Measurements of reactivity coefficients in magnox reactors and in the Windscale AGR have been carried out using a steady state technique, usually as part of a reactor start-up. The reactor power is increased in steps, allowing time for the temperatures to equi­librium at each stage. The change of reactivity is assessed from the movement of the (previously cali­brated) control rod group used to induce the power change. The temperature change is derived from in­stalled instrumentation. The technique has many lim­itations. It requires the use of a calibrated control rod bank. Since the reactivity insertion rate is depend­ent on the axial flux shape, the calibration should be carried out with an appropriate flux shape which may not be easy to arrange. The measurements of temperature become more difficult to interpret as the power level is increased; at low powers the reactor is essentially isothermal and mean temperature changes are easily related to measurement, at higher powers this is no longer true. The technique only gives an overall reactivity coefficient for the reactor, i. e., a combined fuel and moderator coefficient. In order to derive the moderator coefficient, the effect of the fuel coefficient must be calculated and subtracted. Whilst this is satisfactory for magnox reactors where the fuel coefficient changes only slowly with irradia­tion, the technique is not appropriate to AGRs as both fuel and moderator coefficients are strong functions of irradiation.

Because of the importance of the fuel component of reactivity feedback in the safety of the AGR, a new transient technique was developed which relied on the differences between the time constants of the various components to differentiate between their con­tributions to feedback. The method uses small per­turbations around a steady state to measure the fuel feedback coefficient directly. Calculated moderator coefficients are used as correction factors in the ana­lysis route, but the sensitivity to errors in these values is small.

In the measurement, the reactor is perturbed from a steady state by withdrawing a bank of control rods for a short time, sufficient to insert about 50 mN of reactivity, typically 10-20 seconds. The rods are then held in this position for 30-50 seconds then reinserted to their original positions. It is important that any auto-control loops which would affect reactivity in any way (e. g., reactor inlet temperature, grey rod position) be set to manual during the test. The result­ant flux transient is analysed using a specially written computer program APRECOT [2] to give the fuel coefficient. Such a test gives two estimates of the feedback coefficient, one after the rods are withdrawn and one after they are reinserted. Any disagreement between these values is indicative of errors in the moderator feedback correction factors.

The technique has been used extensively at Hinkley Point В power station to measure the fuel coefficient as a function of core irradiation. The good agreement between measurement and theoretical prediction has allowed the uncertainty allowances placed on the feed­back coefficient in safety calculations to be reduced substantially.