IF Band-Pass Filter Bank Switching Network for Wireless Infrastructure
Insertion loss versus frequency in the IF band must be flat to achieve optimum system performance. The ADG904-R switch is ideally suited for this circuit because it has very flat insertion loss versus frequency. The insertion loss also remains flat over the supply voltage and temperature range. Switch off isolation is another very important parameter to maximize the performance of this circuit. The ADG904-R has excellent off isolation of greater than −50 dB up to 200 MHz.
The third SAW filter (Triquint 856656) is connected to Pin 7 (RF3) of U1 and Pin 14 (RF4) of U2. The 856656 filter has a centre frequency of 140 MHz and a typical 1 dB bandwidth of 11.82 MHz.
It is important to use the PCB land layout pattern recommended by the manufacturer of the SAW filters. To achieve the maximum isolation from the input port to the output port, a plated slot was positioned under the filter. Improving this isolation enables the full out-of-band attenuation to be achieved and minimizes the in-band ripple. Each filter is matched to 50 Ω on the input port and output port using 0603 size inductor and capacitor lumped elements. This is an important aspect for achieving good performance from the circuit evaluation board. To further improve isolation, the 50 Ω transmission lines that connect the filters to the switches and to the RF edge connectors are designed as coplanar waveguides.
The standalone insertion loss vs. frequency for the 11.82 MHz
bandwidth filter is shown in Figure 3. This data was measured
with the filter on a separate circuit evaluation PCB with the
same transmission line design and matching components, but
with no switches in the RF path. This response is used as
reference data. The insertion loss at 140 MHz was −9.17 dB,
which is consistent with the data sheet value of −9.2 dB typical.
Figure 1. IF Band-Pass Filter Bank Switching Network (Simplified Schematic: All Connections and Decoupling Not Shown)
Figure 2. EVAL-CN0211-EB1Z Circuit Evaluation Board.
Figure 4 shows test results from the circuit evaluation PCB when the ADG904R switches are selecting the 856656 SAW filter. The response compares well to the standalone response, which is also included in the plot for reference. Due to the loss of the two ADG904-R switches in series, the insertion loss of the filter has increased by 1 dB. Ripple in the pass band is of importance in wireless infrastructure applications. The ripple across the band is 0.47 dB from the minimum to maximum peak within the pass band of 135.5 MHz to 144.5 MHz. This level is well within the data sheet specification for ripple, which is 0.8 dB max. The high off isolation of the switches is critical in this application to minimize leakage from one filter channel to the next. Poor switch off isolation would increase the in-band insertion loss and ripple of the filter. No evidence of this is visible in these test results.
Figure 5 and Figure 6 show the 856592 and 856684 SAW filter channel test results, respectively. The results shown include the loss from the two ADG904-R switches in series. This loss is approximately 1.0 dB. Both filter channels show good flat insertion loss, plus high out-of-band attenuation levels. The results shown are within the manufacturers data sheet tolerances and clearly show that the ADG904-R switches are not introducing any ripple, mismatch, or affecting the flatness of the filters in any way.
Figure 3. Standalone 140 MHz (Center Frequency), 11.82 MHz (1 dB BW), 856656 SAW Filter Frequency Response, No Switches
Figure 4. 140 MHz (Center Frequency), 11.82 MHz (1 dB BW), 856656 SAW Filter Frequency Response Measured on Evaluation Board with and without Switches
Figure 5. 140 MHz (Center Frequency) 20.7 MHz (1 dB BW) 856592 SAW Filter Frequency Response Measured on Evaluation Board with Switches
Figure 6. 140 MHz (Center Frequency) 16.18 MHz (1 dB BW) 856684 SAW Filter Frequency Response Measured on Evaluation Board with Switches
The circuit can be easily scaled to include more or fewer filter channels. Depending on the number of channels, other switches from the ADG9xx family may be better to use and offer more flexibility with layout. The ADG936 is a dual SPDT switch, the ADG918 is an SPDT switch, and the ADG901 is an SPST switch. All of these switches are available in both absorptive and reflective configurations.
Figure 7. Test Setup Functional Block Diagram
- Vector Network Analyzer, Hewlett Packard 8753E with Hewlett Packard 85033D 3.5 mm Calibration Kit, or equivalent.
- EVAL-CN0211-EB1Z Circuit Evaluation Board
- Power supply (2.5V)
- Coaxial RF cable with 3.5 mm type connectors
Setup and Test
The test setup is shown in Figure 7. The VNA is connected to the circuit evaluation board through 3.5 mm connector type cables. Full two port calibration of the VNA to the end of the cables will be required before testing of the circuit evaluation board. Table 1 shows the logic states to enable each individual filter channel.