Orientation: RF is a chain of 50-ohm blocks
The earlier guides in this rung changed how you see a circuit. Once a connection is a fair fraction of a wavelength, the wire is a transmission line with its own characteristic impedance, reflections appear at every mismatch, and the cure is impedance matching read off the Smith chart. The deep shift for this guide is that an RF engineer mostly stops thinking about individual resistors and capacitors and starts thinking in building blocks — self-contained boxes, each designed to look like 50 ohms at its ports and described not by a schematic but by S-parameters: how much it passes, how much it reflects, how much it amplifies.
Wire those boxes in a line and you have a radio. A receiver's signal chain runs: an antenna catches a faint wave, a band filter keeps only the slice you want, a low-noise amplifier makes that faint signal bigger, a mixer slides it down to a convenient frequency, more filtering and gain clean it up, and a demodulator finally recovers the music or the bits. Each block does one job; the craft is in the joins — keeping every interface near 50 ohms so signals flow instead of bouncing, and tracking the signal level through the whole chain.
Because levels span a colossal range — from a fraction of a femtowatt at the antenna to milliwatts at the output — RF speaks in logarithms. Gain and loss are quoted in decibels: dB = 10 times log10(P_out / P_in), so +3 dB is roughly double the power, +10 dB is ten times, and a chain's blocks simply add their dB. Absolute power is quoted in dBm, decibels relative to 1 mW, where 0 dBm = 1 mW and -30 dBm = 1 microwatt. This is not decoration: counting a design in dB is what lets you do the whole signal chain in your head.
The LNA: amplifying a whisper without adding hiss
The first active block after the antenna is the low-noise amplifier (LNA), and its job is subtler than "make it bigger". Every amplifier adds a little hiss of its own — irreducible thermal noise from the random jiggle of electrons in its parts — and an incoming signal that is already terribly weak cannot afford much extra hiss. The number that captures this is the noise figure: it measures, in dB, how much worse the signal-to-noise ratio is at the output than it was at the input. A perfect, noiseless amplifier has a noise figure of 0 dB; a good RF LNA manages around 1 dB.
Here is why the LNA sits first, right after the band filter, and why it gets the careful design. Noise in a chain follows Friis's rule: the total noise factor is F_total = F1 + (F2 - 1) / G1 + ..., where each later stage's noise is divided down by the gain ahead of it. Put real numbers in: an LNA with 1 dB noise figure (F1 = 1.26) and 20 dB gain (G1 = 100), feeding a mixer with a poor 10 dB noise figure (F2 = 10). The mixer's contribution is (10 - 1) / 100 = 0.09, so F_total = 1.26 + 0.09 = 1.35, a total noise figure of 10 times log10(1.35) = 1.3 dB. The whole receiver is barely noisier than the LNA alone — the first stage, with high gain, dictates the noise of everything behind it.
But more gain is not simply better, and this is where RF stays honest. Pile on too much LNA gain and you overload the stages behind it: strong nearby signals start mixing with each other inside the amplifier (intermodulation) and create false signals that no later filter can remove, so the LNA must balance gain against linearity. There is a second honest twist from guide 3: the source impedance that gives the LNA its lowest noise is generally not the conjugate match that gives maximum power transfer. RF designers deliberately choose a slight mismatch — a "noise match" — accepting a touch of reflection to win the quieter front end. Matching here is a design lever, not a single right answer.
The mixer: moving a signal up and down the dial
The mixer is the block that performs the deep magic of radio: it shifts a signal from one frequency to another without disturbing the information riding on it. The trick is multiplication. When you multiply two sine waves, the trig identity cos(a) times cos(b) = half of [cos(a - b) + cos(a + b)] tells you the product contains two brand-new tones — the difference frequency and the sum frequency. Feed a mixer the incoming radio signal at f_RF and a clean tone from a local oscillator at f_LO, and out come copies of your signal sitting at f_LO - f_RF and f_LO + f_RF.
Make it concrete with an FM radio. Your station broadcasts at f_RF = 100.1 MHz. The receiver runs its local oscillator at f_LO = 110.8 MHz. The mixer outputs the difference, 110.8 - 100.1 = 10.7 MHz, and the sum, 210.9 MHz; a filter throws the sum away and keeps the 10.7 MHz copy. Now the clever part: to tune to a different station you only change f_LO. Move it so the new station's difference again lands on 10.7 MHz, and the rest of the radio never has to retune. The mixer has slid the whole dial onto one fixed working frequency.
The oscillator and the superheterodyne idea
The mixer needs a clean tone to multiply by, and the oscillator is the block that makes it — the local oscillator the superheterodyne receiver builds for itself. You met oscillators in an earlier rung: an amplifier wrapped in frequency-selective feedback that satisfies the Barkhausen condition (loop gain at least 1, total phase zero), with some gentle amplitude stabilization so it settles into a steady swing instead of growing forever. At RF the frequency-selecting element is an LC tank or a quartz crystal, and its Q sets how pure the tone is — the sharper the resonance, the cleaner the crystal oscillator's single note.
Now a real tension appears: you want the LO to be both tunable (so you can pick a station) and rock-stable (so it does not drift off and lose it). A crystal is gorgeously stable but stuck at one frequency; an LC oscillator tunes freely — wiggle its capacitance with a varactor diode, a diode whose capacitance changes with the voltage across it, to make a voltage-controlled oscillator (VCO) — but it drifts with temperature. The elegant marriage is the phase-locked loop (PLL): it continuously compares the tunable VCO against a stable crystal reference and nudges the VCO's control voltage to keep it locked, giving you a tone that is tunable, accurate, and clean all at once. Be honest about the limit, though — no real oscillator is an infinitely thin spectral line; its phase noise smears the tone slightly, and that smear sets how close two channels can sit before they bleed together.
ANTENNA BAND-PASS LNA MIXER IF FILTER IF AMP DEMOD
\ / FILTER (low ( x ) (sharp, (AM/FM
| (rejects noise) | fixed freq) -> audio)
+-------[ BPF ]-----[ >>> ]-----[ X ]------[ |||| ]-----[ >>> ]----[ )) ]---> out
image) |
|
[ LOCAL OSC ] f_LO (tunable: a VCO locked
by a PLL to a crystal)
example: f_RF = 100.1 MHz , f_LO = 110.8 MHz
IF = | f_LO - f_RF | = 10.7 MHz (the difference; sum 210.9 MHz filtered off)
tune a new station: just move f_LO -- the 10.7 MHz IF stage never changes
image trap: a station at f_LO + IF = 121.5 MHz would also land on 10.7 MHzStep back and the architecture is breathtakingly economical. Instead of building a sharp, high-gain amplifier that must retune across the whole band — nearly impossible to keep selective and stable everywhere — the superheterodyne shifts every station down to one fixed IF (10.7 MHz for FM, 455 kHz for AM) and lavishes all its careful filtering and gain there. The last block, the demodulator, then strips the information back off the carrier. What kind of information? That depends on the modulation: AM (amplitude modulation) nudges the carrier's height up and down with the audio, while FM (frequency modulation) wiggles the carrier's frequency instead and shrugs off the amplitude noise that plagues AM. The mixer moved the signal; the demodulator unwraps it; the same block kit, rearranged and run backwards, becomes a transmitter.
Practical RF: decibels, the antenna, and honest limits
At the very front sits the antenna — the boundary where a guided wave on the 50-ohm line becomes a wave flying free through space, and (by reciprocity) the same structure both radiates and receives. What it hands the receiver is astonishingly faint. A strong-ish 1 microvolt across 50 ohms is a power of V squared over R = (10^-6) squared / 50 = 2 x 10^-14 watts, which in dBm is 10 times log10(2 x 10^-14 / 10^-3) = -107 dBm. That tiny number is the whole reason the low-noise chain exists; adding up exactly how much signal arrives and how much noise it must beat is the link budget, the subject of guide 5.
This is also where the noise figure earns its keep. Thermal noise sets a hard floor of about -174 dBm per hertz of bandwidth at room temperature; spread across a 200 kHz FM channel that floor is -174 + 10 times log10(200000) = -174 + 53 = -121 dBm. Our -107 dBm antenna signal therefore arrives about 14 dB above the noise — a workable signal-to-noise ratio — and the LNA's 1.3 dB noise figure lifts that floor by only a hair. Tot it up and you can see, before building anything, whether the station will come through clean or buried in hiss.
End on the warning that runs through this whole rung: your DC and audio intuition will mislead you at RF. A wire is a transmission line; a capacitor's lead inductance makes it turn inductive above its self-resonant frequency, an inductor has parasitic capacitance, and so every part has a finite, frequency-dependent Q rather than the ideal behaviour you assumed at low frequencies. A solderless breadboard's stray capacitance and lead inductance make it useless above a few tens of megahertz — real RF is built on a controlled-impedance PCB or dead-bug style over a solid ground plane, with tiny current loops and honest shielding treated as design, not polish. And remember that simulation only knows your netlist, never your layout: at RF the board itself is a circuit element. With these blocks, the dB bookkeeping, and the antenna in hand, you are ready to add it all into a link budget.