Design and implementation of a high-performance V-band CMOS bandpass filter

✍ Scribed by Jin-Fa Chang; Yo-Sheng Lin

• Publisher: John Wiley and Sons
• Year: 2010
• Tongue: English
• Weight: 662 KB
• Volume: 52
• Category: Article
• ISSN: 0895-2477
• DOI: 10.1002/mop.24935

✦ Synopsis

Testing is performed by comparison of results from the RBFNN model and the ADS circuit model with the test frequency points (2.41-2.535 GHz, step-size ¼ 0.025 GHz), and input powers (À9-33 dBm, step-size ¼ 1.5 dBm). Figures 4(a) and 4(b) show the AM/AM and AM/PM distortion versus input powers at different test frequency points, respectively. Results in Figure 4 indicate that the RBFNN model reproduce these nonlinear distortion accurately at different test frequency points.

An additional comparison of PAE, Pout, and Gain performances among the RBFNN model, the ADS circuit model, and measurement is presented in Figure 5. As shown in Figure 5, the PAE, Pout, and Gain curves from the RBFNN model march well with the circuit simulations at the operating frequency of 2.4 GHz. For example, the 1-dB compression point is 31.0 dBm for the RBFNN model which agrees well with the circuit model of 30.9 dBm. The fabricated GaN Doherty power amplifier is shown in Figure 6. 5. CONCLUSION One GaN Doherty amplifier was designed for 2.4 GHz and a RBFNN model is proposed for predicting this amplifier's dynamic nonlinear characteristics. Comparisons of AM/AM, AM/PM, PAE, Pout, and Gain curves among the RBFNN model, the ADS circuit model, and measurement are given. After 75 epochs, the convergence of this RBFNN model becomes slower and the training error reaches a lower value (below 1.2% error level). The results indicate that the proposed RBFNN model can reproduce the nonlinear transfer characteristics efficiently and accurately.