Power Divider Network for TerraSAR-X and Tandem-X Earth Observation Satellites
IMST is component supplier for the 1:32 power divider networks for the SAR-satellites TerraSAR-X (launch: 15 June 2007) and TanDEM-X (planed launch in 2009).
For use in high-resolution X-band SAR satellite accurate (in terms of amplitude and phase) RF power distribution and combination is required to achieve optimum beam forming. Inaccuracies would influence beam shaping and thus degrade the performance of the earth observation satellite.
A planar power distribution network was developed to replace the waveguide arrangement used in previous missions. Among various planar power divider concepts the Wilkinson power divider/combiner was selected for its superior isolation in combination with good symmetry in amplitude and phase. The complexity of a binary (1:32) divider and the requirement to distribute the signal also laterally led to a housing, which is large in terms of wavelength. Waveguide modes will allow signals to propagate not only along the designed transmission lines but also in any direction within the housing thus causing unwanted cross-talk and resonance. These effects degrade amplitude and phase performance. These waveguide modes can be avoided by either absorbing them with suitable material in the housing or by reducing the effective cross-section of the interior housing (operation below cut-off frequency for the waveguide mode).
The spatial arrangement of the TR-modules defines the dimensions of the module: 700 mm x 72 mm x 8 mm. Compared to the wavelength lamda = 19.7 mm @ 9.65 GHz, the housing is oversized. Cavity resonances can and will be excited at all discontinuities. If one or more of the cavity modes is excited, the influences of this mode on amplitude and, even more critical, phase balance cannot be predicted or compensated. There are two approaches to suppress or prevent those unwanted modes:
- Suppression by absorption,
- Prevention by subdividing the housing in channels.
The first approach has the advantage of a fairly simple housing but it absorbs unwanted modes rather than avoiding them. A prototype has proven the effectiveness of the concept and its good-natured behaviour but the losses are about 3 dB higher than with the second method. In space, power consumption is an issue. Consequently a housing with channels has been preferred and realised. Substrate, housing and top cover of the final divider is shown in the figure above.
Subdividing the housing in channels
The classic method would be to mill grooves into the housing and insert the substrate, which is contoured to fit in this shape. Handling the delicate structure of this binary tree from contour milling to assembly would complicate the manufacturing process enormously. The concept presented here leaves the substrate in one rectangle and represents the channels with via chains ("fences") in the substrate and grooves in the lid. Conductive elastomer gaskets provide electrical contact between substrate and lid. This technique becomes obvious in the following photo.
These walls subdivide the housing in channels whose cut-off frequency is well above 9.65 GHz. Conductive elastomer gaskets compensate for thermal mismatch of the materials involved and for the inevitable mechanical tolerances. After optimising the microstrip line in the channel for impedance, two microstrip lines in adjacent channels separated by a wall with a conductive elastomer gasket were simulated with IMST's 3D field simulator EMPIRETM to verify the effectiveness of the shielding. A triple via chain resulted in 120 dB de-coupling between the lines and was considered to have sufficient safety margin. The next step was to optimise the Wilkinson divider by simulating the complete structure taking into account not only the PCB-layout but also the real chip resistor with all parasitics and the influence of housing and lid.
Based on these simulations and the results of some test structures, the complete 1:32 divider was designed, manufactured and tested. Insertion loss and amplitude balance turned out to be well within the ambitious specifications. See the table below for a summary of parameters.
All 32 transmission paths showed very smooth phase vs. frequency, but the phase balance i.e. the phase difference between these 32 paths failed to fulfil specifications. The structure is designed to be absolutely symmetric; geometrical line lengths are diligently kept equal. Nevertheless, there is a phase imbalance of about ±17°. A closer look at the substrate specification reveals a very likely cause for this difference in electrical length. The dielectric constant of this Rogers' material is specified to be in the range of k = 2.92 ± 0.04. Due to the large size, this tolerance can occur across one single circuit board. For the transmit path this translates into a phase variation of phi = ±40°! This worst case scenario assumes two paths with extreme values for the dielectric constant over the complete length, which is extremely unlikely but it explains the measured deviation.
|X-Band SAR Power Divider Network
||> 25 dB
||< 2.9 dB
||± 0.26 dB
|Output Port Isolation
||> 28.9 dB
|Return Loss: Input Port 0
||> 19.4 dB
|Return Loss: Output Ports 1-32
||> 21.8 dB
Artwork of TerraSAR-X Satellite,
Courtesy and Copyright of EADS-Astrium