1 / 20

Exploring the Limits of Digital Predistortion

Exploring the Limits of Digital Predistortion. P. Draxler, I. Langmore*, D. Kimball*, J. Deng*, P.M. Asbeck* QUALCOMM, Inc. & UCSD – HSDG *University of California, San Diego, HSDG September 14 th , 2004. Predistortion with Memory Model. Original measurement. with DPD incl. memory.

newman
Download Presentation

Exploring the Limits of Digital Predistortion

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Exploring the Limits of Digital Predistortion P. Draxler, I. Langmore*, D. Kimball*, J. Deng*, P.M. Asbeck* QUALCOMM, Inc. & UCSD – HSDG *University of California, San Diego, HSDG September 14th, 2004

  2. Predistortion with Memory Model Original measurement with DPD incl. memory Blue points – instantaneous Vout vs. Vin Purple line – gain target Green line – expected value of gain

  3. Outline • Introduction • Contraction approximation for nonlinear systems • Memory effect compensation – model based • Error Vector Magnitude (EVM) metric • Memory effect compensation – measurement based • Results from 2 RF Power Amplifiers • Conclusions

  4. System Block Diagram • DPD is the digital predistortion block • PA is the power amplifier (model or device) • Ideal Gain block sets system performance target

  5. Notation and Relationships • n is the sample index • i is compensated waveform iteration index • x: vectors are denoted with underbars • {} curly brackets denote multiple signals in an ensemble • yn=Goxn is output of the “Ideal Gain” block (the target output of the system) • y’n=Gn(xn) is the output of the “PA” block (with memory)

  6. Waveforms Identified • xn is the input waveform • xpni is the input waveform after digital pre-distortion • y’ni is the output waveform • yn is the target output waveform • eci is the current error waveform • ec(i-1) is the past error waveform

  7. Contraction approximation Memoryless gain Gain with memory effects xpni correction equation Δx adjustment equation

  8. Model Specific Application – Model Based • Generate xpni • Evaluation of model • Compare modeled vs. measured for xpni • Quantify the predictive accuracy of the model

  9. Specific Application – Model Based

  10. Error Vector Magnitude • Over all sample points, n, of a single measurement: • Normalize average power of signals to unity: xα, yα • Generate the rms difference between the normalized vectors

  11. Experimental values of alpha: α • Identify vector Δxn • Sweep α and evaluate for optimal EVM. • Function of: • Memoryless nonlinearity • Memory effect nonlinearity • Noise and chaotic amplifier behavior • Baseband envelope DAC/ADC quantization

  12. Ensemble Average Error Vector Magnitude • Perform an ensemble average over many measurements: E{.} • Over all sample points: n • Normalize average power of both signals to unity: xα, yα • Generate the rms difference between the normalized vectors

  13. Typical EVM histogram with Ensemble EVM (N=16) • Ensemble EVM is typically in the lower range of the histogram members. • As E{eci} becomes small, more ensemble members are needed to have confidence in the ensemble means and variances.

  14. Simple Test Amplifier • Inexpensive catalog amplifier. • WCDMA waveform used – amplifier configured for narrowband operation. • Severe ACPR asymmetry which switched sides and didn’t improve after memoryless predistortion.

  15. Specific Application – Experiment Based Memoryless correction Original I/O performance

  16. Specific Application – Experiment Based Correction with memory compensation Original I/O performance

  17. Non-optimal RF Power Amplifier

  18. EER Amplifier • Power Amplifier • Motorola LDMOS • Vdd amplifier included • PAE: 31.5% • Signal • WCDMA signal • >9dB peak to average • Pin: 3.35 Watts • Pout: 29.0 Watts

  19. RF Power Amplifier using Envelope Elimination and Restoration (EER)

  20. Conclusions • A new metric – ensemble average EVM – has been defined to separate out the deterministic EVM components from the random EVM components. • An measurement based algorithm has been realized that enables one to compensate for deterministic components of the output waveform. • This metric and compensation technique is insightful during: • component evaluation and characterization of amplifiers, • amplifier modeling and model evaluation, • identification of optimal performance targets, • in support of development of real time adaptive blocks…

More Related