Submission to Electronics World: Circuit Ideas. 9th February 2003

1 of 3 Leslie Green CEng MIEE

ABSOLUTE HARMONIC FILTER FOR RF

Whilst it is easy to buy harmonically pure signal generators for operation at audio

frequencies, harmonically pure RF signal generators for use at 1MHz and above do not

exist. The best I have found has harmonics at –60dBc.

The widely quoted solution to this problem is to use a low-pass filter (or notch-pass filter)

on the output of the signal generator to filter out the harmonics. However, if the filter itself

generates harmonics, which very easily happens, the generator/filter harmonics cannot

be predicted by using the measured filter attenuation characteristics and the measured

generator harmonic levels. Furthermore, UKAS accredited labs cannot calibrate

harmonic distortion at 1MHz, and neither can the UK National Physical Laboratory

(NPL).

A simple answer to this problem is to use a quarter-wave transformer as a harmonic

filter. A low-loss open-circuit transmission line will act as a simple filter with negligible

harmonic distortion of its own. Although this will not make a particularly good filter, in

terms of the amount of rejection of an individual harmonic, it does make an absolute

standard. This absolute standard can then be used to verify any low-pass filter that you

make. Beware of SPICE simulations predicting 70dB attenuation of the harmonic; in

practice you may only get 15dB attenuation due to the non-zero attenuation of the line.

The technique is to drive the signal generator through your “home-made” low-pass filter

into a wave analyser (spectrum analyser or oscilloscope with FFT capability) and view

the harmonics. The absolute filter is then shunted across the low-pass filter output. If the

harmonic amplitude seen on the wave analyser is constant, then the signal

generator/low-pass filter combination is not making a significant harmonic contribution to

the measurement.

It is to be expected that the use of the absolute filter will also give some attenuation at

the fundamental frequency. This loss needs to be measured and an equivalent pad

(resistive attenuator) used when the absolute filter is not in circuit. This additional pad is

essential to maintain a constant amplitude of the fundamental into the wave analyser.

Low-loss coax should be used for the absolute filter, since better attenuation is achieved

with low loss cable. At 1MHz, RG58 is acceptable. Figure 1 shows the length of the

transmission lines in nanoseconds for use with a 1MHz fundamental; values can be

scaled for higher frequencies so that at 10MHz (fundamental), for example, an 8.333ns

line is needed to attenuate the third harmonic. When cutting the cable to length, I

strongly suggest that you deliberately cut it too long in the first instance. Measure the

notch frequency, calculate, cut and iterate; cutting the cable down to the right length in

steps.

Submission to Electronics World: Circuit Ideas. 9th February 2003

2 of 3 Leslie Green CEng MIEE

Figure1:

Figure 1 also shows a useful in-line transmission line absolute filter which attenuates the

second, third and fourth harmonics simultaneously. This absolute filter may give enough

attenuation for your purposes, without using a separate low-pass filter at all. Of course to

test whether or not this simultaneous filter is good enough, you could use it in

conjunction with the quarter wave absolute filters discussed above.

For microwave work, the filter sections can be made in microstrip or coplanar waveguide.

The optimum line impedance in terms of output reflection coefficient for the simultaneous

harmonic filter is 115.5Ω (in a 50Ω system), giving an output VSWR of 1. Using 50Ω

lines this filter otherwise gives an output VSWR of 5.3. However, since this application

does not involve amplitude accuracy, this high level of mismatch should not cause

problems, and in any case the filter could be followed by a pad if needed.

For the final filter configuration, combining the simultaneous harmonic filter and the

quarter wave filters, the optimum line impedance for matching is 106Ω, giving an output

VSWR of 1.21.

Submission to Electronics World: Circuit Ideas. 9th February 2003

3 of 3 Leslie Green CEng MIEE

When using coax, I would not bother using anything but 50Ω cable (in a 50Ω system).

Even trying to put 75Ω coax into 50Ω plugs involves merging parts from 50Ω and 75Ω

plug kits and can be a nuisance.

Note that badly made cables can themselves introduce harmonic distortion and for RF

transmitters in particular, passive intermodulation distortion (PIM) can be a problem.

However PIM levels of -120dBc to -160dBc are achievable, so a well made cable should

be ok at -100dBc.

Published in Electronics World Mar 2004

Reformatted for PDF Apr 2012.