Click here to go to our main page on filters
Click here to go to our reflectionless filter page
New for December 2019. This trick has been around a while, but it took an email from Gray to to remind us that it needs its own page. Here is Grey's comments
Another type absorptive filter that sees considerable use as a basic component in multi-kilowatt power combiners at VHF and UHF is the constant Constant Impedance (or sometimes called Balanced) Combiner Module. (These aren used for combining the outputs of multiple amplifiers operating at different frequencies into a common output.)
It consists of back-to-back 3 dB, 90° hybrids with a pair of identically tuned BPF's interposed between. The isolated port on the input hybrid is connected to a dummy load, into which any out-of-band signal is absorbed as will as any signal resulting from BPF tuning imbalance.
The isolated port on the output hybrid can be terminated. Any out-of-band signal ingressing from the output port will be absorbed here, thus affording increased isolation to the source. Or it can be used as a wideband input to combine another signal into the output, as long as it is out of the passband of the module BPF’s.
Thanks! we'll add some analysis below.
Filters are usually reactive. That means that in their passband they are well matched, but in rejection band(s) they have atrocious reflection coefficients. If you shove a pair of filters between quadrature couplers, you can reduce the reflection coefficient in the rejection band(s).. There are some considerations: the quad coupler needs to have more bandwidth than the filters.
Below is an ideal Butterworth bandpass filter centered at 10 GHz. It has five poles and bandwidth 0.8-1.2 GHz. Outside the passband the reflection coefficient (S11) rapidly trends toward 0 dB... In some cases, this can cause trouble in other parts of your design. For example, in a receiver you might have a conditionally stable LNA that could start oscillating if you put this filter in front or after it.
Below, the filter is placed between a pair of ideal 3dB quadrature couplers. You can learn about coupled line couplers here. The quality of the terminations determines how well the reflection coefficient is improved. In our case they are ideal, in your case, they won't be. Also, the coupling coefficient have a big say in the bandwidth...
When we say ideal couplers, we mean in. Below is a plot of the coupler's response, the two outputs are equal at the center frequency.
Below, the response of the balanced filter is shown. If you compare it to the original filter you will see that S11 may not be a perfect match outside the band, but at least it is no longer 0 dB return loss. Your conditional stable amplifier will stay quiet and happy.
What happens when you tweak the coupler to be over-coupled or under coupled? Let's throw in a sweep variable:
Now consider the return loss... you can get a lot more bandwidth for a given return loss: you can get 20 dB over an octave! But there is a cost. Then you deviate from perfect 3dB, you will add loss (see S21 in the plot, scale on the right). If your coupling coefficient is 2 dB (over coupled) you will loses 0.3 dB. If your coupling coeffiicient is 5 dB, you lose 0.6 dB. There is no free lunch!
In any case, the balance filter is one of many tricks in the ol' microwave toolbox. Especially in cases you don't care about a little extra loss.