Continuously Variable Pneumatic Control by Mark J E Bellis, 10th October 2009 Addtional description and applications 20/10/09 The design of this system aligns with standard industrial control theory and practices. A few things are important with this system: - The amount of dither uses a 0.5M crank. More without longer valve lever extensions would make the system more jumpy. Less with shorter levers might be better, but then it's a proportion of valve lever load to motor speed. - The speed of dither should be significantly higher than the speed at which a cylinder can oscillate (except when used in an engine!) - If the dither is too fast, the valve seals might wear out quicker. - The amount of feedback should not be too high or the system will oscillate. This may be used intentionally, perhaps in a show car that has oscillating suspension. Some theory: When powering pneumatics, the valve, if used on the edge between "on" and "off", can vary the pressure to the cylinder port, but this can change only the force per unit area in the cylinder and hence the acceleration of the cylinder against the load (F=ma). This is the double integral of the actual cylinder position, which we want to control. The feedback uses the position of the cylinder to vary the valve position, so the cylinder differentiates the pressure to produce the feedback signal. If the feedback gain is sufficiently low (40% on the turquoise lever in this case), the system will be stable and the cylinder will hunt for the desired position because of the integral in the loop. This is why too much feedback gain causes the cylinder to oscillate, because the hunting is too great. The simplest things that this mechanism should be good for driving are: - vehicle steering, either Hand of God or PF remote with motor worm drive to the red sliding beam. - one or more pneumatic valves, including stopping in the middle. More applications: - Truck suspension with centre-lift or tag axle - apply more force to the other rear axle when the wheels are lifted, but distribute the force to both axles with the lifting axle down. Allow the axles to move and the system will push back on the deviation from level. - Variable height suspension on a 4x4. - Anti-roll suspension, by applying more pressure to the outside wheels when cornering. - Power Steering - add direct mechanical connection to avoid disparity between steering wheel and wheels. - Walking robots - leg pressure balancing, body movement to balance a biped. Either the red beam or the blue beam or both could be swinging rather than sliding, which might be useful for a steering mechanism. The cylinder precision needs the dither and compressor (if used) to be at sufficiently high frequencies so that their cycle times are shorter than the response time of the cylinder to a change in air pressure. Lower frequencies might make the cylinder oscillate, but this could be useful for particular applications, such as intentional chaotic motion, which might make an exhibition model appear more life-like. The oscillation would be subtly different from the kind induced by too much feedback gain. The closed loop system also needs excess air. I realise compressed air is harder to come by in LEGO models because we usually use tiny compressors, but this follows the industrial principles. Industrial closed loop oil and fuel systems follow this principle and have excess fluid pressure, feeding back the excess. Feeding back is less easy for us because we don't have sealed 4-port valves. I had thought of using 4 valve switches, but it would double the dither motor load, slowing it down. The extra switches might help in making a sealed system, feeding expelled air back to a reservoir, but it is the differential pressure that moves the piston, so that could work against the objective. A better use for 2 more valves would be for a separate air loop to a separate cylinder, with the feedback from both cylinders being added or averaged back to the blue sliding beam. I was asked whether mini compressors could be added to the dither motor crank. These would be fixed to the red beam, with the whole lot sliding by 2M. If so, a pair should be used like a flat-2 engine so that air pressure peaks equally on both halves of the dither cycle. Otherwise the uneven pressure would bias the cylinder one way or the other. An air supply that is uneven and synchronised to the dither cycle could be used as a fail-safe, as in industry, to bias a function to a safer state. Given the excess air required, a separate compressor of larger capacity is preferable. If more cylinders are driven from the same air output (2 pipes), the feedback cylinder should be at the far end of the pipes. Otherwise the performance of the extra cylinders is not so good. Movement of extra cylinders will creep to one end or the other unless their position is fed back to the valve positions. With 2 cylinders both fed back, equal pipe distances are recommended. Easier to control is a single larger set of cylinders using just the one feedback connection. The 8421 crane jib could work well with this system, as long as the feedback used the correct proportion of jib height (~2M full travel). A set of parallel cylinders could lift a heavier load with the same control system, giving greater amplification and showing how a simple low-power input moving the red beam by 2M could drive something more powerful. See the 10-minute tutorial video at: http://www.youtube.com/profile?user=mbellisbrickmocs#p/u/u-all/0/9jv73J8-4Zw Mark J E Bellis