ISA has been developing and deploying high performance, missions
critical processing systems for the US Government since 1997. The
systems built by ISA constitute the "heavy lifting" portions of our
customer's processing chain in terms of data rates and compute
intensity. The company's expertise includes:
ISA has deployed dozens of real-time processing systems designed to
detect and characterize unknown pulse and continuous-wave (CW)
waveforms in time series data at rates ranging from 25 million samples
per second (MS/s) all the way up to 1200 MS/s. Example DSP systems are
Floodlight (very wideband parameterizer), Spotlight (wideband
parameterizer), and BeachComber (CW processor). For each DSP system,
the company uses a combination of general purpose computers, FPGA-based
processors, and advanced software architectures.
ISA has extensive experience in automatic parameter estimation and
feature extraction for real-time systems. ISA's pulse and CW waveform
detection and characterization algorithms have been developed to
provide high sensitivity detection, high accuracy parameter estimation,
and low false report rates in a purely automatic mode. These algorithms
have been optimized using both simulated and real data and are capable
of dealing with many difficult real-world conditions such as
interfering signals and non-stationary effects. Below are intermediate
time and frequency data processing products from the automated
algorithms that are used for estimating signal parameters.
ISA utilizes OSP when a DSP solution is either not available or is cost
prohibitive. The Optical Energy Detector (OED) was the company's first
OSP application. The OED used an acousto-optic modulator and the
Fourier transform property of a lens to generate the Power Spectrum
(PS), and employed a high-speed camera and framegrabber to digitize the
PS. ISA also developed an Optical Correlator for performing a
million-point correlation on wideband signals in real time. The OC
module is rack-mountable and has a throughput of 0.6 TFLOPS.
ISA is developing two systems that can provide significant intelligence
products to the defense, intelligence, and homeland security
communities. The first is a Tactical SIGINT System (TSS) designed to
provide Force Protection / Force Multiplication for the War-fighter in
theater. The second system is the Wide Area Video Camera (WAVcam) that
has application in Border Security as well as a military Force
Protection role. The WAVcam also has civil security applications for
large areas such as airports, industrial facilities, or public events.
ISA implements decoupled architectures through extensive use of: the
messaging paradigm, the publish-subscribe model, the service
configurator pattern, shared object libraries, and pipelining.
ISA's product line of graphical user interface displays provide
creative visualization of real-time, high bandwidth data streams.
High performance computing is achieved by parallelizing the problem,
running on multi-core multi-blade servers,
mapping work to processes with the right form of IPC,
leveraging thread pools within processes,
embedding data acquisition,
and optimizing Command and Control.
ISA partitions the problem into proper and achievable building blocks.
Its architectural approach defines the interfaces between these
blocks, and then uses the right technology for each block and boundary.
ISA continually tracks technology advances in order to maintain the critical
edge necessary to compete in our fast paced industry. Whenever possible ISA
will incorporate proven technology into our solutions to ensure we produce cost
effective, on time, and high performing systems. Examples of some of these
commercially developed products are: UNIX-based compute systems, PCI-based A/D
cards, 10baseT, Gbit and HiPPI network fabric, Matlab, high speed area/line
scan cameras, lasers, optics, acoustic/optic modulators, spatial log and light
ISA uses MATLAB (a high-level technical computing language) to
optimize data visualization, algorithm development, and data analysis.
MATLAB allows rapid prototyping of complicated signal
processing algorithms. Below is our streaming RF visualization tool:
amplitude vs time (top),
frequency vs time (spectogram, bottom),
frequency vs power (power spectral density estimate, right).
ISA pulse parameterizers detect and characterize potential signals in real
time. The resultant "descriptor words" have been overlaid below, with color
representing modulation type.
ISA's single pulse plotting MATLAB tool provides parallel displays of
amplitude, phase, and phase error.
ISA's pulse doppler parameterizer processes a stream of digital samples, and
outputs a set of segmented data words (SDWs) that contain the estimated
parameters for the pulse doppler frames detected. Our visualization
capabilities include the spectrogram on the left, and the output of a single
FFT on the right.
ISA signal processing relies on the "pipes and filters" architecture: each
processing algorithm ("filter") contributes its piece to the aggregate
pipeline. The tool below provides a visual quality check on the evolution of a
vector of center frequencies. The top graph is the input to our center
frequency tuning module, and the bottom graph is its output. The analyst can
readily inspect the fidelity of the results.
To assist the signal analyst in evaluating statistical performance and
generating statistical reports, ISA has developed a MATLAB tool that allows an
entire directory of test data to be visualized.