Inferensys

Glossary

Shaping Function

A deterministic mapping function, often implemented as a look-up table, that translates the instantaneous baseband signal magnitude into a target supply voltage for the power amplifier to optimize efficiency and linearity.
Knowledge engineer constructing knowledge base on laptop, document hierarchy visible, casual office setup.
ENVELOPE TRACKING

What is Shaping Function?

A shaping function is a deterministic mapping that translates the instantaneous baseband signal magnitude into a target supply voltage for a power amplifier, optimizing the trade-off between efficiency and linearity in envelope tracking systems.

A shaping function is a deterministic mapping, typically implemented as a look-up table (LUT) or polynomial, that converts the instantaneous envelope magnitude of an RF signal into a corresponding drain voltage for the power amplifier. Its primary purpose is to define the dynamic supply trajectory that keeps the PA operating near compression while avoiding excessive nonlinearity, directly determining the system's overall power-added efficiency (PAE) and spectral purity.

The function is designed using the PA's iso-gain contours—constant gain curves plotted against supply voltage and input power—to maintain linear operation during dynamic modulation. A poorly designed shaping function introduces ET-induced AM/AM and AM/PM distortion, while an optimized one minimizes supply modulator slew-rate demands and prevents clipping at signal peaks, enabling the ET-DPD co-design process to achieve maximum efficiency without violating spectral mask requirements.

ENVELOPE TRACKING FUNDAMENTALS

Key Characteristics of Shaping Functions

The shaping function is the critical deterministic mapping that translates instantaneous baseband signal magnitude into a target supply voltage, balancing efficiency gains against linearity degradation in envelope tracking systems.

01

Deterministic Magnitude-to-Voltage Mapping

A shaping function is a static, memoryless transfer function that maps the instantaneous envelope magnitude |x(n)| of the baseband signal to a target drain voltage V_dd(n) for the power amplifier. This mapping is typically implemented as a look-up table (LUT) indexed by the signal's instantaneous power or magnitude. The function is designed offline using iso-gain contour analysis of the PA's characteristic curves, ensuring that for every input envelope value, the selected supply voltage maintains the required gain while minimizing DC power consumption. Unlike adaptive algorithms, the shaping function itself does not change during operation—it is a fixed, pre-characterized curve.

1D LUT
Typical Implementation
|x(n)|
Input Index
02

Efficiency vs. Linearity Trade-Off

The shaping function embodies the fundamental efficiency-linearity trade-off in envelope tracking. An aggressive shaping function that deeply modulates the supply voltage down to the ET efficiency knee maximizes power savings but introduces significant supply-dependent gain compression and ET-induced AM/PM distortion. Conversely, a conservative shaping function that maintains higher minimum voltages preserves linearity at the cost of reduced efficiency. The optimal shaping function is a compromise, often designed with a voltage floor (V_min) below which the supply is not allowed to drop, preventing the PA from entering its highly nonlinear compression region while still capturing the majority of efficiency gains.

V_min
Voltage Floor Parameter
>30%
Typical Efficiency Gain
03

Iso-Gain Contour Derivation

Shaping functions are derived from the PA's iso-gain contours—constant-gain curves plotted on the two-dimensional plane of supply voltage (V_dd) versus input power (P_in). These contours are obtained through load-pull measurements or nonlinear vector network analyzer (NVNA) characterization under varying DC bias conditions. The shaping function is constructed by selecting a target gain value and tracing the corresponding iso-gain contour across the input power range. For each input power level, the contour specifies the minimum V_dd that maintains the desired gain, directly yielding the shaping table entries. This process ensures the PA operates at the boundary of compression for maximum efficiency.

V_dd vs P_in
Characterization Plane
NVNA
Measurement Tool
04

Interaction with Digital Predistortion

The shaping function directly influences the complexity required of the digital predistorter (DPD). An aggressive shaping function that pushes the PA deep into compression generates strong nonlinearities with significant memory effects, requiring a high-order DPD model such as a generalized memory polynomial (GMP) or augmented Volterra series. The shaping function and DPD must be co-designed: the shaping function determines the operating trajectory across the PA's nonlinear surface, and the DPD must be trained on data that captures this specific trajectory. A poorly chosen shaping function can produce nonlinearities that exceed the correction capability of any practical DPD, resulting in residual spectral regrowth.

GMP
Required DPD Model
Co-Design
Development Approach
05

Bandwidth and Slew Rate Constraints

The shaping function's output—the target supply voltage waveform—must be physically realizable by the supply modulator. The envelope bandwidth of this waveform is typically 3-5× the RF signal bandwidth due to the nonlinear magnitude-to-voltage mapping. If the shaping function generates voltage transitions that exceed the modulator's slew rate capability, the actual supply voltage will lag behind the target, causing envelope-bandwidth mismatch distortion. Shaping functions are therefore often bandwidth-limited through filtering or designed with smooth transitions to ensure the resulting voltage waveform stays within the modulator's tracking bandwidth, trading some efficiency for realizable tracking fidelity.

3-5×
Envelope Bandwidth Expansion
V/μs
Slew Rate Unit
06

Generalized Shaping with 3D LUT Structures

For advanced ET-DPD systems, the simple 1D shaping function is extended to a 3D look-up table (3D LUT) indexed by both instantaneous input power and the current supply voltage. This structure captures the supply-dependent nonlinear behavior of the PA, where the optimal predistortion correction depends not only on the input envelope but also on the instantaneous operating voltage. The 3D LUT effectively combines the shaping function and the memoryless DPD into a single unified structure, enabling joint optimization of efficiency and linearity. Each table entry stores a complex gain correction value, and interpolation is used between indexed points for smooth operation.

3D
LUT Dimensionality
Complex Gain
Stored Value Type
SHAPING FUNCTION ESSENTIALS

Frequently Asked Questions

Clear, technical answers to the most common questions about envelope tracking shaping functions, their design, and their critical role in optimizing power amplifier efficiency and linearity.

A shaping function is a deterministic mapping function, typically implemented as a look-up table (LUT), that translates the instantaneous baseband signal magnitude into a target supply voltage for the power amplifier (PA). Its primary purpose is to optimize the trade-off between power-added efficiency (PAE) and linearity by ensuring the PA operates near its compression point without saturating. The function defines the trajectory that the dynamic supply voltage follows in response to the RF envelope, directly influencing gain compression characteristics and adjacent channel leakage ratio (ACLR).

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.