Excessive processing time with Phase Array/Radar toolboxes: bug or 'feature'

Created a phased array aperture using 576 replicated subarrays, with a total of 36,862 radiating elements. Created a thinned version of the full aperture by setting 124 subarray positions to '[ ]'; this leaves 452 subarrays and 28,928 radiating elements.Both arrays are listed by Matlab as '1x1 ReplicatedSubarrays'. A call to 'beamwidth' for the subarray returns in about 0.2 seconds; the unthinned array averages 2.3 seconds to complete. A call to beamwidth for the thinned array averages 463 seconds. Calls to 'directivity' take 3.4 seconds for the subarray and 180 seconds for the unthinned array. I'm still waiting (>1 hour) for the directivity call for the thinned array to finish.

Anybody have any idea why calculations for the thinned array are taking so much longer?

[Frank]

OK, I let the directivity call for the Thinned Array run overnight. It took 31,210 seconds to complete. And as expected, the result showed slightly less directivity than the Unthinned Array so I think the calculation was correct; it's also consistent with my hand calculations. The directivity calls for each array were made for a single azimuth/elevation (zero degrees for each), but it looks like the calculations for the Thinned Array were made across the entire field of view.

Answer 1

[Arka]

Hi @Frank,

I understand the call to "beamwidth" is taking considerably longer in the case of a thinned array as compared to an unthinned array.

It might be difficult to pinpoint the exact cause of this without taking a closer look at the arrays, but in general, there might be a few factors that could be causing this phenomenon, some of which are:

1. Irregularities in the Array: Thinning an array leads to irregularities in the element spacing. This can lead to a more complex radiation pattern, thus leading to more compute-intensive processes.

2. Increased Side Lobes and Grating Lobes: A thinned array often has increased side lobes and grating lobes compared to a fully populated array.The algorithms have to evaluate a larger number of lobes and nulls in the radiation pattern, thus making it more computationally-intensive.

Hope this helps!