|The Microstrip Breadboard ™
ConstructionThe construction of the board is totally stupid, it's just two coaxial cables with BNC plugs screwed to a thin (3mm) wood (or whatever this is?) plate, the backside of the plate being covered with tinfoil. The cables are mounted with screws, because soldering tinfoil sucks quite badly. The ground leads of the cables are connected to the backside of the board (or rather, the tinfoil there). The top side has two thin copper sheets connected to the signal wires of the cables, which can be used as ports for any structure you build onto the board.
For connecting two tinfoil snipptes, I use (isolated) paper clips. This works pretty well, an alternative solution would possibly be using pins or nails. The main problem with paper clips is that the tracks don't stay flat on the board, but instead tend to hover a bit, which screws up their impedance. If it gets too bad you can put heavy, non-conductive objects on top of them, or just tape them to the board temporarily.
For measurement, a spectrum analyzer with tracking generator is being used. A vector network analyzer would be optimal for this purpose, but this is just fine too. The analyzer works for frequencies up to 1.5 GHz, and let's be honest -- you don't want to construct anything from tinfoil and tape for more than 1.5 GHz anyways ;)
The tracking generator output is connected to the left side of the board, and the spectrum analyzer input is connected to the right side. I'll nickname the left port "1" and the right one "2", thus "S21" will mean "the part of the signal that went from port 1 to port 2".
Transmission linesThe width of a 50 Ohm line on my board (3mm thickness, wood) is about 1cm (probably a bit more, but I don't know the dielectric constant of the wood accurately enough). To calculate that, you can use any program which has a microstrip calculator, such as qucs. You can also calculate it by hand, but why would you?
Here's what you get when you connect a 30cm length 1cm width line to the two ports of my board:
|S21 of a 30cm length, 50 Ohms line on the breadboard.
You can consider calibrating your analyzer to treat this loss as zero loss. I mostly chose not to, but for some structures it might be a good idea.
Building filtersNext, I built some filters. Note that altough I comment on how well those filters perform, this is just describing the experiment, not trying to rank the design ideas or anything. I'm well aware that a tinfoil implementation of the filters is not enough to judge the designs in any way.
First is the simplest design from the first image (top of the post). It's just five quarter-wavelength stubs connected to the line. The expected result is a band-stop filter.
|Screenshot of a quick simulation of the filter measured in the plot above, explaining the drop of S21 close to 1.5GHz. The simulation only has 3 stubs, but adding more does not change the location of the peaks, just their width and depth.
Next, a bandpass filter made of some parallel lines. The length of the lines is about 12 cm, so the pass frequency should be somewhere around 900 MHz.
|900MHz Bandpass filter made from some parallel transmission lines.
|Frequency response of the filter shown above
Next, a lowpass filter made from two "inductors" and a "capacitor":
|900 MHz lowpass filter
|Frequency response of the lowpass filter shown above
Next is another bandpass using a resonant patch between two feed lines:
|Another bandpass filter designed for 1GHz frequency
|Frequency response of the 1GHz filter shown above
And one thing must not be missing, of course... the magic arcs:
|"Magic arc" bandstop filter, with three arcs for a "sharper" response pattern
|Frequency response for the arc filter
Directional CouplerOkay, what else could we do with this thing... how about a directional coupler? Here you go:
|S41 of the coupler, aka what arrives at the coupled port. This plot belongs to the first of the above images.
|S31 of the coupler, aka what arrives at the isolated port. This one should optimally by -infinity everywhere, but obviously isn't. This plot belongs to the second of the above images.
It is worth noting that when removing the 50 ohm termination from port 2 (the right one of the upper line in the picture, where the main part of the signal should go out again), then S31 looks quite exactly like S41 (sorry I don't have a picture). This is correct, since everything is reflected from port 2 and then coupled into port 3.
That's it for now, if I come across some more microstrip stuff which is nice, I'll write about it.