Hi Neil,

I was really excited to see geppetto (I missed seeing it for a month, but that's what I get for living under a rock I guess.) :-)

In 10 minutes I was able to design a board that was probably 90% of what I needed which is pretty cool.  For what it's worth, I do think this sort of thing is going to be the 'next big thing' for people who want to build and make things themselves.  Let the designer still make all the higher level design choices, but give all the low level grunt work over to a program to figure out.  And the really nice thing is if later I decide I want to add/remove/change a component it might be just a minute or two of rework versus a week of evenings in the more traditional tools (with the opportunity to make tons of mistakes for any manual change you make.)

(I realize I'm just one data point in a huge range of needs out there, but thanks for responding.)

For our specific project I prototyped the system using a Pinto-Th and then jumpered over to the other non-gumstix hardware.  What we ended up doing after we proved the functionality of the prototype was to design a relatively simple carrier board that the pinto-th (the 60-pin header) plugged into.  We soldered the male 0.1" headers to the bottom of the pinto-th and the female connectors to the top of our carrier board so the pinto-th just dropped right in.  We did something similar for the other major components and then our carrier board mostly just made all the connections at the board level so all the rats nest of wiring went away.

For this sort of project we would need to be able to precisely position rows of 0.1" header holes on the board and then specify which overo signal line they connected to.  This could be similar to how you set up LED's in geppetto -- drop in an LED and then pick from one of about 40 different signal lines it could be driven from.  If the dimension of the 0.1" through hole was 0.1" box, then you could line up as many of these as you need.  I don't know how hard that would make your backend software work though if there were tons of pins and lines that needed to be routed around.

The only other major thing that we would need for our project would be the ability to connect two 0.1" header holes to each other (not to an overo signal line.)  This would allow non-gumstix component #1 to talk directly to non-gumstix component #2 (assuming they both matched up with 0.1" header interface model which was the case for us.)  I guess I would also include bonus points if you could throw in TTL level translation between any of the 0.1" hole connections to external devices.

Grouped header pins: this is an idea I wondered about and looked for.  I think for many people, being able to drop in a 40-pin header (or 60-pin header) or at least 1 row of the header somewhere on the board could be highly useful.

Another specific thing from our project.  The main things that we need direct connections to are things like the uart's and the 4 high resolution pwm's.  I could see future interest in being able to hook up the i2c or spi lines as well.  If you could pull out uart0 and uart2 lines individually to grouped connectors I could see that being useful.

I think that if the only option is to pull out groups of pins or perhaps the whole 40 or 60-pin header, then for us it would probably make sense to go back to using the Pinto-TH and just jumpering around from part to part rather than redesigning our own custom pinto-th variant.  In our case we already have a custom built solution that meets our needs, but it was done the old hard way -- and it would be awfully cool to be able to quickly design variants of board for different 3rd party components and different types of installations.

I understand the tradeoff here is to keep it simple so software guys like me can make their own designs without needing to hire out to expensive hardware guys that know how to use the traditional tools.  Yet if guys like me pepper you with all sorts of specific requests, you could easily be back to a complex tool that guys like me wouldn't be able to use anyway. :-)

Anyway, I do think this is a brilliant tool.  I think it tracks with an overall brilliant trend to make building complex things easier.  I really hope you get enough people using it to make it worth future development and expansion of the tool's capabilities.



On Mon, Apr 15, 2013 at 2:23 PM, Neil MacMunn <neil@gumstix.com> wrote:
Hi Curt,

See responses inline.

On Sat, Apr 13, 2013 at 8:59 AM, Curtis Olson <curtolson@flightgear.org> wrote:
For one of our projects, we had a simple board designed the old fashion way -- it was basically just a bunch of interconnects between a pinto-th board and other plug in components.  It would be pretty cool to redo the design in geppetto and be able to drop the pinto-th (which we only used to keep our own board design simpler.)  What seems to be missing for our project would be the ability to arrange rows of 0.1" pins in specific places (so other components could plug directly into the board) and then connect those 0.1" pins to arbitrary pins on the overo (or connect one 0.1" pin to another 0.1" pin.)  If we could do that I think we'd be pretty close to being able to do everything we need ... especially if the pins coming out of the overo had optional ttl level translation logic you could add.

This is an interesting idea that we are looking into. Electrical design within Geppetto is based on buses, rather than signals (though GPIO, for example, are one-signal buses), so there is currently no way to specify a pinout right now. Would grouped pins be an acceptable alternative?

This is a great idea if it lowers the entry barrier to designing custom embedded hardware!  $2k setup fee seems steep, but by the time you count up the hours to design a board with traditional tools and all that is involved there, it's probably a pretty reasonable deal for people who are actually serious about building real boards.

We are working on improving the community section so that users can better collaborate and split the setup fee. In the meantime, ad-hoc crowdfunding will work too -- the UAV community seem particularly adept at this.
Is geppetto something gumstix is developing themselves, or is it just an application of a really cool 3rd party app that I never knew existed?

Geppetto is our own service that we've developed for customers who have asked us for custom expansion boards. I'm happy you like it, we're really excited about what it can do and the long list of features in the pipeline... stay tuned!


Precog is a next-generation analytics platform capable of advanced
analytics on semi-structured data. The platform includes APIs for building
apps and a phenomenal toolset for data science. Developers can use
our toolset for easy data analysis & visualization. Get a free account!
gumstix-users mailing list

Curtis Olson:
http://www.atiak.com - http://aem.umn.edu/~uav/
http://www.flightgear.org - http://gallinazo.flightgear.org