The era of expensive custom signal processing systems is ending, as commercial standards such as VME and Raceway join hands with broadly accepted digital signal processing chips in new sonar systems that are more sensitive and powerful than ever before
By John Keller
Sonar signal-processors have long been the custom designs of engineers whose chief concern is performance at virtually any price. This heavy emphasis on system capability with cost barely an afterthought traces its roots to the Cold War, where the ability to detect, classify, and track quiet nuclear submarines was so paramount as to be a defining icon of post World War II national rivalries. Leaders of the U.S. and former Soviet Union knew well that a single ballistic missile submarine had the potential to devastate any country on earth. A Russian Typhoon-class submarine, for example, carries 20 missiles, each one fitted with six to nine independently targeted nuclear warheads. Each of these 120 to 180 warheads is about six times as powerful as the atom bombs that destroyed Hiroshima and Nagasaki in 1945.
It stands to reason that maintaining the ability to know the locations of these destructive vessels was a top national Cold-War priority. Since sophisticated sonar systems are virtually the only reliable way to detect and track submarines hiding in the open ocean, designers were told to spare no expense in building the best devices possible.
Today, however, the Cold War`s end brings with it a new host of underwater threats, budgetary imperatives, and technologies. Aging custom-designed sonar signal processors not only are growing too expensive to build and maintain, but their performance is lagging behind commercially developed computing. At the same time, there is little letup in the demand for sonar performance. The threat of ballistic missile submarines still looms, and extremely quiet diesel/electric attack submarines are proliferating throughout the Third World to menace the world`s merchant fleets. Yet designers must build sonar systems on shoestring budgets as all facets of the U.S. military complete for dwindling resources.
The apparent answer is cost-effective and high-performance commercial off-the-shelf (COTS) technology, which is on the verge of sweeping away all significant implementations of custom sonar signal processors. New generations of high-performance digital signal processors (DSPs), open-systems modular computing architectures where VME printed circuit boards reign supreme, standard software architectures that accommodate new and old algorithms, and multiprocessing designs that hold forth a wide variety of innovative ways to analyze sounds in the water are the hallmarks of the sonar paradigm shift from custom designs to COTS.
"To a very large extent with commercially available DSPs, we can implement virtually all our sonar signal processing on commercial platforms," explains Jerry Bradshaw, technical director of submarine signal programs at Raytheon Electronic Systems in Portsmouth, R.I. Raytheon is involved in several U.S. and international sonar programs, including the AN/BQQ-5 sonar for U.S. Los Angeles-class attack submarines (SSN-688); the AN/SQS-56 hull-mounted sonar for several U.S. Perry-class frigates (FFG-7); the AN/SQQ-89 onboard sonar trainer; as well as for systems aboard naval surface ships in Australia, Saudi Arabia, Turkey, and Italy.
COTS drawbacks
Sonar designers should take heed, however, that COTS is not a panacea. Engineers must understand and plan for the drawbacks of COTS, Raytheon experts warn. U.S. sonar systems must remain operational for 20 years or more, while many COTS components these days show outstanding longevity if their manufacturers still build and support them 20 months after introduction.
"Typical military systems are supported for 15 to 20 years beyond their design, and we used to be able to guarantee their availability for that long, but it`s getting harder and harder because these [COTS] products go obsolescent," says David Reynhout, manager of submarine signal programs at Raytheon. "Long-term supportability must be thought out up front," Reynhout says. "With strictly COTS components, none of these manufacturers wants to support a line beyond what`s practical - two to three years."
Accurately gauging how and when to improve COTS-based systems to improve capability and avoid obsolescence requires a close reading of industry trends, a good guess of what components will cost several years hence, and a bit of luck.
"In order to support a design that is COTS oriented, you must be able to refresh technology more frequently than you would with a traditional custom-designed system," Reynhout advises. "You factor in cost up front, maybe five to six years downstream. You just make the assumption that some of those components won`t be available then, so you make a guess at what will be available and hope that you`re right."
Choosing vendors with reliable track records over time whose officials pay attention to building upwardly compatible families of products that require little software rewrite is also important, he says. Overall, Reynhout says COTS is not a cure-all, and "for those who think it is will have some rude awakenings downstream. You must be wise about how you use COTS and in the long run you will incur additional costs for accommodating the COTS environment."
Sonar beginnings
Sonar, short for sound navigation and ranging, uses sound waves to detect and determine the location, size, and relative motion of objects in the water. Specially equipped fixed-wing aircraft, helicopters, surface ships, and fixed sensor sites use sonar not only to detect and classify submarines, but also to communicate with submarines. The submarines themselves, meanwhile, use sonar to detect and classify surface ships and other submarines. Modern torpedoes use sonar to guide themselves to their targets.
First developed during World War I to combat submarines, sonar as a means to detect ocean vessels essentially comes in two flavors - active and passive.
Active sonar uses a sound receiver and acoustic projector called a transducer to generate a sound wave under water that spreads out and reflects off of targets back to the receiver. Transducers may be on a floating sonobuoy, attached to a vessel`s hull, or dangled in the water from a helicopter. Until the recent advent of sophisticated computer signal processing, active sonar was about the only sure way to detect submarines. Human sonar operators would "ping" with the transducer and count the seconds until they heard the reflected sound to determine the approximate distance to the target.
Passive sonar, meanwhile, consist solely of receivers that detect the sounds of ships or submarines - usually engine and/or propeller noises. Computers and human experts analyze received sounds to identify the type of passive sonar receivers, most often mounted to a vessel`s hull, to sonobuoys, or to a towed array reeled out well behind the propeller wash of its host vessel.
Submarine sonar
Attack and ballistic missile submarines are among the most intensive users of sonar, particularly of the passive variety. Crew members of the so-called "silent service" are always listening to the waters around them for potential targets and for potential predators. The captain of the ballistic missile submarine, or "boomer," must avoid detection while on patrol. The attack submarine captain also must remain as stealthy as possible, but must find and shadow the adversary`s missile subs - and destroy them in the event of war. Submarines give away their presence to anyone within listening distance when they employ the "ping" of active sonar, so sub captains rarely use it.
A powerhouse in the design of U.S. submarine sonar systems is Lockheed Martin Corp. in Manassas, Va., which has become a center of sonar excellence with the rapid consolidation of the U.S. defense industry. Lockheed Martin-Manassas started as IBM Federal Systems Division, the designer of the AN/BSY-1 combat system for the Los Angeles-class attack boats. Folded into the Manassas site today are many of the sonar activities of the former General Electric Co. in Syracuse, N.Y., designer of the AN/BSY-2 for the Seawolf-class attack submarine (SSN-21), and of the former Martin Marietta Aero and Naval Systems in Baltimore, which built several towed array sonars and the Wide Aperture Array hull-mount sonar for attack submarines.
Over the past several years of defense industry consolidation, IBM became part of Loral, which later became part of Lockheed Martin. General Electric became part of Martin Marietta, which later became part of Lockheed Martin when that company merged with Lockheed.
The foundation of the sonar work at Manassas is the BQQ-5 on the Los Angeles-class attack boats. The AN/BQQ-6 sonar for the Ohio-class missile submarines, which engineers at Lockheed Martin Manassas also build, is based on the BQQ-5, and the two systems contain many identical components. The Manassas group is also developing the sonar and combat system for the New Attack Submarine, also called the NSSN, which will use BQQ-5 technology.
Engineers at Manassas are in the midst of a major sonar technology overhaul called Acoustic Rapid COTS Insertion, or ARCI, which seeks to create variants of a single system design to fit the BQQ-5 and BQQ-6, says Robert McCaig, the company`s technical director of NSSN sonar.
One architecture
The core of ARCI is the Multi-Purpose Processor (MPP), a design of engineers at Digital Systems Resources (DSR) of Fairfax, Va. MPP is an open multiprocessing architecture that relies on commercial off-the-shelf microprocessors, software "middleware" to make different processors run the same software, and the fiber distributed data interface (FDDI) high-speed serial data bus. The ARCI architecture uses an MPP version based on the COTS 21060 Sharc DSP from Analog Devices of Norwood, Mass., and the Raceway multiprocessor architecture from Mercury Computer Systems Inc. of Chelmsford, Mass., McCaig says.
The Los Angeles-, Ohio-, and NSSN-class submarines are "slightly different when it comes to sensors and sensor interfaces, so you need to precondition the sensor interface for the MPP," McCaig says. "There are always elements of hardware and software that are unique to the platform, but the fundamental architecture is the MPP."
This architecture not only processes signals from the submarine`s hull sonar array, and from its TB-16, TB-23, or TB-29 towed sonar arrays, but it can process signals from three sonar arrays simultaneously. "You can have processing in one box, data from the hull array, the TB-16, and one of two variants of the TB-29," McCaig says. Each signal processing box contains four VME card cages of 21 slots each - three dedicated to beamforming signal processing, and the fourth for signal conditioning for the hull array, he says.
Besides the Sharc DSP, the Mercury Raceway-based MPP architecture uses a Motorola 68040 microprocessor controller, and runs the VX Works real-time operating system from Wind River Systems in Alameda, Calif. "The software is about 75 percent commercial, and 25 percent tactically unique," McCaig says. Manassas engineers are reusing 30 to 40 percent of the software engineering methodology of the BSY-1 combat system.
DSR engineers devised the MPP with about 180 microprocessors configured in several different chassis and linked with the Mercury Raceway. DSPs range from the Intel I860 to the Sharc, and may include general-purpose processors such as the Sun Sparc or IBM/Motorola Power PC.
"MPP has a software layer called middleware that protects application software from specific uniqueness of the hardware being used," explains Rich Carroll, a DSR spokesman. "When new technology comes along, we don`t have to change the software."
A PCI mezzanine card interfaces sensor data to the DSPs on Raceway to provide real-time sensor communications between sonar arrays and the parallel-processing computer. "It scans up to 180 nodes of 80 Mflops each," says Barry Isenstein, advanced project planning manager at Mercury. "It can play in many of the systems for sonar and anti-submarine warfare. The whole idea is to demonstrate a reconfigurable, programmable, flexible, and scalable system that can be adapted to various parts of the ASW job." MPP can run not only conventional software applications, but also can be configured as artificial neural networks where each microprocessor can home in on separate sonar signals.
The MPP-based ARCI architecture will begin appearing on Los Angeles-class submarine upgrades as early as this October, which will be the first time U.S. attack submarines will receive COTS signal processors. All the 688 subs at sea today have proprietary sonar signal processing, and these vessels always have since the first submarine of the class was commissioned in 1976.
Analog to digital
The ancestry of 688`s sonar system began with the analog Spectrum Analyzer Unit, built in 1970 at Sanders, of Nashua, N.H., now a Lockheed Martin company, McCaig explains. Navy leaders authorized Manassas engineers (then with IBM) to improve the Spectrum Analyzer Unit in 1977 to a proprietary system called the Multi-Interface Unit Digital Spectrum Analyzer, which was one of the first Fourier transform sonar signal processors, he says. The most recent upgrade to the 688`s sonar processor came in 1983 when Manassas engineers designed the custom Tri-Advanced Signal Processor, which is at sea today, McCaig says.
Choosing the MPP-based ARCI has more advantages than simply high-performance hardware at a bargain price. Manassas engineers have a plan to circumvent the obsolescence problems of COTS because the architecture is designed to accept upgrades easily. "The software is hardware independent" because of its Posix operating system interface, McCaig explains. "We are not locked into Mercury, and Mercury fully recognizes that. We are continually looking at hardware and software from a cost and availability perspective to define the moment in time to see when to transition from one technology to the next."
Software is particularly difficult to reuse from one generation of technology to the next, but Manassas engineers believe they are solving the problem. "As the operating system changes with new releases, those applications calls can still be honored, so the tactical software still runs. We are adhering very strictly to open systems and widely used standards."
By using only industry standards, Manassas experts believe they will get the support they need to bridge their new COTS architectures with older systems still in the fleet. Industry standards level the playing field and open sonar opportunities to many companies - some of which have rarely done business with the military before. "Commercial industry will not give up on their legacy systems," McCaig says. "But with custom or unique architectures, commercial industry does not help you make the transition."
A shipyard worker cuts up a Russian Oscar-class ballistic missile submarine in Severodvinsk, Russia, last May as part of the Nunn-Lugar/ Cooperative Threat Reduction Program. Although the Cold War may be over, the worldwide submarine threat still exists, and places continuing demands on improved sonar technology.
FPDP data bus helps speed sonar design
The so-called "front end" of sonar sensor processing - digital filtering, beamforming, and demodulation - is a well-defined yet esoteric task that doesn`t lend itself easily to commercial off-the-shelf (COTS) solutions.
Yet engineers at a Canadian company just outside of Ottawa are helping fill the demand for off-the-shelf front-end sonar processing, and are using an emerging standard of their own design called Front Panel Data Port (FPDP) as the chief enabling technology.
"Our motto is `integrate by lunchtime,`" says Dipak Roy, president of Interactive Circuits and Systems Ltd. (ICS) of Gloucester, Ontario. The FPDP, which uses the straightforward approach of linking circuit cards and card cages with parallel ribbon cables, enables engineers to assemble and optimize sonar sensor processing systems in hours or days, rather than weeks or months, Roy says.
ICS is involved in several high-profile sonar programs, including the Lockheed Martin/Northrop Grumman AN/SQS-53C hull-mounted active and passive system on U.S. Navy Burke-class destroyers (DDG-51) and Ticonderoga-class (CG-47) cruisers, a passive sonar upgrade to the Lockheed Martin AN/SQQ-89 anti-submarine combat system, as well as a Northrop Grumman program for corvettes and patrol boats of foreign navies called 21HS, short for 21st Century Hull Sonar.
On the SQQ-89 upgrade project, Roy says his engineers were able to provide the signal processing for sea trials only six months after winning the contract. "The is the advantage of COTS," he says. Completing a system for field demonstrations often can take a year or more.
The FPDP is a 32-bit parallel data bus that runs at 160 Mbytes/s. It connects circuit cards in the front where they are readily accessible, rather than in the back like a traditional backplane does. The VME International Trade Association in Scottsdale, Ariz., is adopting FPDP as a VITA standard, Roy says.
The FPDP is not as flexible a data bus as something like a VME backplane bus, Roy admits, but explains that front-end sonar sensor processing does not require the complexity of a VME bus. "We have a cable that moves data from the sensor to the processor," he says. "It`s a very well-defined direction."
ICS is responsible for most of the signal processing on the 21HS program, which is similar to the AN/SQS-53C, but with half the number of sonar elements so it will fit on ships smaller than U.S. Perry-class frigates (FFG-7).
"Our box is practically the whole sonar, except for the array," Roy says of the 21HS. "It is a totally COTS sonar, completely out of the catalog, and is more powerful than any sonar in the U.S. Navy today."
The 21HS signal processing uses the 24-bit fixed-point Sharp BDSP 9124 digital signal processor chip, which is not as flexible as an Analog Devices 21060 Sharc or Texas Instruments 320C40, but handles fast Fourier transform processing at acceptably high rates, Roy says.
For more information on the FPDP, contact ICS by telephone at 613- 749-9241, fax at 613-749-9461, or on the World Wide Web at http://www. ics-ltd.com. - J.K.
The Front Panel Data Port from Integrated Circuits and Systems simplifies intra-system communications with ribbon connections between circuit boards.
Static sonar designers face upgrade challenges
Designers of the premiere U.S. fixed-site sonar, the Integrated Undersea Surveillance System (IUSS), have been using commercial off-the-shelf (COTS) technology for the past decade, yet now face a host of issues that center on how to carry out upgrades.
U.S. Navy experts say they are satisfied with the existing IUSS signal-processing architecture, which is based on the Intel i860 digital signal processor chip and the Sun Sparc 32-bit RISC microprocessor, but must decide how to make upgrades as manufacturing support for these components wanes.
IUSS is a collection of several underwater listening posts located at strategic "choke points" throughout the world where fixed-site arrays of hydrophones monitor for the passage of potentially hostile submarines.
The IUSS consists of the Fixed Distributed System (FDS), a fiber optic multi-prong hydrophone array able to determine locations and bearings of submarines; the Sound Surveillance System - better known as SOSUS - another network of hydrophones; shore data processing stations for the Surveillance Towed Array Sonar System called SURTASS-Ashore; a transportable FDS called the FDS Advanced Deployable System (FDS-D); and a classified passive sonar device called the Surveillance Direction System.
"We are 100 percent COTS today on IUSS `dry-end` signal processing, and the transition started 10 years ago," says David Morin, senior IUSS systems engineer at the U.S. Naval Command, Control, and Ocean Surveillance Center RDT&E Division (NRaD) in San Diego.
IUSS data processing centers use all VME-based open-systems equipment. The single-chip i860 DSP boards are the CSPI Supercard II and III from CSPI in Billerica, Mass., and the Sparc microprocessor boards are from Force Computers Inc. of San Jose, Calif.
"The next iteration would go to the CSPI Supercard IV, which has four DSPs per card," Morin says. "It is a programmatic issue to stay with CSPI so we don`t have to throw away millions of dollars in cards."
The software operating system is Sun OS, but naval experts are trying to switch to Sun Solaris, Morin says. "This is happening slowly," he admits.
In the near term, Morin, his assistants, and contractor engineers must decide on a signal processing architecture for the FDS-D. "This is a bigger problem," Morin says. "It is not clear if this architecture could solve the problem because it must mount in a van and will be constrained in space."
Morin says he is confident, however, that COTS boards will be appropriate even for a mobile application like FDS-D, and he will go with special ruggedized boards "only as a last resort."
Looking father ahead confronts Morin and his people with more serious challenges. The first is the VME backplane data bus, which he points out is "constrained in bandwidth," prompting him to consider the Raceway DSP interconnect from Mercury Computer Systems in Chelmsford, Mass., the Skychannel front panel data port from Sky Computers Inc. of Chelmsford, Mass., or CompactPCI. This decision he says he plans to make within the next three years.
The i860 DSP, meanwhile, has diminishing support from Intel because the chip is obsolescent technology beside state-of-the-art DSPs. Morin says he isn`t concerned about a lack of i860 support from Intel "for several more years," he knows that eventually he must move to a newer chip, such as the Analog Devices 21060 Sharc or IBM/Motorola/Apple PowerPC. He admits that DSPs from Texas Instruments are not under consideration.
Yet moving from the i860 could be difficult because Morin says he needs a DSP that somehow will be able to run existing i860 sonar-processing software. "We have all this code that is i860 based, and we want to choose a vendor who knows what to do with that." - J.K.
The Surveillance Towed Array Sonar System, better known as SURTASS, feeds sonar information to a shore processing station. The SURTASS-Ashore system has been using COTS signal processors for years, but designers now must consider upgrading to new technology.
Upgrades are paramount in new airborne sonar processor
Boeing engineers are blending their experience in sonar system design with commercial off-the-shelf (COTS) components to build an airborne anti-submarine warfare system they hope will command attention in the U.S. and international navies.
The designers understand, however, that using COTS requires them to put a big priority on upgrades - even before the original design enters the prototype stage - because the lifecycles of COTS components are so much shorter than custom or mil-spec parts.
The system is called the Airborne Acoustic Processor System (AAPS), a 32-channel sonar from the Autonetics and Missile Systems Division of Boeing North American Inc. (formerly Rockwell Autonetics) of Anaheim, Calif.
Boeing engineers are designing the AAPS with the company`s own money in hopes of selling in the U.S. or overseas. Company executives have already sold one AAPS in the United Kingdom for post-mission sonar analysis, and expect to complete the AAPS in the upcoming U.S. Navy program to upgrade the Lockheed Martin P-3 Orion antisubmarine patrol aircraft, says Paul Brewer, the company`s manager of market development for acoustic and signal processing products.
In the P-3, the 100-pound AAPS has the potential to replace a 500-pound suite of existing sonar processing gear, Brewer says, adding that this approach would reduce electronic system volume from 17 cubic feet to 3.5 cubic feet, and power consumption from 3 kilowatts to 3/4 kilowatt.
The inspiration for the AAPS came from AN/BQQ-9 towed array processor and the Towed Array Broadband Interim Display Unit (TABIDU) on Ohio-class ballistic missile submarines, both of which Boeing Autonetics engineers designed, Brewer explains.
"AAPS is an adaptation of the TABIDU, and is 100 percent COTS," Brewer says. "It was clear we had the opportunity to migrate this technology into the airborne realm."
The system uses the VME backplane, and VME boards - based on the Intel i860 digital signal processing chip - from Mercury Computer Systems in Chelmsford, Mass. It uses the Sun Sparc 32-bit general-purpose microprocessor for control, data display, and other tasks on a VME board from Force Computers Inc. of San Jose, Calif.
Designing the AAPS shallow-water multistatic software algorithms for the active sonar mode is BBN Systems and Technologies in Cambridge, Mass.
Multistatic sonar uses several transmitters and receivers to enhance accuracy. "The AAPS has up to 20 times the range of traditional sonobuoy systems," Brewer says. "Targets are proliferating and are getting quieter, in blue water and shallow water. We designed AAPS to counter all threats."
It has 32 analog receiver inputs, and as many as eight digital signal inputs for digital sonobuoys. In addition, AAPS has four video outputs for displays and plots. It has built-in diagnostics, and all VME cards, power supply, and fans are removable for flight-line maintenance, Brewer says.
Upgrade decisions
The original AAPS design was an eight-board system with only one i860 DSP, Brewer says. Today, Boeing engineers are prototyping a scalable system with quad-i860 cards with 16 to 20 DSPs in each box. "The i860 still offers the best overall capability," he says, but acknowledges that Boeing engineers must consider a replacement as industry support for the DSP fades away.
Candidate replacements are the Analog Devices 21060 Sharc DSP or the IBM/Motorola/Apple PowerPC, but not any of the DSP architectures from Texas Instruments, Brewer says.
Reusing existing sonar processing software in an upgraded system should not be a big challenge because the existing AAPS uses the Posix operating system interface, which enables software to run on virtually any hardware, Brewer says.
The emergency of the PowerPC as a DSP as well as a general-purpose microprocessor may offer Boeing engineers the opportunity not only to make high-capability, cost-effective upgrades, but to simplify the AAPS architecture as well.
This approach would involve switching out the i860 DSP and Sparc microprocessor, and replacing both kinds of chips with the PowerPC, Brewer says. - J.K.
An SH-60F Seahawk from U.S. Navy Helicopter Anti-Submarine Squadron 5 lowers its dipping sonar in the Mediterranean last June. Boeing North American engineers are designing an airborne sonar system for helicopters or for fixed-wing aircraft.
