Comments? email to
These
filters are nothing new, but they do offer performance
comparable to "tin-box" filters, and
except for being bigger, stack up pretty well with those expensive little
helical jobs. You just have to be able to make your own PC boards...
You'll have to put up with a little insertion loss; how much will depend on how sharp you want the filter to be, and the material you make it from. For example, a properly designed 5-loop filter on FR4 (middle) has about 6db loss; the same filter on Teflon/glass (left) is about 2.5 db. The last filter on the right is made from Rogers 3006, chosen for it's low velocity factor and loss, allowing it to be more compact with losses comparable to Teflon. In most applications, this is not a problem, as these losses can easily be made up with MMIC amplifiers.
These filters also have a peak at 1/2 the design frequency. This peak happens due to the resonators being 1/4 wavelength at that frequency, and the peak is only about 40 db down from the main pass-band response. However, this response is quite narrow (the filter is seriously under-coupled at that frequency); another response will happen at 2x the design frequency...the filter resonators are 1 wavelength there. Just be aware that these responses are there, and design the rest of your circuits accordingly.
Below is an LO chain using two 1080 MHz filters; the first was done on PTFE, the second on FR4. Note the use of a component 540 MHz notch filter at the input to the first band-pass filter, and the exceptionally clean signal after the second one at 1080 MHz.
I was able to fit both of these filters (and their MMIC amplifiers) into the small enclosure shown (right). The first multiplier/filter board (below) is back-to-back with the second one (band-pass amplifier) inside the box, making a very compact unit overall.
If one were to need a higher output level (+17 dbm for example), selection of different MMIC's is all that is necessary; the ones shown were selected to provide just +7 dbm, the correct drive level for the transverter it was designed for.
To the right is the second filter/amplifier board, done on .031 FR4.
For other frequencies, the length of the resonators is adjusted accordingly. Slight changes in coupling may also be necessary;
Some additional background information, and artwork for these filters is posted below.
As time permits, I'll post additional artwork for several of
the more commonly used frequencies.
Here's an example of one of the 1080 MHz filters experimented with. The scope image is a linear display of the pass-band response, swept from 880 to 1280 MHz. There is a slightly flattened peak with steep slopes on either side, indicating optimum coupling. This was achieved with careful adjustment of the spacing between the filter elements.
Getting the spacing too close (bottom left) or too wide (bottom right) will over-couple or under-couple the filter, producing double-peaks, broader or narrower responses, and higher Insertion losses than necessary.
The Q of the PC board material also affects the performance of
the filter; generally, the more lossy material (FR4) produces a wider filter
with broader skirts than the same filter on low-loss material like Ro3006.
The
final copy of this 1080 MHz filter on FR4 produced the data plot to the right.
Artwork, in .pdf format for this filter is here...you'll need to scale it as
shown:
 |
 |
| If you'd like to try this on Rogers 3006 (right), here's the
artwork for that one:
|
 |
This one was centered for 1152 MHz on .031 PTFE, and
is +- 1db from 1115 to 1175 MHz
 |
 |
This is on Ro3006, and is centered at 1268 MHz; it
was designed for L.O. service at 1268, but is wide enough to cover most of the
23cm band before dropping off rapidly. At 1296, it's only down an additional
half a db from center.
 |
 |
Here's another 1268 MHz filter, but on FR4. This one
has higher insertion loss, and is a bit broader; +- 1db from
1220-1300MHz, and +-3db 1200-1320MHz before falling off.
 |
 |
| This one (on Ro3006) is optimized for 1296, yet will
still work well for L.O. service at 1268.
|
 |
You're going to laugh when you see this next picture
(below)...these are some of the various cut-and-try attempts that helped me
learn what makes these things tick.
