High-fidelity PWM inverter for digital audio amplification: Spectral analysis, real-time DSP implementation, and results

César Pascual, Zukui Song, Philip T. Krein, Dilip V. Sarwate, Pallab Midya, William Bill J. Roeckner

Research output: Contribution to journalArticlepeer-review


A complete digital audio amplifier has been developed, implemented and tested. The process is entirely computational, and the output load and filter are the only analog components in the system. The process makes use of digital signal processing and a switching power stage to provide both high fidelity and high efficiency, beginning with a digital audio data stream. The advantages of naturally-sampled pulse-width modulation (PWM) are discussed in depth, including spectral analysis and comparisons to uniformly-sampled PWM. It is shown that natural PWM does not introduce audible distortion at switching frequencies consistent with power electronics practice. Interpolation methods for sample data conversion to natural PWM are discussed, and error analysis is presented based on Lagrange's Expansion Theorem. Noise-shaping processes are used to support high fidelity with practical values of time resolution. A counter conversion process enforces switching dead time in the inverter gate signals. The experimental full-bridge inverter implementation demonstrates that miniaturization is possible. A complete test system delivered more than 50 W into an 8 Ω load with an efficiency of 80% and total harmonic distortion plus noise of 0.02%.

Original languageEnglish (US)
Pages (from-to)473-485
Number of pages13
JournalIEEE Transactions on Power Electronics
Issue number1 II
StatePublished - Jan 2003


  • Class D amplifier
  • Digital audio amplifier
  • Noise-shaping
  • Pulse-width modulation
  • Uniform and natural sampling

ASJC Scopus subject areas

  • Electrical and Electronic Engineering


Dive into the research topics of 'High-fidelity PWM inverter for digital audio amplification: Spectral analysis, real-time DSP implementation, and results'. Together they form a unique fingerprint.

Cite this