What do the axes of a BER curve represent?

The x-axis is Eb/N0 in dB: the ratio of energy per information bit to the noise spectral density. Eb/N0 normalizes performance across different data rates, bandwidths, and modulation orders, making it the universal comparison metric. Eb/N0(dB) = SNR(dB) - 10·log₁₀(spectral efficiency in bits/s/Hz). For BPSK at 1 bit/s/Hz: Eb/N0 = SNR. For 64-QAM at 6 bits/s/Hz: Eb/N0 = SNR - 7.78 dB. The y-axis is BER on a logarithmic scale: the probability that a received bit is decoded incorrectly. Typical targets: Voice (AMR codec): BER ≤ 10⁻³ (errors masked by codec). Video (H.264/H.265): BER ≤ 10⁻⁶ to 10⁻⁸ (visible artifacts above this). Fiber backbone: BER ≤ 10⁻¹² (pre-FEC), 10⁻¹⁵ (post-FEC). Satellite (DVB-S2): BER ≤ 10⁻⁷ after LDPC + BCH. The 'waterfall' region is the steep transition zone where small Eb/N0 increases produce orders-of-magnitude BER improvement. The 'error floor' is the region where BER decreases slowly despite increasing Eb/N0, caused by code structure limitations (trapping sets in LDPC, weight distribution in turbo codes).

How do you derive the BER for common modulations?

Theoretical BER expressions for AWGN channel: BPSK: Pe = Q(√(2·Eb/N0)). The Q-function: Q(x) = (1/2)·erfc(x/√2). At Eb/N0 = 9.6 dB: Pe = 10⁻⁶. This is the best possible uncoded performance for binary modulation. QPSK: Pe = Q(√(2·Eb/N0)), identical to BPSK in Eb/N0 (QPSK sends 2 bits per symbol but at half the symbol rate, so Eb is the same). However, QPSK has twice the spectral efficiency (2 bits/s/Hz vs. 1 bit/s/Hz). M-QAM (square constellation): Pe ≈ (4/log₂(M))·(1 - 1/√M)·Q(√(3·log₂(M)·Eb/(N0·(M-1)))). For 16-QAM (M=16): Pe ≈ (3/4)·Q(√(4·Eb/(5·N0))). Eb/N0 for 10⁻⁶: 14.4 dB. For 64-QAM (M=64): Eb/N0 for 10⁻⁶: 18.8 dB. For 256-QAM (M=256): Eb/N0 for 10⁻⁶: 23.4 dB. Each doubling of constellation order (2 more bits/symbol) costs approximately 3-5 dB more Eb/N0 for the same BER. M-PSK: Pe ≈ (2/log₂(M))·Q(√(2·log₂(M)·Eb/N0)·sin(π/M)). 8-PSK at 10⁻⁶: Eb/N0 = 14.0 dB.

What is coding gain and how does it shift the BER curve?

Coding gain is the reduction in Eb/N0 required to achieve a target BER when forward error correction (FEC) is applied. It shifts the BER curve to the left on the Eb/N0 axis. Examples at BER = 10⁻⁶: Uncoded BPSK: 9.6 dB Eb/N0 required. Convolutional code (R=1/2, K=7, Viterbi): 4.8 dB → coding gain = 4.8 dB. Reed-Solomon (255,223): 6.2 dB → coding gain = 3.4 dB (RS is better at burst errors). Turbo code (R=1/2, 8-state): 1.5 dB → coding gain = 8.1 dB (near Shannon limit of -1.59 dB). LDPC (R=1/2, 5G NR BG1): 1.2 dB → coding gain = 8.4 dB. Shannon limit: Eb/N0 = -1.59 dB for error-free communication at capacity. Modern LDPC codes operate within 0.5-1.0 dB of this limit. The BER curve shape also changes with coding: coded curves have steeper waterfall regions (BER drops faster with Eb/N0 increase) but may exhibit error floors at very low BER (10⁻⁸ to 10⁻¹² for LDPC). Concatenated codes (e.g., LDPC + BCH in DVB-S2) push the error floor below 10⁻¹⁵.

Digital Communications

BER Curve

/bee-ee-ar kurv/

Plot of bit error rate vs. E_b/N₀ characterizing digital modulation/coding performance. BPSK: P_e = Q(√(2·E_b/N₀)), 9.6 dB for 10⁻⁶. 16-QAM: ~14.4 dB. 64-QAM: ~18.8 dB. Coding gain shifts curve left: LDPC (R=1/2) achieves 10⁻⁶ at 1.2 dB, 8.4 dB gain. Shannon limit: E_b/N₀ = −1.59 dB. Waterfall region = steep BER drop. Error floor = code-limited BER plateau.

BPSK 10⁻⁶: 9.6 dB

LDPC gain: ~8.4 dB

Shannon: −1.59 dB

Understanding BER Curves

The BER curve is the Rosetta Stone of digital RF engineering. Every link budget, every modulation selection, every coding decision ultimately comes down to the question: at what E_b/N₀ does the system achieve the required BER? The curve answers this directly and enables comparison across modulation schemes, coding rates, and channel conditions on a common basis.

The normalization to E_b/N₀ (rather than raw SNR) is what makes the curve universal. A system transmitting at 100 Mbps needs more total power than one at 1 Mbps, but if both use BPSK over AWGN, they have identical BER curves. This normalization separates the modulation/coding performance from the system's power and bandwidth budget.

Modulation BER Formulas

BPSK / QPSK:
P_e = Q(√(2·E_b/N₀))
Q(x) = ½·erfc(x/√2)
10⁻⁶: E_b/N₀ = 10.5 dB (uncoded)

M-QAM (square):
P_e ≈ (4/log₂M)·(1 − 1/√M)
× Q(√(3·log₂M·E_b / (N₀·(M−1))))

Spectral Efficiency Conversion:
E_b/N₀(dB) = SNR(dB) − 10·log₁₀(η)
η = bits/s/Hz (1 for BPSK, 6 for 64-QAM)

E_b/N₀ Required for BER = 10⁻⁶

Scheme	Uncoded	With LDPC R=1/2	Coding Gain
BPSK	10.5 dB	1.2 dB	9.3 dB
QPSK	10.5 dB	1.2 dB	9.3 dB
16-QAM	14.4 dB	4.5 dB	9.9 dB
64-QAM	18.8 dB	8.5 dB	10.3 dB
256-QAM	23.4 dB	13.0 dB	10.4 dB

Common Questions

Frequently Asked Questions

What are the axes?

X: E_b/N₀ in dB (normalized per bit). Y: BER on log scale (10⁰ to 10⁻¹²). E_b/N₀ = SNR − 10·log(η). Targets: voice 10⁻³, video 10⁻⁶, fiber 10⁻¹².

Modulation BER formulas?

BPSK/QPSK: Q(√(2E_b/N₀)). M-QAM: scales with √(3·log₂M/(M−1)). Each doubling costs 3 to 5 dB. 8-PSK: sin(π/8) term gives 14.0 dB at 10⁻⁶.

Coding gain?

Leftward BER curve shift from FEC. LDPC R=1/2: ~8.4 dB gain (1.2 dB at 10⁻⁶ vs. 9.6 uncoded). Shannon limit: −1.59 dB. Error floor at 10⁻⁸ to 10⁻¹²; concatenated codes push below 10⁻¹⁵.

Related Terms

← Bent-Pipe Transponder Berlekamp-Massey →

Digital RF

Precision RF Components

RF Essentials provides precision terminations and custom waveguide assemblies for BER test equipment, BERT systems, and digital communication link characterization.

Request a Quote