Perhaps, nobody else has pointed this out but..

How on earth would AR applications take off on the n900 if it doesn't have a digital compass built in? Assuming if you did get a digital compass on the n900 working then you have the problem with you have to assume the developers of such AR applications also have an n900 + digital compass. Or that the next Nokia device comes with a built in digital compass and that the application can be back-ported.

The n800 did not have a GPS built in so there's some precedent for peripherals being used which are not built into the device. Granted the cost is !!!!! way more than I want. At this point I am skeptical though that usb won't work with the right SW and if a cell phone battery can be used instead of the sparkfun stuff that would also help a lot.

Additionally, the cool things we build for ourselves can be just that, for ourselves.

I don't know if there is continuing interest in this topic, but I can contribute a few bits of information and opinions.

NXP has a decent application note on building an electronic compass:
http://www.nxp.com/acrobat_download/...22_COMPASS.pdf

Note that to do tilt compensation you'll either have to combine magnetic field data from the magnetic sensors with tilt information from the N900's accelerometers or just put a 3-axis accelerometer on the external board. Assuming that there is access to raw 3-axis tilt info, the former is obviously preferable cost-wise (an ADXL320 runs $15 in single quantities).

The Honeywell HMC5843 seems to be the best candidate for a three-axis magnetometer, since it handles temperature compensation and offset correction by itself. There exist cheaper magnetoresistive sensors, e.g., the KMZ51 and KMZ52 mentioned in the above app note, but then you also need a preamp, ADC, temperature sensor, et cetera. Beyond that, to get three axes out of the KMZ5x parts, we need to mount one of them perpendicular to the other!

If it's possible to connect the HMC5843 directly to an I2C bus on the N900, we're done with the hardware. Otherwise, it's USB or Bluetooth depending on what resources are available. There's a lot of rumor right now, so we'll probably have to defer this discussion until the phones are actually in our hands. Clearly, not having to add a microcontroller and its consequent cost and power draw to the bill would be preferable.

The software to actually produce a compass reading will take the magnetic and gravitational data and do some math to turn it into a heading. From there, you can even use GPS data to easily convert from magnetic north to true north if desired.

I'm certainly still interested, though I would prefer to pay more and just get a pre-built Bluetooth unit (usb not working and all).

Anyway do all keep up your investigations, should make for some cool apps if these things can be made

Prebuilt? My google-fu has failed me.

Mine too. And the google-fu* is strong in me.

*google-fu means "excellence of searching"

Well yeah, I'd just like to buy something rather than make it. Not that said device actually exists....

Maybe those who make the hack will share schematics MAKE-style, or start a business from it, or Craves1 passes by...

there are still a few days to post proposals to the Nokia Push N900
they do not have to be finished models, just the idea to build one and a bit of a plan.

http://blogs.nokia.com/pushn900/

have a go, draw up a components list, find someone with some circuitry skill to help with layouts and tell them how its going to work and what you want to do with it.