jpg[/img]

[Eaglezsoar]

I enjoyed reading that but I have a question about doing it that way or using ordinary wire sheathing.
Is there an advantage to your method?

[geneb]

I know my reaction to an iphone on my shoe would be, *STOMP* followed by, "Eewwww! Scrape it off!"

Nice trick kbob. You know they actually make that sleeve material FOR doing wiring harnesses right? I'll have to see if I can find it online.

I like the diy-ness of your method though. Works great if you can't find the "real" wire sleeve material.

g.

[lordbinky]

Yeah you can buy braided wire sleeving, but for the impatient DIYer I can think of four brick & motor stores I know off off-hand that carry paracord which is still similarly priced as the cable sleeving, and I live in a tiny town in the middle of nowhere so alternatives to waiting for mail delivery are great.