Any detector can process signals over a bandwidth. If you need to reject a section of a bandwidth, you must use a suitable band-pass filter.
Regenerative detectors like those used by radio amateurs, achieve a narrow bandwidth by applying a feedback; such detectors are able to select only one sideband, e.g. of an AM or SSB signal. Their problem is that their performance depends on a manual adjusting the feedback of an oscillator circuit so it operates close to starting oscillation, or already oscillates.
Since there is not downconversion, there is no "image" to be suppressed. But, yes, some filtering is wise, either infront of the detector, or as in a tank circuit to limit the bandwidth of the regenerative amplifier.
It seems to me as the ideal solution for these QRP low-part-count amateur constructions. No image signal to filter out, Q multiplication on a single selectable frequency, ability to detect CW/SSB/AM depended on the feedback amount and low components count. With the output of the detector being both detected audio as well as amplified RF (for further processing), what more a homebrew radio amateur could ask for?
Of course a front end amplifier is needed to block the unwanted rf from passing out to the antenna.
Hello, I had another conversation with a member claiming:
>>> Having only one mixer does not affect audio quality. Regens have image noise, its just that the image is 3kHz away instead of 2x455kHz away in an ordinary superhet.
Is that really the case? Based on the posts here, I thought that the amplifier selects and amplifies a quite narrow bandwidth so that image is rejected, even the close-to-carrier unwanted sideband.