Page 430 - Fundamentals of Radar Signal Processing

over several range bins

(5.144)

This estimate of then replaces the actual S in Eq. (5.142). Since the
I
coefficients used to combine the fore and aft data streams are computed from the
data itself, this is now an adaptive DPCA processor. This method for estimating
S is analogous to cell-averaging CFAR interference estimation to be discussed
I
in Chap. 6 and revisited in Chap. 9.
Equation (5.144) implicitly assumes that the covariance matrices in the
range bins adjoining bin k are all the same as the covariance matrix in bin k

itself. Even if the physical clutter is the same over the averaging interval, this
assumption also requires that a preprocessing gain control step compensate for
3
9
the expected R variation in clutter power with range. Noise power does not
vary with range or Doppler.
Combining Eqs. (5.136), (5.137), (5.142), and (5.143) gives the output of

the DPCA system. Assuming that is a good approximation to S and absorbing
I
all constants into a single constant α gives

(5.145)

While complicated, the structure of a two-pulse canceller is clearly present in
the subtraction of Y [l, k] from Yf [l, k]. If the interference is clutter-limited so
a
that and also highly correlated across phase centers so that ρ[k]→ 1

(implying that the coarse alignment was very successful), the output simplifies
to

(5.146)

The two-pulse canceller structure is clearer here. The complex exponential in

the Doppler index k is equivalent to a time-domain shift of M samples, in
s
accordance with the DPCA condition discussed earlier. The factor β[k]
provides a Doppler-dependent weighting factor that can be optimized to
maximize cancellation in each Doppler channel.
The matched filter design assumes that is an estimate of the covariance

of the interference only, i.e., it should not contain any target signal components.
A practical system must take steps to ensure this is the case, perhaps by skipping