Soft Compression

Unlike hard compression where gain drops abruptly at a distinct saturation point, soft compression exhibits a progressive reduction in incremental gain over an extended input power range. This smooth transition begins well before the 1-dB compression point (P1dB) and continues gradually toward saturation.

• GaN HEMT devices inherently demonstrate this behavior due to their wide bandgap material properties
• The gain compression curve shows a gentle slope rather than a sharp knee
• This characteristic reduces the severity of AM-AM distortion at moderate drive levels
• Allows the amplifier to operate closer to saturation without catastrophic nonlinearity

Soft compression characteristics are inherently more amenable to digital predistortion (DPD) than hard compression profiles. The gradual nonlinearity can be accurately captured by polynomial-based behavioral models without requiring extremely high-order terms.

• Memory polynomial models converge more rapidly when fitting soft compression curves
• Reduces the computational complexity of the predistorter implementation
• Lower-order Volterra kernels can achieve equivalent linearization performance
• Enables wider linearization bandwidth due to simpler coefficient estimation

The soft compression mechanism in Gallium Nitride High Electron Mobility Transistors originates from the device physics of the two-dimensional electron gas (2DEG) channel. As the input drive increases, the gradual depletion of the channel under the gate contact causes a smooth reduction in transconductance (gm).

• Unlike MOSFET hard saturation, GaN HEMTs show progressive channel pinch-off
• Field-plate designs further smooth the electric field distribution at the drain edge
• Reduced knee voltage effects contribute to the extended linear-like operation region
• The absence of abrupt avalanche breakdown mechanisms preserves the soft characteristic

Hard compression occurs in technologies like LDMOS and some bipolar devices where gain remains nearly constant until a sharp saturation point, after which it collapses abruptly. This creates a highly nonlinear discontinuity that is mathematically challenging to model.

• Hard compression generates stronger harmonic distortion and spectral regrowth
• Requires higher-order nonlinear terms in the predistorter model
• Look-up table (LUT) predistorters need finer granularity near the compression knee
• Soft compression distributes the nonlinearity across a wider dynamic range, simplifying correction

In Doherty power amplifier architectures, the soft compression characteristic of GaN HEMTs directly benefits the back-off efficiency profile. The carrier amplifier's gradual compression allows smoother load modulation transitions when the peaking amplifier activates.

• The AM-PM distortion also exhibits a softer phase transition, reducing phase discontinuity at the Doherty combining point
• Enables asymmetric Doherty designs with extended back-off range while maintaining linearizability
• The smooth gain characteristic reduces memory effects associated with rapid impedance changes
• Facilitates broadband Doherty operation where compression behavior must remain consistent across frequency

While soft compression simplifies DPD, accurate behavioral modeling must still capture the extended nonlinear transition region. Generalized memory polynomial (GMP) models with carefully selected basis functions are typically employed to represent the smooth compression curve.

• Cross-term memory effects remain significant even with soft compression due to thermal and trapping time constants
• The Rapp model and Saleh model can approximate soft compression but lack memory effect representation
• Neural network linearization benefits from the smooth activation-like compression characteristic
• Model extraction requires characterization across the full compression range, not just at P1dB

The contrasting nonlinear behavior to soft compression, characterized by an abrupt, sharp saturation of output power as input drive increases. Common in silicon LDMOS and some bipolar technologies, hard compression creates a steep gain roll-off and severe phase discontinuities that are fundamentally more difficult to linearize with digital predistortion. The sudden clipping generates high-order intermodulation products and spectral splatter that polynomial-based DPD models struggle to capture without extremely high nonlinear orders.

The amplitude-to-amplitude modulation distortion that quantifies the nonlinear relationship between input envelope magnitude and output envelope magnitude. In soft compression, the AM-AM curve exhibits a gradual, monotonic roll-off from linear gain rather than a sharp knee. This smooth characteristic can be accurately modeled with lower-order polynomials or neural networks with fewer parameters, reducing the computational burden on the DPD coefficient estimation engine.

The amplitude-to-phase modulation distortion representing the input envelope-dependent phase shift introduced by the amplifier. GaN HEMTs exhibiting soft compression often display a more benign AM-PM profile with reduced phase spreading at high power levels compared to devices with abrupt saturation. This predictable phase behavior simplifies the predistorter's phase correction function and improves the error vector magnitude (EVM) after linearization.

Dynamic nonlinear distortions where the amplifier's current output depends on past signal envelope values due to thermal, electrical, and charge-trapping time constants. Soft compression in GaN HEMTs is often accompanied by distinct long-term thermal memory from self-heating and gate lag from trap effects. While the compression onset is smooth, these memory mechanisms introduce hysteresis in the AM-AM and AM-PM curves that must be captured by the memory depth of the DPD model.

The Gallium Nitride High Electron Mobility Transistor is the primary semiconductor technology exhibiting soft compression characteristics. Key properties enabling this behavior include:

• Wide bandgap for high breakdown voltage and power density
• Low knee voltage extending the linear voltage swing range
• Gradual transconductance reduction at high gate voltages
• Superior thermal conductivity on SiC substrates mitigating abrupt thermal runaway These intrinsic material properties make GaN HEMTs inherently more linearizable than silicon alternatives.

The standard metric marking the onset of gain compression, defined as the output power level where the amplifier's gain drops by 1 dB from its small-signal value. In devices with soft compression, the transition through P1dB is gradual and extended over several dB of input power, making the exact compression point less distinct. This contrasts with hard-compressing devices where gain collapses rapidly after P1dB. DPD designers use this metric to set the target linearization range.