The multiple-tap phase shifter could also be built using a lumped-element version of a transmisson line. In either case, there will be an upper frequency limit to its operation because of the capacitance of the amplifiers at each tap. If we assume 3 pF at each amplifier input and that a reactance of 100 ohms is allowed, that limit would be about 500 MHz. Each multiplier/PA will have a slightly different phase shift so there would still have to be phase shifters associated with each antenna element.
73,
John KD6OZH ----- Original Message ----- From: Tom Clark, K3IO To: AMSAT Eagle Sent: Wednesday, April 25, 2007 21:56 UTC Subject: [eagle] Another idea on phased array configuration
I have been concerned about the complexity of doing the phased array. Most of the discussion so far has been along the lines of this set of premises:
1. We assume an X-element array (where X is somewhere in the 30-45 range to get enough gain to get a good signal to the ground) 2. For a lot of reasons (minimizing losses, redundancy, generating RF power, etc) we like the idea of distributing the PA function to have one PA at each array element. This means that in addition to X antennas, we have X amplifiers. 3. There exist several nice PA chip amplifiers that can each make ~1 watt (I'll use that value because one watt's as good as another). All these breeds of amplifier have a gain in the 10-20 dB (i.e. the numeric gain G is in the [10-100] range, corresponding to a drive level D in the 10-100 mW range). 4. With a PA at each element requiring ~D watts of drive, then we need a driver capable of X*D watts, assuming there are no losses in the X-way power splitter and all the intervening coax cable. 5. It is not hard to imagine that the power splitter+cable losses will be at least 3 dB (it's C-band, remember) so the drive requirements rapidly grow to the 2X*D range. OUCH! -- the driver is now at least as big a deal as the PA at the antenna element! 6. We still need some way to generate the phase shift necessary to point the beam in the desired direction, and be able to update the pointing to compensate for the fact that the spacecraft spin axis doesn't point at the earth.
Several recent additions to these thoughts have included:
7. From Franklin Antonio: How about distributing a lower frequency phase reference and add a PLL to generate the microwave frequency? 8. Also from Franklin: Put a programmable phase shifter along with the PLL at each antenna to obtain the necessary antenna phasing? 9. From me: The phasing can be made easier if the array consists of collinear elements. The required phasing over the whole array needed to keep the beam pointed to the earth is a simple, linear gradient. If the array consists of several linear arrays, then the projection of the needed phase on each of the linear arrays is also linear.
Pardon me while I digress, but I think this story is relevant. When I did my thesis many eons ago, I built a large 10 MHz radio telescope made up of a lot of east-west dipoles. For the north-south arm of the telescope, I made a km-long 450 ohm terminated transmission line made from #12 copperweld supported between a horizontal 2x4 which was held up by fence posts. And I chose 0.5 wavelengths as the between element spacing. The elements were supported in the air by a 20' pressure treated 4x4s at the middle and both ends.
Each antenna element was 3 wavelengths long -- central full-wave dipole fed with a half-wave of tubular 300 ohm TV cable; then at the end a shorted quarter-wave stub, and another full wavelength of wire on each side (this is sometimes called a Franklin collinear). The central feed-point impedance was high (several kohms, as compared with the 450 ohm open wire t-line. The T-line was marked off in 0.1 wavelength units; since the elements were spaced a half-wave, every 5th mark was at an antenna pole. To phase the array to a given declination might have required (as an example) a phase increment of 0.15 wavelengths. The first element was tapped onto the line at position 0. the second (a half wave away) was at position 0.515 but with the connection block flipped to get a 180° phase reversal. then the 3rd element should have been at 1.30 with no reversal, but it was closer to tap at 0.80 with a reversal. And so forth thru all 64 elements. It took about an hour to go thru the array to re-phase it for a different declination. Because the taps ended up contributing a (nearly) randomized set of reflections, the inter-element interactions were quite small unless the array was phased to the zenith (when all the individual ~1.1:1 VSWR phasors added up).
Thinking back on these sins of my youth, I came up with a new idea based on points 7 & 9, which is seen in the first attached drawing "tapped delay line".
As a variation on Franklin's idea #7, my thoughts are to use the Nth sub-harmonic of the carrier. We might make N=8 so that we distribute 730 MHz (corresponding to 5840 MHz center downlink); at this frequency, one wavelength ~ 411 mm. For a 7-element array, I show a 6*L piece of coax, with taps uniformly space L and terminated in its characteristic impedance. There a 7 uniformly tapped steps and at each tap is a high impedance buffer amplifier (think coax Ethernet here). The buffer feeds a *N multiplier (I suggested N-8 because Hittite has some really nice active microwave doublers); alternatively, it could involve a PLL, perhaps with a DRO "puck" as the resonator. The microwave signal from the multiplier feeds a double-balanced mixer to generate BPSK, followed by a PA and the patch antenna element. Note that the BPSK modulation could be done at a lower frequency inside the multiplier, in which case the modulation phase shift is < 180° by some integer divisor.
However, at this point we have not phased the array -- the interelement phase is determined by the tap interval L and the frequency f/N. If we could make the tapped delay line from rubber, then we could get an incremental change in the phase by simply mechanically stretching it. Since we can't change the physical length, can we tune the delay line electronically? I think that the scheme shown in the 2nd "rubber" drawing will work. At each tap point, we add a varactor and set the bias on all the varactors with a D/A. As we change the voltage across the diode string, each diode's capacity will be changed by the same amount, making the equivalent of computer-controlled "rubber". Voila -- we can generate the smooth phase gradient needed to point the antenna.
Note that the phase "swings" from the end where the f/N oscillator injects its signal. But IMHO, we want to "lock" the phase of the central element, half-way down the tapped line. I suggest that, in addition to programming the DC voltage necessary to point the antenna, the computer adds a "DC" constant to each string based on keeping the central element as the phase reference. Note that some of the delay line taps will not have an PA/antenna element attached to it; the degenerate case is the central element which can only be physically present in one of the intersecting arms.
The antenna geometry I tend to favor is the 43-element "12-spoke" ("two bits" * 43 = $10.75) version which has 3 9-element arms (with a single common central element) and 3 6-element arms (with the central 3 elements missing). All 6 would use an 8*L delay line, with the central phase of 5 of the lines slaved to the 6th.
As usual, please enter into a lively, feedback-provoking discussion -- 73, Tom
------------------------------------------------------------------------------
------------------------------------------------------------------------------
_______________________________________________ Via the Eagle mailing list courtesy of AMSAT-NA Eagle@amsat.org http://amsat.org/mailman/listinfo/eagle