The decibel: why radio lives in logarithms
By guide 4 you have all the parts: a line that behaves, a matching network, an LNA, a mixer, an oscillator. The moment you chain them into a real radio, you face an awkward arithmetic problem. A transmitter might pump out 1 watt; the signal that crawls back into a receiver across a city might be a femtowatt — that is fifteen zeros of difference, and along the way every amplifier multiplies and every cable, connector, and kilometre of air divides. Tracking that as raw watts is a nightmare of tiny decimals. The decibel fixes it by working in logarithms, where multiplying becomes adding: dB = 10 times log10(power ratio). A doubling of power is +3 dB, ten times is +10 dB, a hundred times is +20 dB — and a loss is just the same number with a minus sign.
A decibel is a pure ratio, so on its own it says how much bigger, never how big. To pin down an absolute power, radio engineers tack on a reference: dBm is power measured in dB relative to 1 milliwatt. So 0 dBm is exactly 1 mW, +30 dBm is 1 W, +20 dBm is 100 mW, and -30 dBm is 1 microwatt. Now the whole signal chain becomes one running sum: start at the transmitter in dBm, then for every stage just add its gain in dB or subtract its loss in dB, and the number you land on is the power at that point, still in dBm. A -50 dBm signal through a +20 dB amplifier is simply -30 dBm.
The antenna: a wire set loose on purpose
In guide 2 you learned that any wire longer than a fraction of a wavelength becomes a transmission line, and the whole craft there was to stop it leaking — to keep the energy travelling neatly down the line and not radiating away. An antenna is that same physics turned deliberately inside out. It is a structure built to leak on purpose: a graceful impedance transition that hands a signal off from the 50-ohm world of your line to the impedance of free space itself (which works out to about 377 ohm). Done well, the energy slides smoothly into a travelling electromagnetic wave; done badly, it bounces back as a reflection, exactly the VSWR problem from guide 3.
Size is set by the wave, not by taste. An antenna couples best when its length is a natural fraction of the wavelength — a half-wave dipole, or a quarter-wave whip standing over a ground plane — because then it rings in resonance with the field, just like the LC tanks you met earlier. Since wavelength = c / f = (3 x 10^8 m/s) / f, the frequency directly sets the metalwork. At 100 MHz (FM radio) a half-wave is 1.5 m, which is why car aerials are long; at 2.4 GHz (Wi-Fi) the wavelength is only 12.5 cm, so a quarter-wave is about 3 cm — small enough to vanish inside a phone. High frequency is what makes antennas shrink to something you can hide.
One word causes endless confusion: antenna gain, quoted in dBi. It is not free energy and it makes no new power — it is purely directivity, the same total power aimed into a narrower cone, exactly like a flashlight reflector versus a bare bulb. A 6 dBi antenna sends four times the power of an isotropic radiator toward where it is pointed, and correspondingly less elsewhere. By reciprocity, the same antenna transmits and receives with identical patterns, so a dish that shouts in one direction also listens keenly in that direction. And honour the limits: a very small loop is honest but feeble — its radiation resistance is tiny so most drive current just heats the wire, fine for receiving but poor for transmitting. A real antenna must also be matched to the line, and it detunes the instant a hand, a wall, or a battery comes near.
Noise figure and the noise floor: the receiver's hard limit
How faint a whisper can a receiver actually hear? There is a hard floor, and even a flawless receiver cannot dig beneath it: thermal noise. Every resistor, every antenna, everything warm jiggles its electrons and generates a tiny random power that grows with temperature and with bandwidth. The number worth memorising is that at room temperature this noise is -174 dBm in each hertz of bandwidth. Open up a 1 MHz channel and you collect a million times more, so the noise floor rises to -174 + 10 times log10(10^6) = -174 + 60 = -114 dBm. That is the quietest possible hiss in that channel, set by physics, not by your budget.
Real parts are worse than ideal, and noise figure (NF, in dB) measures exactly how much each stage spoils things — it is the amount, in dB, by which a block degrades the signal-to-noise ratio passing through it. A perfect stage has NF = 0 dB; a typical receiver front end might be 6 dB, which lifts the effective floor to -114 + 6 = -108 dBm. But you cannot decode a signal sitting right on the noise; you need it to stand above the hiss by a comfortable signal-to-noise ratio, maybe 10 dB. Add that and you have the receiver's sensitivity: -108 + 10 = -98 dBm — the weakest signal it can usefully recover. Everything in a link budget is a fight to keep the received power above this line.
This is why the low-noise amplifier from guide 4 sits first, jammed right up against the antenna. The noise figure of a chain is dominated by its very first stage: that stage's gain divides down the noise contributed by everything after it. Put a quiet, high-gain LNA at the front and it sets the whole receiver's NF; let the signal arrive at a noisy mixer first and no amount of later amplification can ever recover the SNR you threw away. Gain after the damage just amplifies the damage. First impressions, in a receiver, are everything.
The link budget: the whole radio added up in dB
Now assemble everything into the single most useful spreadsheet in radio: the link budget. It is one running column of decibels that predicts, before you build a thing, whether a link will close. Start at the transmitter power in dBm, add the gains (the antennas), subtract the losses (cables, connectors, and above all the long stretch of air), and the number you land on is the power delivered to the receiver. Set that against the receiver sensitivity you just computed, and the gap between them is your margin — the breathing room you have left.
LINK BUDGET -- 2.4 GHz Wi-Fi, 1 km line of sight
------------------------------------------------------------
Transmitter power +20 dBm (100 mW)
+ Transmit antenna gain + 2 dBi
- Cable / connector loss - 2 dB
- Free-space path loss -100 dB (1 km @ 2.4 GHz)
+ Receive antenna gain + 2 dBi
------------------------------------------------------------
= Received power - 78 dBm
Receiver sensitivity - 98 dBm (NF 6 dB, SNR 10 dB, 1 MHz)
------------------------------------------------------------
LINK MARGIN (fade margin) + 20 dB -> the link closes
free-space path loss (dB) = 20*log10(d_km) + 20*log10(f_MHz) + 32.4- Write down the transmit power in dBm. 100 mW is +20 dBm. This is the top of the column, the budget you have to spend.
- Add the antenna gains and subtract the wired losses. Each transmit and receive antenna adds its dBi; every cable, connector, and filter on the way subtracts its dB. These are usually small, a few dB each.
- Subtract the free-space path loss — the dominant term. Use 20*log10(d_km) + 20*log10(f_MHz) + 32.4. Doubling the distance costs 6 dB; doubling the frequency also costs 6 dB.
- Compute the received power, then subtract the receiver sensitivity. What remains is the link margin. A positive margin means the link works; aim for 10 to 20 dB so rain, multipath, and a hand over the antenna do not break it.
That path-loss term deserves a second look, because it carries an honest surprise. Free-space loss rises with frequency: at the same distance and the same antenna gains, a 5 GHz link suffers about 6 dB more loss than a 2.4 GHz one. That is not the air absorbing more — it is bookkeeping. A fixed-gain antenna has a smaller physical catching area at higher frequency, so it intercepts a smaller slice of the spreading wave. The 20 dB we kept in the example is the fade margin: not slack to be trimmed away, but the reserve that keeps the link alive through rain fades, a passing truck, a reflection that arrives out of phase, or simply a unit at the dull end of the production spread.
Honest limits, and the road you have climbed
Here is the deepest honest caveat, and it ties straight back to guide 3. A matched load drinks the most signal power, which is exactly what you want at a faint receiver where every microwatt is precious. But maximum-power transfer is only 50% efficient — at a conjugate match, half the available power is burned in the source itself. That is a fine trade for a whisper of a received signal, and a terrible one for a transmitter's power stage, which would cook itself and waste half its battery. So a transmit power amplifier is tuned for efficiency, not maximum power transfer. The same word, matching, means opposite goals at the two ends of the link: catch every photon at the receiver, lose no heat at the transmitter.
And yet none of it is magic. The RF frontier is the very same physics you have carried up the whole ladder, simply pushed until the wavelength shrinks to the size of your board and the old approximations crack. The decibel keeps the enormous numbers tame; the antenna trades your line for the sky; the noise floor sets the faintest signal worth chasing; and the link budget, one honest column of plus and minus signs, tells you whether the whole thing will work before you spend a cent. Master those four and you can look at a datasheet's sensitivity, an antenna's gain, and a distance, and simply know — that is the quiet confidence the art of electronics is built to give you.