Yea, that would also be a valid option, though if you want the easy soldering option, you would also have to get the big style microchip, and you still have to find a way to hold it all together, and it's getting a lot bigger. On the other hand I'm probably overestimating my own soldering skills too, the smallest thing I've ever soldered is a ps2 chip years ago.
As for making a separate bluetooth one, that sounds even harder to me, and more hassle getting batteries and chargers and stuff. Also it means that you would have to hold your phone and bluetooth box in alignment or it would be completely meaningless.
Since I've been assured that it will be possible to force the usb into host mode, that would be my choice of interface/attachment.

Oh, right - I didn't realize that the other chips you mentioned are also surface mount. Well, I did see this link the other day about DIY surface mount soldering using a toaster oven, but I'm not sure I'm ready to pop electronics into kitchen appliances yet (my wife might also have something to say about that...)

If/when I get around to trying this, I'll probably use the bigger chips to ensure greater chances of me actually being able to build the thing. I imagine that as long as the device is small enough to be pocketable, so it can be taken out when needed, it would still be usable - though of course, smaller would be better.

EDIT - This seems to be a feasible way to solder the surface mount chips that would be involved: http://www.sparkfun.com/commerce/tut...torials_id=100

Bad news. Looks like the n900, unlike its predecessors, does not have hardware support for USB host. Scratch the USB plans, looks like the DIY compass will have to use IR or Bluetooth to communicate with the n900.

That is only for external use cases. But the OMAP inside N900 DOES support USB OTG. So, this may still work.

I really think you're going to want an external bluetooth unit. You really don't want to be hacking your brand new (expensive) phone by carving away the inside of the plastic back, or something equally risky.

Especially since, as Bratag pointed out, AR apps just aren't there yet in terms of maturity or usefulness.

It will be fun to build, and fun to goof around with, but maybe a BT unit stuck to your N900 with some velcro is your best bet right now...

I also agree with using bluetooth.

I use a cover for my n800: http://www.amazon.com/gp/product/B00..._ya_oh_product

and I think it's reasonable to stick this device into a cover which would give you a known orientation.

My gut says you could do this for under a hundred bucks, if not half that, if you spin your own board. People charge you an arm and a leg for breakout boards.
I don't think spinning your own is as bad as it sounds, and I'm willing to buy an extra toaster oven for surface mount.

Though prototyping with the breakout boards makes sense to do first.

Biggest concern on my part is power which may make the widget infeasible to use.

Bluetooth sounds good, but what chips/components would you use to connect e.g. the HMC5843 compass chip (which has an I2C interface) with a Bluetooth module? SparkFun has a BlueSMiRF Gold Bluetooth module that is pretty small, but it seems to take RS232 as input, not the I2C that the HMC5843 has. How would you interface the two? I found an I2C-to-RS232 module, which could then connect the I2C device to the BlueSMiRF, but the adapter module is as huge as a brick. That company says they are considering making an I2C-to-RS232 chip, but they haven't yet. I also found a supposed I2C-to-Bluetooth module, but I couldn't find it for sale anywhere.

There also seems to be an RS232 compass module which could perhaps directly be hooked up to the BlueSMiRF. Minimum order quantity is 50, though.

Then there's also the issue of power - hooking up some batteries to this will increase the size even further...

I've got another really, really, REALLY low-tech idea about how to do this, but I'll only consider that if all else fails. Idea: use the low-res, user-facing camera to do image processing on an analog compass and determine its orientation This then becomes more a problem of mechanical rather than electrical engineering...

EDIT: Another potential compass module with RS232 interface that might be connectable to BlueSMiRF: TD-CMP03. Kind of large at 40x41mm. It has tilt compensation; I wonder if it can be held vertically upright against the N900 body, or if it has to be held horizontally. No word on its power consumption. Again, comment from electronics gurus would be welcomed... is it really as simple as connecting the RX/TX lines from the compass to a BlueSMiRF, and connecting the power?

Is this chip a requirement?

Yes, I mentioned that.

As to the actual hardware, again I don't know exactly without going through all the options, but magnetometer tied directly to a bluetooth modem is probably not the best one from a size / cost perspective. You are also looking at modules which are expensive and bulky.

I am suggesting a custom board if there was going to be use beyond R&D.

Anyways, I'm going to wait for the n900 to be in my hands just in case this is some kind of bizarre easter egg (wasn't the radio tuner an easter egg in the n800?) In mid october maybe I take more time to look at it if nobody beats me to it..

Another question I have is if there's an existing library to take raw counts from the digital compass and turn it into a heading. I know there are differences also between true north and magnetic north and I don't know the magnitude of those differences.

You can call it like that. The Bluetooth module included (a quite common one, by Broadcom) also had FM tuner. Free, it just was by default disabled due to lack of driver IIRC. The likelihood of such happening with digital compass... well, I wouldn't hold my breath.

Sigh. I'm way too obsessive.

Let me know if I screwed anything up. Attached is costed bom + misc for board / prototype in OO and csv format.