While the RF stage _does_perform some filtering, its main purpose is to develop receiver sensitivity.
Selectivity is, as you mentioned, delevoped in the IF stage(s).
More info here: https://www.electronics-notes.com/articles/radio/radio-receiver-selectivity/basics.php
Antenna tuning is basically for Image Frequency Rejection
The Band-Pass Filtrering is done in IF Coils
.
Quite right. But the image is displaced from the desired signal freq by the IF freq x 2.
So in a receiver that has say a 10.7 MHz IF (common FM radio) or 30 or even 60 MHz
(typical of UHF/SHF receivers) The chances of a signal on the image freq are quite small.
(AM radio is another story due to an IF of 455 kHz.)
When signals that are close together, say only a few kHz or tens of kHz, are the ones under consideration, the IF strip is where that separation is done.
Image frequency = dial frequency + (or -) 2 × IF frequency
Yep. Forgot the “x 2” part. 
Because the image is twice the IF away from the desired signal, it doesn’t take much selectivity in the RF stage to get rid of it. (provided the IF isn’t a low freq like in a common AM radio) That’s one of the reasons a receiver front-end only has a stage or two. (the cost of low noise amplifier components being another factor)
Above ~30 MHz, the RF stage is a source of noise, and hence has a part in setting overall system noise figure.
FWIW the 820T type tuners in a rtlsdr-style dongle produce a low-IF output (5 MHz? somewhere around there), and the final filtering and conversion to I/Q baseband happens in the digital domain within the 2832. Some of the other tuner types (I think the E4k, from memory) produce zero-IF I/Q output directly, so those work more like a direct conversion receiver. The LO is a NCO (uh, maybe not exactly the right terminology - a fractional-N PLL, anyway)
The filter is software tunable too.
https://www.rtl-sdr.com/testing-built-filter-r820t-rtl-sdr/
Is VCO the term you were “looking for?” 
(Voltage Controlled Oscillator)
Yeah, a little.
Not a VCO (though IIRC it uses one internally). It’s basically a PLL where you feed it a controllable divisor value N and it produces a LO such that LO / N = fixed reference frequency (e.g. 28.8MHz from an external crystal)
edit: see e.g. http://www.ti.com/lit/an/swra029/swra029.pdf
Right. Sounds like maybe “Divide-by-N” might be the term.
Got it. Thanks for the link!
Reminds me of the Heath HW-2036 synthesized 2M transceiver I built in the 70s.
.
.
This reminds me that one of these days I hope to find the time to walk through some of the dump1090 code. I’m interested in how the sampling and decoding is performed and curious if it as good as it will ever be or if there are newer/more complex algorithms that would perform better at the expense of more processing power.
) : An Interesting Omni-Directional Horse-shoe AntennaI found an interesting design built by galenthurber, who posted it here.
I then did simulations on 4nec2.
Here are photos/screenshots
2 Horizontals = 1/4 wavelength each
2 Verticals = 1/2 wavelength each
As its a 1.5 lambda dipole that should easily match to 50 or 75 ohms, presumably the high VSWR in the modelling is down to the bends?
The distance between the close ends is too small IMO. Also, this will be a symmetrical antenna, so… not 50 ohm impedance.
It will be worth making this simple Horse-shoe antenna ᒪᒧ, then swap it with a running Spider, and see if the swap has caused performance to drop, increase or remain same. I am planning to give it a try next weekend.
Adding an impedance matching stub to the horse-shoe antenna will improve its SWR from 3.5 down to possibly < 1.5
As you are already feeding into 1/4 sections, why a matching stub? Adjusting the feed point taps should resolve that issue.
??? pray tell why it could not be matched to 50 ohms?
Not Feeding into 1/4 sections. Feeding into 1/4 + 1/2 = 3/4 sections.
