This is a copy of N2PK's website. Paul has decided to delete his website (See N2PK-VNA on message #5, Jan 2020) from ~Feb 2020.

Paul has made availalable a ZIP file of his entire web site, which has been uploaded as is, no changes made except for this preface. <vk3pe>

Contact Paul at

This example shows the insertion loss in dB and phase in degrees vs. frequency for an 80m bandpass filter, as plotted by a VNA transmission test program. This filter was designed to reject the AM broadcast band. The filter was swept from 0.26 to 5.24 MHz in 20 kHz steps. This plot, at 20 dB per vertical division, shows that the filter has attenuation approaching 120 dB in stopband and about 0.8 dB insertion loss in passband.

This plot also shows two 'ghost' passbands below the main passband, which is approximately centered on 3.75 MHz. These two ghost passbands are centered on frequencies, 1/5 and 1/3 of the main passband. The one centered on about 0.75 MHz is about 90 dB down and the one centered on about 1.25 MHz is 82 dB down.

Bandpass Filter Insertion Loss with Ghosts

While these ghost passbands could be due to harmonic generation in the filter, that is not the case here. In this case, the ghost passbands are the spurious, i.e. undesired, result of the combination of harmonics in the VNA's two DDS RF signal sources and harmonic mixing in the VNA's detector due to overdriving the Detector's local oscillator (LO) input.

While hardware harmonic filtering of the DDS signal sources could be done, it is generally cumbersome. Instead, there is a simpler solution that simply involves a reduction of the LO drive to the Detector LO input to its linear region.

Here's the result of a re-test of the above filter, but now with a 20 dB reduction in the Detector LO drive level. Note that ghosts have disappeared into the noise floor with a displayed range of about 100 dB.

Bandpass Filter Insertion Loss without Ghosts

While one could always use the reduced LO drive level, the loss in dynamic range, here seen to be nearly 20 dB with some data averaging, is not desirable in narrowband test cases where the ghosts would not be present anyway.

Recently there have been some exciting new conceptual extensions to N2PK VNA data collection. I call one of these extensions "Harmonic Suppression" which amounts to harmonic filtering in software.

Here is a test result of the above filter with the full detector LO drive level, but now with "Harmonic Suppression" in the software. Note that ghosts have virtually disappeared into either the filter skirt or the noise floor with a displayed range of about 140 dB.

Bandpass Filter Insertion Loss with Harmonic Suppression


Please feel free to contact me with any comments or questions at: see above

Paul Kiciak, N2PK

Last updated: 10 June 2009
Copyright © 2003 - 2009