The interrogator array antenna is a combination of technologies: pseudo-monopulse and lobe-switching, microstrip circuitry and frequency-scanning.   By clever design, only one axis of the antenna needs RF processing; all the rest is done by changing LO frequencies and software processing.
 

Antenna

In elevation, a beam-forming network allows pseudo-monopulse tracking.  A corporate feed network combines the eight subarray outputs into two halves, each composed of four subarrays. These two signals are fed into a hybrid, producing the sum and difference at RF. The sum channel is switched between the transmitter and the rest of the receiver circuitry using a GaAs MMIC switch.   The sum and difference are fed into another hybrid, which produces s+d (sum plus difference) and s-d (sum minus difference). These are switched sequentially into the downconverter using another MMIC switch.  The s+d and s-d signals, when combined are sufficient to extract the off-boresight angle and the sign of that angle, even though no absolute phase information passed through the system.
 

Receiver

The downconverter output, at 70 MHz, feeds a log amp with over 80 dB of dynamic range.  The log amp output is sampled by a fast A/D converter, and passed on to the controller.

Interrogator Mechanization

Since the s+d and s-d signals are both sampled each pulse, the antenna is essentially monopulse in elevation, although it would technically be called pseudo-monopulse.

The pseudo-monopulse outputs are processed through a log amplifier, a form of normalization. This essentially isolates the two axes of the antenna, and the frequency-scanning in azimuth does not affect the pseudo-monopulse operations in elevation.
 

Due to the wide bandwidth of the receiver and its analog-to-digital converter, a different monopulse output can be sampled every 2 microseconds.  Since the transponder downlink pulse width is either 12.8 or 25.6 uS (depending on telemetry data), that means the controller has enough time to read both monopulse outputs during one pulse.  Consequently, this technique is functionally equivalent to monopulse, so long as the signal levels stay constant within the pulse.

By processing the s+d and s-d signals in software, the controller can determine if the antenna is pointed at the transponder, and the direction and amount of error if it is not. By processing that error information and sending commands to the servos, the controller works to keep the antenna pointed directly at the transponder.

With a 2 ms pulse repitition rate, boresight error estimates are available at a 500 Hz rate.  The controller can update the position, velocity, and acceleration parameters of both azes as quickly as they can be calculated from the boresight error estimates.  Given the processor speed, the entire calculation, from raw antenna signal strength data to updated motion parameters, could be accomplished in less than fifty microseconds.
 

Transponder Mechanization

Each yellow "pip" is a pulse-width-modulated RF signal, created by multiplying up a lower-frequency modulated signal.  Shown in red is the amateur 5.7 GHz band.  The transponder signal is actually spread fairly thinly over the band.
 

Beam Selection / Reception

"Now I've got you!  You can't possibly seperate those two beams/signals!  You'll have a mess!"  Please follow me through it step-by-step.

First, by chosing an IF that is lower than the transponder signal spacing, I can select the beam I want to "hear" by simply choosing the right LO frequency to downconvert it into the IF passband.  One it's in the IF passband, the IF processor can digitize the signals's relative amplitude.

"But," you say, "all of the signals from the transponder are still impinging on the antenna, and the antenna is converting them to signals of various amplitudes."  True, but only the selected signal will appear in the IF.

Since the RF passband is fairly wide (about 100 MHz), the noise floor is much higher than in a narrowband system.  But, other system factors (e.g., a good receive preamplifier) can be brought to bear to offset this problem.

Another potential point of objection with this scheme is image rejection.  Since the image frequency isn't bandpass filtered out, a conventional mixer would wrap it into the passband, increasing the SSB noise figure by 3 dB.  But, we're not using a conventional mixer.  An image-reject LO fed by a phase-shifting network allows the noise at the image frequency to be rejected without the trouble of a tracking RF bandpass filter.
 

Tradeoffs

Not all of the forward input power can be radiated in the main beam, and conversely, not all of the incident main beam power can be converted to forward output power.  A matched load (resistor) must exist on each subarray to absorb this reflected power. Therefore, the subarray feedlines are terminated by chip resistors.  This loss of power reduces the efficiency of  the subarrays to around 70%.  If this power were not absorbed, it would be re-radiated in the form of spurious beams and high sidelobe levels. However, other system gain factors overcome this loss, and the simplicity and robustness of this antenna architecture makes it worth the efficiency tradeoff.
 
However, the flip side of this is that the subarrays are very wideband.  Since each element in the subarray is lightly coupled to the transmission line, and  their  mismatch contributions aren't phased and essentially add fairly randomly, the subarrays have a very low (<1.2:1) VSWR over the entire amateur 5.7 GHz band.  This characteristic alone should make them attractive for other uses.

The corporate feed loss contributes to the overall system loss, but it isn't as great a contribution as the traveling-wave array efficiency loss.

Evolution of the Array

The traveling-wave array replaces both the offset-fed dish and its feed, present in earlier incarnations of DARTS.  It was not my original intention to replace the offset-fed dish.  In fact, I considered an offset-fed dish fed by a small patch antenna array to be an optimal combination.  Frequency-scanning of a small array feed seemed to be the perfect solution for reducing microwave parts count and simplifying the system.

However, I found that it is difficult to analyze small arrays, for a number of reasons.  To analyze small arrays, one has to strictly evaluate Green's functions for them, a computationally and analytically expensive proposition.  In addition, in traveling-wave arrays, the assumption that only a forward power wave excites the patches falls apart for a small array.

So, I decided to model a large array.  Well, it seemed that a 9-element subarray was about the smallest traveling-wave array I could design without the equations/assumptions blowing up.  That size array cannot efficiently feed an offset-fed dish, for the beam is too narrow.  In fact, the gain was high enough to use an array for the entire interrogator antenna.  I added subarrays to get an 11x8 (88-element) array to smooth out the sidelobes and increase the gain a little.

Among the many advantages of the array is that it is light and easy to move.  It will allow the interrogator to use smaller motors and gears, and be itself smaller and lighter.

Sometime in the near future, I hope to be able to design a phased-array antenna that sits on top of a box, and can scan horizon-to-horizon in both X and Y axes.  This will eliminate the azimuth and elevation motors and their encoders, as well as all the gears and mechanics.  This would eliminate a major cost and unreliability factor in DARTS.
 

Frequency-Steered Array Antenna
Specifications

Number of elements	88 (11 x 8)
Array confguration	Corporate-fed array of end-fed series traveling-wave linear subarrays
Elevation (subarray) spacing	3.2
Azimuth (subarray element) spacing	2.8
Material	Rogers RO-4003 0.020"
Traverse (azimuth) beamwidth	9.6 deg
Elevation beamwidth	12 deg
Directivity	25 dB
Efficiency	70%
Subarray feed impedance	100 ohms
Excitation profile	Chebychev

 

<this is the older stuff>
 

The Frequency-Steered Array Antenna is a new development in DARTS that I am quite excited about.  It is a high-gain,  planar patch array antenna composed of an series-fed array of traveling-wave linear subarrays.  Through a design that optimizes patch and feedline geometry, the Array will generate five different "virtual" beams, depending on the transmit and receive frequencies.  This completely eliminates any semiconductor components from the antenna, making it simple in concept, easy to manufacture, and robust.
 

Building Blocks

1. End-fed traveling-wave subarrays
2. Subarray feed network

The subarray feed network combines the signals from the subarrays in a similar manner to generate a steered beam in azimuth.
 

Frequency Steering

The array is designed to radiate broadside (zero beamshift in either axis) at some center frequency Fo.  At a lower frequency, the wave propagating along the subarrays (and the subarray feed network) is phase-shifted differently that at Fo, causing the beam to be deflected upward and to the left.  At a higher frequency, the opposite occurs, and the beam is deflected downward and to the right.

With five discrete frequencies transmitted from the transponder, the net effect is to create five beams that lie in a diagonal centered on the array's geometric center:

<beam diagram>

The beam shifts more rapidly with frequency in the elevation plane than in the traverse (azimuth) plane.  This makes it possible to seperate the traverse (azimuth) error signal from the elevation error signal.

Interrogator Mechanization

A short time after the reply rising edge, the interrogator LO changes frequency, selecting one of the squinted beams.  The IF records the signal level, and the LO changes frequency again, selecting the next beam. This process is repeated for the four squinted beams.

The interrogator CPU combines the four signal levels to determine how far off boresight the transponder is, in both azimuth elevation.  The process is akin to solving a system of two equations with two unknowns--the elevation and azimuth off-boresight angles.  The coefficients of the equations are determined by the signal strength measurements.

<equation solution diagram>
 

Transponder Mechanization

Each yellow "pip" is a pulse-width-modulated RF signal, created by multiplying up a lower-frequency modulated signal.  Shown in red is the amateur 5.7 GHz band.  The transponder signal is actually spread fairly thinly over the band.
 

Beam Selection / Reception

"Now I've got you!  You can't possibly seperate those five beams/signals!  You'll have a mess!"  Please follow me through it step-by-step.

First, by chosing an IF that is lower than the inter-pip spacing, I can select the beam I want to "hear" by simply choosing the right LO frequency to downconvert it into the IF passband.  The LO, being both synthesized and sweepable, can "slide" the beams one-by-one into the passband, and the IF processor can digitize each one's relative amplitude.

"But," you say, "all of the signals from the transponder are still impinging on the antenna, and the antenna is converting them to signals of various amplitudes."  True, but only the selected signal will appear in the IF.

Since the RF passband is fairly wide (about 250 MHz), the noise floor is much higher than in a narrowband system.  But, other system factors (e.g., a good receive preamplifier) can be brought to bear to offset this problem.

Another potential point of objection with this scheme is image rejection.  Since the image frequency isn't bandpass filtered out, a conventional mixer would wrap it into the passband, increasing the SSB noise figure by 3 dB.  But, we're not using a conventional mixer.  An image-reject LO fed by a phase-shifting network allows the noise at the image frequency to be rejected without the trouble of a tracking RF bandpass filter.
 

Tradeoffs

Not all of the input power can be radiated in the main beam, and conversely, not all of the incident main beam power can be converted to forward output power.  A matched load (resistor) must exist on each subarray to absorb this reflected power. Therefore, the subarray feedlines and subarray feed network are terminated by chip resistors.  This loss of power reduces the  efficiency of  practical arrays to around 70%.

If this power were not absorbed, it would be re-radiated in the form of spurious beams and high sidelobe levels.

However, other system gain factors overcome this loss, and the simplicity and robustness of this antenna architecture makes it worth the efficiency tradeoff.
 

Evolution of the Array

The traveling-wave array replaces both the offset-fed dish and its feed, present in earlier incarnations of DARTS.  It was not my original intention to replace the offset-fed dish.  In fact, I considered an offset-fed dish fed by a small patch antenna array to be an optimal combination.  Frequency-scanning of a small array feed seemed to be the perfect solution for reducing microwave parts count and simplifying the system.

However, I found that it is difficult to analyze small arrays, for a number of reasons.  To analyze small arrays, one has to strictly evaluate Green's functions for them, a computationally and analytically expensive proposition.  In addition, in traveling-wave arrays, the assumption that only a forward power wave excites the patches falls apart for a small array.

So, I decided to model a large array.  Well, it seemed that a 9x9 (81-element) array was about the smallest symmetric traveling-wave array I could design without the equations/assumptions blowing up.  That size array cannot efficiently feed an offset-fed dish, for the beam is too narrow.  In fact, the gain was high enough to use an array for the entire interrogator antenna.  I moved up to an 11x11 (121 element) array to smooth out the sidelobes and increase the gain a little.

Among the many advantages of the array is that it is light and easy to move.  It will allow the interrogator to use smaller motors and gears, and be itself smaller and lighter.

Sometime in the near future, I hope to be able to design a phased-array antenna that sits on top of a box, and can scan horizon-to-horizon in both X and Y axes.  This will eliminate the azimuth and elevation motors and their encoders, as well as all the gears and mechanics.  This would eliminate a major cost and unreliability factor in DARTS.

 

Frequency-Steered Array Antenna
Specifications

Number of elements	121
Array confguration	End-fed array of end-fed series traveling-wave linear subarrays
Size	11.8" x 11.8" (30 x 30 cm)
Material	Rogers RO-4003 0.020"
Traverse (azimuth) beamwidth	13.67
Elevation beamwidth	13.67
Efficiency	71%
Subarray feed impedance	100 ohms
Excitation profile	Cosine-on-a-pedestal

 

DARTS tracks the rocket using a technique called "single-pulse sequential scan." This means that during the reception of a single pulse from the transponder, all the feed antennas are sampled sequentially. This yields a "snapshot" of the rocket's position with respect to where the antenna is currently pointed. The antenna position, velocity, and acceleration are then updated to track the rocket more closely.

Because DARTS is designed to track rockets, the dynamics of the antenna system are biased to the vertical plane. There are two aspects to this:

1. Because the vertical motion changes fastest, the controller spends more time updating the vertical position of the antenna. The horizontal position update is a background task.
2. The antenna null-tracks (tries to center itself) in the vertical plane, but interpolates antenna positioning in the traverse plane because the traverse feeds aren't centered in the antenna's boresight.

The downconverter is a mixer and a MMIC amplifier, followed by a low-pass filter.

The feed provides are three different, switchable antenna lobes. Note that no movement of the mount is necessary to perform a scan; all three antennas are electrically switched, under command of software running on the controller.   The feed pattern is:

                            L1

                            L2 A1

Due to the wide bandwidth of the receiver and its analog-to-digital converter, a different antenna can be sampled every 2 microseconds.  Since the transponder downlink pulse width is either 12.8 or 25.6 uS (depending on telemetry data), that means the receiver can scan all three feeds during one pulse.  Consequently, this technique is functionally equivalent to monopulse, so long as the signal levels stay constant within the pulse.

By comparing the magnitudes of the L1/L2 and A1 antenna signals, the controller can determine if the antenna is pointed at the transponder, and the direction and amount of error if it is not. By processing that error information and sending commands to the servos, the controller works to keep the antenna pointed directly at the transponder.

With a 2 ms pulse repitition rate, boresight error estimates are available at a 500 Hz rate.  The DSP can update the position, velocity, and acceleration parameters of both azes as quickly as they can be calculated from the boresight error estimates.  Given the processor speed, the entire calculation, from raw antenna signal strength data to updated motion parameters, could be accomplished in less than fifty microseconds.