 |
Jerry McGuire Roll Over Fourier
Undoubtedly, today's signal processing capabilities would make the founding father of DSP, Jean Baptiste Joseph Fourier, spin in his grave with wonder and excitement. Even Cooley and Tukey, who put the "F" in FFT back in 1965 (with a record 1.2 seconds to complete a 2k complex transform), must have watched in astonishment as their algorithm zoomed to incredible speeds over their remaining lifetimes. The performance of today's DSPs redefine the word "fast" perhaps beyond anything Cooley and Tukey ever imagined could become reality.
I'll give you just three examples among the many I know are in the market or under development today. Fractal Audio uses a TigerSHARC to run some incredibly dense and complex algorithms that they call "natural processing." The goal that these guys set out to achieve is to actually replicate the audio response patterns just as they occur in nature. The Fractal folks don't talk about "modeling;" they talk about "replicating" nature to imbue an all-digital guitar processor with smooth, "analog" sound. Fractal geometry (the math, not the company) is closely associated with the deep, mysterious and powerful world of chaos theory. I'm told that fractals have Hausdorff-Besicovitch dimensions that strictly exceed their topological dimensions. The resulting "organic" sound is not found in other products and it is the immense signal processing power of TigerSHARC that lets Fractal achieve their goals at a price point supported by the musical instrument market. Another ADI customer, Masimo, has this SHARC-based thing called the Rainbow SET Rad-57 Pulse CO-Oximeter. This device is an incredibly accurate diagnostic machine that measures carbon monoxide in the blood based on Masimo's Signal Extraction Technology (SET), a method for acquiring, processing, and reporting arterial oxygen saturation and pulse rate that "combines proprietary signal processing algorithms with innovative sensor technologies." SHARC gives Masimo the precision and accuracy they need to non-invasively extract a rich stream of physiological data from a patient's blood in real time. A technician places a clip from the Masimo unit onto a patient's toe, finger, or earlobe. The unit's sensors generate light at two wavelengths: 650 nm and 805 nm. The light is partially absorbed by blood hemoglobin (Hb), depending on whether it is saturated or de-saturated with oxygen. The SHARC then computes the proportion of Hb that is oxygenated by calculating the absorption at the two wavelengths. Innov-X uses Blackfin for elemental analysis, powering their portable X-ray systems that are used in situations ranging from the examination of the hull of a Civil War submarine to the search for WMDs in Iraq. The computationally challenging theory behind the Innov-X system uses the interaction of electromagnetic radiation with matter to help determine the identity and atomic composition of elements in compound mixtures. While Innov-X previously used a proprietary FPGA-based pulse analysis � an effective but costly approach � they now use Blackfin to inexpensively and portably perform on-the-spot X-ray fluorescence (XRF). First, Blackfin enables the bombardment of atoms to create electron transitions that cause X-ray energies or wavelengths specific to the compositional elements of a substance to be emitted. A "fast filter" is then run on the raw data stream to locate peaks that exceed a certain energy threshold � these peaks are time stamped and recorded in a separate array. A secondary Blackfin process then runs a more accurate and more computationally expensive filter on just those peaks, giving a better estimate of the energy from the individual X-ray events that were detected. The results are then compiled in a multi-channel analyzer histogram, maintaining detailed descriptions of the composition of the materials under test. I've only discussed a few applications here, but you get the idea. DSPs are doing increasingly heavy lifting for applications that either could never have been done before, or could never have been done at accessible price points or previously unattainable levels of portability. Not only that, but the innovation curve driven by young engineers (of all ages!) leveraging the explosive potential inside DSPs today shows no sign of abating. And that is probably the biggest reason we have such a deep list of customer success stories on this web site!
Jerry McGuire is a Vice President of Analog Devices' General Purpose DSP Group. His column appears here regularly. |
At this point in time, commercial processors designed with the computing power, memory architectures and I/O hooks needed for signal processing have been with us for about a quarter century. And for those of us who have watched the DSP era unfold from its infancy, it is still literally an awesome thing to witness the steadily increasing kinds of applications enabled by these devices. That's because DSPs, perhaps unlike any other semiconductor device ever invented, make things happen that could never have been done before.
The list of exceedingly cool stuff enabled by DSPs is both broad � covering myriad vertical markets � and very deep. Would CES (the Consumer Electronics Show) even exist without DSPs, for example? CTIA? ESC? VON? Mobile telephones and computer disk drives are often cited as the most compelling successes for DSPs, and deservedly so. But in my role at ADI, the applications that come across my desk, enabled � and I mean truly enabled � by DSPs are sometimes too esoteric to make the big headlines, but they nevertheless represent the huge list of applications that could never move beyond conceptual dreams without the existence of DSPs.