TABLE OF CONTENTS Surveillance radars usually combine multiple observations per scan before thresholding to detect the presence of targets. Essentially, this can be done in two ways, either by coherent processing of a sequence of M pulses, or by processing each separate pulse and then combining the separate observations in some way (normally simply by averaging). The latter case is called incoherent processing. The use of the word coherent here refers to whether the phase relationships (at the carrier frequency) between successive pulses are known and stable.
If they are, then we can regard a sequence of M pulses as one single known waveform. Hence we can construct a matched filter for this waveform. For a non-fluctuating target, the energy returned from a sequence of M pulses will be M times greater than from a single pulse, while the noise power is unchanged. Hence the SNR and peak response of the matched filter will increase by a factor M. The effect that this has on detectability is given by Eq. (4.45).
Worked example A radar system combines eight pulses coherently before making a decision about the presence of a target. We have already seen that this requires a SNR (after coherent processing) of 15.6 if P fa = 10 - 6 and P d = 0.8. Hence the SNR on a single pulse is 1.95. This gives a detection probability of only 0.003 for a single pulse.
