Skip to main content

Bit error probability of coherent BFSK

Short answer

Coherent BFSK uses orthogonal tones with energy E_b per bit. The decision variable is Gaussian with mean ±AT_b/2 and variance N_0T_b/2. Bit error probability is P_e = ½erfc(√(E_b/2N_0)) = Q(√(E_b/N_0)). This is 3 dB worse than BPSK because orthogonal signals are √(2E_b) apart versus 2√E_b for antipodal signals.

Derive the bit error probability for the coherent binary FSK system.

In binary FSK, bit 1 and bit 0 are sent as two different carrier frequencies f1f_1 and f2f_2 with the same amplitude.

Coherent BFSK signals over 0≤t≤Tb0\le t\le T_b:

s1(t)=Acos⁡ω1t  (1),s2(t)=Acos⁡ω2t  (0),Eb=A2Tb2s_1(t)=A\cos\omega_1t\ \ (1),\qquad s_2(t)=A\cos\omega_2t\ \ (0),\qquad E_b=\frac{A^2T_b}{2}

The tones are chosen orthogonal over TbT_b: ∫0Tbcos⁡ω1tcos⁡ω2t dt=0\int_0^{T_b}\cos\omega_1t\cos\omega_2t\,dt=0 (e.g. ∣f1−f2∣=n/2Tb\lvert f_1-f_2\rvert=n/2T_b).

Coherent receiver: two correlators, one per tone; their outputs are subtracted and compared with zero.

        +->(x)--> integ 0..Tb --> y1 --+
        |   ^ cos w1t                  |(+)
 r(t) --+                            (sum)--> l = y1 - y2
        |   v cos w2t                  |(-)   l > 0 ? 1 : 0
        +->(x)--> integ 0..Tb --> y2 --+

Signal part of ll.

  • 1 sent: y1=ATb2y_1=\frac{AT_b}{2}, y2=0y_2=0 (orthogonality) so l=+ATb2l=+\frac{AT_b}{2}.
  • 0 sent: y1=0y_1=0, y2=ATb2y_2=\frac{AT_b}{2} so l=−ATb2l=-\frac{AT_b}{2}.

Noise part. n1=∫wcos⁡ω1t dtn_1=\int w\cos\omega_1t\,dt and n2=∫wcos⁡ω2t dtn_2=\int w\cos\omega_2t\,dt each have variance N0Tb4\frac{N_0T_b}{4} and are uncorrelated (orthogonal tones), so n=n1−n2n=n_1-n_2 has

σ2=N0Tb4+N0Tb4=N0Tb2\sigma^2=\frac{N_0T_b}{4}+\frac{N_0T_b}{4}=\frac{N_0T_b}{2}

Error probability. ll is Gaussian with mean ±ATb/2\pm AT_b/2 and variance σ2\sigma^2; threshold 0. For 0 sent, error if l>0l>0:

Pe=12erfc⁡(ATb/22 σ)=12erfc⁡(ATb/2N0Tb)=12erfc⁡(A2Tb4N0)\begin{aligned} P_e &= \frac12\operatorname{erfc}\left(\frac{AT_b/2}{\sqrt2\,\sigma}\right)=\frac12\operatorname{erfc}\left(\frac{AT_b/2}{\sqrt{N_0T_b}}\right) \\ &= \frac12\operatorname{erfc}\left(\sqrt{\frac{A^2T_b}{4N_0}}\right) \end{aligned}

With Eb=A2Tb/2E_b=A^2T_b/2:

Pe(coherent BFSK)=12erfc⁡(Eb2N0)=Q(EbN0)P_e(\text{coherent BFSK})=\frac12\operatorname{erfc}\left(\sqrt{\frac{E_b}{2N_0}}\right)=Q\left(\sqrt{\frac{E_b}{N_0}}\right)

Check with the general formula: Ed=∫(s1−s2)2dt=Eb+Eb=2EbE_d=\int(s_1-s_2)^2dt=E_b+E_b=2E_b, so 12erfc⁡Ed/4N0=12erfc⁡Eb/2N0\frac12\operatorname{erfc}\sqrt{E_d/4N_0}=\frac12\operatorname{erfc}\sqrt{E_b/2N_0}.

Comparison: BFSK has the same PeP_e as ASK (for equal average energy) and needs 3 dB more Eb/N0E_b/N_0 than BPSK, because orthogonal signals are 2Eb\sqrt{2E_b} apart while antipodal ones are 2Eb2\sqrt{E_b} apart. Non-coherent FSK gives Pe=12e−Eb/2N0P_e=\frac12e^{-E_b/2N_0}, which is slightly worse but needs no carrier phase.

✓ Answer checked by Claude Opus 5.5 (high effort)

All Communication Systems chapter 9 questions →