Waterfilling Power Allocation in MIMO and OFDM
The waterfilling algorithm is used in wireless communications to optimally allocate power across multiple channels (or subcarriers) to maximize data rate under a total power constraint.
Intuition
Imagine pouring water into vessels with different floor heights: - Each vessel = a channel - Floor height = inverse of channel gain - More power (water) goes to channels with better gains (deeper vessels)
Mathematical Formulation
Objective:
Maximize total data rate:
Subject to:
Optimal Power Allocation:
Where: - \( \mu \): Water level (chosen to meet total power constraint) - \( (\cdot)^+ \): Zero if negative
Application in Wireless Systems
OFDM Frequency-Domain Waterfilling
- Apply across subcarriers with different gains \( |H_k| \)
- More power to strong subcarriers, none to weak ones
MIMO Spatial Waterfilling
- Use SVD: \( \mathbf{H} = \mathbf{U} \boldsymbol{\Sigma} \mathbf{V}^H \)
- Power allocated to eigenchannels \( \sigma_i^2 \):
Summary Table
Use Case | Domain | Channel Basis |
---|---|---|
OFDM | Frequency | Subcarriers |
MIMO | Spatial | Eigenmodes (SVD) |
General | Any | Parallel AWGN channels |
Benefits
- Maximizes capacity
- Skips bad (deep fade) channels
- Dynamic and adaptive power use
Does the Transmitter Need CSI?
? Yes ? Channel State Information at the Transmitter (CSIT) is required for waterfilling.
Why?
Waterfilling allocates more power to better channels: - In OFDM, this means knowing each subcarrier's gain \( |H_k|^2 \) - In MIMO, this means knowing the SVD of the channel matrix \( \mathbf{H} \) to extract \( \sigma_i^2 \)
Without CSIT, the transmitter cannot: - Differentiate between good and bad channels - Apply waterfilling (no way to adjust \( P_i \))
Result:
- Use uniform power allocation if no CSIT is available
- Waterfilling is not possible without CSIT
_Last updated: June 06, 2025