by Kyle Chang
Introduction
Marki’s MMIC Filters offer excellent repeatability, smaller form-factors, and higher operating frequencies with rapid development, reasonable prices, and first-pass design success. As noted in Doug Jorgesen and Christopher Marki’s technical article, MMIC Filters’ Time Has Come, GaAs MMIC filters offer the best combination of performance, cost, and consistency – recent testing further indicates that these filters maintain consistency with low variability over temperature and high power handling capabilities.
Figure 1. Sample of 24 typical MMIC filter from a lot of > 8000, showing fabrication repeatability.
We’re often asked, “can these filters handle at least 1W?”, but we’ll do you one better and find the max power handling for our current MMIC filters. This experiment explores the MFBA and MFBC filter families to demonstrate how extreme power and temperature conditions affect different filter topologies.
Power Handling Test Set Up
Power handling tests are conducted by applying increasing levels of RF power and measuring the s-parameters of the device after each power level. The test configuration is seen below in Figure 2.
Figure 2. Test system for RF power handling tests.
A signal generator inputs a continuous wave into a pre-amplifier to drive the power amplifier. This is followed by an isolator, which prevents reflected power from damaging the amplifiers if the filter fails as an open circuit. An input power meter branches off the thru-line using a directional coupler to monitor the power going into the DUT while an output power meter with a 30dB attenuator monitors the output power after the DUT. The difference between the measured powers is read as the power dissipated by the DUT, or the insertion loss.
Starting with 0.5W input, the power is increased by 0.5W increments, held for 10 seconds, and powered off to remove the DUT and take s-parameters on the VNA. We then resume at the last highest power input to the device and increase the power by 0.5W, repeating up to 10W unless we observe performance degradation.
Power Handling Test Results
The first filter in our test lineup is the MFBA-00004PSM (8.4-12.5GHz, 8th order BPF in QFN package) which exhibits excellent stop band suppression and roll-off. Figure 3 shows typical performance for EVB-MFBA-00004P with a +5dBm input power. A 10GHz, 1W signal is applied through the power handling test bench and held for 10 seconds.
Figure 3. S-parameters of MFBA-00004PSM with +5dBm input power.
Figure 4. Insertion loss measured after 1W input.
As depicted in Figure 4, the filter shows zero degradation with a 1W input. Now that we’ve confirmed the device can handle a 1W input signal, we’ll keep pushing the power to see when it breaks.
Figure 5. Insertion loss measured after 10W input.
VNA measurements (Figure 5) confirms that after experiencing a 10W, 10GHz CW signal for 10 seconds, the filter performs with zero degradation to insertion loss.
The next line of filters in our investigation are the MFBC bandpass filters, specifically the MFBC-00001PSM (7.4-9.9GHz, 5th order BPF).
Figure 6. Insertion loss after 1W input.
Figure 6 shows insertion loss with a nominal +5dBm input and the insertion loss after 1W was input to the device. As expected, there is zero insertion loss degradation within the filter’s passband. Considering the power handling and complex topology of the MFBA, we expect the MFBCs to handle better than the MFBAs.
At 10W input, the filter did not catastrophically fail and shows approximately 3dB insertion loss (Figure 7). Additional VNA measurements (Figure 8) confirm there is zero performance degradation on the MFBC.
Figure 7. Input and output power readings at 10W input to DUT.
Figure 8. Insertion loss after 10W input.
Compared to the MFBA filters, the MFBC topologies have larger elements and wider gaps, allowing improved power handling capabilities and fewer points of failure. However, test results show that both filter families are capable of handling up to 10W CW.
Dissipating over 5W is no small feat for the 5x5mm QFN filter – the evaluation board was warm to the touch, so our next question is, “how well do MMIC filters perform when things get toasty?”
How Does Temperature Affect Performance?
We specify the QFN packaged MMIC filters to operate between -55°C and +100°C, so let’s find out well these operate at the temperature extremes. S-parameters are taken actively at each temperature setpoint starting with +25°C and increasing to +100°C, then resetting to 25°C and decreasing to -55°C. Each temperature is held for at least one minute before data is taken.
We first investigate the MFBC-00007PSM on its evaluation fixture: Figure 9 shows s-parameters taken at laboratory temperature while Figure 10 demonstrates insertion loss over varying temperatures.
Figure 9. S-parameters of MFBC-00007PSM at laboratory temperature.
Figure 10. MFBC-00007PSM insertion loss over temperature.
Figure 10 shows insertion loss varies up to 2dB across the passband between the two temperature extremes. Increasing insertion loss with increasing temperature is expected since resistance (especially in the filter’s inductive elements) increases due to heat. Let’s take a closer look at how the -1dBc and the -40dBc points vary over temperature:
Figure 11. MFBC-00007PSM temperature variability at -1dBc and -40dBc points.
Figure 12. Percent change in -1dBc and -40dBc points over temperature, normalized to 25°C.
At increased temperatures, the filter’s insertion loss exhibits red shifting, shown in Figure 11. Figure 12 shows the percent change for the -1dBc and -40dBc points over temperature – the cutoff and rejection points decrease in frequency with increased temperature.
Post-test measurements at room temperature also confirm the part did not see irreversible change due to extreme temperatures:
Figure 13. MFBC-00007PSM insertion loss before and after temperature testing.
Additional temperature testing was conducted on MFBA-00002PSM shown in Figure 14 below.
Figure 14. MFBA-00002PSM insertion loss over temperature.
The MFBA filter also shows approximately 2dB variation and red shifting over a varied temperature range. Both filters observe increased insertion loss with temperature increase due to the same phenomena seen in the MFBC filter. The bandwidth also decreases slightly as temperature increases – the quality factor deteriorates from increased resistivity in the inductive elements. The ripples in the passband and the suck-out at 20GHz are likely due to the test fixture, so we will run temperature tests on the bare die in the future – be on the lookout for our next installment!
As an added bonus, we temp tested the very MFBC-00001PSM used in the power handling test. Overall, the MFBC-00001PSM is a robust filter, capable of handling at least 10W input power and varying less than 1dB in insertion loss over its maximum operating temperatures:
Figure 15. MFBC-00001PSM insertion loss over temperature.
Conclusion
Recent power handling and temperature testing on Marki’s MMIC filters delivered promising high-frequency, high-power designs using our refined GaAs MMIC design flow. Our MFBA and MFBC filter families show zero performance degradation up to 10W input power. Temperature testing also demonstrates <2.5dB variation in insertion loss across operating temperatures with excellent insertion loss even when things get hot.
Marki’s MMIC filter portfolio is constantly expanding with low-pass, high-pass, and notch filters soon to follow our current band-pass filters and diplexers. Our optimized design flow offers quick-turn, high-volume, and consistent filter designs to meet all your filter needs. We’re open to custom designs, so feel free to reach out to Marki’s support team at [email protected] to discover or design your next filter!
