Over the years I've built many filters for use in 24 GHz narrow and wideband equipment.

Below I will describe several filter styles that have been used or are still in use at my 24 GHz station.

I've always focused on constructing waveguide filters because microstrip filters are too difficult to reproduce when using PCB manufacturing techniques available to amateurs at a reasonable cost. There are several reasons for this:

Most 24 GHz stations are built using a single frequency conversion from 145 MHz to 24 GHz. This means that the Local Oscillator (LO) and the mirror frequency at the output of the transmit mixer are very close to the wanted signal, namely only 145 and 290 MHz lower. If one wants to achieve a suppression of these unwanted signals that is acceptable by modern standards, one ends up with a very lossy filter. Because of the close spacing of the wanted and unwanted signals, one needs a sufficiently large number of poles or resonators (at least 6 or 7). And due to the low Q-factor of the printed resonators, the insertion loss will be very high and can easily be equal to the gain of a GaAs-Fet amplifier stage!

It is very hard to achieve the required etching accuracy. Since printed filters consist of a number of copper traces forming individual resonators, and their spacing is basically determining the resonator coupling, it will be clear that one has far too many dimensions where things can go wrong. It is, of course, possible to produce etched filters but one has to deal with film and etching tolerances in the order of 10 or 20 µm maximum.

Combining the knowledge explained above I've also found it easier to build all other microwave building blocks in waveguide so it was a logical step to construct the filters by also using a "mechanical" process.

This brings us to the following types of filters that have been constructed and used extensively.

This filter is based on a design originally published, as far as I know, by OE9PMJ in the early 90's but presented with some mechanical modifications to improve the performance. It was, and perhaps still is, one of the most used filters for simple single conversion 24 GHz stations.

The filter is made up of two coupled "over-mode" cavity resonators. Due to the higher order mode used, the selectivity is really excellent. The response of this filter is given in figure 2.1.1. The LO at 24048 MHz is suppressed by more than 55 dB and the image is suppressed by more than 70 dB.

Making such a filter requires the use of at least a good quality drilling machine and some experience in metalworking. However, most people will be able to find someone who can do this based on the drawings in figures 2.1.2a and 2.1.2b.

I designed this filter model for a fellow OM who wanted to build his 24 GHz station using ex-equipment material that he collected. Most of this equipment was originally designed to operate above 26 GHz forcing the use of WR-28 waveguide instead of WR-42 waveguide.

Perhaps I should clarify these waveguide numbers first.

The WR waveguide standard is an old but still widely used designation that has the broad waveguide size built into its type number. A waveguide like WR-42 actually has a broad (large) inside dimension of 0.42 x 1" or 0.42 x 25.4 mm. This is equal to about 10.7 mm.

Similarly the WR-28 waveguide has an internal dimension of 0.28 x 25.4 mm or 7.11 mm.

WR-229 waveguide, which is used at 4 GHz, has an internal dimension of 2.29 x 1" or 58.2 mm.

The cut-off frequency of WR-42 (this is the lowest frequency that could possibly propagate through a given waveguide) is calculated as 300 / (2 x 10.7) or approx. 14 GHz.

For WR-28 we find 300 / (2 x 7.11) or approx. 21.1 GHz. This also shows that a waveguide size like WR-28 that is only specified for use between 26.5 GHz and 40 GHz could find use at 24 GHz.

Making use of this, I came up with a 3-pole design shown in figure 2.2.1a. What the actual filter looks like, you can see in figure 2.2.1b.

 The filter is made up of three "chambers". Sawing small slots in the waveguide and inserting small pieces of thin brass sheet in these slots makes these chambers or resonators. This can easily be done using readily available hand tools. No expensive tools are needed here. When everything is in place all joints are soldered with regular soft solder.

One can notice that a large number of M2 tapped holes are left open. The "three in a row" holes were intended to improve the matching of the filter but this proved unnecessary. This was also the case for the other two holes that are open. These could be used to modify the coupling between the resonators, but this was needed neither.

Finally, only five screws were needed to get this filter going. The middle three tune the individual resonators or filter sections and the outer two improve the coupling from the straight waveguide into the filter sections. Having access to a network analyser certainly helps to optimally tune such a filter. However, with more modest equipment and some extra patience good results can also be obtained.

The measurements are given in figure 2.2.2. The markers 1 and 2 show that the filter is properly tuned and the other markers 3 and 4 show the suppression of the LO (about -57 dB) and image (about -70 dB) frequencies. Yes, this OM was using a high side LO!

 It should also be noted that the insertion loss could be improved by silver plating this piece of brass waveguide and the tuning screws. I used stainless steel screws because I had these available at the time but this should certainly not be your first choice if insertion loss is a prime consideration. 

After my first experiments with the 2-meter IF station, I decided to make a more versatile transverter using a dual conversion frequency plan. The idea was to convert the 2-meter signal to somewhere in the 23-cm band and then convert this 23-cm signal to 1.3 cm. This would allow me to easily switch between 24048 MHz and 24192 MHz, respectively the satellite and the terrestrial segment of the band, by changing the LO that converts the 2-meter signal to 23-cm.

Also, this would enable me to produce an ATV signal at 24 GHz by simply injecting the low-power output of my 23-cm ATV transmitter but I still have to find someone to receive my signal there, hi.

Given the fact that I wanted to make something "broadband" and that the LO was some 1.2 GHz lower in frequency, I decided to make use of the cut-off properties of the waveguide. This cut-off basically behaves as a high-pass filter and the longer the piece of "cut-off waveguide", the more rejection one obtains. After some calculations I started with a piece of WR-42 waveguide that was filled with tight fitting brass strips to get the broad dimension reduced from 10.7 mm to about 6.5 mm over a length of some 50 mm. This increased the cut-off frequency from 14 GHz to slightly above 23 GHz, thereby heavily attenuating the unwanted LO and image frequencies. This high-pass filter needs only slight tuning at the input and output in order to get a good return loss or VSWR. A drawing that explains this construction is given in figure 2.3.1.

3. Conclusions

I hope this article will provide sufficient information to get some people more interested in exploring the wonderful techniques at these frequencies.

Look forward to working you on 24 GHz.

Best 73, Peter ON1BPS

UBA µWave Manager