Published Aug 27, 2026 · Last updated Oct 18, 2026 · 5 min · IndieRF
dB Mental Math: dBm to Watts, kTB, and 10log vs 20log
+17 dBm is 50 mW because +10 dB is ×10 and +7 dB is ×5. Five remembered ratios, the −174 dBm/Hz noise floor, and the 10log versus 20log rule cover most RF sanity checks without a calculator.
+17 dBm is 50 mW. That's +10 dB (×10) and +7 dB (×5), done in your head before the spectrum analyzer finishes its sweep. Five ratios, one noise number, and one rule about voltage cover most of the checks in a design review.
Five ratios make every dB value
| Step | Exact power ratio | In your head |
|---|---|---|
Peel off the 10 dB decades, then build the remainder from 7, 3, 2, and 1. Going negative, divide: halves.

- : .
- : .
- : . Enough to leave a mark on whoever grabs the output connector to check it's tight.
- : halve it twice, a quarter.
Calling +7 dB "×5" is off by 0.01 dB. A link budget already carries a dB or so of cable and connector slop, so nothing one level up can see it. Save the exact math for cal limits and tolerance stacks.

Noise floor: −174, plus bandwidth, plus NF
A receiver's noise floor in one line, at 290 K:
The bandwidth term climbs 30 dB per factor of 1000, so the ladder is easy to carry around:

| Bandwidth | dB-Hz | at |
|---|---|---|
A 20 MHz channel is 60 dB-Hz for the megahertz plus 13 dB for the factor of 20, so 73 dB-Hz. Thermal noise is about , and a receiver with a 5 dB noise figure sits near . Anything you want to demodulate has to clear that by its required SNR.
Three constants are in circulation. dB Power Converter uses , RF Cascade Analyzer uses the CODATA value at exactly 290 K, , and everyone says out loud. The 0.1 dB spread is smaller than the uncertainty of any noise figure you'll measure. Off 290 K, compute for your temperature instead.
Total power to density
Spread a power evenly across a bandwidth and the density is the total minus the bandwidth term:
+17 dBm over 20 MHz is , or . That's 2.5 mW in each megahertz, and twenty of those slices add back to 50 mW.
| Total power | Bandwidth | Per Hz | Per MHz |
|---|---|---|---|
At 1 MHz, dBm and dBm/MHz read the same number, which is a handy way to check you didn't drop a 60.
This is average density, right for flat noise and a fair first pass on a spread waveform. A CW tone has no bandwidth to divide by, and a regulatory mask cares about peak density, so measure those with the analyzer's RBW set the way the mask says.
10log for power, 20log for voltage
Power goes as voltage squared, , so voltage ratios get twice the dB:

- Twice the power is +3 dB.
- Twice the voltage is +6 dB, and four times the power.
Into 50 Ω, .
The 20log form only holds when both voltages sit across the same impedance. Compare a 50 Ω port to a 75 Ω one with it and you'll be 1.8 dB wrong () with total confidence. Convert each voltage to power first.
Bench reference
| dBm | Power | Decomposition |
|---|---|---|
| three decades | ||
| two decades | ||
| one decade | ||
| reference | ||
In the clip, typing +43 dBm over the default 0 dBm reads 19.95 W, and the converter's mental shortcut builds the same row from four +10 dB steps and one +3 dB step, ×10⁴ × 2 ≈ 20 W.
When to open the cascade
These shortcuts check a budget. Once stages interact, order matters more than any single number: with 20 dB of gain in front, a 10 dB NF second stage adds about 0.3 dB to a 1 dB NF front end, and with only 10 dB in front it adds 2.3 dB. That's Friis, and it's where mental math starts guessing.
Questions? Contact
Discussion
No comments yet. Start the thread with a measurement, a correction, or a worked example.