Prusa i3 problems, a story of madness

So in the last month or so, I have been experiencing feed problems on my RepRap Prusa I3. At first tops didn’t fill in correctly, then I started to notice wall issues and eventually large prints would not complete and every second print had to be restarted because the filament just chewed up.

It became incredibly frustrating to print anything. I assumed (incorrectly) that the problem lay with the new version of Mattercontrol (the software I used to use for slicing). Initially I thought it was the built in slicer, so I switched to their implementation of Cura. This helped initially in terms of finish, especially when I moved from 0.4mm layers to 0.1mm layers. But it wasn’t a week again until even that didn’t work nicely anymore. Switching to Cura itself did nothing to resolve the issue. Madness set in and I kept playing with settings, but with no improvement.

I did some research, and did find that their may be an issue with the hobbed bolt that the Gregs Wade extruder uses on this machine. Honestly though, I ignored that. I mean, come on, its been working all this time, why now all of a sudden. I kept playing with feed settings, new filament, but yes, eventually it sank in. I pulled the bolt form the extruder and found that inside the carved channel some of the ridges had bent because they were so thin.

My only conclusion is that over time they wear down. Strange considering you’re feeding plastic and its supposed to be a hardened M8 bolt. So I searched for a quick resolution. YouTube abounds with quick fixes using a tap mounted to a drill. Most looked kind fo dodgey, but hey it was worth a bash. I ran out to the workshop with gay abandon, fumbled a rig together in the vise and churned a new hobbing into the place where the current one was. The idea is to just deepen the “teeth” and so get a better feed.

YA RIGHT!!! Honestly this MIGHT work if you use the same tap that was used when it was first made, and its not going to work if you don’t have the right tap on hand. I didn’t. Not realising this, I brought the bolt back to its old home and again failed to print successfully. Bummer!! So I looked online for a replacement bolt. Another one would cost me 5X what a normal bolt would in the hardware store. Ok, so maybe its done properly and the issues would go away. But shipping is double the price of the bolt. Bugger that, I wont pay R150 for a R10 bolt unless I REALLY need to.

And so, the plan was hatched. Missery wasn’t going ot be my friend and my life wouldn’t be empty of 3D printing sweetness. After much thought on how to accomplish this, and many hours of real life and work, the plan was boldly implemented one Sunday morning.So looking at the bolt, I could see that many, many washers were used on either side to space the gear in such a way that the current hobbing was inline with the hotend hole.

A quick note before continuing, I wish I had taken pictures of all of this. The thought only occurred ot me after I had completed the rig to help me, and so those are the only images you will get here. Sorry. It is what it is.

Looking at those spacers, I thought I could just move those around and get free unhobbed space where I could start a fresh on the same bolt. I took a few hours to cut and massage some aluminium channel and plate. The channel was drilled to accommodate the bolt in a similar fashion as it works on the extruder, exposing the area I wanted to hob. This was then mounted to some wood, and I built a mount for my cheap Chinese knock off  rotary tool. This took some fettling but I eventually got something that I could hinge on the side of the wood. And so with bolt in the jig, and the rotary tool mounted with a cutting disc, I forged into new territory and cut line after line of tooth into the fresh unhobbed steel.

Channel piece for bolt
Mount for rotary tool
Tool mounted and ready for work
Good and bad parts

5 minutes later I was admiring the job and wondering if it was going to work. In it went and I set a few small prints off. The difference was remarkable. From poorly filled and walled prints, to nice even and closed prints with no feed problems in a Sunday morning. Needless to say I was very pleasantly surprised.

So what are my thoughts at this point? If the hobbed bolt is so prone to failure, then surely there must be a better feed design. Surely even doubling up on the bolt count should help the situation. I will be thinking long an hard on an alternative. Another thought was that I wish I had taken other possible causes into account early on. This would have brought me to the solution quicker. At least I now have a way of creating as many new bolts as I want without the cost. I am sure the jig will come in handy again.

A slight footnote here. While I was working on the bolt in the extruder, I noticed that the x-carriage had broken where one of the bearings resides, with that now sort of floating free. So while I have improved print-fu and the bearing is hacked into place, a new x-carriage is being printed.

Hope you enjoyed my tale of woe. Caio.

Model train lighting Part 3

Greetings again.

In the last post I discussed the circuit used in the model train lighting. Today well look at board layout.


Above is a completed board with the CR2477 battery, MSP430, resistors and tilt sensor wired in. This all went into a nice neat little box I designed for this project which can be found here. I used a socket for the mcu because I would like the ability to reprogram, reuse or replace it at will. Both the RESET and LED resistors were wire under the socket to save space on the board, While I didn’t need to do that for this board layout, I might want to modify the board at some point in the future.

I put the tilt sensor on a board of its own, with solid wires running into the main board. This gave me  the required horizontal alignment, but also allowed me to fine tune the angle by being able to twist the whole board ever so slightly. The original design called for two tilt sensors, but I found that I didn’t need the second one. I removed it in the code, but you could just as well keep it and bridge it or attach it to some other sort of normally closed sensor. I wired the LEDs up in parallel, an to be honest was probably a bit too arty when I did it.

I don’t have an image, but these were then glued into the roof of the unit with some tinfoil backing to provide some reflective. The wires were glued off to one side and fed through two small holes drilled in the base of the coach. They were then cut to length and glued into position and then soldered onto the board. If you look at the first image, you an just see the white wire glue into the base behind a very small wall. This way none of the wiring can be seen.

Below is a time lapse video of the test. To be honest it works exactly as I wanted it to.


Now to the lessons learnt.

  1. I need to add a switch between battery and board for longer storage times when I wont be operating the units.
  2. The units I have built so far seem to turn off the lights at slightly different times, even though the code is EXACTLY the same. I can only assume that its dependent on things like temperature, voltage, and differences between individual packages. To fix this, the design might need to include a more accurate oscillator.

Well that’s it, thanks for checking this set of posts out.