Nera 10GHz Receiver
This project was the redesign of an existing product. A receiver operating from 10.7-11.7GHz was to be shifted to a new operating frequency of 10.0-10.7GHz. Some of the active devices used in the original design were obsolete, and so the redesign included the introduction of new active devices.
The project was a major success.
The first iteration of the new design was exactly correct and met all of the design requirements, first pass. The customer was of course very happy.
The Original Receiver
The original receiver was first evaluated for overall compliance with the original specification. This was done to ensure the specification was being met, and also to verify test arrangements and methods.
The receiver was constructed as a single layer of tracking for the RF and D.C. circuitry on a "Duroid" substrate backed with rigid Aluminium for grounding and mechanical stability. This type of construction is particularly effective, and from approximately 5GHz upwards it becomes increasingly necessary to employ construction methods and materials that are appropriate. Materials commonly used at lower frequencies (e.g. FR4) become lossy and less predictable as the frequency increases.
The lower image shows the original receiver in our test lab undergoing some basic initial testing. The receiver is mounted on a test platform which mates the RF connections with the push-on connectors in use, and also connects the required D.C supplies to an easily accessible connector on the side of the test platform.
The computer at the left is being used to program the local oscillator frequency.
To ensure a high degree of confidence in the new board design, the first task is to create an accurate simulation of each circuit block affected by the changes. These would be all of the microwave parts of the circuit; LNA's, Attenuators, Mixers, and the L.O. Buffer and Frequency Multiplier.
A good working receiver was carefully sawn up (on a milling machine), to provide examples of the required circuit blocks.
To each such block, test connectors were mounted as shown here. These connectors are located on 50 ohm tracks which originally connected between the circuit blocks. The result is, for example, an isolated LNA circuit with 50 ohm input and output connections.
We can now do 2 things: Firstly, we can measure the performance of the circuit block without interference (either electrical or mechanical) from surrounding circuitry. The coax-microstrip transitions are of good quality and provide highly predictable performance. The network analyser is calibrated to the reference plane of the transition, and the s-parameters are extracted and saved.
Secondly, we create a simulation in ADS which includes only the parts that are actually present on the test circuit, including the exact (measured) length of any 50 ohm lines right up to the reference plane.
The measured s-parameters and the simulation should at this stage be identical.
A crucial step at this point is to make sure that the simulation is EXACTLY the same as the measurement data. If there is a difference, the simulation should be adjusted carefully until the best possible correlation is achieved.
When the simulation closely matches the measured hardware, design changes can be introduced with much more confidence than can ever be possible when starting out from scratch. Line lengths and widths, stub positions etc can be altered by small increments, in this case to shift the optimum performance from the 11GHz range down to 10GHz.
Agilent ADS (Advanced Design System)
The top image here shows a representative example ADS schematic. The schematic must be carefully tailored to the task at hand. Experience leads to insight into what's important, and what is not. A schematic that includes too little detail will lead to inaccurate results, but you can very definately include too much detail which becomes both a waste of time, and a potential source of error.
The lower image shows a representative linear simulation result. On the left we can see the gain (S21) and the magnitude of the input and output reflection coeficients (S11 and S22). On the right, we see a smith chart representation of the input and output impedances.
TRL Test Jig
Testing of the re-designed board was performed in a similar way as for testing the original board, being first "dissected" into the particular areas of interest.
However, to expediate the testing and to ensure the utmost accuracy of calibration, a new test jig was constructed.
The new jig provides a pair of high quality coax-microstrip transitions which can be moved around on a base plate, and also raised and lowered to bring them into contact with the microstrip lines. Once constructed, this new jig provided a major increase in test speed and accuracy.
A furthe benefit is that the jig may be used without soldering the RF connections. They can thus be re-connected many times without damage to the jig, or the item under test.
The lower image here shows a section of the new design mounted in the jig and connected to the network analyser. A pair of oposing back-stops have been used to hold the board in place.