Microwave Beacon Finder
Polymax 9000 won the 2010 AHRC Robot Rally. However, it sucked at beacon-finding and dead reckoning due to bugs not found in last years debugging session. The dead reckoning problems were caused by slight differences in the diameter of left and right wheels. Glue used to hold the tire on the wheel was the culprit. But, that's not very interesting. However, microwave beacon tracking is. Here on this web page I will discuss what I hope will solve the beacon finding issues once and for all.
As a beacon, unmodulated white light sucks. There is too much unpredictable stray light around that can interfere. It can work well in controlled environments but not so much in the real world. It's fine for windowless rooms but once you have sunlight coming in through windows there is potential for weird and unpredictable behavior.
An alternative is modulated light. Infrared light modulated at 38 kHz is often used because of plentiful TV remote control IR receiver modules. Since there is no IR light modulated at 38 kHz in the natural world interference is not a problem - if you are indoors. Outside sunlight can swamp the IR detector rendering useless. Operation near windows can also be problematic.
Interference to and from other modulated IR sources can also present problems. Polymax uses pulsed IR to locate objects in its path. If the beacon emitted modulated IR it would create interference with the obstacle avoidance sensors.
So, what to do? After rejecting a few stupid ideas I finally got interested in using X-Band microwave signals as a beacon. Oh my. Sounds pretty intimidating huh. 10,000 mHz radio is black magic. Or is it?
Way back in the early 70's I built myself a pretty good X-Band police radar detector based on the design of a really crappy one sold by Radio Shack. (like this schematic) The 10.525 GHZ parts consist of a horn antenna, wave guide and two 1N23 microwave diodes. The rest is just low frequency audio circuitry. Basically there are two diodes in the wave guide. One is the detector and the other works as a modulator shorting the wave guide when current is applied. A 1khz square wave is applied to the modulator diode and the detector diode is connected to a high gain audio amplifier. When a 10 GHZ signal enters the wave guide it gets chopped at 1 kHz and detected and converted to 1kHz audio by the detector diode. Pretty simple.
A trip to "old crap storage" out in the unheated "barn" yielded a box full of X-Band waveguide items from the 70's. (Yes, I'm a packrat.) Experiments began.
Fortunately I had a 10.8 GHZ microwave oscillator I'd obtained at a hamfest some 20 years ago. I used it, the wave guides and 1N23 diodes to find out what was possible.
To simplify the receivers I decided to modulate the transmitter at 1 kHz. Receivers only need a single diode per waveguide since, unlike police radar, my beacon is modulated.
Experiments proved a microwave beacon finder would be pretty easy to implement but I'd have to find a better microwave source than the big oscillator I was using. A search of EBay turned up a nice $12 solution. It's a HVA121T-6 VCO module with 10mw output power. It will run nicely on 12 volts (100ma) and can be modulated at 1kHz by switching the power lead. It's also tunable over the entire 10GHZ range by changing the voltage on the vco pin. I bought several. [pdf datasheet of the similar -2 part]
The rest of the beacon consists of a 12 volt AA cell battery pack and the modulator board. I used a 555 timer chip to generate the 1kHz square wave and a transistor to boost the current to switch the X-Band oscillator power lead. I put a small mercury tilt switch in series with the power lead to cut power when it falls over. The bot detects loss of signal and stops automatically. [Schematic here]
To track a beacon the bot needs 2 receivers with antennas pointing to the right and left. The antennas receive equal signal levels when the bot is pointed directly at the beacon. As it turns away, the left and right antennas will receive unequal power and the processor can correct the course.
I used 1N23 diodes in short sections of surplus WR90 X-Band aluminum waveguide as antennas. I sawed off the waveguide segments with a band saw. The back end was then closed with aluminum duct tape. The diodes are located about 1/4 wave from the closed end. I used my plastic bender to make hold down brackets for the diodes. A piece of sheet brass on double sided tape makes contact with top end the diode. The other end is grounded to the waveguide.
The microwave diodes are connected to the receiver board mounted on the back of the waveguide antenna assembly. There are two identical circuits for the left and right sides. [See schematic] The diodes are biased at 8 uA which boost their sensitivity. The signal is AC coupled to a current to voltage converter and then to a 1000 Hz band pass filter. The final stage is a AD8307 log amplifier/detector chip which converts the 1000 Hz signal to a DC voltage porportional to the log of the input level. This chip was designed for signal strength displays in communications equipment. It works from 10Hz to 500 MHz.
The output of the log amps go to A/D converter inputs on the bots AVR processor chip. The log chip compresses more than 90 db of amplitude range into a non-linear 0.5 to 2.5 volt range. This is required because of the large changes in signal strength the bot encounters when traveling from 10 feet to within 4 inches of the beacon. The maximum range seems to be about 20 feet. Unlike light, the 10GHZ signal will pass through cardboard and other light thin non-conductive materials.
The upper green line plots the left+right signal level against the bots angle relative to the beacon. The blue line is left-right vs angle and is used to steer the bot towards the beacon. If you are wondering why the plots are a little "noisy" and asymmetrical it's because this was plotted in a small room with many nearby surfaces for microwaves to bounce off. The beacon was 4 feet away from the bot.
The numbers are the A/D converter values in the AVR chip. To convert to voltage divide by 51.
I modified parts of the white light beacon code so it would work with the microwave hardware. The correlator is no longer needed. On start-up the bot spins 360 degrees while checking the signal level every 5 degrees and storing the signal level and angle. Then it rotates back to the angle of maximum signal and starts moving forward adjusting the left and right motor speeds based on the difference between left and right signal levels. Obstacle detection overrides beacon steering. When the signal level reaches a preset high level it assumes it's very close to the beacon and has cleared all obstacles. The lost signal flag is enabled at this point. If the signal drops to zero the bot assumes it has knocked over the beacon and stops.
Beacon schematic Rev B
UPDATE Oct 2011:
I've dumped the waveguide/1N23 diodes in favor of patch antennas and Skyworks SMS7630-079LF diodes. DigiKey sells them. They work better and are smaller and lighter. Photos below. Note the center of the patch is grounded to the back side ground plane with a feed through wire. Also the copper frame around the outside is only a cutout guide and is not part of the antenna.
Patch antenna pdf layout. 1:1 scale.
Hit Counter = 12702