Decibel (dB): Voltage, Current, and Power Ratios

Every datasheet you read expresses gain, loss, noise floor, or dynamic range in dB — op-amp gain, filter attenuation, RF power budgets, ADC SNR, cable loss. The decibel is not a unit of measurement in the physical sense; it's a logarithmic ratio, and the factor in front of the log₁₀ changes depending on whether you're comparing power or a field quantity like voltage or current. Getting that factor wrong (using 20 instead of 10, or forgetting an impedance mismatch) is one of the most common numerical errors engineers make when moving between spec sheets, simulation output, and bench measurements.

Why a Logarithmic Ratio at All

Electrical quantities in real systems span enormous ranges — a receiver's dynamic range might cover 120 dB, meaning a power ratio of 10¹². Linear scales make this unusable: you either lose resolution at the small end or need absurdly large axes. Logarithms compress multiplicative relationships into additive ones, which has two direct payoffs:

• Cascaded gains/losses add instead of multiply. An amplifier stage with 20 dB gain followed by a cable with −3 dB loss and a filter with −6 dB attenuation gives a total of 20 − 3 − 6 = 11 dB, no multiplication required.
• Wide dynamic ranges fit on a compact, readable scale. A 1,000,000:1 power ratio is just 60 dB.

This is the same principle behind Bode plots (covered separately): once a quantity is in dB, series stages of a system sum their contributions.

The Two Definitions

The decibel is fundamentally defined on power:

dB = 10·log₁₀(P₂ / P₁)

Because electrical power is proportional to the square of voltage or current (P = V²/R or P = I²·R), if you want to express a voltage or current ratio in a way that's consistent with the power definition, the square has to come out of the log:

dB = 10·log₁₀[(V₂/V₁)²] = 20·log₁₀(V₂ / V₁)
dB = 10·log₁₀[(I₂/I₁)²] = 20·log₁₀(I₂ / I₁)

Critical assumption: the voltage/current formulas with factor 20 are only equal to the underlying power ratio when both voltages (or currents) are measured across (or through) the same impedance at both points 1 and 2. If the impedance differs between the two measurement points, the 20·log₁₀(V₂/V₁) figure is still a valid statement about the voltage ratio, but it no longer equals the power ratio in dB — you'd need to include the impedance ratio explicitly:

dB(power) = 10·log₁₀[(V₂²/R₂) / (V₁²/R₁)] = 20·log₁₀(V₂/V₁) + 10·log₁₀(R₁/R₂)

This is exactly the situation in RF systems where impedance is matched to 50 Ω everywhere, which is precisely why the voltage-ratio shortcut works cleanly there — and why it can silently mislead you in circuits where impedance changes, like an op-amp output driving a different load than its input reference.

Quick Reference Values

Certain ratios recur so often they're worth memorizing:

Note the doubling asymmetry: doubling power is +3 dB, but doubling voltage (into the same impedance, which quadruples power) is +6 dB. This is a frequent source of confusion — "3 dB down" on a filter response means half power, which corresponds to a voltage ratio of 1/√2 ≈ 0.707, not 0.5.

Worked Example: Op-Amp Gain Stage

An inverting amplifier has input Vin = 50 mV RMS and output Vout = 2.5 V RMS, both referenced to the same ground with high input impedance (negligible loading, so no impedance-ratio correction needed).

Step 1 — voltage gain ratio:

A_v = Vout / Vin = 2.5 / 0.05 = 50

Step 2 — convert to dB:

dB = 20·log₁₀(50) = 20 × 1.699 = 33.98 dB ≈ 34 dB

Step 3 — check via power, assuming both ports see the same resistance R:

Pin = Vin²/R,  Pout = Vout²/R
Pout/Pin = (Vout/Vin)² = 50² = 2500
dB = 10·log₁₀(2500) = 10 × 3.398 = 33.98 dB

Both routes agree at ≈ 34.0 dB, confirming the factor-of-2 relationship (20·log₁₀(50) = 10·log₁₀(50²) = 10·log₁₀(2500)). This cross-check is the standard way to catch a dropped factor of 2 — if your voltage-ratio dB and power-ratio dB disagree by roughly 2×, one of the log factors is wrong.

Cascaded Stages: Why dB Simplifies System Budgets

Consider a signal chain: an antenna feeds a low-noise amplifier (+15 dB), a length of coax (−4 dB), and a mixer with conversion loss (−7 dB).

Total gain = 15 + (−4) + (−7) = +4 dB

In linear terms this would require multiplying gain factors: 10^(15/10) × 10^(−4/10) × 10^(−7/10) ≈ 31.6 × 0.398 × 0.200 ≈ 2.51, and 10·log₁₀(2.51) ≈ 4.0 dB — matching the additive result. For any chain of N stages, this additive property turns a multiplication of N numbers into a sum, which is far less error-prone by hand and essential for RF link budgets, audio gain-staging, and noise-figure cascading (Friis' formula relies on the same log structure).

Absolute Reference Levels

dB alone is a ratio, dimensionless. Adding a reference turns it into an absolute level, denoted with a suffix:

Example: 0 dBm into 50 Ω corresponds to P = 1 mW, so V_rms = √(P·R) = √(0.001 × 50) ≈ 223.6 mV. A signal reading +10 dBm is 10× that power (10 mW), corresponding to V_rms = √(0.010 × 50) ≈ 707 mV — note this is √10× the voltage, not 10×, consistent with the 10·log₁₀ vs 20·log₁₀ distinction.

Practical Implications

• Always check what's being ratioed. A "gain of 20 dB" for a voltage amplifier and a "gain of 20 dB" for a power amplifier imply different underlying linear ratios (10× voltage vs. 100× power) unless impedances are stated and matched.
• Filter cutoff (−3 dB point) is a power-half point, i.e., voltage amplitude at 1/√2 ≈ 70.7% of passband value — this is the same −3 dB used in Bode plot corner frequencies and RLC bandwidth definitions.
• Sum dB values for cascaded stages, but only when each stage's gain/loss figure was computed consistently (all power-referenced or all voltage-referenced with matched impedance).
• Watch mismatched impedances — a probe measuring voltage gain across a high-impedance node feeding a low-impedance load will not match a straightforward 20·log₁₀(Vout/Vin) power interpretation without the impedance correction term.
• Negative dB means attenuation, not "negative power" — it's simply a ratio less than 1.

Key Takeaways

• dB is a logarithmic ratio: 10·log₁₀ for power, 20·log₁₀ for voltage or current — the factor-of-2 difference comes from power being proportional to the square of voltage/current.
• The voltage/current formula only equals the true power ratio in dB when both measurements share the same impedance; otherwise an explicit impedance-ratio term is needed.
• Doubling power is +3.01 dB; doubling voltage (same impedance) is +6.02 dB — a common source of off-by-2× errors.
• Cascaded gains and losses simply add in dB, replacing multiplication of linear ratios with addition — the basis of link budgets and Bode plot analysis.
• Absolute levels (dBm, dBV, dBu, dBFS) require a stated reference; dB alone is always dimensionless and relative.