Silicon Labs provides a range of attractive receiver and transceiver chips in a QFN20 package for telemetry applications in the sub-GHz range with its EZRadioPro series. Additionally, complete modules equipped with this IC can be found for a low cost, notably on platforms like AliExpress. In the latter case, these modules typically feature the Si4463 or the somewhat outdated Si4432. For home automation applications operating at 433 or 866 MHz, I usually opt for the comparable RFM69 from HopeRF. However, what sets the Si4463 apart is its vast continuous frequency range (142-1050 MHz), including coverage of the VHF band, and the fact that it includes the maritime AIS channels (161.975 and 162.025 MHz).
The examples in the AN643 Application Note demonstrate how to construct an LC-Balun with 3 or 4 components to match the differential input of the receiver to 50 ohms. Because I found the receiver sensitivity disappointing in practice, I became increasingly suspicious of this network. Experimenting with fixed capacitors and inductors in a 603 footprint proved challenging. Instead, I replicated the 169 MHz example from the application note in LTspice to assess the criticality of the setup. This article addresses the potential cause of the poor reception sensitivity and attempts to provide some guidance towards a potential solution.
Above is the recommended matching network from the Application Note for 169 MHz. At this frequency, the nominal input impedance of the receiver's differential input is 536 ohms, with a parallel capacitance of 0.98 pF. The network is an LC balun with two tasks: first, impedance transformation from 50 ohms to 536 ohms, and second, conversion from single-ended to differential. Alongside the schematic is a Bode plot of the sum and difference of both input signals. It becomes immediately clear that achieving ±90 degrees phase shift and precise impedance transformation is particularly critical. While the example aligns nicely at 169 MHz, using the nearest E12 values (5.6 pF, 12 pF, 150 nH, and 220 nH) already poses challenges. But that won't be the only problem encountered!
In the Bode plot above, it can be observed that the two separate input signals at 169 MHz are perfectly out of phase with each other, and the voltage gain due to impedance transformation is approximately 2 * 4.3 = 8.6 dB. This plot ranges from 160 to 180 MHz. However, the situation looks quite different when taking a broader view.
Still the same circuit, but now depicted from 50 to 200 MHz. What becomes visible now is a huge peak around 90 MHz, where both output signals after transformation are approximately 17.5 dB and 21 dB above the input signal, respectively. This coincides precisely with the FM broadcast band. Although the two signals are in phase with each other in this range and are expected to cancel each other out, the common mode rejection of the differential amplifier in the SI4463 is not specified. Nevertheless, little to nothing will remain of it once we enter the regions where blocking and intermodulation occur.
To see what I can find in the ether at my location, I connected the AIS antenna (a simple vertical dipole) directly to the spectrum analyzer.
Only the strongest AIS signals peek above the noise on the set bandwidth, but a local radio station at 90 MHz sits approximately 50 dB above, and a DAB+ signal at 182 MHz is about 35 dB stronger. The unfortunate behavior of the LC balun makes the receiver particularly vulnerable to strong signals, especially in an FM band where most of the danger lies. Adding a preamplifier will only exacerbate the situation. It seems that the receiver is being completely overwhelmed, while the vast majority of AIS stations are considerably weaker than the few visible on the spectrum analyzer. Two possible solutions are: 1. Additional pre-selection and 2. A better alternative to the LC balun. Both are likely necessary. Silicon Labs' recommendation for an LC balun is probably motivated by the desire to provide a simple / uniform / reproducible solution across the entire frequency range, particularly for the myriad of applications in the 433 and 868 MHz bands. The VHF range is somewhat of an outlier in this regard. For this frequency, it is much more appropriate to opt for a transformer on a ferrite core. The impedance transformation is roughly 1:10. With a winding ratio of 1:3.3 (or 1:1.66 + 1.66), we can go a long way. The 1 pF input capacitance is not dramatic and can be compensated for if necessary.
To investigate this, I conducted a simulation using a 1:9 impedance transformer, the Coilcraft WBC9-L_. I chose this transformer because the inductances were available in a datasheet, and I couldn't find Spice models for similar transformers from Murata, Amidon, etc.
This all looks a lot more soothing. 180 degrees phase shift and approximately 2 x 6.5 dB voltage gain over a large range.
Indeed, this may provide a somewhat idealized view, but at least we have addressed the resonance phenomena in the FM band and the critical component selection. The next step is to test it in practice.
As I desire to simultaneously receive the two AIS channels, I have opted to include a Splitter/combiner in the design right away. Typically, this consists of two transformers. The first is an autotransformer that halves the input impedance. The second is bifilar wound, doubling the impedance for each port. Since we aim towards 500 Ω, we can omit the first transformer. The 50 Ω at the input will split into 2 ports of 100 Ω after the splitter. That's a step in the right direction. An uncomplicated trifilar transformer will then bring it to 2 * 200 = 400 Ω for each AIS receiver. The idea is to pass both secondary windings through one of the holes of a double aperture core (pig's nose) once again. With this additional half-winding, we will come reasonably close to the ultimate goal of 532 Ω. The test setup looks like this.
To measure the most critical part, I placed a ferrite core between two "matching pi attenuators". At the input, there is a 20 dB attenuator with an input impedance of 50 Ω and an output impedance of 100 Ω. Behind the transformer, there is a second matching pi-network with an input impedance of 266 Ω and an output impedance of 50 Ω, now at the cost of an additional 10 dB attenuation. The unused secondary winding of the transformer is simply terminated with a resistor of 266 Ω.
The used calculator can be found at href="/https : //chemandy.com/calculators/matching-pi-attenuator-calculator.htm"
A first measurement was conducted with a trifilar winding of 2:2.5+2.5 turns (d=0.2 mm) on a FairRite 2643000301 ferrite core 35 x 1.6 x 6 mm #43 material on the board below. The impedance ratio should then be approximately 1:5, resulting in 500 Ω.
The outcome of the measurement
The measured attenuation at 162 MHz is 34.25 - 30 = 4.25 dB. Since half of the power is dissipated in the dummy resistance of 266 Ω but will ultimately be utilized by the receiver, the actual loss is 1.25 dB, which seems quite reasonable to me.
The proposed complete schematic for a dual-channel AIS receiver is as follows:
Below is the provisional assembly of the splitter and input transformers on an AIS receiver instead of the LC baluns for which the board was designed. This time executed with miniature double aperture cores, in #61 material from FairRite Part no. 2861002302. The measurement results were nearly identical.
Additionally, an experiment was conducted with an AliExpress module equipped with a transformer balun.
Preliminary Findings
Additional preselection with an LC or SAW bandpass filter remains necessary and I may revisit this later, but the transformer balun already provides a 20 dB advantage in preventing blocking from the FM band. Moreover, it is easier to implement than the LC variant with capacitors and coils with exotic values. These are not only difficult to obtain (in reasonable quality for a 603 footprint) but there is also nothing left to verify or adjust once they are on the board.
It also appears that careful layout and complete shielding and decoupling are of paramount importance. Due to its high input impedance, the Si4463 is extremely sensitive to any form of noise and interference on the input circuit. See Silabs' recommendations in this regard.
Extra amplification must be used sparingly. In addition to the 3 dB loss due to the splitter, there may be a 2 dB loss due to transformation. Any additional gain to compensate for this loss only jeopardizes the moderate signal strength characteristics of the receiver chip.