Portland, Ore. Advanced digital wireless standards like those overseen by the 3rd Generation Partnership Project use iterative communications protocols that provide error correction simultaneously with optimal compression.
Until now, the most efficient way to correct errors in iterative codes has been to use rather power-hungry digital A/D converters and hardware multipliers to decode the wireless signals. But while digital decoders consume milliwatts of power, analog decoders use microwatts to achieve the lowest levels of noise interference and data corruption.
Researchers at the University of Alberta have found a way to decode wireless signals using far less power than that consumed by digital A/D converters. "In one of our test chips, we have demonstrated that analog decoders can consume 100 times less power [40 picojoules of energy to decode one bit] while performing the same functions as a digital decoder. In another chip with a block size of 256 we showed that they could use up as little as a tenth of the area on the die," said University of Alberta EE David Nguyen.
Today, digital turbo decoders with a block size of 1,024 consume over 120 nanojoules (120,000 pJ) per bit and take up as much as 81 mm2 of silicon, according to Nguyen. Even next-generation digital decoder prototypes consume 1,260 pJ and measure over 50 mm2. In contrast, the power consumption in next-generation analog decoders drops to 86 pJ per bit and 256-block decoders measure under 3 mm2, he said. (The 40-pJ per bit figure Nguyen cited was for a tiny proof-of-concept chip using a block size of only 8.)
Since 1993, the best communications codes so-called "turbo" codes have been packing more information in a channel by not transmitting bits one by one. Instead, turbo codes depend on algorithms that compress information according to the content of the message. Iterative codes optimally improve turbo codes by sending probability matrices that require less signal-to-noise ratio for the same bit-error rates, thereby approaching the Claude Shannon limit the theoretical maximum information transfer rate possible in a communications channel.
"Ours is the most efficient analog decoder chip in the world today, because we use transistors below their threshold using leakage current to process signals," said Nguyen. "We decode in analog, then encode the result in digital with a simple comparator circuit."
Nguyen designed the chip with EE professors Vincent Gaudet and Christian Schlegel and EE graduate student Chris Winstead. All the work was performed at the High Capacity Digital Communications lab (www.ece.ualberta.ca/~hcdc) at the University of Alberta (Edmonton).
Digital decoders for iterative codes received by antenna are first put through a fast A/D converter. Then an array of multipliers perform matrix multiplication in hardware. Some digital de-coders transform to the logarithmic domain to simplify circuitry, but this still requires arrays of adders which, together with the A/D converter, can consume milliwatts.
Analog decoders solve the problem by performing the necessary multiplications while the signal is still in the analog domain, thereby consuming as little as 40 pJ of energy to decode one bit. To accomplish this, an analog Gilbert multiplier is used. It operates below the transistors' threshold on the otherwise normal CMOS chip. By using oversized analog transistors with tight tolerances, the researchers were able to demonstrate matched transistors performing subthreshold analog matrix multiplication. The final decoding step, digitization, was performed by an array of simple analog comparators.
Previous chips designed by the High-Capacity Digital Communications Laboratory include an analog decoder chip with a block size of 256 that was designed by graduate student Winstead. Other chips, demonstrated by the University of Alberta and by other researchers in the United States, Europe and Japan, foretell a future where "on" wireless devices consume as little power as "off" wireless devices consume today.
According to the researchers, the slim power requirements of its analog decoder could not only enable cell phones to run on disposable AAA batteries, but could also enable medical and diagnostic devices that have been impossible to produce before now. For instance, implantable health monitors and automated drug delivery systems using the new analog decoder could generate almost no heat and be controlled wirelessly through the skin.
This research was supported by the Informatics Circle of Research Excellence (www.iCORE.ca), the Natural Sciences and Engineering Research Council of Canada, the Canadian Microelectronics Corp., the Canada Foundation for Innovation and the Alberta Science and Research Authority.