The radio topology that I have centered on is a direct conversion I-Q receiver seen in many HF software defined radios. The goal at this point isn't to build for high performance, but for functionality, low cost, and ease of design. While the fundamental core is not designed to work at a specific frequency, the filters must be designed around at least the expected band of operation. In multi-band radios, filters are switched in and out based on the frequency of operation, but this adds a lot of complexity. While I don't think avoiding multi-band operation requires a ton of justification, a good reason is that the targeted use case is a mono-band dipole antenna on 20m. With the radio effectively part of the antenna, it wouldn't be necessary to pack multiple band operation when it wouldn't be used.
The baseband inputs and outputs are designed to be by a sound card, resulting in a transmit and receive bandwidth limitation of 50-75kHz. The mixer is repurposing a digital bus-mux switch. The oscillator will be a Silicon Labs Si570 oscillator, which will give a lot of flexibility to a prototype system. A future version might use a simple crystal oscillator as the Si570 is definitely not a budget part. The oscillator operates at twice the center frequency, and the I-Q signal generator creates the quadrature spaced signals required by the mixer.
The RF filters are primarily focused on reducing the second and third harmonic of the radio. FCC regulations require spurious signals to be at least 43dB below the fundamental signal. In order to meet this target, the band-pass filters and the output filter will be designed to have at least 43dB suppression of all harmonics. Proper phasing operation will be critical to achieving the close-in response, as there isn't really an easy way to tackle that.
One other goal is to minimize the number of hand-wound magnetics, as that always ends up being a deterrent to me to get started. Reducing the reliance on these parts also helps improve the manufacturability of the radio.