The MODIS Adaptive Level One Accelerator

Marco Figueiredo, Terry Graessle, Ken Winiecki, Umesh Patel

The MODerate Imaging Spectroradiometer  (Modis) is one of five instruments
aboard the Terra spacecraft. The instrument images the Earth in 36 spectral
bands at a sustained data transfer rate of 6.2 Mbps or 10.8 Mbps at peak
output. The estimated average processing load of the level 1 stage of the
ground data processing system is 1.4 Gflops. The estimated daily data volume
is 308 Gbytes/day of Modis level 1 data. The Adaptive Level One Accelerator
(Aloa) is an FPGA based system which delivers Modis level 1 products at
real-time data rates. The scanning data rate of the imaging device is 1.477
seconds. Within this time frame, the Aloa receives level 0 processed packets
as its input data, and delivers a calibrated RGB image and a fire detection
image on its output.

The current implementation of the Aloa system is composed of a low cost
Windows NT based desktop computer. It receives its level 0 data through a
100BASE-T network connection to a Terra direct broadcast data ingest. Aloa's
desktop computer is augmented by an FPGA accelerator card that allows the
calibration of the required reflective bands at a data rate below the 1.477
seconds. The goal is to be able to create both the visible and fire images
at real time with minimum latency as the Terra satellite crosses over a
ground station.

The Aloa system is a combination of software and adaptive hardware. The
software is written in the Java programming language. The adaptive hardware
utilizes the Annapolis Micro Systems' Wildforce cards. The adaptive
application is written in VHDL. This approach allows the creation of a
platform independent solution. Java permits the migration of the code to any
machine without modifications. The Wildforce is a PCI-bus FPGA acceleration
card. There are Wildforce device drivers available for several computing
systems. Despite the dependency on the architecture of the Wildforce card,
the resultant hardware design developed in VHDL can be migrated to other
adaptive acceleration cards.

The level 1 calibration algorithms are received from the MODIS science teams
and adapted to the Aloa environment. Algorithm modifications are
accommodated accordingly.  The adaptive hardware provides a flexible and
maintainable environment that also reduces overall system cost while
increasing the performance by orders of magnitude. This paper describes the
design of the Aloa system including the adaptation of the MODIS calibration
algorithms, the system architecture, the hardware/software partitioning, the
adaptive hardware design and development, and the system overall performance
analysis.