I’m nearing the end of my time building a reprap prusa mendel. It’s almost working — All the assembly is done, firmware loaded, and first “blobs with rounded corners” printing. However, it isn’t yet usefully printing — I’m having issues controlling the speed of my Prusa’s axis’s vs the extruder — but that’s dang close.
I wanted to give everyone a high level view of what can go wrong building a reprap. This post is, “Here, there be dragons” post.
1. The reprap is too long a physical build. It took maybe 80 hours in total. I spent that time drilling holes in the plastic( technically, I built a repstrap, not a reprap, as I used cast parts ). I spent that time soldering things to circuit boards( badly ). I spent that time cutting pieces of steel to the right lengths. I spent that time aligning things with a jig. I spent that time screwing large numbers of tine screws and nuts together. Most of my time errors came from buying — I bought things at Lowes that I should not have — like washers that were too long and seized my 608 bearings. I got out of that error by cutting some twist ties, and using them as spacing washers. The lesson I learned from all this building time is to buy a hardware kit — it’s just so much faster.
2. I blew my arduino. I still don’t know why. All voltages on all pins are correct. Yet, there lies my mega, in a useless pile. I got another one, replaced my RAMPS kit with a newer version, and viola, I have working electronics. The lesson I learned is that it’s better to buy your electronics mostly pre-assembled and pre-tested.
3. The software doesn’t work. Some firmwares don’t work with my endstops. Others don’t have correct speed control for 1/16 micro-stepping. If you get the firmware working, the various printing packages don’t work correctly. Some just crash, taking their host with them. Some don’t read a single value correctly. There are tons of bugs that you must have the source code, and a willingness to understand other people’s code, just to get it to work.
4. My hot end failed. The Pressure of the filament going through the hot end was too much for one of my screws to bear, and the hot end literally flew out of the extruder, snapping one of the wires attached to the heat core in my makergear superpack hot end. The superpack hot end has a problem — it can’t really be taken back apart well, once it’s been used for a while. The nozzle can — which is very nice — but the heat core and the PEEK mount bind to the threads too tightly, and don’t come back apart easily. I had to use a vise to get those out, and though the threads seemed undamaged on the screw, a new core wouldn’t go on, and the insulator would only screw in part of the way, even under force. I tried to repair the threads with a tap and die set — and after that, was able to get the heat-core and insulator back on, but badly. The insulator simply didn’t seal far enough down, and when I used the “refurbished” hot end, the PLA back-washed into the insulator, and bound to it, essentially ruining the hot end completely. I had to pruchase a replacement hot end, since I can’t source the M6 hollow rod or the insulator anyplace else ( though I can source a new heat core cheaply/easily, as well as a new nozzle. ) I don’t really know how I would have avoided this problem — it seems that the insulator plastic is just too soft, and melts a bit under pressure and heat. I think the insulator should have a steel sleeve, the hollow tube should be steel rather than brass, and the heat core should be steel as well. This would make the threads stronger and provide same-metal materials for all mechanical contacts, keeping the system thermally stable, which should allow for disassembly/reassembly after use.
5. The software is sub-optimized. After mucking around with code, I was able to get the software to work. I don’t know python, so I wrote a G-code post-processing script that did some mathematical corrections to the g-code file to compensate for my prusa’s variances. Afterwards, I was able to print for a little while. I printed about 16 layers before my hot-end failed. One thing I noticed is that the G-Code generated by the stack assumes pretty coarse print parameters. I have a very nice micro-stepping system on linear bearings — my machine can run silent and quickly. It can also run at far higher precision than the G-Code assumes — creating prints with lots of binding lines where part of the plastic overflows with the previous deposition. This is typical of rep-rap prints — and I believe that a software re-write is possible to produce higher resolution parts.
The economist in me see these problems as the problem of horizontal integration. It’s the “Windows Mobile” model of building a product — lots of partners, each solving their own cases and interests, loosely coupled to an end product. As I work on the Prusa, I understand Makerbot’s success. You pay more, but you get a working printer and software stack in one shot — because it’s all been mass customized to their product offering, tested, and validated with feedback.