GB1160271A - Testing the Response of Servo Systems - Google Patents

Abstract

1,160,271. Testing servosystems; correlator systems. SPERRY RAND Ltd. Oct. 20, 1966 [July 28, 1965], No. 32201/65. Heading G1U. [Also in Division H3] In an arrangement for testing a servo system, the response of the servo to a unit impulse is determined by applying a pseudo random binary noise signal derived from a feedback shift register as an input to the servosystem and correlating the system output with the input delayed by varying amounts. The plot of the correlation function against delay time gives the impulse response for real time. Since the correlation integration cannot be carried out over an infinite time period, only an approximate value can be obtained and extraneous noise may result in apparent failure of a servo when the correlation values are compared with reference values related to normal servo operation. A predetermined number of failure in such comparisons is therefore allowed. To test a fast acting servo 11, pseudo random binary noise is generated at 12 by a feedback shift register controlled from clock 13, and applied via a buffer 19, whose level may be set from a programme unit 14, to the system 11. Three of the generator outputs are also selected by a solenoid switch 15 and passed to a delayed noise generator 17. This system provides 11 different delays so that 11 points on the response curve may be obtained. The outputs from the servo 11 and the delay 17 are passed to a correlator 29 whose output integrated over one noise cycle passes to a comparator 25. There the correlated output is compared with a standard value set in accordance with the delay from the programme unit 14. If the deviation from the standard value is within an acceptable range the comparator delivers a positive GO signal to a logic unit 18 which steps on the switch 15 to set up the next delayed noise signal. If a zero or NO GO output is obtained the logic unit causes the system to repeat the test with a particular delay, and will not provide a fault indication at 38 until four such tests have resulted in failure. If four such failures occur for four different delays, overall failure is indicated. The overall correlation curve or specific correlation values, may alternatively be provided as outputs from the system. An alternative system for testing slow servos is described Fig. 4 (not shown), wherein the programme unit is replaced by selector switches and the output from the servo under test is passed to a voltage to pulse frequency convertor. The pulses from this are passed to a counter which runs backwards and forwards depending on the polarity of the delayed noise with which it is compared. Over a test cycle the number in the counter represents the approximate correlation function. Since each test takes a relatively long time, tests for each delay are conducted simultaneously by a parallel arrangement of delayed noise generators and counters.