Archive for March, 2010

Today was spent preparing and presenting my paper (co-authored with my doctoral adviser, Prof. Earl E. Swartzlander, Jr.) entitled A novel technique for Tunable Mismatch Shaping in Oversampled Digital-to-Analog Converters, during a poster session at the 2010 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). The poster session was titled DISPS-P1: Analog and Digital Signal Processing Systems, and provided an opportunity to spend time with researchers working on many other interesting topics. My poster was positioned right next to Professor Mark Arnold, who was presenting his poster on Implementing LNS using Filtering Units of GPUs, and I was quite fascinated to hear that GPUs were now being leveraged for real-time co-processing applications in order to eke out more processor performance.

Located in the heart of downtown Dallas, the Sheraton hotel is as unexciting a conference hotel as they come, with more than adequate facilities for big gatherings, but little else on offer. it didn’t help that the limited selection of restaurants and coffee shops in the vicinity left much to be desired. The conference itself was quite well-attended, and as is to be expected, with a very significant proportion of papers and posters having to do with speech and audio processing. The word of the day was most certainly Compressive Sensing (CS), an area of research pioneered at my alma mater, Rice University, and the talks featuring CS were by far the most heavily-attended.

I enjoyed the experience of presenting my research to others, particularly those whose work was so far removed from mine that I found myself resorting to first principles in order to convey the fundamental ideas. A pre-publication version of the paper can be downloaded here (pdf).

From the abstract:

Over-sampled digital-to-analog converters typically employ a unit-element architecture to drive out the analog signal. Performance can suffer from errors due to mismatch between unit elements, leading to a sharp drop in the achievable signal-to- noise ratio (SNR). Mismatch noise shaping is an established technique for overcoming these limitations, but usually anchors the signal band to a fixed location. In order to extend these advantages to tunable applications, this paper presents a novel technique that allows the mismatch noise shaping transfer function to have an adjustable center frequency.

Edit-2011: The paper is now available on IEEExplore.

Read Full Post »