What is the water-filling algorithm?

Water-filling is the information-theoretic optimal power and bit allocation. It allocates more power to subcarriers with higher SNR and less to weaker ones, like pouring water into a vessel where the bottom represents the noise floor. The total capacity is C = sum of log2(1 + SNR_k * P_k) across all subcarriers. In practice, discrete bit loading approximates water-filling with integer modulation orders.

Rate adaptive vs margin adaptive?

Rate-adaptive bit loading maximizes throughput for a given total power and target BER. Margin-adaptive bit loading maximizes noise margin for a given target data rate and power, providing more robustness. DSL systems typically use margin-adaptive during initialization and rate-adaptive during operation.

Digital Communications

Bit Loading

Q: Where is bit loading used?

DSL (ADSL, VDSL) was the first major application, using DMT (Discrete Multi-Tone) with per-tone bit loading. Wi-Fi 6/7 uses OFDMA with MCS selection per RU. Powerline communications (G.hn) uses bit loading extensively. 5G NR does not use per-subcarrier bit loading but selects a single MCS per resource block group based on CQI feedback.

Bit Loading assigns different modulation orders to individual OFDM subcarriers based on their measured channel SNR. Subcarriers with high SNR carry higher-order modulation (e.g., 256-QAM = 8 bits/symbol); subcarriers with low SNR use lower-order modulation (QPSK = 2 bits) or are turned off entirely. This per-subcarrier adaptation maximizes aggregate throughput while maintaining a target BER. The information-theoretic optimal allocation follows the water-filling algorithm.

Category: Digital Communications

Key Systems: DSL, Wi-Fi, PLC

Understanding Bit Loading

In a frequency-selective channel, different subcarriers experience different attenuation and noise levels. Using the same modulation for all subcarriers wastes capacity on strong subcarriers (could carry more bits) and creates errors on weak ones (need fewer bits). Bit loading optimally distributes bits across subcarriers to maximize the total data rate for a given total transmit power and target BER.

The water-filling theorem shows the optimal continuous allocation: allocate power proportional to (threshold − noise) for each subcarrier. In practice, bit loading is discrete (integer bits per symbol) and uses algorithms like Levin-Campello or Hughes-Hartogs to iteratively assign bits to the subcarrier where the next bit costs the least power.

Water-Filling Capacity

Channel capacity with bit loading:
C = ∑_k=1^N log₂(1 + SNR_k·P_k/P_total)

Water-filling power allocation:
P_k = max(0, μ − N₀/|H_k|²)
μ chosen so ∑P_k = P_total

Bit Loading in Standards

Standard	Subcarriers	Bit Loading	Max Bits/SC
ADSL2+	512	Per-tone	15
VDSL2	4,096	Per-tone	15
G.hn (PLC)	2,048	Per-tone	12
Wi-Fi 6	Per RU	Per-RU MCS	10 (1024-QAM)
5G NR	Per RBG	Single MCS	8 (256-QAM)

Common Questions

Frequently Asked Questions

What is water-filling?

The optimal power/bit allocation: more power to high-SNR subcarriers, less to weak ones. Total capacity = ∑log₂(1+SNR_k·P_k). Discrete bit loading approximates this with integer modulation orders.

Where is bit loading used?

DSL (per-tone DMT), Wi-Fi 6/7 (per-RU MCS), powerline (G.hn). 5G NR uses single MCS per RBG, not per-subcarrier loading.

Rate vs margin adaptive?

Rate-adaptive maximizes throughput for a given power. Margin-adaptive maximizes noise margin for a target rate. DSL uses margin-adaptive at init, rate-adaptive during operation.

Related Terms

← Bit Interleaving Bit Rate →