Re: Coax Probes into Waveguide

Andy G4JNT

A progress update:
Thanks for all the replies.   I made a first breadboard which does what it should right from the off...

In brass waveguide, with the 17, 17.5 and 17mm cavity dimensions of the original 'JVL design, I used 4mm  holes in 0.6mm plate for the two coupling irises, which according to the design software should give about 40MHz bandwidth (when used in the original 4 iris design).   Coax probes were the centre spills from normal two hole SMA connectors, initially cut so they protruded 4mm into the guide. All PTFE was removed, spill diameter 2mm IIRC.  The probes were placed 5mm from the end plate but it was not possible to include any screws opposite the probes to vary coupling in.   Main tuning screws were copper plated M3.

I only have a tunable synth and spectrum analyser to measure with at the moment, so could only start with a quick and easy tune up for maximum smoke. (Which, after all, is how the original 'JVL ransverter used to be set up). That initially showed the probe length gave far too much coupling and one end cavity couldn't even be brought to resonance.  So they were shortened to poke about 2.5mm into the cavity.   All a bit rough and ready but at least the two probes were maintained the same length!

Result, a bandwidth rather narrower than expected.   About 25 - 30dB bandwidth at 3dB points, with a corresponding insertion loss around 3.5dB.  That IL masks any ripple in the (rather narrow) passband so it is impossible to accurately optimise the coupling without adjusting  using return loss .   

Insertion loss at 10224 (LO frequency for a 144MHz IF system) is 35dB and at the image is unmeasurable using my spectrum analyser - which isn't too sensitive at 10GHz.   At this stage I haven't even soldered the SMA connectors, they are held on with M2 screws tapped into the guide so I can remove to adjust the probe . The screws aren't even very tight as those threads will strip far too easily.  So isolation off frequency could be compromised, and that 35dB figure ought to improve when the connectors are soldered..

But nevertheless, even with that rough and ready first test, the resulting filter is more than adequate for its intended purpose - a 144MHz IF based sysytem, and is probably good enough, at least on Rx, for one with a 28MHz IF.

I'm not sure why the bandwidth was narrower than its design value - unless the WGFILTER design software isn't correct.   It is Mike 'JVL's original Basic programme modified by myself for better user input.  3 to 4dB insertion, while not terribly good is quite adequate bearing in mind, these days, gain is cheap.  It wasn't in the 1980's when the 'JVL reigned supreme for narrowband 10GHz

So all-in-all - RESULT !   And construction easy enough for even 'JNT with a general dislike of metalwork to get going.

Andy  G4JNT


Background first:
I'm taking a new look at the old G3JVL iris coupled filters in waveguide  as used in his transverter.  I have all the original design software, and a copy of MYJ where the design theory originally came from.   But I want to do it with a difference; in that the original used WG in and out, but I want SMA in/out

So, for a three cavity design, originally we had four irises;  two outer ones with large holes setting the Q or input/output transformation and two inner small irises defining the coupling, or K values.    I want to replace the outer irises by coax probes, so only two iris plates have to be installed, and the rest is drilling and tapping (about the only sort of metalwork I'm happy doing !)

So, how to design the probes for adjustable coupling.   Coax to WG transitions are a complete field in their own right, and there is no one correct answer; there are more variables than there are needed to do a complex match, so all sorts of responses can be produced.

However, for a filter, particularly a narrowband one, a  reduced coupling is going to be needed over that for a transition, so a shorter probe nearer to the backplane.   Which where it gets complicated if I don't want to make false starts in cutting metal and have to throw away bits of prepared and soldered guide.

IF I insert an SMA probe into the broad face of WG, I get a simple transition.  Now, the $24M question:  If I place a tuning screw on the opposite face of the WG so it would meet the probe, does anyone know how much it changes the coupling, or, indeed, in which direction.  Ie could I use a very short probe at a decent spacing from the backplane (say 20% of a wavelength) and rely on adjustment screws to get it right

Is there a better place for a coupling adjustment screw ?   

Or if I use an offset probe, not on the centreline of the broad face that will have reduced coupling, where then is the best place for a tuning screw ?

Anyone with experience in this area?

The  G3JVL transverter RF input filter, designed for 60MHz bandwidth, was originally targeted with a 28MHz IF in mind.  It would be nice to be able to do a 10GHz system that could feed directly into a high performance SDR radio without having to go via 144/43/1296 etc.   Even a 50MHz IF would suit many modern SDRs.

With just two iris plates to make and install instead of four, the JVL filter becomes an awful lot easier to build for those of us who aren't all that happy with metalwork

Andy  G4JNT

Join to automatically receive all group messages.