Thursday, August 30, 2007

 

Pimpin' Richard

After a few weeks of summer, busy job and a period of shops being closed (for some strange reason the construction branch and the entire supply chain related to it closes shop during the same 4 weeks) I have finally come to upgrading Richard to the next level.

The source of most problems I had were related to inaccuracies in the components (and probably also tools), so I figured that some upgrading was in place. First, I used a small XY table to enable more accurate positioning of holes, and second was moving to aluminum for the corner brackets.

The right shows the first new corner block - both very shiny, sturdier and more accurate then the hardwood blocks. Some M5 taps for the screws to hold the rods; design is the same as the original Darwin, except the blocks are 40x40 mm. It is a bit more effort, but I think it will pay off. And if not, it definitely looks cool ;-)

Sunday, August 26, 2007

 

Some notes about gearmotor/rotary encoder ensembles

I've put in pretty much every spare hour that I've had for the past two weeks trying to get a grip on the issues involved in running gearmotor/rotary encoder ensembles at high accuracy. For a long time now I've been able to accurately extrude relatively long, straight lines of polymer at good accuracy and repeatability. Currently, with the gearmotors that I am using I can get a top diagonal extrusion speed in the xy working plane of about 2.4 mm/sec.

What I haven't been able to do well is to print a sequence of very short line segments (~0.1 mm) rendered as a series of print instructions on Tommelise. For this reason, I can make things like polymer pumps, but am totally pants at making features like holes in said pumps.

(Read the full story)

Saturday, August 25, 2007

 

Mime gets a t-shirt

Arthur Rackham prints are few and far between. The drawing that he did that is the theme for the Tommelise project was the dwarf Mime at his forge, was drawn by Arthur Rackham in 1911 as an illustration for Wagner's Siegfried.

Mid-week I finally ran down a print in high enough resolution to justify printing a t-shirt.

(Read the full story)

Saturday, August 18, 2007

 

Modeling firmware behaviour

I've been running single axis calibrations for about a week now and have a fair idea of the dynamics of the gearmotor/rotary encoder combination now. My old approach of interrupts on and off only when I was trying to print a line of polymer was definitely a bad idea. If you leave them on all of the time you get a repeatable absolute position without any problems as nophead implied in some of his forum postings. For me, that has been a wonderful step forward.

What hasn't been fun, however, is trying to get the extruder head to follow a path properly from one point to another. Now that I have an absolute positioning readout it becomes obvious that my firmware routine for running the x and y axes is just a little short of nice.

Armed with all of my new understanding, on Thursday I started trying to write a new xy movement routine. I have an IDE to check my code against which lets me trap a lot of problems. I even have an API that will let me build forcing functions and quite complicated models of device behaviour to test my firmware.

The IDE, however, simulates every clock cycle of the CPU, which means that if you have an instruction like...

WaitMS 5000

...in your firmware code, you can cook and eat supper without having to seriously worry about missing anything when you are simulating an 18F4610 running at 20 MHz. The net result of all that is that while the IDE is useful for preliminary debugging of firmware it is of limited use for detailed testing of complex firmware.

What I had been doing was to write my firmware code and use the serial link to output data about system operation on the fly. The problem with this is that the time that it takes the 18F4610 to transmit serial data about system performance is large with respect to the operations that I am attempting to do. That makes error trapping of real-time operations operating in the millisecond range more than a bit difficult.

(Read the whole story)

Tuesday, August 14, 2007

 

Heater Barrel Assembly

Now that we have finished heater barrels, and finished PTFE barrels available, I'm able to fully document the process of creating the heater assembly. In this build process, we build the assembly with plastic to protect the PTFE barrier and nozzle, wrap the heater wire, apply the JBWeld, and then let it sit. After it dries, we remove the nozzle and PTFE barrier, remove the protective plastic, and then assemble it back together again.

Voila, a finished heater barrel assembly. More pics on Flickr.

 

Small Hack For Opto Endstops

Here are some build pictures I took while constructing an Opto Endstop board. I added one hack to it: I soldered the wires directly to the board. I used a chunk of old Ethernet cord. I used the green pair to carry the signal (white to Gnd, green to Signal) and the solid blue wire to carry +5v.

To finish it, I'll attach a connector to the end of the wires, and then I can simply plug the opto endstop into the board it is supposed to be connected to. For some odd reason, I really dislike making cables. Probably because I dislike doing extra work (and I'm not very good at making them.) Anyway, here are some build pictures from the whole process.

Monday, August 13, 2007

 

A matter of timing

I took a few hours and began to explore the firmware programming issues associated with shaft encoding gearmotors.

My first exercise was to find out exactly how much time the various parts of the existing firmware was taking. I'm running an 18F4610 at 20 MHz. Here are a few of the things that I've discovered.

Solving a linear equation - ~100 uSec
Solving a quadratic equation - ~150 uSec
Reading a byte of data at 9600 bpi - ~1 mSec

(Read the whole story)

Sunday, August 12, 2007

 

Changing directions (again)

This is a very painful blog entry to write. I feel like such a weathercock doing this, but here goes.

I've collected everything that I need to build Tommelise 2.0. I have been getting on with the design and proof of concept exercises during odd moments for the last week. I've been torn between building Tommelise 2.0 by hand or using Tommelise 1.0 to make big parts of the new Tommelise.

It all came to a head on Friday after I did a major deliverables hand-in for my consulting firm. Some weeks ago I'd designed a new polymer pump that would have a lot more torque and in general be a lot easier to run and maintain than the one I'm using now, which is a lash up of the old Mk II that Adrian designed more than a year ago.

(Read the whole story)

Saturday, August 11, 2007

 

Looking forward

I've been trying to get one of my sample pic chips (18f2550) working using usb. Finally got it recognized and the driver installed. Amazing how many ways there are to wire so few pins. But in the course of my investigations I found some things that may be relevant to where the development team has stated where they want to go. There is a nice little USB to i2c project with linux/windows/mac support at

http://www.harbaum.org/till/i2c_tiny_usb/index.shtml

Seems a simple enough project to provide i2c directly to our devices. There are lots of i2c software out there that would have to be incorporated into whatever device is used for the controller. The other piece is jUSB which would provide the interface between Java and our i2c bus. There are linux and windows components for that referenced at

http://jusb.sourceforge.net/

Of course we could always just talk to our devices directly via usb, but having the i2c access would maybe be the more direct way of doing it. The controllers could still be usb programmable of course to facilitate updating. I like the booting via usb concept which eliminates needing a programmer except to get initial code into chips. But with something like rrrf and a base of reprappers that might eliminate the need for programmers except for those that want to do it themselves. I actually have to do it, but since I have a jdm programmer that works, I just added icsp to it and put connector on it and test circuit. Hopefully I can get to writing some actual software and download it now that I have the circuit recognized :)

Give me your thoughts, cheers and jeers. Happy reprapping :)

Labels: , ,


Sunday, August 05, 2007

 

Virtual Circuitboards

While there haven't been any substantive changes to Tommelise's controller board for 5 months and there appears to be only one or two minor ones needed to make Tommelise 2.0, I'm still building controller boards on stripboard, something pre-Darwin RepRappers were doing a year ago. Even though I have pretty stable board designs, I still don't feel comfortable designing special purpose boards for Tommelise. I just don't want to become a prisoner to set designs and inventories.



I use, essentially, a full-line (stripboard) Eurocard and a full-line half-Eurocard. They're made in Ft. Worth and can be had from...

http://www.the-hdb-store.com/servlet/the-VELLEMAN/Categories

...for $6.90 and $3.50 respectively.

(Read the full story)

 

Trying out gearmotors

I did some experimentation with my gearmotor options for Tommelise 2.0.

For now the tiny GM-17's are out of the running. I tested them to see if they were able to turn the 1/4-20 inch threaded rod and got them to operate at as little as 1.5v.

As a rule of thumb I've discovered that for running the positioning system you can generally expect that you will achieve about half of the listed free-running rpm for your motor. Free-running rpm for the GM-17 is about 341, which gives me a translation speed of about 6 mm/sec under no load. Practically I get about half of that, which means that I can do a 45 degree diagonal with x and y axes run by GM-17's of about 4 mm/sec. Those I will have to run at 6 volts.

Running at 6 volts will mean that I have to use an adjustable voltage regulator like the LM317. I bought one of those. It's rated at about 1.5+ amps, but dissipates 15 watts, which is just a shade wasteful.

(Read the full story)

Saturday, August 04, 2007

 

Buying the bits

I spent the morning at the hardware store, Radio Shack and Potters' electronics. I'm beginning to think that the parts for Tommelise 2.0 can be kept under $150 without a lot of strain.


The biggest, nastiest expense are the magnetic shaft encoder chips which, thanks to the fact that they're made in Europe, are up to $12-14/chip. I'm told that Austria Microsystems has a cut down shaft encoder chip that I can get for under $3, but I'll believe that when I have some in my hand. I'm going to press them on that next week.

The Euroboard stripboards have gone up in price as well, something like 50% to $15 since the last one I bought in April. I'm thinking that that is more Potters' than a real bit of inflation, though.

My son returns on the 15th and asked if I'd have a prototype running by then. I'm going to see if I can manage that. My client's web services server went south last night so I can't do consulting work till Monday, so I'm going to have a good run at it between now and then. :-D

Friday, August 03, 2007

 

Mocking up the x-axis

I'm waiting on a project manager to repair some errors in a data file he sent over a few days ago this morning, so I used the time to mock up the x-axis stage of the proposed Tommelise 2.0 to see what design issues I could uncover.


While design charettes using a 3D modeling system like Art of Illusion are very useful, there is always additional information to be acquired by assembling a full-sized mockup of the thing you are designing.

The big issue that I am trying to confront at this moment is what kind of linear motion system to employ.

(Read the full story)

Wednesday, August 01, 2007

 

Tommelise 2.0

I haven't been posting much in the past few months. Sadly for my 3D printer work and happily for my overdraft, I'm been doing consulting work in computational linguistics pretty much around the clock. With a little luck I might even manage to get my overdraft under control if this goes on as long as my client says it will.

Yesterday, however, I had a major turn-in of contract deliverables and was able to get a hot meal, a bath and six hours of glorious uninterrupted sleep. My client is busily trying to put what I gave them on their server farm, so I haven't heard from them today and I don't expect to hear too much for several days. This lets me work on the next deliverables at a more civilised pace.

As you recall, some days ago I posted some tentative ideas for a new extruder design. This evening I decided that it was time to start thinking about Tommelise 2.0. Mind, Tommelise 1.0 works fairly well. I've been able to print usuable parts on it with difficulty for about a month now and there is still plenty to be learned from working with it.


In fact, I've already learned a considerable amount from it. Mostly, I've learned that it is a complete lash-up, not that I didn't know that already. While I love the old thing, esthetically it deeply offends the old architect in me. There are so many things about it that could be improved.

(Read the full story)

This page is powered by Blogger. Isn't yours?

Subscribe to
Posts [Atom]