Skip to content
Tutorials Center
My learning

Topics

Lesson 29 of 46 ยท Noise and Error Performance

Bit Error Rate Curves

Read and compute bit error rate curves: formulas for BPSK, QPSK, DPSK, FSK and M-QAM versus Eb/N0, how SNR relates to Eb/N0, and how to find the required Eb/N0.

A bit error rate curve is the most important picture in digital communications. It plots, for one modulation scheme, how often bits go wrong (vertical axis, logarithmic) against how much energy per bit you spend compared with the noise (horizontal axis, in dB). Every modem datasheet and every standards document has these curves, and a link designer's day often starts with the question "what Eb/N0 do I need for a BER of one in a hundred thousand?"

In this lesson you collect the formulas for the schemes from chapter 5, learn to convert between SNR, Es/N0 and Eb/N0, and compute the required Eb/N0 for a target BER.

Four BER curves on a log axis against Eb/N0: BPSK and QPSK lowest, then 16-QAM, DPSK and noncoherent FSK. BPSK reaches 10^-5 at about 9.6 dB.
Curves further left are better: they reach the same error rate with less energy per bit.

What you will learn

  • Write the BER formulas of BPSK, QPSK, DPSK, FSK and square QAM.
  • Convert among SNR, Es/N0 and Eb/N0.
  • Read a "waterfall" curve and find the Eb/N0 for a given BER.
  • Explain why the curves are so steep.
  • Distinguish symbol error rate from bit error rate.

Key terms

  • Eb/N0 (ฮณ): bit energy over noise density; dimensionless, usually in dB.
  • Es/N0: symbol energy over noise density; Es/N0 = Eb/N0 ยท log2(M).
  • Waterfall curve: BER falling rapidly as Eb/N0 rises.
  • Ps and Pb: symbol error probability and bit error probability.
  • Error floor: a BER that stops improving, caused by something other than thermal noise.

Three ratios, one conversion

The same link can be described by three signal-to-noise numbers.

Name Definition Use
SNR S / (N0ยทB) what a spectrum analyzer shows
Es/N0 SNR ยท B / Rs symbol decisions
Eb/N0 SNR ยท B / Rb fair comparison between schemes

They differ only by the ratios of the bandwidth to the symbol or bit rate. For a modulation with k = log2(M) bits per symbol, Es/N0 = k ยท Eb/N0. In dB, add 10ยทlog10(k): for 16-QAM that is 6 dB. If the signal occupies bandwidth B = Rs (ideal Nyquist), SNR = Es/N0.

The formulas

Let ฮณ = Eb/N0 (linear), and Q(x) = ยฝยทerfc(x/โˆš2).

Scheme Bit error probability
BPSK, coherent Q(โˆš(2ฮณ))
QPSK, Gray, coherent Q(โˆš(2ฮณ)) (same as BPSK)
DPSK ยฝ ยท e^(โˆ’ฮณ)
Coherent orthogonal binary FSK Q(โˆšฮณ)
Noncoherent orthogonal binary FSK ยฝ ยท e^(โˆ’ฮณ/2)
Square M-QAM, Gray (approximation) (4/log2 M) ยท (1 โˆ’ 1/โˆšM) ยท Q(โˆš(3ยทlog2(M)ยทฮณ / (M โˆ’ 1)))

For 16-QAM the last row becomes 0.75 ยท Q(โˆš(0.8ยทฮณ)). For 64-QAM it is 0.583 ยท Q(โˆš(0.286ยทฮณ)). The QAM expression counts only the nearest-neighbor errors and assumes Gray coding, which is accurate at the BERs of interest.

What the table says: BPSK and QPSK are the best in this list. Coherent FSK needs 3 dB more. DPSK is only about 0.7 dB worse than BPSK and avoids phase recovery. Noncoherent FSK pays another small penalty for needing no phase information at all. QAM needs more Eb/N0 as M grows, but uses less bandwidth per bit.

Why the waterfall is so steep

Errors come from the Gaussian tail, and the Gaussian tail shrinks like e^(โˆ’xยฒ/2). For BPSK at Eb/N0 = 4 dB the BER is about 1.3 ร— 10โปยฒ; at 8 dB it is about 1.9 ร— 10โปโด; at 10 dB, 3.9 ร— 10โปโถ. Going from 4 dB to 8 dB cuts the error rate by a factor of about 65, and two more decibels cut it by about 50 again. That sensitivity is the lever of every link budget: a small improvement in antenna gain or noise figure is worth a lot of reliability.

Two practical targets to remember for BPSK or QPSK: BER 10โปยณ needs Eb/N0 โ‰ˆ 6.8 dB; BER 10โปโต needs about 9.6 dB.

Symbol errors versus bit errors

Constellation analysis first gives the symbol error probability Ps. With Gray coding the likeliest symbol error flips one bit among k, so Pb โ‰ˆ Ps / k. Without Gray coding, an error can flip several bits and Pb is larger. Always check which of the two a plotted curve shows.

Beyond thermal noise: error floors

Real curves can level off. Phase noise, quantization noise, nonlinear distortion and interference scale with the signal, so increasing power does not shrink them. Their effect appears as a BER floor, and the cure is not more power but a better component or algorithm.

Common misconceptions

  • "A lower BER means a better modulation." Only at equal Eb/N0 and with the same bandwidth considerations. 64-QAM has a higher BER than BPSK at the same Eb/N0 but delivers 6 times the data per symbol.
  • "BER is the same as SNR." SNR is an input condition; BER is the consequence.
  • "Adding 3 dB always halves the errors." The drop is far larger because the curve is a steep waterfall, not a straight line on a linear axis.
  • "QPSK at the same Eb/N0 has twice the BER." With Gray coding it is identical to BPSK.

Try it

First, a BER calculator. It plots all curves and computes the Eb/N0 each scheme needs for BER 10โปโต.

Python BER curves and required Eb/N0

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Second, test the 16-QAM formula with a Monte Carlo simulation using Gray-coded levels โˆ’3, โˆ’1, +1, +3 on each axis.

Python 16-QAM by simulation

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Check your understanding

Quick check ยท 5 questions

of 5 answered You scored / 5 ยท perfect!

  1. 1. A 16-QAM link has Eb/N0 = 8 dB. What is its Es/N0?

    Es/N0 = Eb/N0 ยท log2(16) = 8 dB + 10ยทlog10(4) = 8 + 6 = 14 dB.
  2. 2. Which scheme has the same Gray-coded bit error rate as BPSK at the same Eb/N0?

    QPSK is two independent BPSK streams on orthogonal carriers.
  3. 3. Roughly what Eb/N0 does BPSK need for a bit error rate of 10^-5?

    Q(โˆš(2ฮณ)) = 10โปโต at ฮณ โ‰ˆ 9.1, i.e. about 9.6 dB.
  4. 4. Why does a Gray-coded system have Pb โ‰ˆ Ps / k?

    With Gray labeling, nearest-neighbor symbol errors flip a single bit.
  5. 5. A BER curve flattens at 10^-6 no matter how much power is added. What is the likely cause?

    Impairments that scale with the signal cannot be beaten by more power.

Summary

  • Es/N0 = k ยท Eb/N0 and SNR = (Eb/N0) ยท Rb/B convert between the three ratios.
  • BPSK and QPSK: Pb = Q(โˆš(2ยทEb/N0)); coherent FSK needs 3 dB more; DPSK is within about 1 dB of BPSK.
  • Square M-QAM needs more Eb/N0 as M grows in exchange for spectral efficiency.
  • BER curves are steep waterfalls; floors reveal impairments beyond thermal noise.

Next: Chapter 6 review