YF-23 Team Sees Savings in Switching From Distributed to Integrated Processor
Northrop and McDonnell Douglas will propose an integrated processor architecture for their YF-23 advanced tactical fighter, saving 50% in processor cost, weight and power compared to the distributed system originally planned.
The big surprise is the change back to an integrated processor. Distributed processors, capitalizing on “computer on a chip” technology, had been heralded as the architecture of the future. Originally the YF—23 team thought so, too, and was developing distributed processor systems. Increased combat survivability was a major reason for distributing systems throughout the aircraft, since a single hit would not be apt to cause critical damage.
In aircraft like the YF—23, where computers direct both the flight and engine controls, survivability of the computer and aircraft are likely to be synonymous.
SIX-FIGURE SAVINGS
However, while preparing to submit a full-scale development proposal, the Northrop/McDonnell Douglas team evaluated a number of trade-offs—commonly called trade studies. One effort examined the processor architecture and came up with a different answer—the integrated system which had been considered passe. The resulting savings will be in the six figure area for each aircraft, according to
Larry Nanney, mission avionics program manager for the YF-23 at McDonnell Douglas. Undoubtedly the savings were a big motivator, but increased performance of the integrated processor was also a factor.
The new technology that enabled greater performance at lower cost is Unisys’ advanced capability 32-bit General Purpose Processing Element (GPPE), Nanney said. The GPPEs are standard type B electronic modules. The YF-23 team will propose an architecture using one common signal processor, packaged in two core areas, with two 75-card racks. The two core areas will be separated for survivability, so that a single 40-min. shell would not destroy both, Nanney said.
The original distributed design, which was evaluated during ATF demonstration/ validation (dem/val), consisted of two common signal processors and three mission data processors. Each of the latter had three 1750 processors. Because they were not collocated, each processor required a module to interface with the data bus, and three power supplies for each mission data processor, The capacity of the integrated processor also will increase compared with the distributed system. The throughput capacity will be 20 million instructions per sec. (MIPS) instead of 2 MIPS. Further, with the new architecture the system will have a general purpose processing capability of 420 MIPS, and 6 billion operations per second of signal processing. Signal processors operate at higher speed than general purpose processors because each is tailored to a specific task.
During the demonstration/validation phase, the distributed system with 1750 processors handled only 2.5 MIPS, with contractors hoping for growth to 5 MIPS. The YF-23 will have 20 times the data processing and 87 times the signal processing
capacity of the F-15C. Nanney used the F-15C rather than the F-15E because the C is also an air-superiority aircraft. The ground attack mission adds memory requirements for the F-15E. “I’d be surprised if any tactical aircraft exceeds the E today since "it is the most recent upgrade in existing aircraft,” Nanney said. The YF-23 will have 17 times the data processing and 20 times the signal processing capability of the F-15E, he said.
There will be 15 different types of modules, with reliability predicted to range from 10,000 to 17,000 hr. The modules will use many very high—speed integrated circuits (VHSIC) chips. Typically between 10% and 15% of each chip is devoted to monitoring its own health.
Because of the predicted high reliability, the core signal processor will carry only one spare of each type of module.
The contractors demonstrated the advanced 32-bit general purpose processors in August. Faults were injected, and the system showed the ability to diagnose problems and reconfigure itself. The team’s response to the Air Force request for proposal (RFP) for full-scale
development of the ATF is due Jan. 2. Following evaluation of the proposals, the Air Force expects to select one team by
Apr. 30, 1991.
In preparing its proposal, the Northrop/ McDonnell Douglas team went through a number of risk reduction activities. They primarily addressed high risk / potentially high payoff items that fit in three categories:
* Trade studies.
* Avionics ground prototype demonstration.
* Flight test.
The team conducted about 103 major trade studies, including a host related to the core processing system. The change in architecture is one result. Other subjects considered included radome designs, whether to use active or passively scanned radars, and the Electro-
Optic Sensor System (EOSS). The Air Force decided that BOSS, which was actually an infrared search and track system, had not made sufficient progress to be included in the baseline aircraft (AW&ST Oct.1, p. 73).
The Northrop/McDonnell Douglas studies also resulted in the elimination of a sensor to detect lasers. Research showed there was only one laser threat in the world, Nanney said, which did not make the sensor cost effective.
The preponderance of avionics testing to date has been conducted in ground prototypes at Northrop’s Hawthorne, Calif, and McDonnell Douglas’ St. Louis, Mo., facilities. Nanney termed the tests in June and August the “final exams” for avionics in dem/val. In the June test, all sensors were operated in the modes that would put the most stress on the software. Threats were simulated and sent to signal and mission data processors as if they had come from the sensors. The August tests used in-flight sensor data to demonstrate the ability of the data fusion algorithm and the displays.
The YF-23 team conducted 103 avionicstest flights in a BAC-111 aircraft, accumulating 279 flight hours between July, 1989, and Sept. 12, 1990. Tests concentrated on the radar, electronic combat system and EOSS. The flight tests were particularly beneficial
for the Electronically Scanned Array (ESA) radar, produced by the Westinghouse and Texas Instruments team. McDonnell Douglas first flew the array in August, 1989. The hundreds of active transmit/receive modules give the radar the ability to rapidly transmit energy in different types of beams and multiple directions.
One concern that could be answered only in flight test was whether vibration and temperature changes would affect the coherence of the radiated signal. That problem and others were solved, stretching the ESA as far as the state of the art could accommodate, Nanney said. He predicts that the processing will stay state-of-the art until the turn of the century.
Reducing the pilot‘s workload has been a major goal of the YF-23 team. That effort has centered on processing and displaying the data so the pilot can spend his time planning his attack and flying the aircraft, rather than interpreting data. In the past, a multiple cathode ray tube (CRT) cockpit has dedicated one display to each sensor. The YF-23 team is taking a different tack and trying to integrate the data so the pilot could get the offensive information on one display and the defensive picture on another. The intent is to combine all attributes of a target, gathered from multiple sources, into one presentation. That takes a significant amount of processing, Nanney said.
To reduce the pilot’s workload, the team is attempting to develop intuitive display formats. The contractors will be working to improve
the data fusion throughout full—scale development. Thus far, they have integrated a Digital Equipment Corp. VAX computer, simulating 20 or more aircraft to stress the system. They also have used recorded in-flight data, but of a lesser number of aircraft.
The goal of increasing the pilot’s situational awareness will be aided by the combination of long-range, very-high accuracy sensors and a great improvement in processing capability. With an avionics suite operating in consonance with the stealth characteristics employed for a particular mission, the pilot should be able to detect and either engage or bypass an enemy before the YF-23 is detected,
according to Mike Major. He is Northrop’s manager of YF-23 operational mission objectives.
ESOTERIC CONCEPTS
To develop the integration, the Northrop/McDonnell Douglas team used a lot of man-in-the-loop simulation. Blue (friendly) pilots flew YF—23 simulators against red (enemy) pilots in MiG-29 and a next-generation threat aircraft. While the contractor provided most of the simulator pilots, the USAF’s Tactical Air Command sent a core group of pilots to act as a “sanity check.”
Some of the more esoteric display concepts, such as three-dimensional (3-D) displays and 3-D sound, were tried and rejected as not ready for operational use. The technology has developed sufficiently for 3-D displays, but the problem was that except for one situation, they were too hard to interpret, Major said.
Another idea being studied uses tactile stimulus to alert the pilot to high—payoff functions. The device being examined is a sleeve for the left arm. One extremely high payoff cue would help the pilot locate threats or targets in the quadrants behind him.
Systems for preventing g—induced loss of consciousness (gloc) in high—performance fighter aircraft have received increasing attention in the past few years. There are many ideas but no consensus on how a gloc system should be implemented, Major said. The ATF specification does not call for a gloc system. The YF-23 team has an assisted positive pressure breathing system that should increase the pilot’s ability to tolerate g forces.
There is a requirement for a ground collision avoidance system, which will warn the pilot to pull up, but will not take over control of the aircraft. To expand the capability to a pure gloc would require additional avionics, Major said.
View attachment 691432
Cockpit of a YF-23 is seen in an air-to-air engagement
at the McDonnell Aircraft simulator in St. Louis. The displays were sanitized to protect presentation details.
Source: AWST 17 December 1990