measurements. Figure 3-19 shows a simple bridge of the sort 

 initially considered for this project. A null will be obtained when 

 R R = R R . Because we have specified R and R as matched 

 resistors, a null will be obtained when the variable (switch-settable) 

 resistor network designated R is set to equal the resistance of the 

 sensor switched in as R . The bridge configuration has the advantage 

 that only the network R must be known precisely; R and R need only 

 be matched, because their absolute values do not affect the measure- 

 ment. The conditions for null across the bridge are independent of 

 the voltage placed across the bridge. Dependence on resistors as 

 standard elements also reduces the amount of long-term drift, be- 

 cause resistors are very stable circuit elements. 



The error-detecting amplifier will affect bridge stability only if 

 its null-sensing point (the point where it recognizes its two inputs 

 as equal) drifts. In fact, this amplifier's drift is less than the input 

 created by a change (in either R or R ) of 10 ohms, which is the 

 smallest resistance change we are trying to resolve. The operation 

 of setting the resistors in the network R . does not require the error- 

 sensing amplifier to be linear; the amplifier is always driven to pre- 

 set limits, indicating whether R . is set too large or too small. It 

 should also be noted in passing that the switching elements used are 

 mercury-wetted-contact relays, which have a life expectancy of over 

 one billion operations. 



The bridge of Figure 3-19 has the disadvantage that it allows the 

 voltage across a sensor to vary. Each time a sensor is switched in 

 or out of the bridge, a voltage transient is created on the signal 

 cable. These voltage changes would not be a problem if the measure- 

 ment interval were much longer than the cable time constant. The 

 actual installation, however, places several miles of cable between 

 the sensors and the Data Acquisition System. This cable introduces 

 over 1. 0|»f of capacitance. As a result of this capacitance, any 

 voltage transient will take several seconds to settle out. 



The problem of capacitive coupling among the lines is solved by 

 providing an individual amplifier for each sensor line. These ampli- 

 fiers by means of feedback,, hold the voltage across each input line 

 nearly constant for all time, whether or not the sensor is being 

 measured. Through this feedback scheme, the voltage variations 

 on each line are reduced to about 0. 02% of the value they would have 

 using the scheme of Figure 3-19. 



Figure 3-20 illustrates the measurement bridge, revised to include 



216 



