Serial-parallel redundancy





Foreword

I never was involved in real Space Exploration but I love to dream and think about it. One thing I like is you have to build devices without being allowed to test them. You can make rough tests in laboratories but your device will never experience the real thing till it's too late. Far away in Space there is no way to get repair. I once got a contract from a factory, to conceive an electronic circuit and assemble a prototype. The circuit had to cope with both digital and analog signals. I decided to work at home, where I had no possibility to test the circuit. I didn't even own an oscilloscope. So I relied on computer simulations. I wrote numerical calculation programs to check the supposed behavior of each part. There were a thirty or so electronic components involved. I soldered them together very carefully and checked thoroughly the Assembler program of the circuit's microcontroller. The sole tests I could perform was to switch the circuit on and check for basic behaviors, like the clock ticking and some electric levels being normal. When I traveled back to the factory, I plugged in the circuit in their system... and it behaved wonderfully. I was very proud. This was just a game I played with myself. I could have traveled to the factory from the first days on and use their oscilloscopes. Later analysis showed there were some parasite little glitches in the signals generated by my circuit. That wasn't a serious problem, it was solved by adding a little condensator part. Probably I would have build the circuit in half the time if I traveled to the factory from the beginning on and used their oscilloscopes. Nevertheless I know my attitude is the right one. Because the people at the factory were trying to build that circuit since two years. I build it in one month time. That's because I've been playing the like since ever.



Techniques

Many techniques an approaches can be used to make a device reliable:



Parallel ailerons for an airplane wing means an aileron is made out of little independent ailerons placed side by side or as parallel strips:







Both approaches can be mixed to create a checkerwork of aileron parts, or one large flexible aileron containing many tiny actuators, with no hinges nor slits. A well-known example of failure of a parallel devices system is the N1 russian Moon rocket. One problem with parallel devices is the failure of one device can cause the failure of all other devices. For example, in a rocket, one motor exploding can cause all other motors to explode. This has to be coped with.







A lot of the approaches mentioned above come to redundancy. I especially like the serial-parallel approach. It may seem to be expensive and heavy since it needs four the same parts or more. This can be discussed. Most parts and devices in Space Exploration are big, heavy, utterly expensive and need a long time to conceive and build. What's more interesting techniques are often neglected because their reliability isn't proved. Suppose four devices are conceived and build by four different factories and put in serial-parallel. Each device is four times lightweighter and was granted ten times less money and time to conceive. Each will be maybe a hundred times less reliable. Anyway if they are put in serial-parallel they should achieve altogether a higher reliability for a some weight. If captors are added, the failures of some devices can be diagnosed and used to improve the next designs. I expect space probes conceived that way to be more lightweight and less expensive.

What would a serial-parallel radio transmission system with time delays look like? On the emitter side, each little emitter would be controlled by a second circuit that stops the emitter signal if the signal seems faulty. Each emitter would use its own frequency. Data would be fed to each emitter with a different delay. The receiver system on the other side would be made of a corresponding set of receivers. Each receiver would not perform the detection of the signal. Instead the received carrier would be multiplied by a reference signal so each receiver outputs a low-frequency carrier with the same frequency. The time delays would be imposed then finally the low-frequency carriers are summed and the data signal is detected. The delay, summing and detection devices are made redundant too. If the emitters are spread across a space probe body, local harm won't destroy all emitters at once.

In some cases it is better to use a "crossed" serial-parallel redundant system. Let's take for example the parallel ailerons above. One approach can be to use a complete chain of systems for each little aileron part. That is, starting from the pilot's control stick, each aileron part would have its own captors on the stick, its own computer, its own inertial platform, its own security unit, its own power unit, its own wiring, its own captors on each aileron parts and its own actuator on its aileron part. The problem with this serial layout is on the whole chain of devices there is quite a chance one of the devices fails. That makes each aileron part has quite a chance to fail. A possible solution is to exchange data between the chains. For example the computer of each chain is allowed to know the measures made by the inertial platforms and captors of all other chains. This allows each computer to produce much more reliable decisions. The security devices can take into account the decisons made by the other chains. The data exchange can be made through redundant and independent networks. If its power unit fails, a chain can be allowed to borrow energy from other chains. The software of each chain is capable to compute out how to compensate for the failure and misbehaviour of one or a few other chains. Even the final actuator on each aileron part can be made to take into account the decisions of all other chains alltogether. If an airplane uses many such crossed serial-parallel chains made each of different devices, it becomes almost impossible for the aileron function as a whole to fail. Such a mesh of chains is allowed by today technologies yet it requires a different approach in funding, conception, manufacture, pilot training and maintenance.



Eric Brasseur  -  February 11 2005        [ Homepage | eric.brasseur@gmail.com ]