BT/WSR-2002 Amateur Weather Radar Project

by Sean Breheny(KA3YXM) and Anish Trivedi

Introduction

Anish and I have worked on a number of crazy projects (such as our model rocketry projects) so far, but this one is the most challenging. I'd like to see if it is possible to make an amateur radio based weather radar of modest capabilities. While I have found a few people on the web who have experimented with detecting rain scatter of microwave signals, and one guy who made a secondary tracking radar for model rockets, I have found no one yet who has tried making a full-fledged weather radar. Our main motivation is that radar (detecting objects at a distance by simply reflecting radio waves from them) is fascinating and challenging. We originally wanted to build a radar to detect aircraft, but after doing some calculations, we decided that at the power level we could afford to use, our maximum range and capabilities would be much greater if the target was precipitation. For one thing, as long as your antenna beam is narrow enough to be entirely intercepted by precipitation (a valid assumption as long as it is pointed at a relatively nearby storm), the echo strength decreases as the square of the radius, rather than the radius to the 4th power, as with a point target (like an aircraft).

System Overview

Above you can see a very high-level block diagram of the system. The plan is as follows: The transceiver generates 10uS long pulses at about 3.46GHz at 0 dBm (1 milliWatt) once per millisecond. These pulses are amplified to 20 Watts by the power amplifier and are transmitted by the TX antenna (a 3 foot parabolic dish). The pulses propagate out (in a beam which is about 10 degrees wide) and interact with whatever is in their path to produce a continuous stream of echos. Some of this echo energy propagates back towards the RX antenna (also a 3 foot dish), gets intercepted, amplified by the low-noise preamplifier, and fed back into the transceiver. The transceiver listens for 500 microseconds after each transmitted pulse. During this time, input energy is filtered by an input filter, downconverted to 550MHz, and then downconverted again to 10.7 MHz, where the strictest filtering is applied (360kHz wide ceramic filter).

This signal is then fed to two detectors, one which has a logarithmic response and another which has a linear response. The log detector output is used for strong signals and the linear detector output is used for the lowest 10dB or so of weak signals. Both outputs are sampled every 5 microseconds and fed to a PIC microcontroller. The PIC has 100 range "bins", which are simply storage locations in memory, one location for each sample which will occur in the 500 microsecond listening period, which corresponds to 50 miles maximum range, with a 0.5 mile sampling interval. In addition to the 500 microsecond listening period, the transmitter waits an additional 500 microseconds before transmitting to prevent very distant echos from being received in the beginning of the next listening period.

For every degree of antenna rotation, the transceiver will clear the range bins and send out about 25 of these pulses. If the signal at a particular range sample is strong,the PIC takes the output of the log detector and computes the anti-logarithm of it. If the signal is weak, it takes just the linear output. This number is then averaged with the current value in its corresponding range bin. After all 30 pulses have been processed, the contents of the range bins is sent to the software on the laptop PC via a serial link.

In addition to the PIC in the transceiver, another PIC also performs feedback control of the speed of the rotation of the antenna platform. This platform carries the antennas and all of the electronics shown inside the light gray box. The only connection to the rest of the system is two wires (serial data out and ground) fed through a pair of slip rings. The platform is kept at a constant 5 RPM and the angular position (bearing) of the antennas is monitored by the speed control PIC.

The sofware on the PC takes the incoming reflection data, as well as the antenna bearing data from the speed control system, and creates a PPI type traditional radar plot (polar range plot), using false color to indicate the strength of echos. The software will also automatically compensate for the fact that echos from a greater distance are weaker to produce an output which indicates primarily the amount of reflectivity at each location, independent of distance.

Separate TX and RX antennas were chosen because high power, fast antenna transmit/receive switches are very expensive and hard to find. The isolation from antenna to antenna should be on the order of 60dB, which is enough to prevent receiver damage during transmit and also enough to prevent the receiver from hearing the output noise of the TX amplifier during the listening period.

Since amateur radio regulations require station identification, we will program the transceiver to stop this continuous pusle transmission cycle once every 9 minutes to send my callsign (KA3YXM) in 18 word per minute CW (International Morse Code).

Calculations predict that this system should have a maximum range (for moderate rain, at about 5 millimeters per hour) of about 75 miles. Probably there will be some additional losses that I am not considering which will bring it down a fair amount, but I am hopeful that it will be able to see moderate precipitation within the 50 mile designed range.

If you know anything about professional weather radar equipment, you may be wondering how I can claim this kind of performance, since professional equipment only tries to achieve about 150 mile range and uses as much as hundreds of kilowatts to achieve it. Several factors come into play here. First of all, my radar is only a rough experiment. It doesn't have to work under all conditions, and it doesn't have to see light precipitation (or snow, which gives a considerably weaker echo) out to its full range. It doesn't have to achieve range or angular resolution which is nearly as good as the ones used for weather forecasting. When you consider all of these factors, you can see why it is possible for a cheap (about $2000 total) 20 Watt radar to work sometimes out to 50 miles, and a several million dollar unit is needed to see the lightest of precipitation, always, out to 150 miles with tenth of a mile and 1 degree resolution. In addition, my calculations (which are based on a textbook on radar meterology that I borrowed from the library) may well be wrong. We will have to wait until testing day to find out!

Progress and Current Status

The following section lists the status of each major component. Please check back soon for further updates! We hope to test this by the end of the summer. We had originally planned to test it in June, but problems with Anish's work schedule have not yet allowed him to come up to Pennsylvania to test the system. In addition, as usual, the project is taking longer than scheduled.

RF Safety

The radar's average output power (which is thought to be the critical factor in RF safety) is about 200mW. This creates an effective isotropic average radiated power, along the antenna boresight, of about 100 Watts. We plan on being inside the car or at the base of the antenna structure while the radar is in operation. The angle between us and the antenna will be far off boresight so the effective power in our direction will likely be about 30dB less, or 100mW, which should be completely safe, especially if we are inside the car, which will provide some degree of shielding. Additional calculations will be done to ensure safety and the operating time will be very limited (only a few minutes at a time), which is also necessary because of limited battery power.

Antenna Platform and Speed Control

VIDEO of dishes rotating (open loop at full speed, will really rotate at about 1/5th of this speed): Dishes Video

I am designing the entire system to be transportable in my 2001 Honda Accord, and to mount in the open trunk, so that we can just drive to a suitable location (on a hill away from houses and other buildings and radio towers) and set it up, using the car as a fixed base. I have the platform completed, but I found in my testing that the main wooden board that the system is assembled on is too flexible, which leads to variations in the drive belt tension. In addition, the vertical piece of PVC pipe is not exactly perpendicular to the base, which causes the dishes to wobble. The combined effect causes the belt tension to vary periodically during the rotation, which leads to belt slippage. I have to fix this before the system can go into operation).

The speed control is in final assembly and although this test was done open loop (with the dishes spinning faster than they will be in the final version), the next test should include the speed control and the repairs to stiffen the board and make the mast vertical.

Power Amplifier

I bought this Toshiba UM2683B 3.44 to 3.68GHz 20 Watt power amplifier as new surplus on eBay for $90. I have tested it using an HP8616A signal generator and HP33330 detector and it works! This is really amazing because such an amplifier (which is designed as a replacement for the still common Traveling Wave Tube (TWT) type amplifier) would normally cost about $1000, I'd estimate. These have apparently become quite common in the amateur radio community, since they cover the upper portion of the 9cm band. If you are looking for one of these amplifiers, you might ask Joe Ruggieri pyrojoe@prodigy.net whom I bought it from.

Last weekend, I did some tests on the power amplifier to verify that it works and to see how fast its electronic enable/disable can work (it turns out now that this is irrelevant because of using two antennas). Below are some photos of my test setup (the HP8616A is the thing under the scope, a government surplus signal generator I got on eBay), the inside of the amplifier, and finally of a scope, showing the enable/disable signal (on the bottom) and a signal which is proportional to the RF output power on the top. The square wave is at about 2 kHz.

Antenna System

I recently found out that circular waveguide can propagate linear polarized EM waves (in fact, the lowest-order mode, TE11, is linearly polarized). I had previously thought that it could only propagate circularly polarized modes. Circular waveguide is really easy to obtain because regular food cans work very well for this purpose :-) In addition, a circular horn antenna is better than a rectangular one as a dish feed because it gives radially symmetric illumination. So, using Campbell's soup cans and Old El Paso Enchilada Sauce cans, I made a feed antenna (shown in the photo below) and I now have to make a second one for the other antenna. I tested the SWR for this feed and get a return loss of about 13dB, which isn't a bad start. I may try to tweak it a bit more, but I think that the little probe wire inside the cans is just slightly too short, so I might also just let it be.

The heart of the antenna system is a pair of 36 inch parabolic dishes from DH Satellite, which can be seen in the rotating platform photo above. I bought one and Anish bought the other. Ideally, at this frequency, we should be using a larger dish (like 6 or 9 feet. The National Weather Service's WSR-88D radar systems use a 9 meter (!) dish, and use a similar frequency (2.7 to 3 GHz)). Because of the need to transport this in my car, I had to settle on a 3 foot model. The disadvantages are decreased range and decreased angular resolution (because of the relatively wide 10 degree beamwidth).

The antenna system also consists of a feed antenna for each dish and the cables to connect the preamp and the TX amplifier to the feeds. I'm planning on using simple, open pieces of waveguide for the feeds, which works well with deep dishes (ours have an f/D of 0.3). These will be metal cans with one end open and with an SMA connector on them, connected to a small probe (piece of wire) inside the box. We want linear polarization (see above to see how I originally thought we needed rectangular waveguide to get this, but circular cans work even better!). To first order, circular polarization yields no echo from precipitation. In addition, the "handedness" of circular polarization is reversed when it is reflected, which would mean that one feed would have to be the opposite sense of the other.

I have bought a 20 foot length of LMR-400DB microwave cable, and also a 2 meter length of LMR-400. These will be used to connect between the preamp and transceiver (the 20 foot length) and between the TX feed and the TX amplifier. A small SMA jumper will be used between the RX feed and preamp. The 20 foot length could be much shorter but that was all I could find at a cheap price at the time. Because of the preamp, it makes almost no difference to the noise figure of the receive system, so that is why it is being used there rather than on the TX side, where it would cause about a 20% greater power reduction.

Preamplifier

I am planning on using a 9ULNA (9cm Ultra Low Noise Amplifier) from Downeast Microwave. This costs only $120 fully assembled and is very impressive, with 15dB minimum gain and 0.8dB maximum (0.5 typical) noise figure. There might be some problems with strong signals causing intermod or desensitizing the receiver with such good sensitivity, but since we can disconnect the preamp if we want, we can play it by ear. It sure will help a lot with hearing those distant echos if we can use it.

Transceiver

NEW Below is a photo of the almost completed radar transceiver/signal processor. I should have a better photo soon.

Here are photos of the inside and outside of the 10.7MHz pulse generation section of the transmitter. I decided to put this in a separate box so that it couldn't leak RF into the sensitive 10.7MHz receiver section.

Click here to see a slightly out of date block diagram of the transceiver, without LO board.

The transceiver is clearly the main component of the whole system, and it is a very difficult design to accomplish. I have done it in two parts: the local oscillator board (the system needs two local oscillators, one at roughly 2.8GHz and the other around 570MHz), and the main board. I completed the LO board several months ago and it works fine, as far as I can tell (I don't have a spectrum analyzer so I can't check for phase noise performance or spurs, but it outputs the correct amount of power at the correct frequencies). The LO board has a PIC16F628 which can be reprogrammed to change the frequencies. The LO uses a dual PLL chip (ADF4213) from Analog Devices along with two voltage controlled oscillators from MiniCircuits.

The transceiver hardware is finished but some adjustments still need to be made and the software is not done yet for the 16F628 that acts as a signal processor.

Below is a photo of the LO board.

Power Supply

The power supply is the most straightforward part of the project. The supplies for the electronics on the platform and those off the platform will be kept separate to eliminate the need for extra slip rings to carry power. The on-platform power supply is finished and I should have photos of it up here soon.

Email Sean(shb7@cornell.edu)

Sean's Personal Home Page

Visitor number: