RE: [Paparazzi-devel] Hello

[Edward Freeman]

>>The servo control is not really a problem. We have offloaded it (ppm
>>decoding too) to a second MCU (atmel mega8), mostly for security reasons
:)

I thought it would reduce chip count and would use very little recourses,
but that would be a way off anyhow.

>>I prefer to use SMDs for many reasons; one is that I build the pcbs myself
>>and it avoids the tedious drilling. Another reason is that we fly a small
>>aircraft (1.5kg and we hope to get under 1kg) and this method allows for
>>smaller and lighter boards.

I agree especially as I have access to good tool at work but also can see it
limiting the number of people that can build the board.

>>For the hard wired Kalman, I think it's a great idea. We have build the
>>autopilot IMU but have discovered that we don't have enough horsepower in
>>the onboard MCU to run the INS kalman filter.

It should be possible to implement a full INS this way cheaply.


>>We have a show on October first and have decided not to go further with
>>the IMU until that date. We will concentrate on our first method, wich
>>uses infrared thermopile to sense horizon (as in
>>http://www.ctie.monash.edu.au/hargrave/horizon_sensing_autopilot.pdf).
>>This method requires almost no computation.

I think the IMU is a better choice in the long term but good luck for the
show!

>>After the show, we intend to work again on the IMU and try to find a way
>>to run the INS filter onboard.  Your FPGA solution might save us at that
>>time.

The timing on that maybe a bit quick as getting the FPGA before Q1 2004.
Although all the tools and information are available to develop products
now.


I intend to use a bigger aircraft standard glow fuel powered size will
probably buy plans but haven’t decided yet. Have my hart set on a Second
World War D.H. Mosquito ;-)

I have only just started and will probably wait a while before I start
building for financial and space reasons. (I live in a small house and don’t
have room to build at the moment)
As for the Kalman filter project I am still trying to find out what's
needed.
I have got my tools sorted out. (Didn't’t have any at home until recently)
and learning to live without all the nice bells and buttons you get with
over priced EDA tools ;-)

>>How easy is it to experiment with FPGA ? Can they be easily reprogrammed ?
>>what performances can be expected from them? are there open source tools ?

They are easy to reprogram most are SRAM based so load from a PROM.
The PROM is programmed via JTAG or PROM programmer.

Performance depends on the FPGA and what your doing.
You are making hardware so it depends on how much area you want to devote to
a task ect. For a better answer in our case see my post to Autopilot dated
15/08/2003 20:25.

 There are two main companies that make them Altera and Xilinx. I used an
Altera device for my Degree dissertation but have used Xilinx ever since at
work. I go to both of there sales/training days each year and in my opinion
xilinx are now better for most uses. (I’m not associated either company them
just a user at work)
As you seem happier with SMD then you could use a bigger FPGA the XC3S400
PQ208 is the biggest chip they do in SMD and that has 141 user IO 16
multipliers and a lot of programmable logic. But I think that would be a bit
over the top. Could embed a microprocessor and have a “system on a chip”
solution. (It’s all the rage in the industry)

As for tools I have been told there are some open source tools but have
never tried them and have heard bad things.

Xilinx give away tools for their smaller chips these are cut down versions
of very expensive pro tools. (Also happens to be what I use at work for the
most part)

Its called Webpack and available from Xilinx website.
http://www.xilinx.com/xlnx/xil_prodcat_landingpage.jsp?title=ISE+WebPack
And you’ll need model sim xilinx addition
http://www.xilinx.com/xlnx/xil_prodcat_product.jsp?iLanguageID=1&iCountryID=
1&title=modelsim_xe
Plus there are some patches.
This is a big download so If you don’t have cable think twice.

I use VHDL at work and am happiest when using it but I do speak some verilog
so I would rather use VHDL.

Is there an integer only (FPGA’s don’t like FP) simple to under stand
version of the filter?
Ie in sequences formula rather that code.
With sizes and bounds of the matrixes ect.

I’m not to good with C and am trying to put a spec together.


Edward Freeman