Tuesday, June 19, 2012


The latest Digital Aggregates project to show up on my radar screen uses the ODROID-A4, an Android reference and development platform from Hardkernel for mobile devices based on the Samsung Exynos family of processors. The A4 uses an Exynos 4210 system-on-a-chip that features dual one gigahertz ARM Cortex-A9 cores, has one gigabyte of memory, and a microSD card slot for its persistent storage. As you can see in the photograph below, the A4 is a handheld battery-powered mobile device the size of a largish smartphone or a smallish tablet. (I believe it's form factor is the same as the Samsung Galaxy S Android smartphone.)

This project is code-named Conestoga. This project is so new, there's no web page for it yet.

(Update 2012-07-06: the Conestoga web page exists now and has a link to the tarball for what I'm working on. Warning: it's definitely a work in progress.)

Here's a photograph of the A4 on my development bench fixture. Clockwise from top: amplified speakers (because my ears are old), blue HDMI cable to an external display (because my eyes are old), the ODROID-A4 and an unconnected accessory I/O board just above it, the debug board with TTA20 debug cable, USB cable, and DB9 serial cable, and a powered USB hub.

ODROID-A4: Bench Fixture

I'm no stranger to Android, having previously worked with several different releases of it on a couple of BeagleBoards for my Contraption project in which I ran a GNU software stack along side the Android software stack on a common Linux kernel. I'm a big believer in working as high as you can on the abstraction ladder even on embedded systems, which makes frameworks like Android of interest in me in a broader context than just mobile devices like smartphones and tablets.

But I blew nearly half a day trying to get the USB-based Android Debug Bridge (adb) to recognize the A4 on my Mac Mini desktop system running OS X 10.6.8.

I found some useful background on the web on Stack Overflow, Intohand, and this discussion on Goggle Code. As usual, software development is a team effort, even when you've never met any of the other team members.

What finally worked for me was two-fold.

  1. I used a separate powered USB hub (visible in the photograph above) instead of connecting the A4 to the hub built into the Cinema Display on my Mac Mini (no clue why, but that's what worked for others too).
  2. I did not use a USB cable to the mini-USB port on the debug board, connecting only the USB-to-TTA20 debug cable and a DB9 serial cable. The TTA20 port on the debug board provides USB access for the Android Debug Bridge. The DB9 port provides console access with a root shell.

To be clear, here's a photograph of both of the A4 and the debug board on the left with all three cables connected, which did not work.

ODROID-A4: Console, USB, and TTA20

Here's a photograph of just the debug cable and the serial cable, which did work.

ODROID-A4: Console and TTA20

Here's a screen snapshot of how the device enumerates on OS X.

ODROID-A4 USB Enumeration on Mac OS X

And here's the results of asking the Android Debug Bridge to list all the visible devices.

$ adb devices
List of devices attached

Here's hoping this saves others some time.

Update (2012-06-03)

While adb on the Mac works just fine, I've had no consistent success getting fastboot, the interface that allows you to communicate with the U-Boot boot loader over USB, to work on the Mac. Web perusal has led me to believe this is a common issue with later versions of Mac OS X. (I'm running 10.6.8 a.k.a. "Snow Leopard".)

After blowing an entire day trying various strategies including enabling USB debugging on OS X and writing my own OS X codeless kernel extension, I finally gave up and attached the A4 to my Ubuntu server (which, for unrelated reasons, I had recently moved from the vast subterranean catacombs to my office on the second floor of the Palatial Overclock Estate), where within an hour or two everything worked just fine.

The weird part is that fastboot devices on the Mac actually worked once out of dozens of tries. What I saw logged by the USB stack in OS X smells like some kind of race condition.

For now I'm sticking with Ubuntu for platform work on the A4, and will return to the Mac (which now entails just a cable swap at the powered USB hub) when and if I do application development in Java on the A4 using the Eclipse Android plug-in.


Anonymous said...

You wouldn't happen to know when that Odroid-A4 was released would you?

Chip Overclock said...

It was first announced on the Hardkernel ODROID forum in March 2012 for availability late March or early April 2012.

I ordered mine early June and received it quite quickly mid-June 2012.