5 Hardware
[See all pictures here]
5.1 Airplane
To fly one needs an airplane.
In 1999 I had begun a prototype airplane and finished it in summer 2004.
Unfortunately it did not survive its maiden flight.
The fact that the cause was probably a pilot error encouraged me to develop a new prototype.
Like the first it should have a backstroke drive and be driven by a .40 combustion engine.
In the backstroke drive I see the advantage that exhaust gases and fuel residua do not stick at the fuselage
and thus the sensors attached there are not affected.
I wanted to use the wings of my first electric soaring plane (Graupner Thermic Sport) that have 2.5 m span
and are simple to fly due to the V-type and wing ears.
This time I used a CAD program for the construction.
The time spent for training, disbursed itself later.
So I could check different solution variants and see the model from all sides and eliminate errors before it was built.
Figure 15: Construction design
The fuselage of the airplane consists of 6mm plywood, which forms a very stable box framework.
Additional bracings serve as mounting from engine, tank or connection to the v-tail.
By the relatively heavy engine (700g with exhaust system) the centre of gravity of the model lies rather far in the back.
For this reason it is not worth to want to save weight by sawing out all too large.
The fuselage is covered with balsa and foil.
Two carbon fiber pipes connect the fuselage with the tail unit, which consists of balsa too.
A 300 ml tank is mounted underneath the wings.
Since tank lies some centimeters underneath the carburetor, there were some problems with the fuel supply and I had to turn the engine over.
Centre of gravity and other dynamic characteristics could not be computed with the CAD program.
After the model was finished, still 350g of lead had to be attached to the front in order to adjust the centre of gravity.
The airplane undertook successful taxi drives. But it shows that weight of 4 kg is an issue.
When I started for the first flight the plane went out of control within a short time.
This might have been caused by lose fixing of the carbon tubes which allowed the tail to be twisted.
The impact was hard but the fuselage is still intact.
I am now thinking about buying a prebuild model instead of fixing the existing one...
Figure 16: Finished Flobo airplane
Controller
As controller I decided for the Gumstix controller board.
It is a very small single board computer based on ARM running embedded Linux.
More information about the board can be found on the Gumstix home page.
The board has an Intel XScale-PXA255 processor with 200MHz system clock.
It offers 3 serial, one I2C interface and a SD/MMC slot.
Additionally there is a set of IO pins, which must be attached however all via a tiny connector.
For beginners it is advisable to also order a so-called breakout board which provides the important interfaces
like I2C and serials.
However there is no AD converter.
If necessary this must be realized externally and be tied up over the interfaces.
Figure 17: Gumstix controller board
Meanwhile I brought the board to life.
I had to build one additional board which offers current supply and a RS232 converter.
After it is set on power the uboot loader starts and then boots the Linux system.
You can watch the boot process over a 115200 Baud serial line and also log in afterwards.
I repaired the initial problem with the transmission quality in the meantime.
The cause were wrongly polarized elcaps at the MAX3232 level transducer.
Using the appropriate it is possible to connect the board to the PC via USB.
As described in the Gumstix Wiki
a network bridge is established on the PC so that the Gumstix can finally connect to the LAN.
After it has obtained an IP address from DHCP it is possible to connect to the board by HTTP or SSH.
This way programming and file transfer can easily be done.
Figure 18: Gumstix console on the Palm PDA
5.3 GPS
After I had compared several products, I decided for the u-Blox module.
The SAM LS seems to be well suitable especially for my purposes.
It has an integrated antenna; so there is no need for additional antenna cables.
The position update rate of 4Hz is particularly favourable.
Item data are thus sent maximally 4 times per second over two serial interfaces either in the NMEA format
or in other binary formats.
Furthermore the companies website provides good documentation and software for evaluation and configuration.
Unfavorable is however the connection of the module.
One needs a so-called FFC cable with 20 wires in 0.5 mm grid.
Module and cable can be ordered directly at u-Blox in Switzerland.
This costs approx. 120 euro plus tax and shipment.
I got it substantially cheaper from the German distributor Pointis.
They have the module for 59 euro plus VAT and shipment.
Unfortunately the cable was not available.
And it is a genuine problem to get such a cable with plug.
Eventually I ordered connectors of various sizes together with matching surface mounts from Digikey
which turned out to be a bit expensive.
Figure 19: U-Blox SAM LS module without leads
The module works pretty good standalone.
A few moments after connecting to the testboard the first satellite and position data were received at the serial interface.
However using the u-Blox close to the running Gumstix caused real problems.
EMI from the Gumstix seemes to jam the satellite reception completely.
That's why the Gumstix has to be thoroughly shielded in order to use it with the GPS
5.4 IMU
The project Do it yourself UAV develops an autopilot for model helicopters.
The associated hardware consists of 2 acceleration sensors, 3 gyro sensors and an ATMega controller.
It is able to drive several servos however these features are hardly documented at all.
To the project belongs the company Rotomotion,
who sell complete modules or kits.
I ordered myself a 6DOF IMU kit
for approximately 300 USD plus shipping and tax.
With the associated software accelerations and rotating motion can be measured and processed.
In the motionless condition thus the orientation concerning the space axes can be measured.
The module was tested so far only on the desk.
Therefore no statements on drift and temperature dependency can yet be made.
Also in what respect a Kalman filter is already implemented internally can only be found out on the basis the source code.
As soon as the airplane completed some successful test flights, I will insert Controller,
GPS and IMU and will accomplish measurements.
Figure 20: Rotomotion 6DOF-IMU
5.5 Sonar
For ground distance measurement the Polaroid Ultrasonic Ranging system is to be used.
My buddy Frank kindly lent me an OEM kit with two modules of the 6500 series.
The module supplies distance measurements within the range of 1 to 10 meters and is to be driven by a small ATMega.
In an experimental setup rangings measurements could already be made and the data be transferred to the PC.
Thereby the measuring range was between 0,5 and 6 meters.
However also disturbances and deviations were observerd.
Here the work on calibration must be continued.
Likewise accurate measurements in target environment are to be accomplished.
Figure 21: Polaroid Sonar test setup
5.6 More Sensors
The measurement of the airspeed is to be made by an anemometer.
The data logging pages of Dietrich Meissner
brought me to this idea.
The anemometer was built from a processor fan for notebooks.
It can be attached to an AD converter of an ATMega with minimum wiring.
The number of revolutions of the engine is measured with the help of a reflex interrupter (SFH 9201).
The hub of the propeller attachment was painted black accordingly, so that the light barrier sends one impulse per revolution.
The impulses are to be likewise seized later with the ATMega and converted into revolutions per minute.
Data is finally sent to the main controller via I2C.
Figure 22: Airspeed and RPM sensors
5.7 Adapter Board
The adapter board serves for the current supply of the Gumstix controller, as well as for its mechanical attachment and of course for the interfacing of all sensors and modules.
The Hirose-60 connector is attached on the back side of the board.
Nearly all other electronic parts are placed on the front side.
The PCB was designed in a way that the Gumstix can be surrounded with a shielding.
As already mentioned above, the EMI of the Gumstix causes problems with the satellite reception, therefore it must be surrounded by as close a shielding as possible.
Figure 23: Complete adapter module in housing
The current version of the Adapter Board has the following features:
- voltage regulation of 3.3V and 5V for the supply with 4 or 5 cell RC batteries
- serial interface to the GPS as well as memory buffering with a 3V coin cell
- interface to sonar, RPM, airspeed sensor, temperature sensor and two switch inputs
- one ATMega8 connected to the Gumstix via I2C responsible for collection most of the sensor data
- two RS232 connections are provided that can be used to access the Gumstix console, both GPS serials as well as the ATMega8
- I2C levelshifter to transform signals between 3.3V an the voltage of the RC system (5V)
- monitoring of battery voltage for both circuits using voltage divider and AD converter
- status display with 4 external LEDs connected via small cable
Figure 24: schema of Gumstix Adapter
Soldering of most SMD parts was simple with exception of the Hirose Connector.
This had caused enormous problems with solder shorts with the first version of the board and made the board finally almost useless.
Therefore I decided to have the connector soldered in a company using the reflow technique.
For the shielding tinplate of beverage cans was used - Jever beer is a good choice IMO.
The sheet metal is however so thin that it should provided with a profile for higher rigidity.
Full completed and packed into a housing the entire module weighs somewhat less than 60 grams (2 oz.).
| 
Flobo1 DXF
