" INERTIAL REFERENCE SYSTEM " Field of the Invention This invention relates to an inertial reference system (IRS) for producing output signals representative of the inertial position of the object to which the IRS is affixed and, more particularly, the present invention pertains to an IRS employing dithered ring laser gyros as rotational displacement sensors, accelerometers for translational displacement sensors, a first processor, Fl, for compensating the sensor-produced signals and synchronizing the gyro-produced signals to a common time interval and a second processor, P2, for performing navigational computations on the PI processor synchronized and compensated data.

Background of the Invention Inertial reference systems are commonly employed in aircraft and other vehicular guidance systems. Such systems are packaged within a standard- size box, which is then mounted within the vehicle to be monitored. Sensors within the. system detect both attitude and velocity changes with respect to three mutually perpendicular axes. These positional signals are then processed to produce output signals corresponding to positional changes of the vehicle with respect to a fixed, reference position. In the past, angle rate sensing has been provided by gim alled, spinning wheel gyroscopes. The cost of such gyroscopes has been high, due to the careful assembly required of the precision machined parts employed. In addition, mechanical gyroscopes suffer many limitations, including mass unbalance, structure instabilities, a requirement for extreme cleanliness, and the need for a substantial "spin-up" time prior to use.

Many modern inertial reference systems now utilize ring laser gyroscopes as attitude sensors. The ring laser gyro employs .an enclosed light path that is oriented in the sensitive plane of the gyroscope. Two laser beams simultaneously traverse the light path, one in a relative clockwise direction, and the other in a relative counterclockwise direction. When the enclosed path is

rotated in inertial space, the clockwise and counterclockwise paths exhibit different lengths. As such, there is a frequency shi t between the two traversing light signals, which frequency shift is representative of the rate of rotation of the gyro in its sensitive plane. Thus, by monitoring the difference in frequency of the laser beams traversing the path, an output signal representative of angular rotation rate may be produced. By providing a cluster of ring laser gyros, each with its sensitive plane oriented with respect to three mutually perpendicular axes, attitude changes in any direction may be accurately monitored.

A clearly identified problem with ring laser gyros is a pheno enon known as "lock-in". As the rate of rotation of the gyro is reduced, the two oppositely traversing light rays will frequently lock to a common frequency long before the rotational rate of the gyro has fallen to zero. As a result, resolution of small angular rate deviations is lost.

One way to prevent lock-in is to dither the gyroscope in its sensitive axis. By applying a sufficient dither motion to the gyroscope, the two oppositely traversing light beams constantly experience rate changes sufficient to prevent the two beams from locking to a common frequency.

In inertial reference system applications, the three ring laser gyros forming a cluster are all mounted to a common structure, designed to orient each gyro in one of three mutually perpendicular axes. If all ring laser gyros are dithered synchronously, this could result in the excitation of resonant frequencies of the gyros and support structure, thereby producing undesired deflections and corresponding error output terms from the gyros. As such, it is imperative that the three ring laser gyros be dithered asynchronously. The signal produced by a ring laser gyro is digital in nature, having a pulse repetition rate corresponding to the rate of gyro angular displacement. To eliminate the dither signal from the produced gyro signal, the count of the gyro pulse output is accumulated over one full dither cycle, and measured at dither zero crossings. Inasmuch as the three gyros are dithered asynchronously, the corresponding gyro dither zero crossings occur asynchronously. The naviga¬ tional algorithms used to process the data and produce positional output information assume, however, that the rate data for each of the three mutually perpendicular inertial reference system axes is taken at the same point in time. Thus, to avoid errors in the produced positional data for the inertial reference system, a means must be provided to resynchronize each of the three gyro signals to a common interval.

Heretofore, inertial reference systems employing ring laser gyros have been relatively large in size, and quite expensive in cost. As such, the

systems lend themselves to application only in larger aircraft. It is desirable to reduce the size and cost of ring laser gyro-based inertial reference systems such that they are suitable for a broad application for all sizes of aircraft or other vehicles. As the ring laser gyros and accelerometers are packaged in smaller enclosures, special attention must be paid to the temperature compensation of the sensors, both at an individual sensor level, as well as a system level. Also, bias and offset compensation both at a sensor and a system level should be provided, along with sculling correction for the accelerometers and coning correction for the gyroscopes. Further, the correction factors used for tempera- ture, bias, and other error source correction must be capable of inexpensive and high-speed recalibration, thereby minimizing the downtime of such devices, while maximizing accuracy. Further, the functions of the inertial reference system should be modularized in such a way that a single basic inertial re erence system configuration may be quickly and inexpensively modified for use in any of a variety of applications.

Summary of the Invention The present invention, therefore, is directed to an improved inertial reference system. This improved inertial reference system may be constructed in a small enclosure and is designed to minimize the use of expensive components, thereby reducing overall cost. In addition, the present inertial reference system may be quickly and inexpensively modified for particular applications. To assure high accuracy, the system employs high-level temperature, coning, and sculling correction, as well as both individual sensor and system correction for bias, misalignment, and other error sources. The correction factors used tύ implement such compensation are easily and quickly changed, thereby providing ease in recalibration.

In addition, a unique resynchronizing system synchronizes each gyro zero crossing output signal to a common time interval, thereby assuring high accuracy in the produced position signals. The resynchronizing is ac- complished by circuitry that is relatively inexpensive and consumes a minimum of space.

Briefly, according to the invention, an inertial reference system comprises an inertial sensor assembly. The inertial sensor assembly includes gyros for producing asynchronous output signals Δθ , Δ6 , and Δθ_ representative of the change in angular position of the inertial sensor assembly with respect to three mutually orthogonal coordinate axes X, Y, and Z, respectively. Also provided in the inertial sensor assembly are accelerometers for producing output signals ΔV . ΔV„, and Δ V, representative of the change

in velocity of the inertial sensor assembly in the X, Y, and Z coordinate axes, respectively.

A first processor, PI, predeterminedly accesses and processes the Δθ _χ , Δθ„, Δθ _z , ΔV _χ , ΔV _y , and ΔV _Z signals. The PI processor includes means 5 . _ : --for:

(a) providing coning correction for the Δθ_ x, Δθ. y,, and Δθ z signals,

(b) providing sculling correction for the ΔV X , ΔV jr , and Δ V z signals, and

10 (c) temperature compensating the Δθ„ x, Δθ y„, Δθ_ z, & V x ,

Δ V , and ΔV _Z signals.

A second processor, P2, performs navigational computations utiliz¬ ing the PI processed Δθ _χ , Δθ , Δθ _z , _S _χ , ΔV , and ΔV _g signals to produce output signals related to the inertial position of the inertial sensor assembly. l- Preferably, the gyros are comprised of a cluster of three ring laser gyros, with each ring laser gyro being mounted to sense angular deviation of the inertial sensor assembly in one of the X, Y, and Z coordinate axes. Each ring laser gyro produces a first output signal + Δθ X , + Δθ V , + Δθ z having a pulse repetition rate proportional to the rate of angular displacement of the ring laser 0 gyro in a reference clockwise direction and a second output signal - Δθ , -

Δθ and - Δθ having a pulse repetition rate proportional to the rate of angular y z displacement of the ring laser gyro in a reference counterclockwise direction.

Dither means are provided for applying an asynchronous periodic dither motion to each of the ring laser gyros. 5 The PI processor preferably includes a gyro counter for accumulat¬ ing the pulse count of each ring laser gyro-produced first output signal and each ring laser gyro-produced second output signal over each dither cycle. Gyro storage is provided for storing the accumulated pulse counts for each of the gyro output signals over each dither cycle. The PI processor logic periodically 0 samples the gyro storage and predeterminedly processes the stored pulse counts to produce angular rate signals synchronized to a common interval. This interval, preferably, is the PI logic sampling time.

The period of each dither cycle is, preferably, longer than the sampling period of the PI processor logic. The PI processor further includes a

dock for producing a periodic signal having a significantly higher frequency than the frequency of the PI processor logic sampling rate. A resync timer accumulates a count of the clock signal during the interval between the end of each dither cycle for each ring laser gyro and each subsequent logic sampling time. The P processor logic processes the stored pulse count and the resync timer stored count for synchronizing each ring laser gyro-produced signal to the logic sampling interval.

The PI processor further includes means for identifying a first condition wherein the sampled pulse count changed from the previous sampling interval and a second condition wherein, due to the dither period being longer than the logic sampling period, the sampled pulse count remained the same as the previous sampled pulse count. In response to the first condition, the PI processor logic interpolates a value for the resynchronized pulse count between the sampled pulse counts for the preceding two dither cycles. In response to the second condition, the PI processor logic extrapolates the value for the re- synchronized pulse count from the sampled pulse counts for the preceding two dither cycles.

The inertial sensor assembly accelerometers preferably comprise a triad of three accelerometers, with each accelerometer being mounted to sense acceleration of the inertial sensor assembly in one of the X, Y, and Z coordinate axes. Each accelerometer produces a first output signal + ΔV X , + ΔV„ j, + ΔV z having a pulse repetition rate proportional to the acceleration of the accelero¬ meter in a reference positive direction, and a second output signal - ΔV A.

- ΔV„, - ΔV having a pulse repetition rate proportional to acceleration in a y ^z reference negative direction of the accelerometer.

The PI processor includes an accelerometer counter for ac¬ cumulating the pulse count of each accelerometer-produced output signal over a predetermined periodic interval. Accelerometer storage is provided for storing the accumulated pulse counts for each of the accelerometer output signals over each of the predetermined periodic intervals. The PI logic periodically samples the accelerometer storage and predeterminedly processes the stored pulse counts.

Further, the PI processor preferably includes temperature com¬ pensation for the ring laser gyro and accelerometer-produced signals, including means for: a) monitoring the temperature of the gyros, b) monitoring the temperature of the accelerometers,

c) storing predetermined coefficients for the gyros and ac¬ celerometers, these coefficients providing correction factors for the gyro and accelerometer produced output signals as a function of temperature. PI processor logic produces temperature-compensated gyro and accelerometer signals responsive to the monitored temperatures and stored coefficients.

Also, the P processor preferably includes means for storing bias, scale factor and misalignment-related correction factors for the gyros and accelerometers and the PI logic produces bias, scale factor and misalignment- corrected gyro and accelerometer signals responsive to the stored bias, scale factor and misalignment-related correction factors.

The stored correction factors for temperature, bias and misalign¬ ment correction are all stored in electrically erasable, programmable read-only memory, whereby these factors may be easily changed, and the inertial refer¬ ence system quickly recalibrated, by writing in new correction factors. Brief Description of the Drawings

FIGURE 1 is a block diagram illustrating the principal features and functions of the inertial sensor assembly, the PI processor and the P2 processor;

FIGURE 2 is a detailed block diagram setting forth the principal components of the ring laser gyro cluster and the digital accelerometer triad; FIGURE 3 is a detailed block diagram of the ring laser gyro and its associated dither drive and signal-conditioning circuitry;

FIGURE 4 is a detailed block diagram of the accelerometer and its corresponding digitizer;

FIGURE 5 is a detailed block diagram setting forth the principal components of the PI processor

FIGURE 6 is a block diagram depicting the resyncing and com¬ pensation hierarchy performed by the PI processor;

FIGURE 7 is a detailed block diagram illustrating the arrangement of the counter banks used in the PI processor; FIGURE 8 is a logic flow diagram illustrating the processing of ring laser gyro produced pulse counts through the gyro counters;

FIGURE 9 is a logic flow diagram illustrating operation of the resync timer counter;

FIGURE 10 is a logic flow diagram illustrating the processing of the accelerometer produced signals through the accelerometer counters;

FIGURE 11 is an overall flow diagram illustrating the sequential steps performed by the PI processor in resynchronizing the gyro data for all three axes;

FIGURE 12 is a detailed logic flow diagram illustrating the sequential steps performed by the PI processor in resynchronizing gyro data for a single axis;

FIGURE 13 is a timing chart illustrating a resynchronization case 1 condition wherein past and present gyro data are valid;

FIGURE 14 is a timing chart illustrating a resynchronization case 2 condition wherein the present gyro data is not valid;

FIGURE 15 is a timing chart illustrating a resynchronization case 3 condition wherein the present gyro data is valid but the past gyro data is not valid;

FIGURE 16 is a graph illustrating the second-order polynomial fit used to interpolate and extrapolate resynchronization data;

FIGURE 17 is a flow chart illustrating the procedure performed by the PI processor to provide bias, scale factor misalignment, and coning correc- tion to the gyro-produced output signals;

FIGURE 18 is a flow chart illustrating the procedures performed by the PI processor to provide accelerometer bias, scale factor misalignment, and sculling corrections; and,

FIGURE 19 is a detailed block diagram of the P2 navigational processor.

Detailed Description FIGURE 1 is a block diagram illustrating the preferred arrange¬ ment of the present inertial reference system (IRS). The IRS is divided into three principal modules, including an inertial sensor assembly (ISA) 12, a PI processor 14, and a P2 processor 16.

The ISA unit 12 contains the positional sensors. Attitude sensing is provided by a cluster of three ring laser gyros 20. Each ring laser gyro (described in more detail with respect to FIGURE 3) is mounted with its sensitive axis normal to one of three mutually orthogonal coordinate axes X, Y, and Z. The output from each gyro is a digital signal having a pulse repetition rate proportional to the rate of angular displacement of the ISA 12 about the gyro's sensitive axis.

As discussed above, ring laser gyros are subject to a condition known as "lock-in" when operated at low angular displacement rates. To prevent gy _r0 iock-in, a dither drive applies a sinusoidal dither motion to each ring laser gyro in gyro cluster 20 at a periodic rate. To prevent cross-coupling among the three ring laser gyros in the cluster 20 the dither drive 22 drives the ring laser gyros asynchronously.

The angular displacement-related signals out of the ring laser gyro cluster 20 are processed through signal conditioning electronics 24. In addition, status signals from the ring laser gyro cluster, such as operating temperature, are provided through the signal conditioning electronics 24. An accelerometer triad 30 includes three accelerometers, each having its sensitive axis aligned with one of the three mutually orthogonal coordinate axes X, Y, and Z. Preferably, each accelerometer in accelerometer triad 30 is comprised of a digital accelerometer, having an output pulse repetition rate related to the rate of change of velocity of each gyro along its sensitive axis. A more detailed description of each accelerometer is given with respect to FIGURE 4.

The output signals from the accelerometer triad 30 are processed through signal conditioning electronics 32. Also provided by signal conditioning electronics 32 are status signals for the accelerometer triad 30, such as operating temperature.

The three ring laser gyro cluster 20 signals Δθ x , Δθ y , and Δθ z , as well as the three accelerometer 30 signals ΔV A . ΔV y , and ΔV z_ are passed to the inputs of the PI processor 14.

The functions of the PI processor 14 are to acquire and store, as well as provide compensation for the sensor data. Inasmuch as the gyro and accelerometer-produced signals are digital in nature, the data acquisition function of the PI processor 14 involves the accumulation and storage of the input pulses over periodic cycles. A detailed description of these functions is provided with respect to FIGURES 5 and 6. To correct for various errors in the gyro and accelerometer signals, the PI processor 14 performs sensor signal compensation. By empirical testing, temperature-induced errors in the gyro cluster 20 produced signals, and the accelerometer triad-produced signals 30 are modelled as a second-order poly¬ nomial. The coefficients for this polynomial are stored in memory provided in the PI processor 14. As discussed with respect to FIGURE 5, this memory is preferably electrically erasable, programmable read-only memory (EEPROM) which facilitates convenient loading and changing of the coefficients without effecting a hardware change.

As mentioned hereinabove, due to the fact that the ring laser gyros in the ring laser gyro cluster 20 are dithered at asynchronous rates, and it is desired to accumulate the digital outputs -of each ring laser gyro over a full dither period to thereby integrate out the dither signal, the Δθ , Δθ , and Δθ

* y signals supplied to the PI processor 14 from the ISA 12 are necessarily

asynchronous. It is desirable to synchronize the three signals to a common time interval to both provide coning correction and to perform navigational computa¬ tions on positional signals taken at the same instant in time. In a manner described in great detail with respect to FIGURES 10 through 15, the PI 5 processor 14 performs gyro signal resynchronization.

The gyro signals Δθ _χ , Δθ , and Δθ^ must also be provided with coning correction. Coning correction is required as a result of errors induced in the gyro-produced signals due to an environment having a spectrum of vibration which is detectable by the gyros. Specific techniques for providing coning

10 correction are well known to those of ordinary skill in this art and, for conciseness, will not be described in detail herein.

The accelerometer signals ΔV^, ΔV , and ΔV ₈ must be provided with sculling correction. Sculling correction is required due to the fact that linear and angular vibrations in the environment are sensed by the accelero-

15 meters. Specific techniques for providing sculling correction are well known to those of ordinary skill in this art and, for conciseness, will not be described in detail herein. The coefficients required for sculling correction are similarly stored in EEPROM.

Further, the PI processor 14 provides bias, scale factor and

20 misalignment correction for both the gyro cluster 20 and the accelerometer triad

30. The gyros and accelerometers are subject to offset biases and scale factor error that are corrected by the bias and scale factor compensation. Also, known

_. misalignments of each sensor with respect to its corresponding coordinate axis

. _{. .} ^' create sensor output errors. Such misalignment errors may be small for gyro

25 duster 20 whereby correction is not required, whereas misalignment correction for the accelerometer triad 30 is essential for minimizing system error. Bias, scale factor and misalignment errors at the individual sensor levd, and at the system levd are provided by coefficients stored in EEPROM within the PI processor 14. This bias, scale factor and misalignment correction is described in

30 detail with respect to FIGURES 16 and 17.

The fully compensated and resynchronized gyro signals, as well as the compensated accderation signals are then passed to the P2 processor 16.

_

The function of the P2 processor 16 is to process the attitude rate and linear accderation signals through navigational computations to produce output posi- 35 tional data. This positional data is then passed to utilization equipment, such as the aircraft avionics onboard an airplane. . ^* '

A particular feature of 'the present invention is that in being divided into the three main components (i.e., the ISA 12, PI processor 14, and P2

processor 16) the IRS is designed in such a way that it may be easily modified for various applications and easily serviced. If, for example, one of the gyro clusters 20 or accderometer triads 30 requires replacement, new compensating co¬ efficients may be easily loaded into the EEPROM of the PI processor 14. Further, the navigational equations used by the P2 processor 16 may be conveniently changed or, in the alternative, a replacement, fully compatible P2 processor 16 may be easily installed. Thus, the functional arrangement of the individual modules 12, 14, and 16 for the present IRS renders a design that is very easy to modify or repair. FIGURE 2 is a detailed block diagram of the inertial sensor assembly 12 of FIGURE 1. Shown is a duster, indicated generally at 40, of three ring laser gyros 42, 44, 46, each having its. sensitive axis aligned with one of three mutually orthogonal coordinate axes X, Y, and Z, respectively. To prevent lock-in of each laser gyro 42, 44, and 46, dither drive is provided by three dither drive units 52, 54, and 56, respectively. The dither drive units 52, 54, and 56 apply asynchronous, periodic dither motion signals to each of the corresponding ring laser gyros 42, 44, and 46 about their sensitive axes. The three dither sig ^β nals D_ z _^ c-„x _t D _β z _Λ c- „y, and D„ z _Λ c-z „ are provided on output lines from each of the three dither drive units 52, 54, and 56, respectivdy. Each ring laser gyro 42, 44, and 46 produces an output digital signal having a pulse repetition rate proportional to the rate of angular displacement of each gyro 42, 44, and 46 about its coordinate axis. These signals are processed through three corresponding ring laser gyro sensor electronics units 62, 64, and 66. The ring laser gyro sensor dectronics units 62, 64 and 66 output for each ring laser gyro comprises a first line having a signal with a pulse repetition rate related to angular displacement of the ring laser gyro in a reference clockwise direction, and a second line having a signal with a pulse repetition rate representative of ring laser gyro angular displacement in a reference counter- dockwise direction. Thus, the ring laser sensor dectronics unit 62 for the X axis includes a first output line 62a with a signal having a pulse repetition rate proportion^ to the rate of angular displacement of the X axis ring laser gyro in a dockwise direction, this signal being designated + Δθ„ _. A second ring laser x gyro sensor dectronics 62 output line 62b carries a signal - Δθ _χ , having a pulse repetition rate representative of the rate of angular displacement of the X axis laser gyro 42 in a counterclockwise direction. Corresponding signals + Δθ , -

Δθ y„, + Δθ z_, and - Δθ z_ are provide _' d by the Y and Z axes ring laser gyro sensor dectronics 64, 66, respectively. '

As is discussed in greater detail with reference to FIGURE 3, during operation the ring laser gyros 42, 44, and 46 experience mode hops in their frequencies of oscillation. Each ring laser gyro sensor electronics unit 62, 64, and 66 produces an output mode hop signal corresponding to a mode hop condition in its associated ring laser gyro 42, 44, and 46, respectively.

Due to the fact that the signals produced by each ring laser gyro

42, 44, and 46 are temperature dependent, one or more temperature sensors monitors the temperature of each ring laser gyro 42, 44, and 46. The ring laser gyro sensor dectronics units 62, 64, and 66 process these temperature sensor signals, providing corresponding temperature sensor output signals:

Also provided in the ISA is an accderometer triad, indicated generally at 70. The accderometer triad 70 is comprised of three analog accderometers 72, 74, 76, each having its sensitive axis aligned with one of the three mutually orthogonal coordinate axes X, Y, and Z, respectively. The analog accderometers 72, 74, and 76 produce andog output signals corresponding to the rate of change of vdocity of each accderometer dong its sensitive axis. These andog signals are converted to digital signals by corresponding digitizers 82, 84, and 86.

Each digitizer 82, 84, and 86 produces a pair of output signals + ΔV x„, - ΔV x . + ΔV„ y, - ΔV„ y, and + ΔV z__, - ΔV,_., respectively. The signals

+ ΔV * . + ΔV y , and + ΔV_ z are digital signals having a pulse repetition rate

•corresponding to accderations of each corresponding accderometer 72, 74, and 76 in a reference positive direction. The second outputs - Δ ^V _X» - ΔV , and - ΔV _B are digitd signals having a pulse repetition rate rdated to accelerations of the corresponding accderometers 72, 74, and 76, respectivdy, in a reference negative direction.

Inasmuch as the accderometers 72, 74, and 76 and the digitizers

82, 84, and 86 produce outputs that are frequency dependent, each has associated with it a temperature sensor 72a, 74a, 76a, 82a, 84a, and 86a, respectively. The temperature sense signals from the temperature sensors 72a, 74a, 76a, S2a, 84a, and 86a are provided as output temperature sensor signals.

The various indicated output signals from the ISA are routed to the inputs of the PI processor, described in detail with respect to FIGURE 5.

FIGURE 3 is a detailed block diagram illustrating a ring laser gyro and its associated dither drive and output processing circuits. The laser gyro 100 includes a dosed light path for two laser beams, each of which traverses the light path in an opposite direction. .Rotation of the laser gyro in the plane containing the light path effectively varies the path length for each of the

traversing beams, causing a corresponding frequency difference between the two light beams traversing the light path. This frequency difference is detected by means of producing interference patterns between the two light beams and monitoring the interference patterns via photodetectors 102. The photo- detectors 102 produce an output signd A representative of dockwise rotation of the laser gyro, and an output signd B representative of, counterclockwise laser gyro rotation. These signals are passed through corresponding photopreamps 104, 106 and are supplied as inputs to the signd conditioning circuit 108. Signal- conditioning circuit 108 processes the A and B input signals to produce, on a first output line 110, a signd + Δθ having a pulse repetition rate proportional to the dockwise angular rotation rate of the laser gyro 10 in its sensitive axis. On a second line 112, the signd-conditioning circuit 108 produces an output signd - Δθ having a pulse repetition rate proportiond to the counterdockwise angular rotation rate of the laser gyro 100 in its sensitive axis. The signals produced by the laser gyro 100 and its associated circuitry 102, 104, 106, and 108, are known to be temperature dependent. The temperature of the laser gyro is monitored by one or more temperature sensors (not shown) which produce output gyro temperature sense signals on a line 120.

To prevent "lock-in" of the two, oppositdy traversing laser beam signals at small angular displacement rates of the laser gyro 100, a piezoelectric dither drive motor 130 dithers the laser gyro 100 in its sensitive axis at a periodic rate.

The controlling signals to the dither drive motor 130 are provided by the dither drive circuit 132, which also produces an output signd D c corresponding to the zero crossing times of the dither drive motor.

As the laser gyro 100 heats, the dosed light path expands causing a corresponding change -in the path length for the two traversing light beams. A photodetector 140 monitors the laser beams, producing an output signd that is amplified in a preamplifier 142 and fed to the input of a mode hop circuit 144. The mode hop circuit, in response to the signd from the amplifier 142, produces an output mode hop warning signd in the event that the light beam path has changed to the point that a shift in the frequency mode of the light signals is

imminent. In this way, the PI processor can monitor the mode of the laser gyro 100.

The various output signals produced by the ring laser gyro shown in FIGURE 3 are routed to the PI processor inputs for processing, as is described in detail with respect to FIGURE 5.

FIGURE 4 is a detailed block diagram illustrating a single digital accelerometer for use in the inertid sensor assembly. Shown is an analog accderometer 150 that produces an andog output signd proportiond to the rate of accderation of the accderometer 150 dong its sensitive axis. Preferably, a "Q-FLEX" type accelerometer, available from Sundstrand Data Control of Redmond, Washington, is used.

The output from the andog accderometer 150 is passed to the input of a digitizer 152. The digitizer 152 processes the andog output signal from the accderometer 50, as well as an input AC reference signd on a line 154, to produce a pair of output signals + ΔV, - ΔV. The + Δ V signd is a digitd signd having a pulse repetition rate proportiond to accderation of the accelero¬ meter 150 in a reference positive direction dong its sensitive axis. The - ΔV signd has a pulse repetition rate proportiond to acceleration of the accelero¬ meter 150 in a reference negative direction dong its sensitive axis. The + ΔV and - ΔV signals are provided on output lines 156, 158, respectively.

Inasmuch as the signals produced by the accderometer 150 and the digitizer 152 are temperature dependent, temperature sensors (not shown) monitor the temperatures of the accderometer 150 and digitizer 152 producing corresponding temperature sensor output signals on lines 160, 162, respectivdy. The various output signals from the digitd accderometer of FIGURE 4 are fed to the PI processor, as is discussed in detail with respect to FIGURES 5 and 6.

While various implementations of the digitd accelerometer shown in FIGURE 4 are possible, the preferred embodiment is described in copending U-S. patent application Serid No. 620,441 invented by Douglas MacGugan et a , entitled "DIGITAL ACCELEROMETER", filed on the same date as the present application (Attorney Docket No. SUNC-1-1290) and assigned to the same assignee.

FIGURE 5 is a detailed block diagram illustrating the principal components of the PI processor, indicated generally at 200. The inputs to the PI processor are the above-described outputs from the inertid sensor assembly.

Provided as inputs to an X , channel counter bank 202 are the gyro and accderometer signals + Δθ _χ , - 'Δθ , +V _χ , and -V _χ , respectively. The corresponding gyro and accderometer signals are provided as inputs to the Y

channd counter bank 204 and the Z channd counter bank 206. The function of the channel counter banks 202, 204, and 206 is to accumulate the various digitd signals from the inertid sensor assembly over each dither cycle for each ring laser gyro, and over a reference time intervd for each digitd accelerometer. Once the counts have been accumulated for the various sensor signals, these vdues are then loaded into storage provided within the counter banks, 202, 204, and 206. These stored counts are then made available over an address and data bus 210, providing access by the PI processor CPU 212. A detailed discussion of the counter banks 202, 204, and 206 is given with respect to FIGURES 7 through 10.

Coupled over a multiline bus 214 to an input andog multiplexer 216 are the various inertid sensor assembly "vitd functions". These vitd functions are status signals indicating, for example, a mode hop condition of any of the ring laser gyros. Other typicd status signals, such as a fault condition in a ring laser gyro or a digitd accderometer may also be provided on the multiple-line bus 214.

The various aforedescribed inertid sensor assembly temperature sensor signals are provided over a multiple-line bus 218 to a second analog multiplexer 220. The andog multiplexers 216, 220 have address inputs 216a, 220a that are connected to the PI processor address and data bus 210. Thus, the CPU 212 accesses any of the ISA vitd function signals on bus 214, or any of the ISA temperature sensor signals on bus 218 by appropriate addressing on the address and data bus 210. The sdected ISA vitd function or ISA temperature sensor signals are then output from the andog mdtiplexers 216, 220 and passed to the input 224a of a 12-bit andog-to-digitd converter 224. In the known manner, the 12-bit andog-to-digitd converter 224 converts the andog signal at its input 224a to a corresponding digitd signd, which' is then applied at its output 224b to the address and data bus 210.

At predetermined intervals, the CPU 212, preferably comprised of a commercially available type Z8002 microprocessor, accesses the various sensor signals from the channd counter banks 202, 204, and 206, as well as the ISA vitd function signals and the ISA temperature sensor signals as processed through the andog multiplexers 216, 222, respectively, and the 12-bit andog-to-digital converter 224. The operating rate of the CPU 212 is determined by a reference dock signd produced by a dock 230. This, Wgh-frequency dock signd is also passed as an input to a resync timer 232. As is described, in great detail with respect to FIGURES 7 and 9, the resync timer 232 produces a signd

representative of the time between a sampling period by CPU 212 of a counter bank 202, 204, and 206 and the corresponding end of a ring laser gyro dither cycle. The resync timer 232 stored value is used by the CPU 212 to synchronize the various ring laser gyro signals to a common time interval. . The sampling period time, as well as the dither end of cycle signal are provided to the resync timer 232 over the Pi processor address and data bus 210.

The programming for the CPU 212 is stored in a pair of erasable, programmable read-only memory (EPROM) chips 240, 242. In one construction of the invention, sufficient programming capability was provided by each EPROM 240, 242 having a memory capacity of 16kx8.

While the detailed program stored in the EPROMs 240, 242 is not given herein, such programming is manifestly apparent to one of ordinary skill in this art based upon the discussion provided herein of the operation of CPU 212. During the course of performing computations on the input data, the CPU 212 makes use of two random-access memories (RAMs) 250, 252. In one implementation of the invention, each RAM 250, 252 had a provided memory capacity of 8Kx8.

As is described in detail with respect to FIGURES 17 and 19, the PI processor stores coefficients used for temperature, scale factor, bias, and misalignment correcting the various gyro and accelerometer input signals. These coefficients are stored in two electrically erasable, programmable read-only memory (EEPROM) chips 260, 262. A characteristic of the EEPROM chips 260, 262 is that a user may easily change the stored coefficients without requiring a hardware change of the inertial reference system. Thus, recalibration of the disclosed inertial reference system is accomplished quickly and inexpensively. In one embodiment of the invention, each EEPROM chip 26Θ, 262 was provided with a 2Kx8 memory size.

The CPU 212 accesses the EPROMs 240, 242, the RAMs 250, 252, and EEPROMs 260, 262 over the system address and data bus 210. The resynchronized, compensated gyro signals, and the compensated accelerometer signals are routed by the CPU 212 over the address and data bus 210 to the navigation processor P2, described in detail with respect to FIGURE 19.

An interrupt controller 268 applies eight periodic interrupt signals to the PI processor data and address bus 210 causing CPU 212 to perform the procedures corresponding to the particular interrupt

FIGURE 6 is a block diagram illustrating Pi processor activity in response to the three types of interrupts.

The highest frequency interrupt, at block 270, causes the Pi processor to enter the high frequency loop. In this loop, the Δθ x , Δθ y and

Δθ _χ asynchronously produced gyro signals are all resynchronized to the sampling intervd of the PI processor CPU 212. The resynchronization process is described in detail with respect to FIGURES 12-16.

Also performed in the high frequency loop 270 is the coning compensation accumulation for the three resynchronized gyro signals. The accumulated coning compensation signals are used in the coning compensation as described hereinbelow. The resynchronized and accumulated Δθ , Δθ , and Δθ _ signals are passed to the compensation loop 280.

The temperature sense signals from the various inertid sensor assembly temperature sensors are monitored in a low frequency loop 290. Stored within the EEPROMS 260, 262 are temperature dependent coefficients for bias, sede factor and misalignment corrections used in compensating the Δθ . Δθ _f Δθ _z , ΔV _χ , ΔV and Δv _z signals.

At the intermediate frequency loop 280, the PI processor performs data compensation. The input resynchronized and accumulated gyro signals

Δθ . Δθ and Δθ from the high frequency loop 270 are compensated for bias and sede factor errors in compensation loop 280 by means of temperature dependent fourth order corrections to bias and sede factor provided by low frequency loop 290. Also, coning compensation for the gyro signals is produced by use of the accumulated coning compensation as provided by the high frequency loop 270.

The PI processor, in low frequency loop 290, also draws on the fourth order polynomid temperature dependent models stored in the

EPROM 240, 242 for the accderometer signals ΔV A , ΔV y and ΔV ___» to provide bias, sede factor and misalignment correction for the compensation loop 280 accelerometer signals. In addition, sculling correction is provided for these signals to compensate for vibration induced outputs from the accelerometers. Inasmuch as the accderometers are subject to output signals due to angular displacements, the sculling correction draws upon the gyro produced signals Δθ _χ , Δθ _y , and Δθ _z .

The compensation loop 280, therefore, also produces output com¬ pensated accelerometer signals ΔV _χ , ΔV , and ΔV _* FIGURE 7 is a detailed block diagram illustrating the construction of one of the channel counter banks 202, 204, 206 of FIGURE 5, as well as the resync timer 232 of FIGURE 5.

As is discussed with respect to FIGURE 3, each ring laser gyro 300, and its associated signd-conditioning electronics 302 produces two distinct output signals. The first output signd, designated + ΔΘ , has a pulse repetition rate proportiond to the angular displacement rate of the laser gyro 300 in a dockwise direction, whereas a second output signd, - Δθ , has a pulse repetition rate proportiond to angular displacement of the laser gyro 300 in a counter dockwise direction.

The + Δθ and - Δθ signals are passed to the input of counter gate control and synchronization circuitry 310. In a manner described in detail with respect to FIGURE 8, the counter gate control and synchronization circuitry passes the + Δθ signals to a counter 312, whereas the - Δθ signals are passed to a counter 314. Each of the counters 312, 314 may be prdoaded by a vdue stored

^' in a prdoad register 316, 318, respectivdy, and the stored accumulated count in the counters 312, 314 may be hdd in associated hold registers 320, 322, respectivdy.

The prdoaded vdues in the prdoad registers 316, 318 are con¬ trolled by CPU interface circuitry 330. Also, the CPU accesses the vdues stored in the hold registers 320, 322 via the CPU interface circuitry 330.

As is discussed with reference to FIGURE 4, the accderometer 340 and associated digitizer 342 produce a pdr of output signals + Δv, - ΔV. The

+ Δ V output signd has a pulse repetition rate proportiond to accderation of the accderometer 340 dong its sensitive axis in a reference positive direction. The

- ΔV output signd has a pulse repetition rate corresponding to acceleration of the accderometer 340 in a reference negative direction dong its sensitive axis. The + ΔV, - ΔV signals are passed to the inputs of counter gate

-control and synchronization circuitry 350, which also receives a CPU sampling intervd signd over line 351. In a manner described in detail with respect to

FIGURE 10, the counter gate control and synchronization circuitry 350 routes the + ΔV signd to the input of a counter 352, whereas the - Δ V signd is passed to the input of a counter 354. Each counter 352, 354 may be prdoaded with a vdue determined by associated prdoad registers 356, 358, respectively. The accumulated counts in the counters 352, 354 may be loaded into, and stored within associated hold registers 360, 362, respectivdy. The prdoad registers

356, 358 may be activated to load a predetermined initid count into their corresponding counters 352, 354 under control of the PI processor CPU, via the

CPU interface circuitry 330. The PI processor accesses the hdd accumulated counts in the hold registers 360, 362. via the CPU interface circuitry 330.

Also shown in FIGURE 7 is the circuitry utilized for the resync timer, indicated generally at 370. The resync timer 370 includes timer gating circuitry 372 that receives as inputs the dither zero crossing signd, provided from the signd-conditioning electronics 302, the CPU sampling interval, which is the time at which the CPU accesses the various hold registers, such as hold registers 320, 322, 360, 362 to identify the status of the position sensors, and a reference dock signd. In a manner described in detail with respect to FIGURE

8, the timer gating circuitry loads the reference dock signd into a counter 374.

Counter 374 may be prdoaded through an associated prdoad register 376. The accumulated count in counter 374 may be loaded into a hold register 378. The prdoad vdue for the counter 374 hdd in preload register 376 is activated by the PI processor through the CPU interface circuitry 330. In addition, the PI processor accesses the hdd resync timer count in the hold register 378 by means of the CPU interface circuitry 330. FIGURE 8 is a logic flow diagram illustrating the operation of the counter gate control and synchronization circuitry 310 (FIGURE 6) and its associated counters 312, 314, preload registers 316, 318 and hold registers 320, 322.

At start-up, block 400, all counters and prdoad registers are initialized. Each prdoad register 316, 318 is prdoaded with a count of 1, with each corresponding counter 312, 314, being loaded with an initid count of 1.

At the beginning of a dither cyde (i.e., at the negative zero crossing of the dither signd) at block 402, the counters, such as counters 312, 314, are allowed to count their corresponding + Δθ , - Δθ , respectively, gyro pulses at block 404. The counting at block 404 continues until the end of a dither 'cyde (i.e., at the negative zero crossing of the dither signd) at block 406. At this point, the accumulated counts in the counters 312, 314 are loaded into the corresponding hold registers 320, 322, respectivdy, at block 408.

At block 410, each prdoad register 316, 318 then preloads the count of 1 into its corresponding counter 312, 314. The purpose for preloading a count of 1 is that in one embodiment of the invention, the first pulse coupled through to the counters does not get counted.

At block 412 the PI processor CPU reads the vdue stored in the hold registers 320, 322 at the CPU sampling rate. The cyde then repeats with the beginning of the dither cycle at 402.

FIGURE 9 is a detailed logic flow diagram illustrating the se- quentid steps performed by the timer gating circuitry 372, preload register 376, counter 374, and hold register 378 of FIGURE 7. At start-up, block 420, each

resync counter 374 is initialized with a count of 1, as is the prdoad register 376. At each CPU sampling intervd, at block 422, counter 374 is allowed to count the reference clock signd at block 424. At 426, at the end of a dither cycle, i.e., at the negative zero crossing of the dither signd, the accumulated count in the counter 374 is loaded into its corresponding hold register 378, at block 428. At block 430 the prdoad count of 1 is then loaded from the prdoad register 376 into its associated counter 374. A prdoad of 1 is loaded into the counter 374 due to the fact that in one implementation of the invention the initid pulse applied to the counter 374 is not counted. Finally, at its subsequent sampling intervd the CPU, at block 432, reads the vdue of the hold register 378, the system then returning to repeat the cyde at 422.

FIGURE 10 is a logic flow diagram illustrating the sequentid steps performed by the counter gate control and synchronization circuitry 3-50, counters 352, 354, prdoad registers 356, 358, and hold registers 360, 362 of FIGURE 7.

At start-up, block 440, the counters 352, 354 and corresponding prdoad registers 356, 358 are prdoaded with a count of 0.

Then, upon the beginning of an accderometer sampling time intervd at 442, the counters 352, 354 are allowed to count the + ΔV, - ΔV digitd accderometer pulses, respectively, at block 444. At the end of the accderometer sampling time intervd, at 446, the accumdated counts in the - counters 352, 354 are loaded into the corresponding hold registers 360, 358 at block 448. At block 450 a prdoad pulse causes a prdoad of zero to be loaded into the counters 352, 354. At its sampling intervd the CPU, at block 452, reads the hdd vdue in the hold registers 360, 362. The system then returns to repeat the cycle at 442.

Returning to FIGURE 7, at each sampling period the PI processor CPU (212 of FIGURE 5) accesses the + Δθ accumdated count from hold register 320 for each gyro, and the - Δθ accumulated count in hold register 322 for each gyro. Inasmuch as the accumulated + ΔΘ , - Δθ counts in the hold registers 320, 322 are taken over one complete dither cyde of the gyro, a simple subtraction of the count in hold register 322 from that in hold register 320 by the CPU produces a count proportiond to the rate of angular displacement of the particular gyro, with the contribution due to gyro dither cancelled.

As discussed above, the three gyros in the gyro duster are dithered asynchronously to prevent cross coupling between the gyros. Thus, the net

Δθ count taken by the CPU at each sampling period represents change of displacement rates for each gyro at a different point in time. The navigationd computations to be performed on the positiond data in the P2 processor assume that each data point is taken at the same instant in time. Thus, to avoid error in the output of the IRS, the various gyro Δθ pulse counts must be synchronized to a common time intervd. This time intervd, in the preferred embodiment of the invention, is sdected to be the sampling intervd of the PI processor CPU.

FIGURE 11 is an overview flow diagram illustrating the sequentid operations performed by the PI processor (212 of FIGURE 5) in performing the gyro data resynchronization. First, at block 500, the X axis gyro data is resynchronized to the sampling time intervd of the PI processor CPU. This is followed at blocks 502 and 504 by resynchronization of the Y axis, and Z axis, respectivdy, gyro data to the same PI processor CPU sampling time interval.

Once the resynchronization for all three axes has been accomplished, the CPU at 506 cdculates the actud synchronized data vdue for each of the three axes at the previous CPU sampling time. The resynchronized X axis gyro vdue, GXSYC, is cdculated as a previous net pulse count, GYROXP, for the X gyro over the previous dither cycle, plus an incrementd count vdue, GXSTR, minus the incrementd count vdue for the previous pass GXSTRP. Similar expressions for the resynchronized Y vdue, GYSYC, and resynchronized Z vdue, GZSYC, are performed at 506.

In block 507 the coning compensation accumulation is performed as:

SGXCNS=SGXCNS+GXSYC SGYCNS=SGYCNS+GYSYC

SGZCNS=SGZCNS+GZSYC.

As is described with respect to FIGURE 6, the accumulated coning compensation vdue is used to provide coning compensation for each gyro signal. At block 508, the resynchronized position data is then available for additiond bias, misdignment and coning correction, as is fully described with respect to FIGURE 18.

To assure that the stored gyro counts over each dither period are processed, the period of the PI processor CPU sampling time must be shorter than the gyro dither period, i.e., the frequency of the PI processor CPU sampling rate must be greater than the gyro dither frequency. Thus, for each CPU sampling intervd, there are three different possible cases: case 1 - the present stored gyro data is valid (i.e., by valid, it is meant that the Δθ and - Δθ vdues stored in the hold registers have been

updated since the previous CPU sampling intervd) and the past gyro data was valid, case 2 - the present gyro data is not valid, and case 3 - the present gyro data is vdid but the past gyro data was 5 not valid.

The above three cases will repeat at varying rates dependent upon the ratio of the sampling period to the dither period. If the ratio of the sampling to the dither period is dose to one, the sequence of the three cases will be rdatively rare with case 1 occurring most of the time. However, if the ratio is 10 rdativdy large, the frequency of the three cases will be high. Note that the ody sequentid cases allowable are, case 1 followed by case 1 or case 2, case 2 followed by case 3 and case 3 followed by case 1.

FIGURE 12 is a logic flow diagram illustrating the sequential steps executed by the PI processor CPU (212 of FIGURE 5) for resynchronizing the 15 X axis gyro data. Thus, this flow diagram corresponds to block 500 of FIGURE 10. The same sequentid steps of FIGURE 12 are then performed for the Y data, corresponding to block 502 of FIGURE 10, and the Z data, cor¬ responding to block 504 of FIGURE 11. As such, detailed logic flow diagrams for the resynchronization of the Y and Z data are not given herein. 20 . Referring to FIGURE 12, the X axis resynchronization begins at

600. The test is then made at block 602

DTMX ≠ 0?

The vdue DTMX is the time intervd determined by the resync timer 370 of FIGURE 7. FIGURE 13 is a timing diagram depicting the various

25 time intervals used in a case 1 situation. Here, with time as the X axis, shown are sequentid PI processor CPU sampling time intervals i - 1th, ith and so forth.

The period between sampling intervals is designated DT.

Also shown are the X gyro asynchronous dither crossing intervals j - 1th, jth and so forth. The period of each dither cycle is designated T2MX. •30 i will be recalled from FIGURE 9 that the resync timer interval

DTMX begins at each CPU sampling intervd and ends at the successive dither crossing time. The previous resync timer period is designated DTMXP.

If the vdue DTMX does not equd zero, this means that the present gyro data is valid, i.e., the present gyro data vdue has been updated since the 35 previous CPU sampling time. A test is then made at block 604 to determine whether or not the past data flag is set, i.e.,

PGXFLG = 1.

The vdue PGXFLG is a flag which is set, as will be understood herein below, to indicate whether or not the previous sampled data value was vdid. If the previous sampled data vdue was valid, the PGXFLG flag is set to 1.

Thus, assuming for the present pass that PGXFLG = 1, the system identifies, at block 606, that it is in a case 1 condition, i.e., valid past and present gyro data.

Referring again to FIGURE 13, the ^' net Δθ. _χ gyro count at the j - Ith dither crossing is designated GYROXP. The net Δθ count at the jth dither interrupt is designated GYROX. Since, in case 1, both the GYROXP and

GYROX vdues are valid, the PI processor uses these vdues to interpolate, at the ith sampling intervd, a vdue GXSYC which is a cdcdated Δθ _χ vdue for the X gyro at the i - 1th PI processor sampling intervd. The PI processor uses a second order polynomid fit to cdcdate the vdue GXSYC. A full derivation of the terms used to interpolate this vdue is given with respect to FIGURE 16. The discussion with respect to FIGURES 13-15 simply uses the results from this derivation.

At block 608, time T3MX, corresponding to the period between the j -1th dither crossing and the i - 1th sampling intervd, is cdculated as T3MX = DT - DTMXP.

Since the dither period T2MX is subject to some change, the precise dither period for the j - ith to jth dither crossing is cdculated at block 610 as

T2MX = T3MX + DTMX.

The rate of change of gyro counts, GXSTRI, over the previous dither period is then cdculated at block 612 as GXSTRI s ((GYROX - GYROXP)/T2MX)/2.

Then, at block 614 the ddta gyro count, GXSTR, is cdculated as GXSTR = (GYROX + GXSTRI x (T3MX - T2MX)XT3MX/T2MX). The present resync timer intervd DTMX is then stored, at block 616, for use in the next pass as the previous vdue DTMXP. This completes the X axis resynchronization procedure at block 618, and the system then returns to block 506 to cdedate the interpolated vdue GXSYC from the expression

GXSYC = GYROXP + GXSTR - GXSTRP.

Note that from this expression the ddta term GXSTR is added to the past gyro count GYROXP to cdculate the synchronized gyro count GXSYC. The past ddta vdue GXSTRP is subtracted to conserve the totd number of gyro counts.

Returning to the decision at 604 of FIGURE 12, if the PGXFLG flag does not equd one, thereby indicating that the past gyro data was not vdid, the system, at block 620, identifies a case 3 condition, i.e., vdid present gyro data but past gyro data not valid. FIGURE 14 is a timing diagram illustrating a case 3 condition.

Here, at the i + 2th sampling intervd, the vdue DTMX does not equal zero, indicating that the present data vdue is valid. Notice, however, that for the ith through the i + 1th sampling interrupt times, there was no dither zero crossing. Thus, at the previous PI processor sampling intervd i + 1th time, the gyro data had not been updated and, thus, was not valid.

The procedure in a case 3 condition is to use the previous gyro data at the j and the j + 1th dither crossing periods, GYROXP, GYROX, respectively, to interpolate, using a second order polynomid fit, the intermediate vdue GXSYC at the i + 1th CPU sampling intervd. The first step, at block 622, is to cdculate the dither period,

T2MX, between the jth and j + 1th dither crossings. This is accomplished by adding the current resync timer period DTMX to the previously cdculated (i.e., cdculated from the previous pass) T3MX signd T2MX = DTMX + T3MX _QLD . Then, at block 624, the new vdue T3MX intervd between the jth dither crossing and the i + 1th sampling intervd is cdcdated T3MX _NEW * DT ₊ T3MX _OLD .

Notice now that the resync timer intervd DTMX is equd to the interrupt intervd DT plus an incrementd intervd between the i + 1th sampling interrupt intervd and the j + 1th dither crossing intervd. If this DTMX vdue is used in the next pass of the PI processor at the i + 3rd intervd, an erroneous cdculation will resdt from the use of this DTMX. Thus, a correction at 626 is made

DTMX - DTMX - DT. The rate of change of gyro counts, GXSTRI, over the previous dither period is then cdcdated at 628

GXSTRI * ((GYROX - GYROXP) T2MX)/2.

At block 230 the past gyro data count GYROXP is set to zero to preserve counts due to the fact that the preceding case 2 pass already added these counts to the accumdators (block 507 of FIGURE 11). Also, the past gyro data not valid flag, PGXFLG, is set to one, thereby indicating for the subsequent pass that the past data vdue was valid

The ddta gyro count, GXSTR, is cdculated, at 614, as GXSTR = (GYROX + GXSTRI x (T3MX - T2MX))(T3MX T2MX). The vdue of DTMX is stored as DTMXP at 616 and the X axis resync is completed at 618. Returning to 602 of FIGURE 12, if

DTMX ≠ 0 is not valid, the system identifies a case 2 condition at block 640. FIGURE 15 illustrates the case 2 condition wherein the present gyro data is not vdid. Here, the previous gyro data GYROXP and GYROX at the previous j - 1th and jth, respectivdy, zero dither crossing intervals are used to extrapolate the resynced gyro data GXSYC at the ith PI processor sampling intervd.

Notice that for the i + 1th sampling intervd, there is no resync timer period DTMX because the j + 1th dither crossing has not yet occurred. The first step, therefore, at block 642 is to cdcdate the time between the jth dither crossing and the ith sampling intervd as T3MX = DT - DTMXP, where DTMXP is the resync timer intervd cdculated for the previous re¬ synchronization pass.

The dither period between the j - 1th and jth dither crossings, T2MX, was also cdcdated in the previous pass. Also, the rate of change of gyro counts, GXSTRI, over the previous dither period was previously cdcdated.

Therefore, the ddta gyro count, GXSTR, is cdculated at block 644 as

GXSTR = (GRYOX + GXSTRI (T3MX + T2MX)) (T3MX/T2MX). Then, at block 646, the past gyro data valid flag, PGXFLG, is reset to zero, thereby indicating for the subsequent pass that the past gyro data was not valid.

This completes the case 3 X axis resynchronization at block 618.

As discussed above, the equations used for cdculating the resynced gyro count vdue are based upon a second order polynomid fit. The derivation of these equations is understood with reference to FIGURE 16. Shown in

FIGURE 16 is a plot of a graph 700 having ordinate vdues x _Q , x. and x ₂ , with an unknown vdue, labeled x, to be approximated between ₁ and x ₂ . The coordinate vdues of these points are, correspondingly F(x _Q ), F(x and F(x ₂ ⁾ , with the vdue F(x) being the unknown vdue for the ordinate x.

The second order estimate of the polynomid defining the graph 700 is

P(x) ΔF( _Xl ) + ΔF(x _χ ) (x-χ.) Δ'F(x,) (x-xJ (x-xJ li < ^χ ₉ - ^χ ,r 21 (x ₂ -x _χ )

where iF(x _χ ) = F(x ₂ ) -F _χ (x) <*2 ^{" x} l ^}

and

Δ^F( _Xl ) __ F(x ₂ ) - F( _X]L ) Ε(χ.) ^~ Ε{x _Q ) ^x 2~ ^x l ' ^x l ^"x 0

It is noted that

^X l ^{" X} 0 = ^X 2 ^" l

By rearranging terms P(x)s F(x _χ ) + (x-x _χ ) +

+ F(x _χ ) (x-X _j ) (F(x ₁ )-F(x _Q )) (x-x ₀ ) (x-x _χ ) ^x r*o 2(x _r x _Q ) ²

Now, rdating the terms of FIGURE 16 to the terms defined with respect to FIGURES 13-14.

P(x) = GXSYC

F(x _χ ) - GYROXP

(F(x ₂ ) - Εj J) (x- ^ (X-X ₂ ? _ _GXSTRI * (T3MX-T2MX) xT3MX 2(x ₂ - x.) ^" T2MX

where F (x ₂ ) - F(x _χ )

2(x ₂ -x ₁ ) ^{= GXS} ™

and

X - X _χ - T3MX χ ₂ - χ _χ = T2MX.

Therefore, x - _χ = T3MX - T2MX then for the present x the estimate P(x) is

GXSYC = GRYOXP + GXSTR - GXSTRP where GXSTR - (GYROX + GXSTRI x (T3MX - T2MX)) x T3MX/T2MX and

GXSTRP is the past GXSTR.

FIGURE 17 is a flow chart illustrating the further compensation performed by the PI processor on the resynced gyro count. Each of the three . resynced gyro counts for the three coordinate axes, X, Y and Z, is processed in the manner depicted in FIGURE 17, the ody difference among channels being a change in compensation factors.

At block 802 the resynced gyro count is scded by the pulse weight Gnwt and appropriate sede factor Gsfc. Each gyro produces a varying number of pulses for a given input. Thus, to get a calibrated weight for each gyro count, a gdn factor Gnwt is applied at 802 corresponding to a nomind pulse weight, this term being divided by one plus the sede factor gdn, Gsfc, a factor dependent upon the temperature of the individud gyro.

Each gyro also exhibits a unique bias characteristic which is also temperature dependent. Thus, a sensor bias term is subtracted from the output of block 802 in a summer 804.

The resultant count out of summer 804 is passed to the input of a system levd sede factor gdn block 806, a misalignment, between sensors gain block 808, and to the input of a summer 810. ' The system levd sede factor block 806 applies a gdn factor Gssf to adjust sede factor at the gyro cluster system level.

A correction factor Mg is applied at block 808 to correct known misalignments between the desired axid location of the gyro and its actual axid position. The outputs from the gain factor blocks 806, 808 are passed to the inputs of a summer 812. Also passed to the input of summer 812 is a factor

representative of a system bias error, i.e., a bias offset for the entire gyro duster.

The resultant count out of summer 812 is combined in summer 810 dong with factors corresponding to quantization correction and coning correc- tion. The quantization correction term compensates for the fact that samples are taken at discrete quantum levels. Such correction is well known in sampling systems.

The coning correction term compensates for error in the output of each gyro due to environmentd vibration to which each gyro is sensitive. As is described in detail with respect to FIGURE 6, the coning correction vdues are accumulated in the high frequency resync routines and applied to the input of summer block 810.

The resultant count out of summer 810 is a count Δθ c, which is a fully compensated and resynced term corresponding to the change of angular displacement of the corresponding gyro about its sensitive axis.

FIGURE 18 is a flow diagram illustrating the compensation per¬ formed by the PI processor on the accderometer count, Δv. Referring again to FIGURE 7, the PI processor CPU (212 of FIGURE 5) accesses the hold register for the counts + ΔV and - ΔV in hold registers 360, 362, respectively, at regdar sampling intervals. The difference between the counts in the two hold registers produces a resultant accderation count Δ V.

In FIGURE 18, the ΔV count for each channd is scded at block 852. For a given accderation, each accderometer produces a dif erent pulse output. To compensate for these differences, each pulse count is mdtiplied by a nomind wdghting factor Anwt, and then divided by one plus a sede factor gdn Asf c.

In this way, the output from block 852 is -a calibrated pulse count.

Each accderometer exhibits its own bias error. To correct for this error, the output from block 852 is summed with a sensor bias compensation signd in a summer 854.

The resultant count out of summer 854 is applied to the input of a system levd sede factor block 856, the input of a misalignment gain factor block 858 and the input of a summer 860.

A gdn factor Assf in block 856 adjusts the sede factor of each accelerometer in the triad.

A gdn factor of Ma in _/ block 858 adjusts the output of each accderometer to correct for misalignments between the actud axis of the accelerometer, and its desired coordinate axis.

The outputs from gain blocks 856 and 858 are passed to the inputs of a summer 862. A further input to summer 862 is a system bias term to correct for an overall bias of the accderometer triad.

The output from summer 862 is summed in summer 860 with a gain factor Aaniso. The gdn factor Aaniso compensates for accderometer output error caused by motion transmitted to each accderometer due to the fact that the accderometer is not located at the center of rotation of the inertid reference system. This is commody known as anisodastic correction.

The resultant output from summer 860 is passed to block 864 wherein sculling correction is applied. Each acderometer is sensitive to vibration in its environment. If not compensated for, this vibration can lead to a net output error signd from the accderometer. Using known techniques, a sculling correction factor is produced and applied as an input to block 864.

The output from the sculling correction block 864 is the fully - compensated accderation term ΔV .

Referring agdn to FIGURE 5, the PI processor CPU 212 cdcdates synchronized and compensated vdues for each gyro count, and compensated vdues for each accderometer count. These vdues, dong with their appropriate labd are then passed to the P2 processor for navigationd odculations. FIGURE 19 is a block diagram illustrating the configuration of the preferred P2 navigationd processor. The PI processor data is received over the PI processor address and data bus 210 which contdns 16 lines. The PI processor presents its data to the P2 processor in the form of a 16 bit labd followed by a corresponding 16-bit data word. These 16-bit labels and data words are received in two first-in and first-out registers 902, 904. The two registers 902, 904 are employed due to the fact that each register has ah 8x256 capacity. The labels and data stored in the registers 902, 904 are accessed by the P2 processor CPU 910 via the P2 processor 16 line address and data bus 912. The CPU 910, preferrably a type Z8002 commercially available unit, receives its timing signd from a conventiond dock 914.

The programming for the P2 processor CPU 910 is contained within four EPROM ¹ !. 920, 923, each of which is tied to the P2 processor address and data bus 912. In the preferred embodiment of the invention, each EPROM 920-923 has a memory capacity 16kx8. Stored within the EPROMS 920-923 are standard navigationd equations which cdculate the pre¬ sent position of the inertid reference ased upon the ddta angular displacement rates and delta vdocity signals provided by the PI processor. Inasmuch as such

navigationd equations are well known to this art, for purposes of conciseness they will not be presented herein.

During the course of its c culations, the P2 processor CPU 910 utilizes scratch pad memory provided by two RAM's 930, 932. In the preferred embodiment of the invention, each RAM 930, 932 has a memory capacity of 8kx8.

Based on its navigationd cdculations, the P2 processor CPU 910 produces on its address and data bus 912 output signals in a format suitable for use in the drcraft's avionics. Thus, these signals might be fed into the aircraft's flight control systems to fly the aircraft on a desired course, or they may be provided to flight deck instrumentation, indicating to the flight crew the location of the aircraft.

In summary, an improved inertid reference system has been described in detail. The organization of the system into an inertial sensor assembly, a PI data acquisition, sensor compensation and gyro resynchronization processor and a P2 navigationd computation processor provides numerous advantages. Compensation for the sensors may be quickly changed by rewriting new coefficients into the EEPROM within the PI processor. A fdlure within any one of the three major organizationd blocks of the system may be quickly corrected by replacing the failed modular unit with a fully compatible replace¬ ment unit. It has also been found that the present organization lends itself to construction of the IRS within a small chassis, thereby making the present IRS sdtable for application in smaller vehides, such as privately owned aircraft. In addition, the resynchronization technique employed by the PI processor yields highly accurate gyro resynced data, while requiring a minimum of hardware components and thereby realizing a substantid savings in both cost of the system and system size.

While a preferred embodiment of the invention has been described in detail, it shodd be apparent that many modifications and variations thereto are possible, all of which fall within the true spirit and scope of the invention.