Accuracy was a seperate statement describing how on AESAs, signal analysis can be done per element, unlike on PESAs. This greatly improves accuracy in look down modes, particularly SAR, by allowing for much better isolation of the target’s returns. Although it should be noted even the most modern AESAs currently in service dont rely on fully digital per element beam forming/signal analysis.
So, uh, multi-beamforming has lots of potential issues, and beam interleaving is basically nearly always better. Which is something PESAs can actually do; early ones didnt, but it was due to processing power limitations, even early enogh prototype AESAs lacked it.
So, uh, firstly, null formations. Basically when you have multiple beams at once, they actually recieve contructive and destructive interfearence from each others sidelobes. This is bad as it adds in significant interfearence. There are ways to filter this out, but they grow exponentially more complex the more beams are active. And without this costly filtering, the range which they can reliably track targets is greatly degraded.
Also, so, radars have things called beam cycles, which are made up of coherent processing intervals. CPIs include pulses sent out specifically to perform checks on velocity, azimuth, and elevation and such. Here things such as PRF comes into play, as a higher PRF mode allows for more checks to be performed thus improving accuracy. Point is in multi-beamforming, without digital per element analysis, the information in these CPIs can basically only be processed one detection at a time. So you’re more or less limited to the accuracy you’d conventionally have in like an LPRF mode, which is not suitable for engaging targets (but is otherwise fine for early warning). There are ways to have it so that way it splits how the information is recieved, but this brings us to what is done instead.
Mode/beam interleaving, is where, CPIs for multiple different track types are interweaved. This is what basically every AESA, including the J/APG-2 and AN/APG-77 use. Basically they use expanded onboard RAM to remember interval azimuth/elevation data between sniffs, allowing for “multiple” instantanious detections, although its really just swapping between them at a sub-cycle level. Which fun fact, modern PESAs would also be able to do.
Japan and britain have been working on a true multi-beam design for GCAP through a program called JAGUAR. Since japan having lots of research into APAARs and per element beam analysis, but yeah that just shows how not a thing currently it is.
If its easier to think about in a more general electrical sense, think of it in a conventional circuit as PWM versus other methods of power control. Instead of reducing the powrr outright (splitting it between two tasks), its instead alternated in frequency between active tasks, as lowering the active power causes sub-optimal performance. This also ties back to stuff such as duty cycles on the radar, but i dont wanna get into that unless the rest of what i said was decently understood.