As for the book itself, it is BIG. For some reason, it gives the impression of being somewhat inflated. I can't pinpoint the exact reason for this, but I suspect it must have something to do with the relation between font size and paper size. The publishers claim there are 2000 references in the book. This may very well be true, as the bibliographies after each chapter generally are 10 to 15 pages long. Unfortunately, there is no overall bibliography in the book. While such a bibliography would inflate the book by somewhere between 100 and 150 pages (it is ~1500 pages in the current version), I think it would be worth adding it.

What the contents is concerned, it is a definite academic angle on the material and the presentation. The practitioner may want to pay attention to the fact that the book title is "_Optimum_ Array processing", not "_Practical_ Array Processing". The "Multiple Signal Classification" (MUSIC) method and "Estimation of Signal Parameters via Rotational Invariant Techniques" (ESPRIT) algorithm are discussed extensively in the later chapters, but some details that turn out to be cruical when putting these methods to practical use appear to be missing in this book. The very motivation for developing the ESPRIT algorithm is that the MUSIC algorithm is extremely sensitive to array calibration data, i.e. that the array calibration matrix must be known with very high presicion. The inventors of ESPRIT, Roy and Kailath, pay meticulous attention to this problem in their journal articles (which are cited by van Trees). The problem of lacking array calibration data has indeed spawned an entire research field known as "blind source estimation". This is not mentioned at all in this book.

The second issue I would like to point out is that alternative, SUBoptimum array processing techniques are very breafly commented in section 9.4 (which all in all is half a page long). I find one sentence in that section somewhat annoying:

"We found that, for estimating the [Direction of Arrivals] of plane-wave signals the [suboptimum] algorithms did not perform as well as MUSIC and ESPRIT"

and then the author defer from discussing these algorithms further. I am sure this decision can be defended from the point of view that the scope of the book is optimum array processing. However, I would like to see how the author tests the suboptimum methods, how they perform, and what he bases his conclusions on. There is a vast difference between tests with synthetic data, generated by the computer, and working with data measured in the real world. My experience from working with real-world data from short arrays at low signal-to-noise ratios, is the exact opposite: The optimum algorithms work bad, if at all, while suboptimum algorithms do the job.