CSAIL Research Abstracts - 2005 link to http://publications.csail.mit.edu/abstracts/abstracts05/index.html link to http://www.csail.mit.edu
bullet Introduction bullet Architecture, Systems
& Networks
bullet Language, Learning,
Vision & Graphics
bullet Physical, Biological
& Social Systems
bullet Theory bullet

horizontal line

Raw Wireless System

Nathan R. Shnidman & Anant Agarwal

Introduction

The goal of this research project is to explore the practicality of, and innovations enabled by, a wireless communication system based on the Raw architecture. The streaming communication patterns and serial data processing aspects of wireless systems suggest a natural mapping of such applications to the high-bandwidth, low-latency communication and concurrent processing capabilities of the Raw architecture. In turn, the implementation of radio functions in software on the Raw architecture allows such functions, and the system in general, to have a greater level of adaptability.

The system is intended to enable three main areas of research: Raw system design evaluation, exploration of software radio adaptability, and novel communications algorithm and system development.

The Raw architecture and handheld systems both involved many design tradeoffs. The Raw wireless system will allow an evaluation of the applicability of such choices to the communications application space. Such information can be used to validate current tradeoffs or to inform design decisions for future Raw processors and systems. In addition, the building of the system allows an exploration of the complexity and design requirements of an actual end-to-end real-time embedded system.

A software radio based design such as this allows adaptation of both the radio operation and the system itself. Such flexibility in the radio design enables parameters, algorithms, and even the dataflow to be changed to optimize for current channel conditions or communication needs. At the system level a software radio design allows the system to morph in order to reassign some or all of the resources used by the radio to other tasks.

Finally, the finished wireless system will provide a flexible research testbed to explore new communications algorithms and systems. The system will provide a large bandwidth (up to 80MHz) which can be converted to/from the analog domain, up/down-converted, and transmitted/received. The entire baseband and any further processing will be performed in software on Raw processors. This enables a flexible software digital baseband. Thus, any changes in the baseband operation of the system can be written in software and quickly implemented on Raw at both the transmitter and the receiver.

Approach

The system transceiver consists of a Raw handheld board connected to an analog RF front-end board. The two boards interface through an expansion connection built into the Raw handheld board. An overview of the full system can be seen in Figure 1.

Diagram of the Raw Wireless System

The base design of the radio frequency front-end board was generously provided by Engim, Inc. The design consists of both 2.4GHz and 5GHz front-ends and can perform both transmission and reception. The front-end board performs transduction, frequency conversion, and conversion between the analog and digital domains. The output/input of the board is up to 80MHz of 12-bit digital data. A modified version of this design has been created to allow the necessary interfacing to the Raw handheld board. It provides signal level conversion as well as buffering and is controlled by the Raw handheld board through a serial interface across the expansion connector.

The Raw handheld board provides an FPGA to perform data reordering and conditioning to reformat the front-end's data for introduction onto Raw's static network. This FPGA also contains a controller state machine to allow Raw to more easily direct the front-end board. Once the data has been injected into Raw's network all further processing takes place on the Raw processor. Such processing includes baseband functions for the physical layer as well as higher layer functions such as MAC processing.

The 802.11a [1] and 802.11g [2] wireless network specifications were chosen for the initial system implementation. This choice was informed by multiple factors: the current and growing popularity of these high speed wireless networking specifications; the ability to utilize the same baseband for either operating frequency of the front-end board (5Ghz for 802.11a and 2.4GHz for 802.11g); and, most importantly, the use by both standards of OFDM for improved datarate. The use of OFDM by 802.11a/g creates an opportunity to directly shape and modify the frequency output of the system in the digital domain. This ability provides an interesting starting point for the development of future algorithms and protocols.

The streaming nature of the 802.11a/g baseband design provides an excellent opportunity to take advantage of the spatial pipelining capabilities of the Raw processor. The 802.11a/g baseband block diagram, along with a 802.11a/g receiver baseband mapping, is shown in Figure 2.

Mapping of receiver onto Raw processor
Progress & Future Work

The Raw handheld board and Raw processor have been completed and are functioning and available for use. The front-end RF board has also been completed and integrated with the Raw handheld board. The entire Raw wireless system has been use to both send and receive messages. The baseband software has been implemented and run on the Raw processor. Both a single tile version of the baseband and a parallelized multiple tile version of the baseband, as seen in Figure 2, are functioning, both in simulation and on the system hardware.

Work has begun exploring extensions to this system to take further advantage of the properties of the Raw architecture. Heavily parallelized versions of critical path system elements, such as the Viterbi Decoder and FFT blocks, have been implemented to significantly improve system performance. The improvements and trade-offs enabled by the ability of the Raw wireless system to arbitrarily replicate system functions and process multiple data streams concurrently, or the same data stream multiple times, is being investigated.

Research Support

This research is supported by DARPA, NSF, and the Oxygen Alliance.

References:

[1] IEEE Std 802.11a-1999. Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: High Speed Physical Layer in the 5GHz Band, July 1999.

[2] IEEE Std 802.11g-2003. Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: Further Higher Data Rate Extension in teh 2.4 GHz Band, June 2003.

horizontal line

MIT logo Computer Science and Artificial Intelligence Laboratory (CSAIL)
The Stata Center, Building 32 - 32 Vassar Street - Cambridge, MA 02139 - USA
tel:+1-617-253-0073 - publications@csail.mit.edu
(Note: On July 1, 2003, the AI Lab and LCS merged to form CSAIL.)