38 replies to this topic

[Corvus Corax]

I got myself one of these:

http://www.ic-board....-MX353-OEM.html


Its a system on a chip 532 MHz AMR11 with 256MB DDR2 RAM - running linux - the board is roghly the same size as CC and weights only 7 gram. Perfect any powerful on-board number high level computation tasks. (It could run the GCS if it wanted)

It interfaces with 4 50-pin Hirose connectors which unfortunately do not interface with anything we have directly, so it needs a breakout board, IC-board offers a variety of dev-boards which are nice for development, but too big and way too heavy for anything airborn. Also needs power supply and stuff in a way compatible with OP/CC

So I designed and built my own - interfaces with CC/OP-PRO via USB. (The mx353 has an inbuilt host controller)  and also features some serial uart's for its own IO, and an on-board micro-SD card for mass storage:

Schematics: http://cybertrench.c...b/mainboard.sch

I used as little SMD components as possible. The big through-components have the advantage of spanning significant board area which helps routing more on the front layer:
http://cybertrench.c...b/mainboard.brd


The guys at http://www.shackspace.de (a hacker space in Stuttgart) have an ink-jet printer capable of printing acid resistant ink directly on copper. I used that to print and edge this little baby

Soldering those 200 tiny pins was a PITA:

But it was worth it:


Since I had almost no things on the back, I skipped the back layer and used bridges/wires instead


And it works!!!!

[Kenn Sebesta]

1) Bravo!
2) What are the relative advantages/disadvantages of this vs. a Gumstix?

[Corvus Corax]

Overall weight of the board - including CPU daughterboard and USB cable is 35.7 g
Board dimensions are 65x65x20 mm

[Corvus Corax]

I think its roughly comparable. At the time I started it had more CPU power than the verdex gumstix available at that time.
Once OP has its own gumstix connector I would go for the gumstix, but the old beta board doesn't.

Both gumstix and ic-board provide a range of cpu boards, which do overlap in capabilities I think. I'd have to compare data sheets and available connectors in detail to tell you the exact con and pro's of each.

[dankers]

Damn Corvus you is impressive!

[Reddog]

Do you regularly wear a ninja outfit? If not you should!

[Berkely]

Every time you pop up with something amazing . great work!

[dankers]

Reddog, on 12 November 2011 - 06:44 AM, said:

Do you regularly wear a ninja outfit? If not you should!

I can confirm he does. You know something big is happening when he is quiet for a while.

[Corvus Corax]

lol - are all crows ninjas then?

[dankers]

No, only the best ones.

[chris1seto]

If direct flight control is possible over that link, think of the possibilities for things like flying machine vision and very advanced navigation...

EDIT: Also, I wonder if it might be a good idea to make a similar PCB, only some more IO broken out, and have it done through some place like BatchPCB. Assembly would not be difficult with a reflow oven and a solder paste stencil.

Edited by chris1seto, 21 November 2011 - 11:49 PM.

[dankers]

So the Revo has an External SPI, it doesn;t share the peripheral with anything else so can be used in master and slave mode, this is for connecting to things like the Overo Range, BeagleBone and other small platform for people wanting to do much more advanced things. The Overo has a carrier that we have made already and it was prefered because it has the camera header input for SLAM etc.

[Corvus Corax]

@chrisseto
Sure, that would actually be the obvious way to go

I did all of it myself mostly for financial reasons (I'm pretty broke still) and to gain experience.
If you have a slightly bigger airplane / copter, that can carry both weight and size, you could also use one of the dev boards sold by the ic-board guys, like this:

http://www.ic-board....va-ADB1003.html

I didn't because it was too big for my plane, doesn't have an sdcard holder (though one could easily solder one onto it) and the two connectors on board - ethernet and USB end device ( are quite useless for the on-board purpose.

However they have the complete schematics for that board and most of their other dev boards for download on that page, which make excellent templates to create your own
see:

http://www.ic-board....ungsboards.html

The eagle lib for the hirose connectors interfacing with their CPU's is here:
http://www.ic-board....guage=en&coID=9

with both of that, its quite easy to do this. I hadn't worked with eagle before at all, and still got a working board on first try
(Though we did learn how to design and edge boards with through connections "by hand" in school, so I was not completely new to the process)

What dankers posts there is important though. With an Openpilot carrier for Overo already available, I might likely switch to one of those newer Gumstix once I got my hands on a revo

Advantage of the USB based solution on the other hand is, it can interface with CC as well. (But I think the Overo has an USB host as well, doesn't it?)

EDIT: here are some in flight logs I created during an OpenPilot (beta) stabilized flight :
http://wiki.openpilo...th on-board log

Edited by Corvus Corax, 22 November 2011 - 09:10 AM.

[Reddog]

Is it worth it to create a logging board about the size of the Revo with an SD card slot and other peripherals that might be of use? I would like the option to log information but I don't have the ninja skills of Covus to get this sort of thing done and I suspect a fair few other people might be the same.

[metRo_]

Off-topic: Give a try to Kicad

[peabody124]

Yay FOSS layout.  I did first ESC rev in gEDA.

[robert b]

i would rather prefer a more powerfull flightcontroller board.
mainly doing navigation. i am still confident that a floating point unit is not of real benefit.
most can be done with integer math.
on the memory side i would opt for more ram
having SRTM elevation data in mind.
and some sort of mass storage is also a must.

linux adds some burdon to it - plus a learning curve which is way to steep for most of us.
me going for c++ with lua

dynamic upload of code ...  

robert

Edited by robert b, 24 November 2011 - 07:33 AM.

[dankers]

robert b, on 24 November 2011 - 07:31 AM, said:

i am still confident that a floating point unit is not of real benefit.

Is this serious?

[robert b]

in front of me i have a cortex m4.
i will do some timing using seb madgwick's sensor fusion code.
i have seb's code flying nicely: http://www.youtube.c...u/1/67io9AkXjZM
standard diydrones hw - a stupid 8 bitter running at 16 mhz
ever looked at the synthetic code handling uint32_t and floats?
from the technical point of view i can nearly do everything correct.
and doing so - the result is nice.

did you do ever do an analysis what performance benefits you get from a floating point unit?
if not, are you serious?

[Kenn Sebesta]

dankers, on 24 November 2011 - 11:34 AM, said:

Is this serious?

From a perspective of raw performance, that's probably a fair claim. However, from the perspective of actually getting things done, quickly, right, the first time, without bugs, the FPU is an incredible boon.