It was found helpful to add parallel inductors, which resonate out mounting/holder capacitance, giving a more symmetrical passband.
CRYSTAL FILTER DESIGN ON LINE SERIES
It wasn't found necessary to use series tuning caps (which can partially correct for mismatched center frequency between crystals). The crystals are arranged in an extended pi network, as per usual. So, that puts the crystal impedance near 1kohm. 100 ohms, and obviously fixed for a given part). Crystals, the bandwidth depends on Zo, where a lower Zo gives a wider bandwidth, at some expense to insertion loss (because ESR is modest, ca. Note, tube amps need high impedances probably plate load is ca. Coils around 32uH (j800 ohm at 4MHz), tapped at 28%. No input filtering (tapped coil steps up from 50 ohm antenna to grid no tuning or matching really as I recall it's pretty wideband and needs an inline bandpass for better image rejection) Mixer: cathode input from LO (a separate triode oscillator), grid input from RF1 (single tuned, no matching) IF1: input from mixer (double tuned, ~200kHz? BW), output to crystal filter Crystal filter: two resonant coils, two crystals. IIRC, the receiver topology is: RF1: preamp at antenna connector. oh, some smart aleck also soldered a shield panel over top the 1st IF coil, I can't even see how it's tuned, let me get the soldering iron too then. hold on, let me take this down from the shelf, what all did I do there. I really should document that properly! The filter at least is pretty self evident. More pictures: No schematics unfortunately. I made a somewhat crude one years ago, as part of a radio receiver - seems to work very well, given the relatively poor stopband rejection from the low crystal count. The previous design added a bunch of even harmonics, only to have to filter them out.
![crystal filter design on line crystal filter design on line](https://d3i71xaburhd42.cloudfront.net/721b0ad00e2546caf45501abc7e026a737e87b5a/3-Table2-1.png)
![crystal filter design on line crystal filter design on line](https://hackaday.com/wp-content/uploads/2016/03/crystalfilterschematic.png)
Since the input clock had a clean 50% duty-cycle (virtually no even harmonics), I redesigned the thing, using a 155 MHz SAW filter and a post-amplifier/squarer. The previous designer had started with a class-C-type multiplier and then had to filter out all the even and odd harmonics. We needed to get 3x (155.52 MHz) from a 51.85 MHz clock. This reminds me of a 51 MHz tripler I re-designed. That looks very useful! Have you seen "Experimental Methods in RF Design" (an ARRL publication)? This has a pretty good section on crystal filters, more practically-oriented than mathematically rigorous.