Run The RISC No. 34
A View from Above
Decoding Satellite Images Part 1

Remote imaging has always been one of my strong interests and I have wanted to construct a satellite image decoder for a long time. Well now I have actually got down to it so let's see how we can receive weather satellite images on our computer. Mind you itÕs not all computers that are up to the job due to display and memory requirements so I think that this is a RISC PC project only. Not that all lesser computers wonÕt work but you wonÕt have the quality of display to get the best out of the project.
There are basically two types of weather satellites, geo-stationary and polar orbital. The geo-stationary type gives a view of the whole of the Earth from a point above the equator, itÕs the type you see on the TV weather forecast. ThatÕs fine for an overall view but at this latitude the view over Great Britain is a bit distorted due to the curvature of the earth and the situation gets worse as you get closer to the poles. So for a more detailed local view of the weather, polar orbital satellites are used. These donÕt actually fly directly over the poles but do pass within a few degrees of them. What they do is to scan the Earth as they orbit it and send down a continuous stream of imaging data, this is shown in figure 1. The image is produced by a rotating mirror which reflects a small spot from the Earth into a light detector. As the mirror spins a whole swathe of land is scanned. Now when the mirror makes its next rotation the satellite has moved slightly in its orbit and a different part of the Earth is scanned. In this way an image is built up of the Earth as seen from directly underneath the satellite.
However, you canÕt receive the satellite all the time because it is in a low orbit and most of the time it is hidden by the curvature of the earth. But when you can receive the signals from it you will get a view of your part of the world. For us in Great Britain this means we can receive images from Iceland to the North coast of Africa but wherever you are, your location is always on the central line of the image.
The orbital period of these sorts of satellites are round about 100 minutes but that doesn't mean you will be able to pick it up every orbit. Due to the rotation of the earth underneath the satellite not all orbits can be received, however, over the course of a day there will be at least four good passes and several other shorter ones. A good pass can last up to 15 minutes and will be close to going directly overhead. A poor pass might last only a few minutes and the satellite will briefly appear over your eastern or western horizon. The length of pass determines the size of image you receive as the satellite is always imaging a swathe of land centred directly below it. There are two American satellites that are always operational with usually a reserve satellite already in orbit in case one should fail. In addition there are a few Russian satellites and the odd one from other countries. An organisation dedicated to keeping a watch on these and using them is RIG the Remote Imaging Group. Their quarterly journal keeps you up to date on what is happening as well as having lots of articles on home construction and techniques, see the box for details.
These satellites transmit both high resolution and low resolution images. The high resolution images are transmitted in the microwave region and take quite an expensive set-up to receive as they have to be tracked by a moving dish. Although if your pocket is deep enough you can do it, it is the low resolution images I want to look at because these are the easiest to receive. However, even at low resolution, a good pass will produce a 5 M byte sprite file!
The low resolution image is transmitted in the 137 MHz band and is quite easy to pick up. Unfortunately, the last government allowed paging transmitters in this band, disregarding international treaties. So, receiving interference-free signals is not as easy as it once was. However, you can use an omni-directional aerial and so no tracking is required. You can get receivers from many places including Maplin, I will talk about them in a later article but for this month I want to look at what to do with the signal they receive.
The image is sent in the form of an amplitude modulated audio signal of 2.4 KHz. Figure 2 shows a signal and what it looks like when it is amplitude modulated. ItÕs easy if you think of it as the loudness of the audio signal depending on the brightness of the image at any one point. What we need to do is to convert this trembling audio tone into an analogue signal and then digitise it. Also we need to generate a clock signal to tell the computer when to read in a pixel. This could be derived from the audio signal itself but in practice it is best not to. This is because, as the satellite passes over, the signal strength varies due to the change in angle and changing satellite distance. This means there are times when the signal is not strong and hiss or interference is heard. If the clock signal were being derived from the audio signal you would lose it at times and the picture would lose synchronism. Therefore, you are best to derive the clock signal from a stable crystal oscillator.
The block diagram of the decoder is shown in figure 3, a reference oscillator is used to generate the handshaking signals for the computer and the audio signal is processed to feed an A/D converter. The heart of the circuit is the precision rectifiers, these detect the negative peaks of the audio signal but because one is fed with the inverse of the audio signal we detect both the positive and negative peaks. When these two signals are mixed we have what is known as full wave rectification.
On to the real thing, shown in figure 4. Now this might look like a complex circuit but, as you get four operational amplifiers in one package, it is very much simpler than it first looks. The amplifiersÕ chips are labelled A and B and they need +7.5 volts on pin 4 and - 7.5 volts on pin 11. These voltages are not too critical but need to be a few volts above 5 volts as most amplifiers canÕt drive the signals close to the supply rails. The diode types are not critical but you should use a small signal diode rather than a rectifier type. The digital circuitry needs 5 volts and a 1.8432 oscillator. You can get oscillator modules of this frequency but if you prefer you can construct a normal type of crystal oscillator. If you do this make sure you include a small capacitor to trim the frequency. The circuit makes use of the computerÕs parallel input routines, at regular intervals determined by the oscillator and divider circuitry the ACK line is driven low by flip flop 2. This causes the computer to read the input bits on the printer port. When it has done this, the computer then generates a strobe pulse, this is used to kick start the A/D converter for the next sample. The ACK line is driven high a short time later by a signal on the flip flopÕs preset input. So data is clocked in at a rate of 4.8 KHz, thatÕs twice the carrier frequency. In order to use this routine, the busy and the select lines must be grounded as shown.
The NE567 is a phase locked loop tone decoder and is not strictly necessary and so could be left out. ItÕs purpose is to detect the incoming audio frequency and light an LED when it sees it. This signal could then be used to tell the computer that an image was being received, but I donÕt use it in my setup. It has an important use if you want to record a satellite pass for decoding later. Straightforward recording is often unsatisfactory due to variations in the tape speed, these will show up a skewed images. So the standard trick is to use a stereo recorder with one track for the audio and the other recording the clock signal, this can be obtained from pin 2 of FF1. On playback the clock track is sent to the input of the phase lock loop, pin 3 via the capacitor and the connection to pin 7 of the amplifier is disconnected. Also the output, pin 5, is used to as a clock for the data, pin 11 of FF2. Note if you do this you will have to have a switch to select pin 5 of the NE567 or pin 5 of FF1 for pin 11 of FF2. In this way any tape speed variations are reflected in changes of clock signal frequency and are automatically cancelled out. A simpler but more expensive solution is to get a mini disc recorder, this does not suffer from speed variations. It does however have an audio compression circuit that could distort the signal but I have found it gives good results.
Well thatÕs about all I have space for, so tune in next month when I will show you the software you need to display the images.