(c) 2000, 2014 Akos Szoboszlay
The harvester manufacturer (Blackwelders, in Rio Vista, CA) needed an automatic way of controlling the sickle height respect to the ground as the harvester moved down the tomato field. Akos Szoboszlay designed the entire servo loop to keep the sickle a constant distance below or above ground level, as specified by a dial set by the operator.
Akos Szoboszlay obtained Patent #4414792 (granted 11/15/83) for the position servo loop using an ultrasonic sensor to measure distance between ground and frame. This measurement had to be non-contacting. The distance from sickle to frame is measured by an LVDT. An operator sets a potentiometer for the desired distance. The distance the sickle is below or above ground is what needs to be controlled, and this is calculated using analog circuitry.
The ultrasonic sensor both transmits a sound pulse and receives two echos per transmit. (Subsequent echos were ignored until the next transmit.) The electronics continuously calibrates for variations in the velocity of sound in air, without which the accuracy requirement would not have been met. Transmitting an ultrasonic pulse simultaneously starts the counting of a counter, where the clock period, an output of a VCO [Voltage Controlled Oscillator], is varied to counter the velocity variation of sound. There is a narrow bar in front of the ultrasonic transducer a short (constant) distance away that gives a small first echo. This echo is converted to a pulse, which goes to one input of a phase detector of a phase locked loop, a type of servo loop. When the counter reaches a certain count, that also generates a pulse. This count occurs, initially after power-up, only approximately when the small echo occurs. Within a short time after power-up, after dozens of transmits, it occurs coincidentally. The output of the phase detector is the difference in time of these two pulses (comparing rising edges), and is known as a servo "error" signal. It is translated from time information to voltage, is compensated, and fed to the VCO input. The VCO output is the clock for the counter. At a specific count of the counter, the phase detector is given a pulse, thus completing the loop of the phased locked loop. As with any servo loop, the "error signal" drives subsequent stages in an attempt to make the error signal zero. Therefore, the counts are always calibrated with distance, and is independent of the velocity of sound in air. A second echo that is received comes from the target (in this case, the tomato bed after removal of the plants), is converted into a pulse, and this pulse loads the count from the counter into a register. This register value is converted to an analog signal (with units of volts per distance), using a DAC. The DAC output is the distance from the ultrasonic sensor to the target, and is used in the motion control servo loop (described next).
There are no electric motors in the control system, as the system is hydraulic (using an oil-like fliud). Motions are linear (as opposed to rotary) by use of hydraulic cylinders. A solenoid-valve controls oil movement into the cylinder, one for each direction. There was an option of using on-off valves or linear valves (which are used in airplanes). I selected the on-off valve because it was 1/10 the cost, saving over $800 per machine. This would have resulted in an unstable control system but for my method of controlling the valve. Off-pulse width modulation (a VCO followed by a retriggerable timer) modulated the valve. This was calibrated to pulse the sickle position in increments of 1/10 inch. Following feedback control system theory, the rate of pulsing, and therefore the hydraulic fluid flow rate through the valve, was proportional to the "error signal", although modified by the compensation network. It gave a velocity command. The hydraulic cylinder mathematically integrated the hydraulic fluid flow, resulting in position control.
There were two position control loops (but only one bidirectional valve to drive the motion). The inner loop one used an LVDT (measures linear distance) for position feedback and kept sickle distance constant respect to average ground distance, and had higher gain then the outer loop. The outer loop used the ultrasonic sensor and was an averaged (low pass filtered), low gain loop that determined the ground distance respect to body frame. The desired result was that
Also necessary was proper compensation of the system, a lead-lag network. Without this compensation, the sickle would oscillate up-down, as predicted by the theory.
I designed this to be an "over-damped" feedback control system, to eliminate any overshoot. An overshoot in the downward direction caused the sickle to scoop some dirt out of the ground, and when the sensor (ultrasonic) measured the ground distance, it caused the sickle to drop. This was a positive feedback, an instability, that had to be eliminated, because the machine would dig itself into the ground. This was achieved by eliminating overshoot and by locating this feedback (the ultrasonic sensor) as close as possible to the sickle itself. The latter reduced the time delay, as the harvester went forward, between dirt height at the sickle (the ideal location point for the measurement), and the actual location point for the measurement, which was about two feet behind the sickle due to mechanical space constraints.