The Floor You Have Been Standing On
Every link budget since M2-L4, Shannon’s capacity, every BER curve — all of it assumed a noise floor nobody derived. One constant and one temperature is all it takes.
Warm Things Hiss
Charge carriers in any resistor above absolute zero are in thermal motion, so a resistor makes noise. What Nyquist proved is what the answer leaves out: not resistance, not material — only temperature and bandwidth.
Memorise −174
kT₀ = 1.380649×10⁻²³ × 290 = 4.0039×10⁻²¹ W/Hz = 4.0039×10⁻¹⁸ mW/Hz, and 10log₁₀ of that is −173.98 dBm/Hz. At 300 K you would get −173.8 — pick one and stay with it.
Bandwidth Is Noise
Widening 20 MHz to 100 MHz is a factor of five, which is 7 dB — and the floor rises by exactly 7 dB. Wide channels buy capacity with noise, which is the trade Shannon priced in M6-L1.
Say Which Ratio You Mean
One radio, three numbers: 20 dB of SNR in 20 MHz is C/N₀ = 93 dB-Hz and, at 4 bit/s/Hz, Eb/N₀ = 14 dB. Quote the wrong one and the budget is out by 73 dB.
Floor, Noise Figure, Sensitivity
Three sliders, four numbers. Watch the ladder climb from thermal floor to the weakest signal this receiver can use.
What the Receiver Adds
NF 1 dB is F = 1.259, so Te = 290 × 0.259 = 75 K. NF 3 dB is F = 1.995, so Te = 289 K — the receiver doubles the noise. F says nothing about gain.
The First Stage Decides
LNA (1 dB, G 20 dB) → mixer (8 dB, G 10 dB) → IF (15 dB): 1.259 + 5.31/100 + 30.6/1000 = 1.3427, i.e. NF 1.28 dB. Put the mixer first and the same parts give 6.31 + 0.0259 + 0.0306 = 6.366, i.e. 8.04 dB.
Sensitivity in Four Terms
−174 + 73.0 + 5 + 20 = −76 dBm. Real radios miss that by an implementation margin — and in a crowded band the floor they actually see is set by neighbours, not by kTB (M9-L3).
What you learned
- N = kTB, and kT₀ at 290 K is −174 dBm/Hz
- Any floor is −174 + 10log₁₀B: −101 dBm in 20 MHz
- SNR = (Eb/N₀)(Rb/B); C/N₀ adds 10log₁₀B
- NF = 10log₁₀F and Te = T₀(F−1): 3 dB is 289 K
- Friis: 1.28 dB LNA-first, 8.04 dB mixer-first