Here is the note I sent to a few of you last month on how the S/C band antenna arrays might be handled. As I see it, one really  critical question is "How do we make a combiner that works?". Comments?

Tom

-------- Original Message --------
Subject: Eagle Microwave Concepts
Date: Mon, 08 Jan 2007 14:44:38 -0500
From: Tom Clark, K3IO <K3IO@verizon.net>
Reply-To: K3IO@verizon.net
To: Bill Ress <bill@hsmicrowave.com>
CC: Jim Sanford <wb4gcs@amsat.org>, Bob McGwier <rwmcgwier@gmail.com>
References: <459FF31C.70300@amsat.org>


Bill -- Jim Sanford just dinged me for not getting back to you:
Tom:
Bill Ress has been attempting to get in touch with you regarding the C-band phased array packages.  Do you have his email?  If not, try bill@hsmicrowave.com
73,
Jim
Lots of excuses for my being uncommunicative, but the main one is that I tend to be a very absent minded old professor who gets easily diverted onto other things!

I've spent quite a bit of time thinking about the needs, on both the RX and TX side. It is most important that we not fall into the same trap that a lot of space-borne phased arrays suffered -- when God decided it was necessary to change from one beam to the next, the switch was made abruptly, resulting in a phase discontinuity. This would mean a discrete level change and a phase discontinuity. If we are spinning 1-2 RPM, we will need to move from one beam to the next every few seconds, and IMHO this would be unacceptable for the digital services.

My initial thought was to generate the array with PA at each antenna (in TX, the same applies to the RX side with an LNA) and feed each element with the proper amplitude/phase. In fact, if we try for maximum gain, then all elements have the same amplitude, and only the phase needs to be tweaked. A linear (in 2 dimensions) phase gradient is what is needed, and this could be generated by phasing the LO to an active up-converter at each element driving each PA. But I worry a lot about the inherent complexity of this solution. We still have to have the computer phase up N LOs.

So I got thinking about other solutions while driving to the new AMSAT lab on the Eastern Shore and I came up with an alternate arrangement. Let's still consider that there are N² PA's, one at each element of an NxN square array, and feed the PA's from a (2 dimensional) Butler matrix. The 2D matrix consists of N 1D "row" matrices feeding another N-1 "column" matrices. This yields (N-1)² independent beams on the sky. (The "minus one" comes from the fact that we may as well throw away the garbage beam at the horizon that results from having elements with alternating +-+-+- phase). Picking a couple of illustrations from my October talk:

----------------------------------------------------------------------------
----------------------------------------------------------------------------
----------------------------------------------------------------------------
OK -- So I'm suggesting using a Butler matrix -- but that yields (N-1)² independent beams and we would experience the phase "click" if we did a discrete switch between the beams. So here comes the new idea I thunk up. Let's first of all go back to the One-D case. It turns out that you can generate the beam that is centered 1/x of the way between two of the discrete beams (and hence (x-1)/x of the way to the adjacent beam position) by feeding using a power splitter and feeding 1/x of the power to the first beam port and (x-1)/x of the power to the second port. After evaluation this idea in detail, I was pleased to prove that the signal fed to the 2 ports are in phase -- i.e. there are no programmable phase shifters needed. Just to make certain I wasn't fooling myself, I "built" a Matlab simulation and was pleased to see that it works.

Some time soon, I need generalize this to the 2D array. I have the gut feeling that we will get no more than ~2 dB of gain ripple by using only a 2 position power splitter. The computer would start with the beam that is closest to the "now" position, and then project the s/c spin to the beam closest to the "next" time step. A simple linear interpolation between "now" and "next" would suffice. When a new "next" beam tap is needed, the "old" beam would play leapfrog to the new "next" position. It would be better if we could concoct a 4-port power splitter so that we can interpolate more smoothly.

So here are some challenges to make something like this work:
  1. Is there some better way to do the phasing that using Butler matrices?
  2. The antenna array phase needed to point at a given location is a simple linear phase gradient. If we don't use a Butler combiner, is there a simple way to invoke a linear phase "tilt" on a bunch of elements.
  3. How complicated is it to fabricate the 2N Butler matrices that we will need in a microstrip structure?
  4. The current baseline calls for 36 elements, arranged as either a 6x6 square or as a 37 element "bee hive" hexagon. What do Butler-like matrices look like for the cases where the array does not have 2^N elements?
  5. Can we invent the microwave widget that allows us to linearly interpolate power between 2 (or preferably 4) of the (N-1)² beams?
  6. Remember that we need a similar structure for the S2 receiver, with LNA's instead of PA's at each antenna.
  7. Have I gone off the deep end with these ideas?
73, Tom