Bit-Interleaved Coded Modulation (BICM)

Definition:

Bit-Interleaved Coded Modulation (BICM)

Bit-interleaved coded modulation (BICM) is a pragmatic approach to coded modulation that separates coding and modulation through a bit-level interleaver:

  1. Encode: A binary code (turbo, LDPC, or polar) produces coded bits.
  2. Interleave: A bit-level interleaver permutes the coded bits.
  3. Map: Groups of m=log⁡2Mm = \log_2 M interleaved bits are mapped to constellation points from an MM-ary modulation (e.g., 16-QAM).

At the receiver, the demapper computes bit-level LLRs for each coded bit, the deinterleaver restores the original order, and the decoder operates on these LLRs.

BICM achieves near-optimal performance by ensuring that coded bits mapped to the same constellation point experience effectively independent fading or noise realisations (after deinterleaving).

BICM is the dominant coded-modulation approach in modern wireless standards (Wi-Fi, 4G LTE, 5G NR) because it decouples code design from modulation design.

,

Theorem: BICM Capacity Approaches CM Capacity with Gray Mapping

The BICM capacity for an MM-ary constellation with bit-labelling μ\mu over the AWGN channel is

CBICM=∑j=1mI(bj;Y)C_{\text{BICM}} = \sum_{j=1}^{m} I(b_j; Y)

where bjb_j is the jj-th bit in the label and YY is the channel output. In general,

CBICM≤CCM=I(X;Y)C_{\text{BICM}} \leq C_{\text{CM}} = I(X; Y)

with equality only for BPSK/QPSK. However, with Gray mapping, the gap CCM−CBICMC_{\text{CM}} - C_{\text{BICM}} is negligible (typically <0.1< 0.1 dB) for all standard constellations (16-QAM, 64-QAM, 256-QAM) across the practical SNR range.

BICM treats each bit position independently, losing the correlation between bits mapped to the same symbol. Gray mapping minimises this loss because neighbouring constellation points differ in exactly one bit, making each bit position's channel nearly independent.

Proof
Chain rule decomposition

I(X;Y)=I(b1,…,bm;Y)=∑j=1mI(bj;Y∣b1,…,bj−1)≥∑j=1mI(bj;Y)=CBICMI(X; Y) = I(b_1, \ldots, b_m; Y) = \sum_{j=1}^{m} I(b_j; Y \mid b_1, \ldots, b_{j-1}) \geq \sum_{j=1}^{m} I(b_j; Y) = C_{\text{BICM}}

The inequality follows because conditioning cannot decrease mutual information. BICM ignores the conditioning.

Near-equality with Gray mapping

With Gray labelling, the conditional distributions p(Y∣bj=0)p(Y \mid b_j = 0) and p(Y∣bj=1)p(Y \mid b_j = 1) are nearly independent of the other bits, so the conditioning terms contribute negligibly. Numerical evaluation confirms the gap is <0.1< 0.1 dB for all practical cases. ■\blacksquare

,

BICM Capacity vs. Coded Modulation Capacity

Compare BICM capacity with coded modulation (CM) capacity and the AWGN channel capacity for different constellation sizes. Observe how small the gap is with Gray mapping.

Parameters

Example: BICM with 16-QAM

A system uses a rate-3/4 LDPC code with 16-QAM and BICM over an AWGN channel.

(a) What is the spectral efficiency?

(b) At what Eb/N0E_b/N_0 does the BICM capacity equal this spectral efficiency?

(c) Compare with the Shannon limit.

Solution
Spectral efficiency

η=Rc×m=34×4=3\eta = R_c \times m = \frac{3}{4} \times 4 = 3 bits/s/Hz.

BICM capacity threshold

From the BICM capacity curve for 16-QAM with Gray mapping, CBICM=3C_{\text{BICM}} = 3 bits/channel use requires SNR≈9.8\text{SNR} \approx 9.8 dB, or equivalently Eb/N0≈9.8−10log⁡10(3)≈5.0E_b/N_0 \approx 9.8 - 10\log_{10}(3) \approx 5.0 dB.

Shannon limit comparison

The unconstrained AWGN Shannon limit at 3 bits/s/Hz: Eb/N0=(23−1)/(3)=7/3=2.33E_b/N_0 = (2^3 - 1)/(3) = 7/3 = 2.33, which is 10log⁡10(2.33)=3.6810\log_{10}(2.33) = 3.68 dB.

The BICM constraint costs about 1.3 dB compared to Shannon. A good LDPC code will operate within 1-2 dB of the BICM capacity, so the total gap to Shannon is about 2.5-3.3 dB. ■\blacksquare

🎓CommIT Contribution(1998)

Foundational Theory of Bit-Interleaved Coded Modulation

G. Caire, G. Taricco, E. Biglieri — IEEE Trans. Inform. Theory, vol. 44, no. 3, pp. 927-946

This seminal paper established the information-theoretic framework for BICM, proving that the pragmatic approach of separating binary coding from higher-order modulation through a bit interleaver achieves near-optimal performance with Gray mapping. The paper introduced the BICM capacity formula CBICM=∑jI(bj;Y)C_{\text{BICM}} = \sum_j I(b_j; Y), analysed performance over fading channels, and demonstrated that the capacity gap relative to coded modulation is negligible (<0.1< 0.1 dB) for standard constellations. This work became the foundation for coded modulation in all modern wireless standards (Wi-Fi, LTE, 5G NR). The follow-up monograph by Guillén i Fàbregas, Martinez, and Caire (2008) extended the framework to iterative decoding and modern code design.

bicmcoded-modulationcapacityfadingView Paper →
🎓CommIT Contribution(2008)

BICM: Foundations and Trends Monograph

A. Guillén i Fàbregas, A. Martinez, G. Caire — Foundations and Trends in Communications and Information Theory, vol. 5, no. 1-2

This comprehensive 153-page monograph extends the original 1998 BICM paper to cover fading channels, mismatched decoding, iterative BICM-ID with set-partitioning labelling, and connections to modern LDPC/turbo code design. It provides the definitive theoretical treatment of BICM, unifying the information-theoretic analysis with practical code design guidelines used in current standards.

bicmmonographiterative-decodingcode-designView Paper →

BICM with Iterative Decoding (BICM-ID)

BICM-ID feeds back soft information from the decoder to the demapper, enabling iterative exchange:

  1. The demapper uses decoder feedback (a priori LLRs on coded bits) to refine its bit-level LLR outputs.
  2. The decoder processes these improved LLRs and sends updated extrinsic information back to the demapper.

BICM-ID can close the gap between BICM and CM capacity, but requires set-partitioning labelling (not Gray) to benefit from iterations. With Gray mapping, the demapper gain from feedback is negligible.

In practice, BICM-ID is used in some DVB standards but not in cellular systems (Wi-Fi, LTE, 5G NR), where the simpler non-iterative BICM with Gray mapping is preferred.

Quick Check

Why does BICM with Gray mapping achieve near-optimal performance despite treating each bit position independently?

Gray mapping makes all bit positions equally reliable

The interleaver creates time diversity

Gray mapping minimises the mutual information loss from ignoring inter-bit dependencies

BICM uses a stronger code than coded modulation

Correction:
Gray mapping minimises the mutual information loss from ignoring inter-bit dependencies

With Gray mapping, neighbouring symbols differ in exactly one bit, making each bit channel nearly independent of the others. The loss CCM−CBICMC_{\text{CM}} - C_{\text{BICM}} is negligible.

BICM

Bit-interleaved coded modulation: a coded-modulation scheme that separates binary coding from higher-order modulation through a bit-level interleaver. Near-optimal with Gray mapping.

Related: Coded Modulation, Gray Mapping Minimises BER, Interleaver

Coded Modulation

The joint design or analysis of channel coding and modulation. Includes trellis-coded modulation (TCM), multilevel coding (MLC), and BICM as different approaches.

Related: Bit-Interleaved Coded Modulation (BICM), Tcm, Constellation Diagram

Polar CodesHybrid ARQ