Test log for DefconBots 2009 target
See DC09TargetDesign for main page.
Disclaimer
This is not a full / proper test in controlled environment. It is just quick check to make sure that the concept works.
Setting up
Once I had the basic design, I wanted to see if target will actually work. Things that I wanted to check are:
- Device does not miss proper hits
- Device is not trigerred by normal car vibrations or by car crashing into the wall
- Device is strong, and is not damaged by bullets
In order to do this, I have wrote firmware for the MCU that works as a simple oscilloscope. It continiously reads ADC, and some time after the shake is detected, it sends recent data over the serial port to PC. Then a simple program would receive data, parse it and save it.
Techincally, I did the following:
- I used internal 8 Mhz RC oscillator
- ADC was running at 10 bits, 9600 samples/second
- Input was differential, 1.1Vref, 1X amplification
- Software had 230 buffer words = data points on graphs
- Each data point corresponded to 5 ADC readings; the highest is stored
- We also counted number of zero crosses
Guns
I have decided to test different types of guns.
From left to right:
- Yellow - got in in $1 store; very unreliable and low-powered
- D-Eagle - spring-powered Desert Eagle replica; most powerful non-gas gun I had
- Auto - mini-AEG, battery powered; the kind I would expect to see on robot
Car
I did not have the proper car at the time of this test, so I have used EdgeRunner instead. It is the same dimensions, and the fact that it does not steer should not affect target performance.
I have attached target on the back with some tape. You can see the serial cable which goes to the computer.
Hit detection
I have hit target a number of times; all of the times, the hits were detected. Here are some represenative graphs from each gun type. The car was not moving while the graphs were being taken.
Yellow low-power gun:
Mini-AEG:
Desert Eagle (high power):
Note the peak is the same in all graphs - this means that the diodes are working and are limiting the voltage. It seems that 100K load resistor I was using was too high (so the device is too sensetive and is overloaded). In my next test I will try 47K and even 20K.
Movement
To check for false detections, I have driven car around and crashed it into the wall.
The represenatitve graphs are:
Regular driving:
Crash into the wall:
Mechanical strength:
The device was holding while I was shooting it, so it has some strength. However, I wanted to see how far will it go, so I got by green-gas powered handgun (the one which was used in RoboGames 2008, and which literally destroyed many plastic targets there)
I was afraid it will damage the car, so I have mounted the device differently:
Then I have started shooting. After about 10 hits, the superglue gave up and piezo sensor detached.
Still, the device continued to work, and LED was still in position even after I have emptied the magazine. The signal was OK, but with more oscillations. The LED output was working.
Detached speaker:
For the next version, I might try a different glue, like Gorilla or contact cement, that will not snap off in one moment.
Conclusion:
It works!!
- Device detects hits.
- There is no automated rejection of movement / crashes, but the graphs look differently engouh, so it should be possible.
- It is strong -- only Green-gas powered gun could damage it, and even than, it kept working mostly.
-- Mikhail
Comments (0)
You don't have permission to comment on this page.