For digital communication, average signal power alone does not account for how quickly information bits are transmitted.

If the average transmitted power is $P$ and the information bit rate is $R_b$, the energy per bit is

$$E_b=\frac{P}{R_b}.$$

For white noise with two-sided power spectral density $N_0/2$, the ratio

$$\frac{E_b}{N_0}$$

compares the energy allocated to each information bit with the noise intensity per unit bandwidth.

$E_b/N_0$ is dimensionless and is commonly expressed in decibels:

$$\left(\frac{E_b}{N_0}\right){\mathrm{dB}}=10\log{10}\frac{E_b}{N_0}.$$

Unlike SNR measured inside a particular receiver bandwidth, $E_b/N_0$ normalizes by bit rate and noise spectral density. It therefore makes it easier to compare modulation and coding schemes that use different bandwidths or symbol rates.

For a given modulation and receiver, error probability is often expressed directly as a function of $E_b/N_0$.