skip to content W3C SVG

Medium Wave Radio - Band Pass Filter

The medium wave radio band runs from 531–1,602 kHz with 4.5kHz bandwidth in Europe, and 530–1,700 kHz with 5kHz bandwidth elsewhere. A bandpass filter should only pass 500kHz – 2MHz approximately for good front-end filtering.

A filter was calculated on rf-tools.com using standard component values and the following parameters:

  • 5th Order Chebyshev Bandpass
  • Conventional, Series First
  • Lower Cutoff Freq. = 500 KHz
  • Upper Cutoff Freq. = 2 MHz
  • Passband Ripple = 0.10 dB

The filter schematic diagram

The filter schematic diagram
The filter schematic diagram

Filter parts list

2 x 4.7nF  3 x 2.2 nF 2 x 5.6 µH 2 x 8.2 µH 1 x 10 µH

Simulation

The filter was simulated in QUCS with these parameters and equations:

S-parameter simulation:

  • Logarithmic
  • Start : 25kHz
  • Stop : 40 MHz
  • Points : 1001

Equations:

  • dBS21 = dB(S[2,1])
  • dBS11 = dB(S[1,1])
  • Group Delay = -diff(unwrap(angle(S[2,1])),2*pi*frequency)

S21 Plot of the filter

Plotting S21 vs frequency gives this very nice filter shape. I suspect the actual filter I build will be nowhere near as good as this, and definitely not reach >100dB of attenuation, but it’ll be interesting to measure.

S21 Plot of the filter
S21 Plot of the filter

Part selection for the filter isn’t critical, as the frequencies involved are low, and the wanted bandwidth is fairly wide. I built a prototype on copper-clad board using polyester film capacitors and some miniature axial inductors. The inductors were a bit of an unknown, but they seem to work well in this application.

The filter constructed on some copper clad board. Input is to the left, output is on the bottom of the board
The filter constructed on some copper clad board. Input is to the left, output is on the bottom of the board

The filter prototype was measured using a nanoVNA and the data captured with the vna-saver software. Unfortunately, the vna-saver only saves in Touchstone files, whereas the QUCS simulation is magnitude vs frequency CSV. I haven’t found a way to overlay the two yet, to directly compare simulation and realisation.

The filter as measured with the nanoVNA
The filter as measured with the nanoVNA

The prototype on the copper board was too fragile for actual use, so it was re-built on some protoboard. It re-measures on the VNA as essentially identical to the copper board version - there’s not too much effect from all the distributed capacitance of the protoboard.

Filter rebuilt on protoboard. Input is top right, ground is the three central pins and output is the bottom right pin.
Filter rebuilt on protoboard. Input is top right, ground is the three central pins and output is the bottom right pin.

Filter plots.svg
Filter plots.svg

The filter has degraded a bit by being built on the protoboard, but it still more than good enough for the purpose