You have this insatiable urge to homebrew a SSB/CW transceiver and I mean totally homebrew! This rig will have no IC's. (Gulp) an Analog VFO, a homebrew microphone and even a home constructed case. The 12 pole QER Crystal Filter, of course homebrewed, is suddenly a decision point. No, the decision is not about 12 poles but what frequency? Did I mention this will be your first ever homebrew project.
Of all of these pieces indeed the IF Filter Frequency is perhaps the 1st item on the critical path. The McCoy Silver Sentinel or Golden Guardian Filters were some of the 1st commercial 9 MHz Crystal Filters. The 9 MHz filter was popular in the single conversion 80/20 rigs because a 5MHz VFO yields two bands, although you would need two BFO crystals.
Had you used a 5 MHz Filter and 9 MHz VFO then because of sideband inversion only one BFO crystal is needed to give USB on 20M and LSB on 80M.
The 9 MHz filter is seen a lot in homebrew and commercial radios (Ten Tec). This all worked pretty cool until new bands were added like 17M where harmonics of BFO's and VFO's operating at 9MHz are difficult to filter out of Band Pass and Low Pass Filters. Ten Tec operated their LO at 27 MHz in the Corsair I with hopes of skirting that problem on 17M.
Then there are specific filter frequencies that create major issues. The Collins 455 kHz Mechanical Filter in a single conversion frequency design is OK at 80 and 40 Meters but beyond that you have imaging problems. The Collins filter in a Dual Conversion design skirts that issue.
In my KWM-4 the 1st Conversion is to 10.7 MHz where I used an FM Narrow Band Crystal Filter as a "roofing filter" and a second conversion stage at 10.245 MHz easily up converts from 455 kHz and likewise down converts all signals to 455 kHz in a simple ADE-1 Double Balanced Mixer with the 10.245 MHz Crystal LO.
The FM Roofing Filter resolves one of the products out of the ADE-1. We of course have 10.245 + 0.455 results in 10.7MHz which will pass through the FM Filter But the other product 10.245 - 0.455 yields 9.79 MHz is outside the Filter Pass Band and is stopped. I mention this because you must look at all possible mixing schemes to avoid issues such as other mixing products -- do this exercise with a single conversion 455 kHz filter at 15M.
At the other end are filters in the 25 MHz range. Friend N2COR has operating a 10/15 Meter rig using a low frequency VFO in the 3.5 MHz range to give two bands. This has been highly successful for Bill. But he does note because of the high frequency of the filter the normal flat pass band of the crystal filter is somewhat rounded. While not a showstopper it is carryon baggage with higher frequency filters.
Many homebrew crystal filter proponents seem to focus on filters in the 4 to 12 MHz range. There may be a lot of simple factors for choice of filter frequencies in this range. Cost may be a huge driver since to build a 12 Pole Filter you need to have a pool of about 50 to 100 crystals where you can find 12 whose total frequency spread is no more than 50 Hertz. While computer crystals in this range are cheap enough -- 100 crystals will still cost you about $50.
But lurking (like the lurkers on hackaday) in the background are parameters such as frequency stability in Parts Per Million. Many of the cheap crystals have a 50PPM stability. At 1 MHz that means 50 Hz stability. But at 12 MHz that means 600 Hz and at 24 MHz is 1.2kHz. Now take that across 12 Crystals and that is a big number. Lower frequencies mitigate the issue of stability.
Some frequencies you might want to avoid are filters at 5 and 10MHz. Unless you provide in the circuitry a lot of shielding, inevitably WWV will always seem to be in the background noise.
Undertaking and understanding the "black art science" of fabricating a homebrew filter must be approached in a rigorous manner. The first piece is understanding how the crystal parameters work to make a filter. The 2nd piece is how to measure those parameters with the 3rd piece of using the parameters in a software suite to determine the coupling capacitors, the filter impedance and the matching to the real world. You also need o think of the filter form factor. The QER (Quasi Equal Ripple) topology seems to be the flavor of the month choice today, amongst filter builders. This is where having a Nano VNA if used properly can aid in the measurement of actual filter performance.
Today's journey highlights there is more to homebrewing a crystal filter and it is more than simply putting one's hand into a poke and pulling out 4 crystals and check off the box. The choice of filter frequency drives much of the rest of the design. Personally, I like 4.9152 MHz as that was good enough for Elecraft in the K2 and good enough for N6QW. That frequency does not have the baggage of 9 MHz for 17M.
TYGNYB.
73's
Pete N6QW