In 1955, I went to work for the North American Aviation Company’s Aerospace Laboratory as a research machinist in Downey, California. A while later, the laboratory was split up into entities that became separate divisions of NAA. One such division was named “Autonetics” and the new division was to be responsible for the development of the, then, new science of “inertial navigation”. I became a “bench machinist” and in that capacity was assigned the responsibility of fabricating the parts of the electro-mechanical system that was employed in the adaption, for submarines, of the inertial guidance system being developed for the, soon to be cancelled, SM-64 intercontinental cruise missile system. These adapted IMUs were the first of the many “ships inertial navigation system” we built for the U.S. Navy over the years and were the ones which were used to guide the U.S.S. Nautilus and the U.S.S. Skate under the arctic sea ice to the north pole. That was my introduction to the world of inertial guidance. (We knew the IMUs were intended for submarines when we were told the IMUs had to fit thru what we were told was a wooden mockup of submarine torpedos hatch that were leaning on the wall in the machine shop. Also, it was clear from the drawings that the IMU was intended for a ship application.)
After a stint as an instrument test technician and obtaining a BSc in physics, I was assigned as the test engineer for the then new Autonetics gyroscope, the G9. ( The gyroscope for the aforementioned IMU was the G2 and the gyroscope for the Minuteman missile IMU was the G6.) The G9 design was aimed at the aircraft market, specifically, the Autonetics N16 IMU , then under design. I was promoted to the position of Senior Responsible Engineer for the G9 gyro. Autonetics was successful in in obtaining a contract for the N16 as the guidance system for the FB-111 strategic bomber and it served as such for the life of the aircraft type. The U.S.Navy, meanwhile, had great success with our G7 gyro and the SINS, of which it was a part, and expressed an interest in a SINS for the then new U.S.S. Los Angeles class attack submarines. The Navy did not need the accuracy of the G7 equipped SINS for the attack boats and Autonetics proposed a modified N16 as an attack submarine SINS. Thus was born the concept of the MiniSINS. The Navy bought the concept of the N16 derived MiniSINS and Autonetics continued as the U.S. Navy’s sole supplier of the Navy’s SINS.
There was a significant difference between the two applications of what was essentially the same system. This difference were centered on the performance demand placed on the gyro controlling the azimuth axis of the system. In the aircraft application, the azimuth gyro must perform within specification for a matter of time measured in hours. In the marine application, this time is measured in days. Thus, it became increasingly apparent that the success of the MiniSINS application would depend in large part on the extent to which the system was over- designed for the aircraft application. A case in point is the stability of the mass unbalance along the spin axis. The mass unbalance along the spin axis is the gyro’s sensitivity to acceleration in the plane normal to the spin axis. The azimuth gyro is oriented such that the spin axis is horizontal, Therefore, since the IMU was of the locally level variety, the azimuth drift rate was principally of a value equal to the unbalance sensitivity times one “g”. In an aircraft IMU, over several hours, the accumulated azimuth error due to changes in the mass unbalance sensitivity was insignificant. Replace “hours” with “days” and one has the making of a problem for the MiniSINS IMU application of the N16.
The G9 gyroscope was a smaller version of the G6 gyroscope with the added feature of continuous rotation of the gyroscope case with respect to the mounting surface. This rotation resulted in the averaging to zero of the gyroscope’s case fixed bias. This was the distinguishing difference between the Minuteman missile G6 and the FB-111/MiniSINS G9 gyroscopes. Both of these gyroscopes were “two axis, spherical gas bearing, free rotor” types. The rotors for these gyroscopes were supported by and were spun upon a spherical, auto-pressurised gas bearing. The stationary part of the bearing was made from a hard chrome plated ball supported on a shaft embedded in the gyroscope housing. The rotating part, the rotor, was made of two hemispherical halves, plated with electroless nickel, which, when assembled over the ball, formed the “spherical gas bearing”. The position of the center of support of the mass of the azimuth’s gyroscope rotor was required to be within a fraction of a microinch of its original value if there was any hope of meeting the repeatability requirement set for the value of the mass unbalance along the spin axis. The amazing fact is this requirement was met for the FB-111 application of the gyroscope.
However, for the MiniSINS application, as system data accumulated over time, it was clear that a performance problem that was best explained by instability of the mass unbalance along the spin axis was making itself known. Since the gas bearings for the G9 gyroscope were identical for the two applications, it was obvious that the root cause of the problem was not the gas bearing. Somehow the center of mass of the rotor was changing from startup to startup and over time for any particular startup.
Any gas bearing needs a pressurised gas in maintain its integrity under load. The G9 gyroscope used Hydrogen gas (least viscosity). The G9 rotor spun at 240 revolutions per second. This rotation rate guaranteed the gas flow inside the gyroscope was chaotic, not laminar. This chaotic gas flow was the principal source of the white noise part of the G9 drift rate. I had been experimenting with operating the gyroscope at a lower rotor speed in order to lower the white noise, so, it was easy to begin to investigate the MiniSINS problem with rotor speed as a investigative parameter. It did not take long to establish that the instability of rotationally induced rotor distortion was the culprit. It seemed obvious that some rotor distortion was a fact of life;the question was: how to eliminate the instability of this distortion?
We needed to first understand the form and amount of the distortion of the rotors as they were normally spun-up. We did the first ever “finite element” analysis of the G9 rotor. This analysis was made more challenging by the existence of a bonding joint that served to join the outer Inconel inertia ring to the beryllium rotor web. Another challenge to the analysis was the way the two rotor halves were joined together to create the spherical gas bearing. The equatorial faces of each rotor half were joined together, face to face, to form the bearing and were held in this position by screws. This made for a rather complex setup as far as the elements of the analysis were concerned. It made the interpretation of the results problematic in the we could not be sure we “modelled” the actual rotor assembly realistically. However, we did gain confidence that we understood well the “shape” response of the rotor to the stress of spinning. We puzzled over the results of the analysis for a while before asking the critical question: what was the purpose of the equatorial faces of the rotor halves? It was to create a dimensionally correct spherical gas bearing from the joining of the two rotor halves, face to face. We did not need entire face to face contact to achieve this; a narrow ring adjacent to the bearing hemispheres would do the job! Once this critical insight had been achieved, it became clear to us what our solution to the azimuth drift problem would be.
The solution to the IMU azimuth drift problem was to allow the rotor to freely assume its final shape under the influence of the stress placed on the rotor by spinning. This was achieved by limiting the face – face contact area to a much smaller annular ring immediately adjacent to the bearing hemisphere. This change in rotor design greatly reduced the chance of the migration of the center of mass of the rotor along the spin axis by creep resulting from the rotor internal stress slowly being relieved and causing the azimuth drift rate to be erratic. The problem of erratic azimuth drift rate was solved.