Waveguide Filters

Hairpin printed circuit filters on the lower microwave bands are popular and Paul Wade has done a great job of explaining how to design these for the rest of us, however I tend to favour waveguide based filters of various types for bands above 13cm. This has as much to do with my building block approach, as in the difficulty of achieving the sort of tolerances required at higher frequencies using home workshop printed circuit etching techniques.  Paul’s characterisation of both Evanescent Mode Waveguide and Pipe-Cap filters is also to be commended.

I’ve had success in following the article “Design and Construction of Waveguide Bandpass Filters” by Dennis G. Sweeney WA4LPR and using his associated WGFIL program.  However for filters requiring  anything larger than WR90, size and bulk becomes an issue in portable gear.

In both plots the horizontal divisions are 50Mhz, centred on 10368Mhz and vertical divisions are 10dB. The insertion loss is less than 2dB and note return loss is in excess of 20dB. For a 144Mhz IF with low side LO injection, the attenuation of LO and image would be equal or greater than 30dB making this a very effective little filter for a transverter.

I’ve also built a 10Ghz filter based on a 3 section design using the WGFIL program by WA4LPR.  This is also fabricated in WR90 waveguide and is a post type filter using hobby brass tube for the posts.

The plots below show both the response of the filter and it’s return loss. Notice this 3 section waveguide filter shows quite steep skirts, the horizontal divisions on both plots are 50Mhz and the vertical divisions 10dB. With low side injection of the LO using a 144Mhz IF we can expect the LO and image frequencies to be some 50dB below 10368Mhz.  After adjustment the return loss is of the order of 20dB, insertion loss is around 2dB. This will vary somewhat from filter to filter based on precision and mechanical tolerances adhered to by the constructor.

Recently I built filters based on Paul’s evanscent mode designs described in the proceedings of Microwave Update 2008 for both 3.4Ghz using WR90 and 5.76Ghz using WR62 waveguide. These filters were quick and easy to construct and while I used a sweeper and analyser to tune them, I suspect they can just as easily be tuned simply looking for a peak on a power meter to give similar results. A better approach for adjusting filters is to tune for RL (Return Loss) however this is not so easy with just a power meter indication to follow.

The response of the 5.76Ghz filter is shown below together with the return loss plot.  A bit “lumpy” but perfectly useable in a 144Mhz transverter, the horizontal divisions on the plots are again 50Mhz/division and the vertical scale 10dB/division. This filter shows a reasonable sharp response, the LO and image are at least 30dB down and more and the return loss is close to 20dB, not bad for such a simple design.

24Ghz Waveguide Filter

OE9PMJ 24Ghz Filter

This is a cavity style filter that exhibits a very sharp response, less than 100Mhz at 24Ghz... as I said very sharp!

Dimensional tolerance is fairly important with this filter, a mill-drill is a handy tool although Neil VK2EI did tell me about his attempts to make this filter using copper pipes and soldering the whole thing together.  I’m not sure that’s the approach I’d want to take given I have a mill to make the component parts out of aluminium, as shown.  Neil got his pipe model to work, I didn’t, so perhaps simple is best?

My first attempt yeilded a filter resonant on 25.058Ghz, close but not close enough. The tuning screws weren’t able to pull the filter anywhwere near 24.048Ghz, so I went back to the mill to widen the cavity dimensions.

I used an idea from Andy VK2AES that replicates the cavities using a square configuration. The original OE9PMJ design describes 18mm circular cavities, however because I can’t drive an 18mm end-mill I decided to try an alternative method. Using a 10mm end mill and moving it 6.5mm in the X and Y planes yields a cavity of the same overall area as an 18mm circular cavity.  Andy VK2AES had modelled this idea using CST Studio and with mill radii of various sizes.

Anyway, the first attempt was interesting. I will now go back and re-fabricate the bottom housing with cavities and try again... I think tolerances are important in filters of this type at the frequencies we’re talking about.

Notice in the right hand photo the ‘square circular’ cavities? The input and output takes WR42 waveguide but notice that the input and output apertures are circular, also the coupling between the cavities is a circular aperture. I was unable to measure the RL of this filter but I suspect it’s reasonable given the insertion loss is around 2dB.

These filters are very interesting for higher microwave frequencies, they certainly exhibit a very narrow response which makes them useful for single conversion down to a 144Mhz IF.

I really enjoyed making this filter, even though it didn’t work first go.  It certainly helped me improve my skills with the mill and made me very aware of tolerances and measurement.  It shows great potential at frequencies above 24Ghz as a simple and effective filter structure.