Frontend - Input Band Pass Filters - 17M

17M Band Pass Filter

Series-Trap, Shunt-C filter topology is used. 4 sections are enough to obtain more than sufficient image rejection. The filter design is based on a 0.1dB ripple Chebyshev response. The schematic is shown in the following drawing:

Because of design constraints, especially regarding IMD, the following theoretical and practical component values are chosen:

17M BPF, Fc=18000KHz, BW=1856KHz, Ql=9.8
Dipole Design Implementation Inductor
R (ohm) L (uH) C (pF) R (ohm) L (uH) C (pF) C (pF) ATC Core # turns wire (mm)
1 50 50
2 180 120 120
3 2.61 35.7 2.35 18 + 18 36 T80-10 24 0.50
4 390 390 390
5 2.61 34.7 2.35 18 + 18 36 T80-10 24 0.50
6 497 470 + 27 470 + 27
7 2.61 34.7 2.35 18 + 18 36 T80-10 24 0.50
8 390 390 390
9 2.61 35.7 2.35 18 + 18 36 T80-10 24 0.50
10 180 120 120
12 50 50

Dipole 1 and 12 represent the filters input and output impedance in the simulation and are not physically present in the real circuit. The practical capacitors are made up from combinations of standard values. The first column shows the values I used with Polystyrene capacitors. The other column shows values when ATC100-B chip capacitors are used. The inductors are fine tuned by changing the distribution of turns around the toroid.

Dipole 2 and 10, that match the filter to 50 ohms, have been reduced by about 50pF in order to compensate for the parasitic capacitance of the complete motherboard system.

In simulation when the coils are modeled with a Q=170 the filter has the following key characteristics:

  • Insertion loss: 1.40dB
  • Image rejection at 36MHz: 124dB (!)
  • IF rejection at 9MHz: 74dB
  • -3dB bandwidth: 2000KHz
  • Center frequency: 18000KHz
  • Loaded Q: 9.0

The 124dB image rejection at 36MHz is theoretical and can probably not be realized in real life. The IF rejection of 74dB at 9MHz is not sufficient by itself. However the H-Mode mixer's symmetry (55dB) and the 9MHz notch filter (80dB) will add more than enough extra attenuation.


Because of the IMD requirement and the space available on the board the following Micrometals toroid is tested: T80-10. Suflex 2.5% polystyrene capacitors are used.

The table below summarizes the filter measurement data. The 2-tones have 20KHz separation. IMD is measured at 0dBm and 5dBm input levels to detect non 3rd order law behavior. MDS is measured within 2.2KHz bandwidth:

17M BPF IMD, FSA3157 + T1-6T mixer, QT roofing filter
Core IIP3 2-tone level 0dBm IIP3 2-tone level 5dBm
(-dB) (-dB) (dBm) (dBm) (dBm) (dB) (dBm) (dBm) (dBm) (dB)
T80-10 1.48 131.5 49.3 50.1 49.1 120.4 50.3 50.1 48.5 120.2

The table requires some explanation. For each core the insertion loss of the resulting BPF and the MDS figure for the complete frontend with that BPF is given. Next, IMD results at 0dBm input tones and 5 dBm input tones are given. The columns marked 'BPF' show the IIP3 of the BPF only. The columns marked '-BPF' show the IIP3 of the frontend without the BPF in front and the columns marked '+BPF' show the IIP3 of the complete frontend including the BPF. Also the IMD3 dynamic range is computed and shown.

A number of additional remarks to the measurements:

  • All measurements are done with the complete BPF motherboard with 10 BPF filters, notch filter and the attenuator board mounted. So the IL and MDS figures include the small but measurable additional losses (about 0.2dB) introduced by the attenuator board (4 relays), the 9MHz notch and the 2 relays on the measured BPF board.

  • On 17M the additional loss in the IF notch filter can be neglected.

  • On 17M the T80-10 matches the expectations and the performance of the mixer!

  • The obtained inductor Q with the tested toroids seems to be around 170. Silver plated 0,50mm enameled wire is used.

17M BPF with T80-10

A picture of the assembled filter PCB is shown below. It measures 160x50mm, a half height eurocard. The assembly is meant to be plugged onto a motherboard with 9 other similar BPF's and a 9MHz notch filter board:

Measurement overview

The following table summarizes some measurements made to the T80-10 17M BPF:

17M BPF results, FSA3157 + T1-6T mixer, QT roofing filter
BPF Filter Center 18000 KHz
BPF -3dB Band Width 1856 KHz
BPF Loaded Q 9.8
BPF Insertion loss -1.48 dB
MDS level without BPF -133,5 dBm
MDS level with BPF -131,5 dBm
IP3 frontend without BPF @ 0dBm +50.1 dBm
IP3 of BPF only @ 0dBm +49.3 dBm
IP3 of frontend with BPF @ 0dBm +49.1 dBm
IMD3 dynamic range @ 0dBm 120.4 dB
Image rejection @ 36MHz -113 dB
BPF box IF rejection @ 9MHz @ 0dBm -116 dB

Single tone spurious responses

The following table shows the relevant single tone spurious responses of the H-Mode mixer frontend with the 17M BPF. Spurious signals are determined by the following equation: | n * Fo m * Fs | = Fif, where n = 0, 1, 2, 3, 4, 5 and m = 1, 2, 3, 4, 5. The number of possible spurs up to 5'th order is 55, but only those that fall within the -40dB pass band of the BPF are investigated. Higher order spurs and spurs outside the -40dB pass band are considered insignificant and have not been measured. The spur test signal is a very strong 0dBm (S9+73dB!) signal.

17M BPF + frontend single-tone spurious responses at 18.118.000Hz, @ 0dBm input level
Description Order: (n,m) F-spur (Hz) level (dBm)
IF (0,1) 9.000.000 < -127
Image (1,1) 36.118.000 -113
(1,2) 18.059.000 -78
(2,4) 15.809.000 < -127
(3,4) 18.088.500 < -105
(3,5) 18.070.800 < -105
(4,5) 19.894.400 < -124

The cyan colored measurements are limited due to the phase noise levels of the signal generator and / or LO. The given values are the best that could be done.

The (1,2) spur at 18.059.000 when listening at 18.118.000 with -78dBm (S8) is the worst spur in the entire frontend. Reason for this is that it is 2nd order spur (strong!) and that 18MHz (2xIF trouble spot!) is very close to the 17M band and cannot be eliminated with the BPF on 17M. In practice it is not that bad. Lowering the test tone to -26dBm (S9+47dB) puts the spur already below the noise floor. The (3,4) and (3,5) spurs also cannot be filtered out by the BPF, but are of much higher order and the mixer has no problem with them!


The next analyzer plot shows the pass band characteristics of the T80-10 17M filter. IL is around 1.48dB over the entire 17M band at a -3dB bandwidth of 1856KHz. The filter is tweaked such that it is not aligned for maximum flatness in the pass band but lightly peaks at the narrow 17M amateur band. The value of the IL indicates that the Q's of the inductors are close to about 170:

The following analyzer plot shows the wideband insertion loss up to the image frequency at 36MHz. At 9MHz the attenuation is about the predicted 74dB. The drastic dip of the notch filter is clearly visible. The stop band is close to the analyzers available 100dB range

The 9MHz IF rejection is measured at < -130dB (not measurable), with the 9MHz notch in place. The image rejection is measured at -113dB.

The following analyzer picture is a wideband plot up to 100MHz. Stop band reduces abruptly beyond 80MHz. Probable causes are self resonance of the big toroid inductors and / or resonance of the entire filter slot as a cavity resonator. With a pure enough LO signal this reduced stop band performance at those frequencies should not be of much concern.


Finally the impedance of the filter is measured with N2PK-VNA and Exiter software:

The measured curve almost overlaps the curve when simulating the theoretical filter and shows a good match to a 50 ohm system.

PCB artwork and schematic for the 4-pole filter board in PDF format.

160M BPF









Back to BPF Boards

Back to the TOC