Erik's Blog: Discussing open source innovation, personal fabrication and a ...

Two weeks ago I made print circuit board for a more advanced stepper motor driver, using the IMT-901. This is a chip that can drive power to the motor and control motion in a very precise way (micro stepping). It can in theory increase the precision of the machine from 0.125 mm to 0.0078 mm, or 7.8 micrometer. Of course the accuracy in practice is not in that range, but it will probably allow smoother motion and better detail of small features.

I use the following files:
http://reprap.erikdebruijn.nl/files/imt901.zip
It can be opened in Eagle (>= 4.2, I guess)

This guide explains the steps I took: http://fablab.marcboon.com/pcb/

A possible problem with the PCB might be that C4 (1000uF) is pretty close to R2, if R2 gets hot (not just warm, I mean hot), it can damage C4 in time (thanks for pointing this out, José!).

I like the IMT chip because it includes the logic an drives the power for the stepper, microstepping included. This board layout was nice because it can be made with single side copper (with just a few bridges needed).

Microstepping could improve precision and remove a lot of vibration in the frame (loosening screws and noise). I hope it will make a difference!

Here are some screenshots of milling the PCB:







This is a good site that explains making g-code to do milling (so you can do it on a RepRap):
http://www.brusselsprout.org/PCB-Routing/

Perhaps I'll be working on it some more this evening!

Update: fixed again. Update 2: fried again.

Remembering from the top of my head that I had some PCBs with the silkscreen reversed, with an incorrect measurement I concluded that it must have been the PWM board. The PWM board gives 12 Volts on the -, or so I thought... It actually always gives 12 V on the + (silkscreen and actual) and I thought that this was the minus. This means that I put 12 V on many inputs of the sanguino, rotational sensor's opto and even the thermocouple PCB! Nooo... the sanguino went really hot. After the damage was done, CPR didn't work, giving it shocks or reprogramming it with an AVR didn't work.... Yes, it was dead, forever.

I had a spare AtMega644p just in case the above happened. But programming it was harder than I thought.

My first attempts were various programs that would make an AVR programmer out of an arduino (which was the inspiration for the sanguino). This didn't work. The software on the arduino couldn't program this target. Apparently everyone used homemade Parallel programmers (which aren't that hard to make, I admit) or expensive AVR programmers (too expensive for me!).

After two iterations of making a Parallel programming cable, this one finally worked, but only by issuing a manual command on an old laptop with a LPT port (a.k.a. printer port).

Doing it directly via the Arduino program just caused the software to hang without much information on what went wrong. Using avrdude from the command line yielded more results. After doing it the utterly slow way (waiting 50 minutes for burning a < 1k bootloader to upload) that didn't work, there was a fast way that I should have tried:

path-to\arduino-0013-win(2)\arduino-0013\hardware\tools\avr\bin>avrdude.exe -c dapa -i 10 -p m644p -U flash:w:ATmegaBOOT_644P.hex -C "path-to\arduino-0013-win(2)\arduino-0013\hardware\tools\avr\etc\avrdude.conf"

avrdude.exe: AVR device initialized and ready to accept instructions

Reading | ################################################## | 100% 0.00s

avrdude.exe: Device signature = 0x1e960a
avrdude.exe: NOTE: FLASH memory has been specified, an erase cycle will be perfo
rmed
To disable this feature, specify the -D option.
avrdude.exe: erasing chip
avrdude.exe: reading input file "ATmegaBOOT_644P.hex"
avrdude.exe: input file ATmegaBOOT_644P.hex auto detected as Intel Hex
avrdude.exe: writing flash (65382 bytes):

Writing | ################################################## | 100% 42.70s

avrdude.exe: 65382 bytes of flash written
avrdude.exe: verifying flash memory against ATmegaBOOT_644P.hex:
avrdude.exe: load data flash data from input file ATmegaBOOT_644P.hex:
avrdude.exe: input file ATmegaBOOT_644P.hex auto detected as Intel Hex
avrdude.exe: input file ATmegaBOOT_644P.hex contains 65382 bytes
avrdude.exe: reading on-chip flash data:

Reading | ################################################## | 100% 41.22s

avrdude.exe: verifying ...
avrdude.exe: 65382 bytes of flash verified

avrdude.exe: safemode: Fuses OK

avrdude.exe done. Thank you.

This was with the Sanguino powered by the USB-to-TTL cable.

And then it was alive again... it's alive, it's alive!

Lesson learned: don't assume that a + output port is actually - (ground). It might actually be 12 V and fry your chips.

I read the Guardian article about "building an open source world", on which I have some comments about this.
In the article this remark is made:
"Open source hardware doesn't have the same power as software if only because the final product, as opposed to the designs, can't be replicated for no extra cost as software can."

I'm involved with the RepRap project (also in another Guardian article). I also built a 'machine that can make almost anything'. I'm now building 6 more machines in a group, and the project has already attracted a new group who want to start building.

While in essence the statement about non-free distribution of physical goods is true, we're working hard to mitigate that. We're working on making sure it can use garbage plastics (since we have an abundance of garbage) and it could in principle recycle its own products. The costs of the plastics that we can use are significantly lower than that of commercial 3D printers (about 20 times cheaper, just like the machine itself). There's also an opportunity for home-grown biodegradable plastics (plastics from starch). An interesting remark about this is:
"In a sense, hardware is becoming much more like software, up to the point where you actually fabricate an object," von Hippel says. "That's why you're starting to see open source techniques in hardware. Design is largely going to shift out from manufacturers to the communities."
Source: Eric von Hippel of MIT

Last year, I attended the Rapid Manufacturing and Mass customization conference. Frank T. Piller, considered the guru of Mass customization, said that manufacturing will start happening closer to the consumer, and that the demarcation between producer and consumer will become more blurred. Also, Terry Wohlers, consultant in Rapid Prototyping (also called 3D printing), used the example of an open source fabricator and highlighted the concept of neighborhood fabrication. Granted, he does have reservations for mass adoption.

In the near future where people have fabricators at or near their home (right now RepRap's diffusion is amazing) products would be distributed digitally and produced locally.

The network effects of the RepRap community and the installed base of RepRaps are compelling. I've documented some of the growth characteristics of the RepRap community on my blog here.

The community has also become more organized and shops have oriented themselves around the concept. Now, for the non-technical people, or those with little time, there is now a kit available for about 750 pounds with which you could build a RepRap (equivalent) in 2-3 days (my estimate).

And an interview with Eric von Hippel:

URL: von Hippel interview

They discuss fabricators reaching the home at some point in time. Von Hippel (in the video, here's another), whom I referred to in my earlier blog, is a did seminal work on the origin innovation.

Here's a Google TechTalk about Innovation. It's told in an amusing, motivating way. It's mostly about being restrictive, overprotective and being too prudent, you would risk ruining a technological revolution. Also, he brings forward the risk of business method patents.


Some remarks that struck me:

• Patent quality and use is very low (also brought forward by Chesbrough and others).


• Offensive use of patents (as opposed to defensive)


• A great empirical argument: "When you're extending copyright terms, don't extend the for dead people. It is unlikely you will spur innovation by doing so.". Can't argue against that!
  
• "Ninety-something percent of all western culture, produced in the last hundred years, is A is under copyrighted and B has no identifiable copyright owner which means that copyright is locking it up while providing absolutely no benefit."

Have you ever been in a foreign country in need of a translation? Or you stumble across a manual that's only in foreign languages... Going to google and typing over all the text tedious. You may not even have a computer at hand (e.g. when you're on the road) or it may not be connected to the internet. When you have a mobile phone with internet access, you could use go to google translate, but typing on a mobile phone is even more of a pain in the ...er... fingers.



For a final year course at the University of Tilburg, Seminar of IT, we have an assignment to create a mashup of various systems including Evernote. Evernote can recognize typed and handwritten text in images, via technology called 'OCR' or optical character recognition. It becomes much more than just a searchable image store, since you can take all your notes with you on an internet enabled phone, synchronize them with your PC, etc. There are many uses of Evernote. After exploring trivial uses of Evernote, we decided to make a more sophisticated application, called TranslateAnywhere, which integrates Evernote's OCR and various other services to create a solution to the scenario explained above.

Below is an overview of what happens. A snapshot is taken and sent to Tarpipe, which is a mashup tool. This tool posts it to Evernote for you and fetches the recognized text. Tarpipe than forwards that text to my scripts that call the Google's translate API for a translation. This translation is then sent via SMS. The result is received on the phone (within a minute under ideal circumstances).


Figure 1: Overview of the system interactions

It's not really a complex system (few interactions, and through clear APIs) but it is a pretty long chain involving many nodes (most are online services). Below is the more detailed picture, which is still a bit oversimplified, but hopefully clear:



Here's a presentation detailing our findings about Evernote (including comparisons):