Aerospace & Defense Technology - August 2022 - 4

Sponsored Content
INSIDE STORY
Vision Research manufactures Phantom high-speed cameras and imaging systems for research
applications in defense, automotive, engineering and much more. In the following interview,
Senior Test Engineer Lee Denaro discusses high-speed camera innovations involving the CXPover-Fiber
standard to improve throughput in high-speed machine vision applications.
A&DT: What is the camera capable of?
Denaro: Taking traditional machine vision
to the next level, the Phantom S991 highspeed
camera features a 9-megapixel
(Mpx) CMOS sensor, a low exposure time
of 5 microseconds (µs) and speeds up
to 52,080 frames per second (fps). As
the world's highest-throughput machine
vision camera, it combines world-class
image quality with direct data transfer
speeds of 9 gigapixels per second (Gpx/
sec) and can capture 938 fps at full 9-Mpx resolution.
Using the standard CoaXPress (CXP) protocol, the S991 transfers
data to backend frame grabbers, where data is processed on the fly
either in the frame grabber field programmable gate array (FPGA)
or in the graphics processing unit (GPU) plugged into the computer
motherboard via the PCI-e bus standard. In a process called stitching,
the camera divides the transmitted images by rows and then
interleaves the images as they arrive in the frame grabber FPGA.
Using a simple algorithm, each image is then " stitched " back together.
This streaming ability avoids the time-consuming process of
saving data to a camera's RAM before downloading it to a computer.
A&DT: How does CXP-over-Fiber work?
Denaro: An add-on to the recent CoaXPress 2.0 specification,
CoaXPress-over-Fiber (CXPoF) provides a way to run an unmodified
CXP-12 protocol over a standard Ethernet connection including
fiber optics. In other words, it uses standard electronics,
connectors and cables designed for Ethernet, but the protocol is
CoaXPress instead of Ethernet or GigE Vision.
Because CXPoF combines the CXP-12 standard with optical fiber,
this upgraded standard eliminates the need for multiple CXP-6
copper cables. As one of the industry's first CXPoF cameras, the
Phantom S991 requires just two fiber cables as opposed to the 16
copper cables required by its predecessor, the Phantom S990.
A&DT: What are the components?
Denaro: Since CXPoF is an add-on to the existing CXP standard,
you can leverage the same programming to upgrade your
camera. Since it uses Ethernet components, CXPoF also only
requires standard Ethernet connectors and cables, keeping your
costs low. At the same time, it takes advantage of the " free "
evolution of Ethernet as it progresses toward higher bandwidths.
In addition to CXPoF frame grabbers and substantial PC power,
you'll need the following components to get started:
* Quad Small Form-Factor Pluggable (QSFP+) transceiver modules.
CXPoF achieves 4 x 10 Gbps on a single QSFP+ module
for a total of 40 Gbps per camera - the same net bandwidth
as four CXP-12 links over four copper coaxial cables.
* Fiber cabling. These cables are small and lightweight and can
4
A&DT: You have three machine vision cameras: the S990,
S710 and S640. Why did you convert this one to fiber?
Denaro: By combining extremely high throughput (9 Gpx/sec),
high resolution (9 Mpx), fast frame rates and CXPoF cable technology,
the Phantom S991 enables extreme high-speed imaging,
expanding machine vision to challenging applications that
require real-time analysis or long record times like space shuttle
launches. In addition, the integration of CXPoF technology
opens the door to new high-speed machine vision applications
while, at the same time, simplifying those applications where
high-speed machine vision is already in use.
Visit Vision Research at www.phantomhighspeed.com
mobilityengineeringtech.com
Aerospace & Defense Technology, August 2022
handle higher bandwidths compared to copper cables. They
are also immune to electrical noise.
Prior to CXPoF, you had to deploy expensive repeaters to translate
the CXP-6 standard into fiber, driving up costs by several thousands
of dollars. This setup also created very bulky cable interfaces with
many potential points of failure. CXPoF simplifies this setup, eliminating
the need for repeaters and their associated failure points.
CXPoF also expands machine vision to applications requiring
recording at longer distances, an undertaking that previously
relied on costly, complex repeaters. Now, you can simply set up
your machine vision camera several miles away without worrying
about the added complications associated with connecting external
equipment. And, the availability of single-mode transceivers
and cables significantly extends cable length. This ability simplifies
the process of bringing machine vision to applications that
involve ballistics or explosives, which, by their nature, require
camera operators to remain far away from the recording process.
A&DT: What are the challenges?
Denaro: To realize all the benefits high-speed machine vision
has to offer, you must have the right technologies in place to
manage the vast amounts of data coming from the streaming
cameras. Although stitching is a vital part of the streaming process,
it takes up computing resources and can require lengthy
processing times if not addressed properly.
A second issue that has emerged, particularly at higher
frame rates, is image " thrashing, " which occurs when the volume
of incoming frames outpaces the CPU of the host PC. In
other words, the computer, as it receives and processes the high
number of frames, can't keep up and subsequently fails.
Thrashing typically occurs at higher frame rates, usually around
20,000 fps and higher.
Fortunately, frame grabber manufacturers like Euresys and
BitFlow offer innovative solutions to the data issues that arise
from stitching and image thrashing at high frame rates, and
their frame grabbers are a natural complement to Phantom
high-speed streaming cameras.

http://www.phantomhighspeed.com http://www.mobilityengineeringtech.com

Aerospace & Defense Technology - August 2022

Table of Contents for the Digital Edition of Aerospace & Defense Technology - August 2022