How does the signal bandwidth affect the memory effects in a power amplifier?
PA Memory Effects and Bandwidth
Memory effects are a critical consideration in modern wideband PA design. As signal bandwidths increase (100+ MHz for 5G), memory effects become increasingly dominant in determining the PA's linearity.
| Parameter | Class A | Class AB | Class F/Doherty |
|---|---|---|---|
| Max Efficiency | 50% | 50-78% | 70-90% |
| Linearity | Excellent | Good | Moderate (needs DPD) |
| P1dB Backoff | 0-3 dB | 3-6 dB | 6-10 dB |
| Complexity | Low | Low | High |
| Common Use | Test, small signal | General PA | Base station, broadcast |
- Performance verification: confirm specifications against the application requirements before finalizing the design
- Environmental factors: temperature range, humidity, and vibration affect long-term reliability and parameter drift
- Cost vs. performance: evaluate whether the application demands premium components or standard commercial grades
Frequently Asked Questions
How do I measure memory effects?
Two-tone test with swept tone spacing: apply two equal-power tones and sweep the tone spacing from 100 kHz to 100 MHz. Measure the upper and lower IM3 product magnitudes versus tone spacing. If the IM3 products are equal at all spacings: no memory effects. If the IM3 products diverge at wider spacings: memory effects are present, and the divergence indicates the memory bandwidth. Modulated signal test: apply a modulated signal (OFDM, QAM) with different bandwidths and measure the ACLR. If the ACLR degrades as the bandwidth increases (beyond the degradation expected from the increased PAPR): memory effects are contributing.
How do I reduce memory effects in the PA design?
Thermal memory: improve the thermal design (lower junction-to-case thermal resistance, lower thermal impedance) to reduce the temperature swing with signal power. Use larger die sizes (lower thermal resistance). Electrical memory: increase the bias network bandwidth by using: low-impedance bias feeds (multiple bypass capacitors at different frequencies), inductorless bias networks (resistive bias provides flat frequency response but wastes DC power), and envelope tracking (the bias supply tracks the signal envelope, eliminating the bias modulation). GaN trapping: use process-optimized GaN epitaxy with reduced buffer trapping (iron-doped vs. carbon-doped, field plates, surface passivation).
What DPD bandwidth is needed?
The DPD must capture the signal content up to 3-5× the signal bandwidth (to correct the 3rd and 5th order intermodulation products). For a 100 MHz 5G signal: DPD bandwidth = 300-500 MHz. For a 200 MHz signal: DPD bandwidth = 600-1000 MHz. This means: the observation receiver (feedback path) must sample at 600-1000 MSPS, and the DPD processing must run at this rate. The computational requirement scales as BW^2 or BW^3 for memory polynomial models.