Microcontroller expert around? - recommendation needed

SuperH would be a good start, maybe an SH2+DSP. Also, a possible fit lower on the cost curve but with lower power, would be an H8S (some of these have MACs), a Motorola M-CORE, or an Atmel or Phillips ARM processor. As far as how to communicate with a PDA, it depends on the PDA. SDIO seems to me to be congealing into a plurality, but developing a SDIO "peripheral" is a major commitment.

Sometimes, a 32-bit timer with input capture winds up being a nice thing to have with GPS designs. These are rare, but they do exist -- look for parts targeted toward automotive markets.

As far as cheap tools: keep an eye out for parts with a JTAG debugger, which

*usually* opens the door a bit for more variety of tools. In my experience, stick to the GNU C or C++ compiler. Using proprietary tools is not only expensive, but if you get into trouble (compiler or tool problems), then it costs real money to get out of it. GNU is not perfect, and it's IDE's arent as nice, but the user base is huge and the output code is generally very good. Microcross is a good place to start for graphical GNU tools. For an ARM processor, for instance, you can easily have a GUI IDE, Compiler, JTAG Emulator, and evaluation board for well under $2000.

If you're using the SH-X or H8S, then you might want to look at the Hitachi C/C++ Compiler. On a H8 platform, our analysis has shown that the Hitachi compiler produces 25% more compact code when compared to GNU. Also we found the Hitachi support to be excellent (and free for us). Although the GNU toolset is basically free, the maintenance contracts usually amount to cost more than the commercial C compilers (which kinda negates the motivation to use the GNU toolset). Of course there are the various support forums, but I found that their helpfulness is not consistent.

So if you don't need support & able to tolerate the inefficiencies of the generated code, then GNU is more than adequate (for Hitachi micros anyway). For Win32 development environments, I would recommend the one from Kpit.

Ken.

+====================================+ I hate junk email. Please direct any genuine email to: kenlee at hotpop.com

Perfect place.

Cheapish development cards for the Hitachi H8 family are available from Triangle Digital Services (TDS2020 card) and Microprocessor Engineering (Powerboard). Both systems come with a wealth of Forth tools and make good use of the on-chip resources. Their websites are linked from the bottom of my Forth page. I have used products from bith companies and their back-up service is excellent.

--
********************************************************************
Paul E. Bennett ....................
Forth based HIDECS Consultancy .....
Mob: +44 (0)7811-639972 .........NOW AVAILABLE:- HIDECS COURSE......
Tel: +44 (0)1235-811095 .... see http://www.feabhas.com for details.
Going Forth Safely ..... EBA. www.electric-boat-association.org.uk..
********************************************************************

You might consider the TI TMS320VC5470 or OMAP series. They combine an ARM7 or ARM9 MCU with a DSP on the same chip. The math is a snap for the DSP and then you have the ARM to do the rest. Has flash, ram, etc. Developer kits available in the $1500 range (I think) the IDE is more, but you may find it bundled with the developer kit.

-- Scott Nowell ExoTech R&D, Inc. Lafayette, CO 80026

which starts like:

"London, 1 April, 2003 : Renesas Technology Europe Limited was launched today marking the beginning of the joint venture between Hitachi, Ltd and Mitsubishi Electronics Corporation. The new company will focus on the design and manufacture of highly integrated semiconductor system solutions for the mobile, network, and automotive industries, digital home electronics and industrial markets. Key products will be: microcontrollers, including smart card ICs; flash memory technologies; and mixed signal devices. These are all areas where the joint company is already a leading supplier. ... "

Regards, Arie de Muynck

simpler

It smells like a poor man's cruise missile, done by someone who doesn't know all that much about Kalman filtering.

18 states: I'm guessing, but I'd say he probably has:

X Y Z position X Y Z velocity X Y Z acceleration Yaw Pitch Roll position Yaw Pitch Roll velocity Yaw Pitch Roll acceleration

80 Hz IMU updates, with Z-axis position/velocity/acceleration, make it a flying machine as opposed to a ground vehicle, surface boat, or submarine. At that, it is a flying machine with some pretty ugly dynamics, if he needs 80 Hz control information. GPS as opposed to a local beacon system make it a vehicle with a fairly long range and a relatively noncritical positioning requirement (because of GPS position errors).

Oh, and GPS also rules out a spacecraft. The velocities involved in space travel, even orbital travel, are too high for GPS navigation.

Put them all together, and "cruise missile" is what comes out.

If you're referring to the size of the matrices, it's because it's a navigation filter. Right now it looks like I'll be doing an error-state implementation of a GPS-aided INS (although details could change). That means anywhere from 12-?? states depending on how detailed I need to get to achieve good performance. Now some of the matrices may be sparsely populated, but they need to be this big simply because of the sheer number of states I have to keep track of (3 position states, 3 velocity states, 3 attitude states, gyro drifts, accelerometer biases, GPS clock error and pseudorange error, etc.).

Nothing that sinister, John. Just a cheap, commercially available MEMS IMU integrated with a GPS receiver to see what kind of low-cost, small-size, low-power performance I can get. Its best use would be as a component in a larger system. This is just masters' level thesis work...

Probably not XYZ acceleration, and almost certainly not attitude velocity or acceleration. The rest of the states would be system errors such as those I mentioned previously.

The IMU update rate is largely out of my control because it's a cheap commercial-off-the-shelf IMU. Of the three IMUs I hope to test with, only one is over $2000.

GPS because it's cheap, easy, and doesn't limit where the system can function.

I'm not expert enough to dispute you on the main point, but I do know that the shuttle gets excellent GPS performance because of the lack of atmospheric error (granted, the shuttle is in LEO).

totally

Thanks for that. I'm reassured ;).

Steve

simpler

No, I was referring to the use of matrices at all. I repeat: what is the objective? Alternatively, what is the nature of the noise that you seek to filter out?

(It's a slightly trick question. GPS has artefacts that can be caught with a smart filter, and others that can only be caught by longterm sampling of the same position i.e. without movement.)

Steve