Let's get a few more dBs sensitivity: Adaptive coherent averaging

Coherent averaging is a technique used in wind profilers to gain sensitivity by averaging the raw returns for many successive pulses. In wind profilers, the Nyquist velocity is generally huge, and the phase of weather targets does not rotate significantly from pulse to pulse. By averaging N successive pulses by doing a vector sum, the contribution of correlated weather signals increase as N but that of uncorrelated noise increase as N^.5, helping to increase sensitivity by 5*log10(N) dB more than can be done otherwise. This approach has never been implemented successfully on weather radars: because the phase of weather targets can rotate rapidly, the vector sum of N successive returns from weather will grow considerably slower than N, or even N^.5.

If what prevents coherent averaging to work on scanning radars is the rotation of the phase, why not try to compensate for it? Instead of adding vectors directly, let us first compensate for the rotation in phase that occurred, and then average signals coherently. This is what I mean by “adaptative” coherent averaging. This process won't affect noise that is completely decorrelated from pulse to pulse, but it will enhance the weather signal.

Experimentation of the technique was done both using simulated data as well as using real data from vertically pointing (X-band) and scanning (S-band) radars. On vertically pointing radars, enhancements of sensitivity of up to 7 dB was observed, and of up to 4 dB on the scanning radar, provided the spectrum width of the signal was small enough.