by Scott Sacks, Director of Advanced Technology; Becker Sharif, RF Engineer; Tom Graves, Principal Test Engineer; and Mike Stein, Program Manager, CAES (formerly Cobham Advanced Electronic Solutions)
Often, high frequency, broadband RF electronics require some amount of tuning to optimize parameters such as input and output VSWR, gain, gain flatness, P1dB, and noise figure. This tuning can be a complex process, requiring very small adjustments in multiple locations in the RF circuit that are conducted in an iterative manner to obtain the best overall performance.
To help reduce time spent on the tuning process as well as increase accuracy and consistency, CAES has made a significant investment in advanced manufacturing by developing an Automated Robotic Tuning (ART) system for fine tuning RF modules and circuit card assemblies (CCAs). The ART system (Figure 1) provides the robotic hardware (high precision motorized stages, high precision laser, high-definition camera, etc.) and the software to control the robotic system and tune the desired RF unit. This article will describe the kinds of tuning problems encountered, how the ART system can address them, and the benefits the system provides compared to manual tuning.
Tuning RF modules and circuit cards helps to obtain optimal electrical performance and is typically performed manually (Figure 2). However, challenges arise from designs that have a large number and mix of discrete and integrated amplifiers, switches, limiters, detectors, filters and more, in addition to all the RF interfaces and matching circuits that accompany these different circuit elements. It is often not economical to design low-to-medium volume RF module and CCA products that use customized MMICs as it results in a higher device count and an increased number of RF interfaces. Differences in manufacturing assembly processes (wire bonds, substrate gaps, substrate material variation, device location differences and others) as well as variation in the semiconductor devices creates the potential for a sub-optimal RF match and RF discontinuities.
Tuning is required to mitigate discontinuities and improve RF matching circuits in high-frequency broadband products. This process typically entails adding or removing small amounts of conductive material from the RF circuit. The addition of conductive material results in increasing shunt capacitance and decreasing series inductance on a microstrip trace. A reduction in conductive material results in decreasing shunt capacitance and increasing series inductance in this circuit. The addition of conductive material can be accomplished by daubing small amounts of conductive epoxy to the microstrip trace, accompanied by removal of conductive material by scribing or cutting a small portion of the microstrip trace.
Determining the location where conductive material must be added or removed and the amount of conductive material that must be added or removed is often found by placing an RF tuning probe at various locations in the circuit. The RF tuning probe is simply a small amount of conductive material at the end of a low-dielectric insulator. If the RF response is improved with the addition of the RF tuning probe, that would suggest more conductive material is needed in this portion of the circuit. A degradation of the RF response with the RF tuning probe inserted into the circuit indicates that less conductive material is needed in this area.
The insertion of the RF tuning probe is done while measuring the electrical response of the device. The electrical response could be the input or output VSWR, gain, gain flatness, output power or noise figure of the product. Very often this process of inserting the RF tuning probe in various locations in the circuit, measuring one or more of these parameters and deciding whether a change in the circuit is needed, is an iterative process and requires balancing and trading one electrical parameter for another. A good example of this is improving the low-frequency input VSWR while sacrificing or degrading the high-frequency input VSWR.
The amount of conductive material and location of the RF tuners are critical to obtaining the desired electrical performance for broadband circuits. This varies considerably depending on the frequency of operation, bandwidth required, and the desired level of performance. To provide an idea of the size and location impact of RF tuning performed on one CAES high-frequency broadband product, tune locations need only vary by 0.005 in. and the area of material added or subtracted need only be different by 20 square micro-inches to cause a significant impact in performance.
This kind of RF tuning is most often a manual process conducted by highly skilled RF technicians whose fine motor skills make such tiny adjustments possible under a microscope. They must be able to hold the RF probe tip steady and change their gaze from the RF measuring instruments to the circuit under the microscope and back. They also need to be able to add and subtract the conductive material in very small amounts and in the correct location. Beyond the need for fine motor skills, they need to have a good understanding of the circuit to be able to find the right tuning locations and determine how much conductive material needs to be added and subtracted from the circuits.
CAES set out to develop the Automated Robotic Tuning (ART) system to address these challenges. The ART system was designed to provide the hardware platform and software algorithms required to perform intricate and precise RF tuning adjustments. The ART platform includes an X, Y, Z motorized stage capable of positioning the various hardware elements to the desired location within a tolerance of +/-0.005 µm.
The system has a rotary stage (Figure 3) that allows for a variety of probe tips including several sizes of RF tuning probes, DC voltage probe, several sizes of conductive epoxy pin transfer/daubing tools, and RF measurement probes. The ART platform also has a height sensor capable of measuring substrate heights to within +/-0.005 µm, a laser and galvo scanner capable of ablating metal traces or printed circuit boards or thin film substrates with a kerf of 20 µm, and a high-definition camera.
A significant portion of the development of the CAES ART system has been the software algorithms utilized to perform the tuning. A wide variety of base subroutines were developed initially so the system could function well. One such base routine was the software developed to find substrate fiducials so that RF tunes could be carried out in the correct locations. Another base routine was the measurement of the substrate heights. Height is critical to ensuring that epoxy daub and laser trim results are those desired. Calibration software to align the high-definition camera, laser, probe tips and height sensor was also essential.
CAES approached the RF tuning algorithm development in two phases. During the first phase of the project, the ART system was used in a semi-automated fashion where a technician continued to make decisions about where and how much to daub or trim. This allowed a wealth of data to be gathered, including the precise location of tunes, the amount of conductive material added or subtracted and the corresponding RF response of each tune. After collecting enough data, the information was analyzed, and several RF tuning algorithms were developed and tested. The RF tuning algorithm development included evaluation and testing of several regression-based machine learning models as well as data analytic approaches. After evaluating different models and approaches, the software team was able to release a viable software RF tuning solution and enter phase two of the project—full automation.
The CAES ART system provides benefits over typical manual RF tuning processes. One advantage is that ART systematizes and automates the RF tuning process, which alleviates the need for many experienced and highly skilled technicians to produce RF modules and CCAs. These high-level RF technicians are difficult to find, creating challenges when quickly ramping up production. The ART system reduces the number of technician labor hours needed to tune RF modules and CCAs, which in turn reduces product cost.
It has been possible to achieve a cost reduction on one high-volume CAES product by 10%. Another benefit is that it results in a more consistent product with less product and performance variation versus manually tuned units. The ART system is 10 times more accurate and 30 times more consistent in placing RF tunes. The ART system also captures a tremendous amount of performance and tune configuration data as it tunes and processes units. This learning provides a path for continuous improvement to further optimize the tune process and product performance.
CAES is currently reaping the benefits of this patented system on several RF module products that are embedded in space and aerospace and defense programs.