Signal Processing
Crossfield employs digital and/or
analog signal processing in almost all of the systems
we develop. From digital FIR filters to very large “n”
FFT’s, Crossfield uses a range of digital signal
processing including off-the-shelf purchased IP
(intellectual property) to custom signal processing
developed for unique requirements. For example,
Crossfield has developed hardware and corresponding
software to implement near real-time FFT’s with
millions of points.
Some applications are more efficiently implemented using
analog signal processing such as the programmable gain and
active filter techniques used to condition many sensor
circuits prior to digitization. Crossfield has developed
many such signal conditioning chains to support its
instrumentation business.
Heterogeneous Clusters for Signal Processing
Crossfield is developing
clustering technologies for communications intense
algorithms, such as those commonly found in real-time
signal processing and hardware-in-the-loop simulation
applications. The centerpiece of this technology is a
full-mesh bridge chip that provides unsurpassed data
communications bandwidth between signal processors.
Crossfield's "multipath" communications technology
provides an order-of-magnitude increase in network
bandwidth through a switch fabric.
Emerging Computing Technologies
In many real-time
signal processing applications, the data rate exceeds the
capability of traditional processors. Reconfigurable
Computing (RCC) in Field Programmable Gate Arrays (FPGAs),
General Purpose Computing on Graphics Processing Units
(
GP GPU), and digital signal
processing on the
Cell Broadband
Engine are examples of advanced computing methods
used by Crossfield to accelerate
computationally-intense algorithms.
Message Passing Interface Optimized for
DSP
Crossfield supports the
Open MPI project's efforts to
provide a common message passing interface across
heterogeneous computing platforms. Crossfield is
deploying a DSP MPI across a 10 Gbps Ethernet network
fabric using
OpenFabrics and
Remote Direct Memory Access
(RDMA) networking technology.
Large-Scale Storage Systems
Large-scale
instrumentation systems need large-scale storage systems
for data capture, processing and archival. Crossfield
Technology is teamed with
Copan Systems to provide
Massive Arrays of Idle Disks (MAID)
technology to the instrumentation community.
High-Speed, Non-Volatile Network Data
Buffers
Advanced sensors produce copious amounts of
data in real-time. Crossfield is developing high-speed,
non-volatile memory systems to capture network sensor data
arriving at Gbps to Tbps.
