Ham Filter Box

2025

SDRs can easily be overloaded by RF in a noisy city which results in higher receiver noise figure and the production of spurious intermod products that can interfere with your signal of interest. This is even more of a problem if you add an amplifier outside of the SDR. To let me quickly switch between different filters for observing the amateur radio bands, I designed this filter bank. Its nothing fancy, but I like the packaging and can't find anything equivalent to it online. Nooelec has an interesting selectable filter bank with RF switches, but it isn't as rugged as I would like, and the RF switches add loss.

Ham Filter Box

Design files can be found on Github.

Theory

All of the filters in this box are SAW (Surface Acoustic Wave) filters. At a high level, SAW filters are two port RF devices which consist of two piezoelectric transducers on a common substrate. The first transducer converts the RF on the first port to mechanical energy which then travels through the substrate to the second transducer which converts the energy back to RF which then exits the device on port two. The frequency selectivity of SAW filters come from the design of the piezoelectric transducers not from a mechanical resonance of the substrate. Here`s a basic transducer design:

Basic Transducer Design from Spectrum Control (formerly APITech) White Paper

The spacing of the positive and negative contacts controls the frequency response of the filter. The principle is pretty straightforward: the positive and negative contacts of the transducer are spaced so that only waves at the desired frequency sum together constructively. This happens when the contacts are spaced \(λ/2\) apart which results in all of the negative electrodes always being 180° out of phase with all of the positive contacts. The velocity of these mechanical waves is 3-5 km/s according to this white paper. Assuming 4km/s, the wavelength of a 1 GHz mechanical wave in this medium will be:

\[\lambda = \frac{v}{f} = \frac{4 \cdot 10^3 m/s}{1 \cdot 10^9 Hz} = 4 \cdot 10^{-6} m = 4 µm\]

This means that each contact could be 1µm wide with 1µm spacing: this is readilly achievable with modern semiconductor fabrication techniques. Real SAW filters are a bit more complicated since various other structures can be added to improve bandwidth, passband flatness, insertion loss, and stopband rejection, and this is touched on in the previously linked white paper. The core principles are easy to understand, though.

Design

This filter bank is designed around a Hammond 1550A box, and all surface mount parts were installed by JLCPCB. All I had to do was solder in the SMA connectors. I copied the SMA transitions from previous projects, but it is clear that there is room for improvement at higher frequencies.

Ham Filter Box Inside

Results

Everything worked as expected aside from the 13cm band filter. The package size was too small for JLCPCB to assemble. I suspect this is due to past not transferring through the tiny aperture of their stencil I verified that the footprint was correct and briefly got one of the filters to work by pressing down on the package with tweezers, but it quickly broke. On a second version of this board, I would use a more manufacturable 13cm band filter and perhaps have two filters to cover different sections of the 23 cm band rather than two overlapping filters for the 33 cm band.

Unmanufacturably Sized F6HH2G350EH75-J 13 cm Band SAW Filter

The grid here is 10 mil, and those are 0402 passives next to the filter. While the assembly yield on this part was 0%, the yield on the slightly larger 33cm band filter was 100%:

No Issues with Larger B39921B2672P810-J 33 cm SAW Filter

Here are the plots for the three bands. Raw data can be found here.

70 cm Band TA0419A Filter Response

33 cm Band B39921B2672P810-J and TA1163A Filter Response

23 cm Band TA0862A Filter Response

The in-band responses for all of the filters generally agree with datasheet numbers. Blocking performance far from the passbands looks fine on all of the filters except for TA0419A which provides minimal attenuation above 2.3 GHz. The datasheet for this part makes no promises about its performance more than 100 MHz from the passband, so this is actually in spec. This sort of performance if something else in your system can handle frequencies far from the passband making it through the filter. For example, if a transceiver with narrowband tuning is placed after the filter, it won't matter that 2.4 GHz WiFi is making it through the filter. If this filter is placed before a wideband LNA with a low P1dB, you may experience desense near WiFi devices that could have been avoided with a better filter.

All Filters