GB2184305A - Propeller speed and phase sensor - Google Patents

Abstract

A speed and phase sensor for counterrotating aircraft propellers is described. A toothed wheel is attached to each propeller, and the teeth trigger a sensor as they pass, producing a sequence of signals. From the sequence of signals, the rotational speed of each propeller is computed, based on the time intervals between successive signals. The speed can be computed several times during one revolution, thus giving speed information which is highly up-to-date. Given that the spacing between teeth, or their shape or composition may not be uniform, the signals produced may be nonuniform in time. On the basis of an ideal period, computed from the period for a whole rotation and the number of teeth, error coefficients are derived to correct for nonuniformities in the resulting signals, thus allowing accurate speed to be computed. Phase can be viewed as the relative rotational position of one propeller with respect to the other, but measured at a fixed time. The invention computes phase from the signals. <IMAGE>

Description

SPECIFICATION Propeller speed and phase sensor The invention relates two rpm sensors for aircraft propellers and, more specifically, to a system which meas ures the rpms of both propellers in a counterroatating propeller pair. The system also measures the phase relationship between the propellers.

Background of the invention It is common to measure the rotational speed of an aircraftpropeller3 in Figure 4 by attaching atoothed wheel 6 (called a "targetwheel ") to the propeller shaft. Each tooth (orflag) 9 produces a signal in a magnetic pickup coil 12 as it passes. Additional circuitry (not shown) processes the signals. The circuitry may operate as follows in measuring rpm.

Ifthetoothed wheel 6 has eight teeth, the circuitry measures the frequency at which teeth pass the coil, and from this frequency infers the rotational speed. For example, ifthe frequency is 160 Hz (i.e., 160 teeth per sec3nd), and there are eight teeth per revolution, then the rpm inferred is 1200 rpm (i.e., 1200 = 60 x 160/8).

One disadvantage to this approach is that it only provides an average speed over the time interval taken by several teeth to pass the coil. Accelerations and decelerations of the propeller during the interval are not detected. Also, this approach does not provide information as to the instantaneous positions of the propeller blades. For example, it may be desirable to knowthe precise instant in time when blade No. 1 on the propeller was located at the 1 :30 o'clock position.

Objects ofthe invention It is an object of the present invention to provide a new and improved speed measurement system for aircraft propellers.

It is a further object of the present invention to provide a speed measurement system for a counterrotating pair of aircraft propellers which measures the speeds of each propeller.

It is further object of the present invention to provide real time data as to the instantaneous rotational positions of aircraft propellers.

Brief description ofthe drawing Figure 1 illustrates one form ofthe invention associated with a pairofcounterrotating propellers.

Figure 2 is a schematic of a pair of counterrotating propellers.

Figure 3illustrates two arcs 125 and 135, which are also shown in Figure 2.

Figure 4 illustrates a prior art sensor for propeller speed measurement.

Figures5and 5A illustrate variations in positioning oftheteeth 45 ofthewheel 30 in Figure 1.

Figure 6illustrates a sequence of pulses produced by the counter 57 in Figure 1,togetherwith the strobe pulses identified in Figure 1.

Figures 7and 8aretiming diagrams used to explain the asynchronous operation ofthecounter57 and the microprocessor 66 in Figure 1.

Figures9and ?Oare plots of simulations which comparethe operation of oneform of the inventionwhich uses error coefficients in computing speed, with another form which does not.

Figures Ii and 12showtwoschematicarrangementsoftheteeth45in Figure 1.

Summary ofthe invention In one form of the invention, a clock storesthe real-times atwhich propeller blades cross a reference point.

From these real-times, the rpm of the propeller can be computed nearly instantaneously.

Detailed description of the invention Figure 1 illustrates a pair of aircraft propellers 15 and 18. They rotate in opposite directions as indicated by arrows 21 and 24, and are thus termed counterrotating. Fastened to the propellers are target wheels 27 and 30 which also rotate in opposite directions. Magnetic pickoff coils 33 and 36, known in the art, produce signals (called strobe signals herein) on strobe lines 39 and 42 in response to the passing of the teeth 45. One such coil is Model No.726452 Fan Speed Sensor, available from Electro Corp., located in Sarasota, Florida.

Strobe lines 39 and 42 are connected to the strobe inputs of latches 48 and 51. The strobe signalsthuscause the latches 48 and 51 to load the data present on data bus 54. Data bus 54 carries the output of 16-bit counter 57.16-bit counter 57 counts from the binary number zero to the binary number 216-1 (commonly called 64K, which is decimal 65,535) atthe rate of 2 Mhz, and is used as a clock. That is, the counter 57 changes 2 million times per second, in sequence, from decimal Oto decimal 65,535, and then starts at zero ("rolls over") and continues.

The autputs 60 and 63 of latches 48 and 51 are fed to a microprocessor 66 indicated bythesymbol ,u P.The data bus 54 also feeds the microprocessor 66. Thus, both the latches 48 and 51, as well as the microprocessor 6t-v, have inputs from counter 57, and thus have access to a real-time signal. The microprocessor 66 is pro gramme according to the flowchart described bythe eight steps listed in the following Table 1.Adetailed description of each step follows the listing. The reader is invited to nowjumpto this description, which refers step-by-step to Table 1.

TABLE 1 1. Calculate time it takes forfull revolution.

Time (full revolution) = AT(1 )+ AT(2)+ AT(3)+ AT(4)+ AT(5)+ AT(6)+ AT(7)+ AT(8) 2. Calculate errorcoefficientforthe tooth opposite the currenttooth. m and n are indices.

m = (n+4) modulo 8 error(m) = #T(m) Time(full revolution)/8 3 In an underspeed condition (lessthan 340.9 rpm),the coefficients are reinitializedto one. They will gradually converge to their correctvalues when the underspeed condition terminates. Underspeed exists when a latch does not change for eight consecutive readings. (This is termed the "eight-run rule.") 4. Calculate speed.

speed = ERROR(n) x min)x(2,000,000 counts/sec) #T(n) (60 sec/8teeth/revolution 5. Select good sensors. (Al and A2 refer to two sensors on one propeller. B1 and B2 refer to two sensors on the other.) 5.1 IF absolute value (sensorA1-sensorA2) 40.0 rpm then speed = (sensorA1=sensorA2) and 2 2 reset the flags indicating that both fore sensors are good If NOT, then do this: 5.2 IF sensor B1 and sensor B2 are both good then 5.2.1 IF absolute value (sensorAl - aft speed) < absolute value (sensor A2 - aft speed) then fore speed = Al and then set flags indicating that fore sensorAl is good and fore sensor A2 is bad.

5.2.2 IF NOT then fore speed = sensorA2 and set flags indicating that sensor A1 is bad and sensor A2 is good.

5.3 IF sensor B1 and sensor B2 are not both good then pick lower.

5A Repeat5.1 - 5.3 for other propeller, replacing Al 'by Bl , A2 by B2, B1 byA1, and B2 byA2.

6. Check for sensors not reading when the engine is running.

IF core speed > 10,000 rpm AND ABS (fore pitch - scheduled fore pitch) < 3.0 degrees AND ABS (aft pitch -scheduled aft pitch) < 3.0 degreesTHEN iF fore sensorA < 350 rpm THEN set flag indicating fore sensor A is bad.

IFfore sensor B < 350 rpm THEN set flag indicating fore sensor B is bad.

IF aft sensorA < 350 rpm THEN set flag indicating that aft sensor A is bad.

F aft sensor B < 350 rpm THEN setflag indicating that aft sensor B is bad.

7. Compute phase phase angle - Time front latch -Time rear latch Time front latch - Last time front latch x45 The flowchart is written based on the assumption that each propeller 15 and 18 in Figure 1 has eight blades, and, correspondingly, eight teeth on each target wheel 27 and 30. However, for ease of illustration, each propel ler is shown as having only fou r blades. i Step lisa summation in which the total time for one r volution of a propeller is computed. Thiscomputation is done for each propeller. The computation is executed as follows. As stated above, when a tooth 45 passesthe pickup 36, the signal produced on line 42 causes latch 51 to load the number presently existing on bus 54. in effect, latch 51 is loaded with the exacttime of day at which tooth 45 passed pickup 36.

The factthat counter 57 counts from zero to 64k and then starts over at zero again does not significantly affectthis concept, as will be explained later. Further,the exact definition of what is meant by "passing" the pickup 36 will be explained in connection with Step 2.

The microprocessor 66, on a continuing basis, reads each latch 48 and 51, and places the real time data into a random access memory (RAM) array 70. One subarrayofthe RAM is indicated byfour boxes 73 for blades 1-40n propeller 15, and a similarsubarray 75for propeller 18.

The boxes in subarrays 73 and 75 are, in fact, RAM memory locations. Each box corresponds to a propeller blade. The usual sequence of operation would be: a tooth passes, changing the number in latch 51 .Tne microprocessor 66 reads latch 51 and stores the numberjust read in RAM 77 in subarray 73. A subsequent tooth passes coil 36, again changing the number in latch 51. The microprocessor 66 again reads latch 51 and then stores the number just read in another RAM 79, and so on,therebystoring the real-time occurrences of the strobe signals. This is tantamount to storing the real-times of tooth passings, which istantamountto storing the real-time occurrences when blades cross a predetermined point, such as a point 82.The latters true because the relative geometries of the propeller 15 and the toothed wheel 30 are known in advance, from the construction of the propeller system.

As art example, for the clock rate of2 Mhz described above, for an eight-toothed wheel and for a constant n'ropellerspeed of 1200 rpm, ata given instantthe numbers contained in the RAM for propeller 15 might be '.se, ;us, such as "t=9,000", as shown. The readerwill note that all numbers differ by 12,500, which isthe number of counts occurring during the 0.00625 second interval between tooth crossings.

The microprocessor 66 also stores data in subarray75forthe other propeller 18 in the same manner. The execution speed (say, 1 million assembly code steps per second) of the microprocessor 66 is so much faster than the strobe signals which change the data on the data bus 54 (say, 160 changes persecond foran 8-toothed wheel at 1200 rpm), that no problem exists for the microprocessorto read and store both latches between latch strobing events.

The real-time information on blade crossings, which is stored in RAM 70, allows the microprocessor 66to compute the time intervals (AT's) between successive blade crossings. The interval is the difference between the stored real-timefortwo successive blades, as shown by symbol ATin Figure 1 near boxes 77 and 79. AT is, in this example, 12,500.

,Tj1 w referstothetime interval (AT) between the crossing oftooth No.8 and tooth No.1. AT(2) referstothe time interval between the crossings oftooth No. 1 and tooth No.2, and soon. Thus, Step No. 1 computesthe total time interval for a single revolution of each blade.

Step No.2 computes an error coefficient. One reason for the error coefficient is explained with reference to Figure 5. During the manufacture oftoothed wheels 6 in Figures 1 and 5, it is almost inevitable that a tooth 9A in Figure Swill not be located exactly at its intended position, but (1) may be displaced to phantom position 85(2) may be oversized as shown by dashed lines 88, or (3) may be undersized as shown by dashed line 89. In any ofthe three cases, edges 90 can be displaced from intended positions 93, and by up to 0.1 degrees, indicated by angles D. Thus, the signals produced by pickup 12will in factoccurat different times than if edges 90 were in their intended positions.As a consequence, thetime intervals measured between theteeth bearing the edges 90 [shown as AT(1 ) and AT(2)] will be different than the time intervals measured between teeth 9C and 9D, [i.e., AT(3) and AT(4)], even ifthe toothed wheel 6 is rotating at a constant speed.

Unlesi, corrected, the data in latch 51 in Figure 1 would indicatethatthe wheel 30, and thus the propellerto which it is attached, is undergoing an acceleration followed buy a deceleration because time interval AT(1 ) is less than time interval AT(4).

Further, even ifthe toothed wheel 30 were perfectly manufactured, nonuniformities in the reluctance ofthe wheel material can induce nonuniformities in the the strobe signals. One reason is thatthe coil 12 istriggered by a given reluctance change in region 95. Both the composition of wheel 30, as well as the wheel geometry, are involved in the reluctance change. It is the given change in reluctance to which the coil 12 responds in order to infer a tooth's passing. Step 2 corrects the deviation in composition and geometry with an error coefficient.

As indicated in Step 2, m and n are indices. ("Modulo 8" means that the highest number used is 8, so that if n -- C, rn is not 10, but 2.9 becomes 1,10 becomes 2,11 becomes 3, and 12 becomes 4. A kitchen clockcould reviewed as "modulo 12." The highest number used is 12. Adding 4 hours toll o'clock does not yield 15 o'clock, but 3 o'clock. The "8" in "modulo 8" refers to 8 teeth.) For example, when m = 1,then n = 5, andthus, with 8 teeth, an error coefficient for the tooth opposite the tooth currently loading latches 48 and 51 in Figure I is being computed. This has significance during accelerations and will be discussed in greater detail atthe end of the Detailed Description. The error coefficient is computed by the equation shown in Step 2. The equation has the effect of normalizing the time interval for the opposite tooth with respect to one-eighth of thet.me interval for a full revolution. For example, if the toothed wheel were perfectly manufactured, and of a perfect material, and if the propeller speed were constant, all AT'sin Step 1 would be identical. If time four a frill revolution were 8 units, then each ATwou Id be 1 unit, and the error coefficient in Step 2 would be unity.

However, iftime interval AT(1) in Figure 5A were 3/4 units, and time interval AT(2) were 1-1/4 unit, then the errnr coefficient for tooth 1 would be 3/4 according to the equation in Step 2

(TheAT's are referenced to lines 99 running through the centers of the teeth ratherthan through the edges for ease ofillustration.)The errorcoefficientis a ratio ofthe actual time interval AT(1) in Figure SAtoan idealized time interval AT(lD) at constant speed. AT(lD) would resu It from perfect geometry and perfect composition. AT(ID) is estimated by dividing TlME(full revolution) by 8, as Figure 5A indicates.

The error coefficeints are used in Step 4, but first the microprocessor 66 inquires whether an underspeed conditions exists in Step 3. One such underspeed is engine idle. Another occurs during start-up. If the underspeed condition exists, all error coefficients are re-initialized to unity. One reason for reinitializing the coeffic eintsto unity is that atsuch lowspeed there is no requirementfor high accuracy of propellerspeed measure- ments. Also, start-up seems a logical time to set variables, such as error coefficients, to nominal values, such as unity. Further, the 641(range of counter 57 places a limit on the slowest speed that one can measure.The error coefficients are thus useless at speeds below the limit, because speed isn't computed. This discussion will briefly digress to consider some problems with speed measurement at low speeds, beginning with reference to Figure 6.

Figure 6shows a pulse train 101 produced by counter 57 in Figure 1. The output of counter 57, while actually a constantly changing binary number, can be viewed forthis explanation as equivalenttothe pulse train 101 in Figure 6, with each pulse separated by 1/2,000,000 sec as shown. If the pulse 103 produced by strobe42, corresponding tothe passage of a tooth 45 in Figure 1, isseparatedfrom thefollowing pulse 105 by a distance which is equal to or greaterthan 64k xl /2,000,000 secs, the microprocessor 66 cannot distinguish pulse 105 from a pulse 107 occurring exactly one T,,II earlier. Both pulses 105 and 107 presentthe same real-time data to latches 48 and 51 on bus 54.The speed computed based on pulses 103 and 107 would bethe same asthatcomputed based on pulses 103 and 105, yet pulse 105 represents a slower actual speed.

Another way to state this is that strobe pulses 103 and 105 must be closerthan 64k counter pulses in train 101 in order to correctly compute the speed. In the case of teeth, a 64k counter, and a clock rate of 2 Mhz,the lowest rotational speed measurable is 228.7 rpm, computed asfollows.

Troii is the maximum time interval between two teeth. For an eight-tooth wheel, Troii corresponds to 228.7 rpm

This limitation could be eliminatedby using a counter larger than 16-bits, such as a 32-bit or largercounter which rolis over less often, thus increasing the time interval Troll in Figure 6, but such would impose increased cost, as well as impose possible hardware availability problems.

This limit on speed measurement just discussed assumes that the data in RAM 70 is continuously updated.

However, if the updating is not continuous, but periodic, a different limit is obtained. The different limit results chIefly from thefact thatthe 2Mhz clock running the counter 57 can be asynchronous with respectto the clock running the microprocessor 66, as will now be explained.

There exists a larger control system (the "primary control system," not shown), forthe engine and aircraft, with which the propellers 15 and 18 operate. The reader need not be concerned with the primary control system exceptto knowthat a largercomputer program (the "primary program") forthe primary control system must run, startto finish, every 10 milliseconds (msec).That is, the primary program repeats everylO msec, as shown by arrows such as 150 in Figure7. The arrows 150 indicate the startups ofthe primary program. The 10 msec requirement is imposed by factors unrelated to the present invention.

The program ofTable 1 herein (the "speed program") is run within the primary program every 4,8, and 10 msec during each run ofthe primary program. The speed program can beyiewed as a subroutine ofthe primary program. The runs of the speed program are illustrated as lines 155 in Figure 7. The time required for one run ofthe speed program is short, say, 50 microseconds, a microsecond being 1/1,000,000 of a second.

This time is so much fasterthan the 10 msec (i.e., 10/1,000 second) intervals between the startups 150 ofthe primary program, thatthe running time of the speed program cannot bye drawn to scale on Figure 7. The running time is too short. The running time would occur,for example, in the 50 microsecond interval be tz'een the times 49.975 and 50.025 shown in the Figure. Such a length oftimewould probably be invisibleto the naked eye underthescaleshown.

Therefore, the speed program runs every 4,8, and 10 msec during each run of the primary program. The runs ofthe speed program areso fast that they can be viewed as instantaneous on the scale of Figure 7. They can also be viewed as instantaneous with respectto Traii, which is 32.8 msec. Each run ofthe speed program updates RAM 70 in Figure 7, as explained above. The asynchronous aspect of the counter 57 and the microprocessor 66 will now be considered.

Four Trolls are shown, beginning at0, 4,8, and 10 msec. Counter 57 in Figure 1 can start at zero (i.e., rollerover) at any of these points, or at any point in-between. Thus, in a sense, counter 57 and microprocessor 66 are asynchronous: the startup time 150 forthe primary program does not necessarily coincide with the start of Tro, nor does the startuptime 150 have any fixed, known, relationship with the start of Trolls In this asynchronous situation, the Inventors' analysis has led to this conclusion: subject to an exception identified later, the following relationship between runs 155 ofthe speed program must exist.

The speed program run immediately before a strobe occurs will be called FIRST. The subsequent speed program run 155 after the next strobe is called LAST. That is, the sequence is the following: FIRST occurs, then a strobe occurs, then zero or more intervening speed program runs, then a second strobe occurs, and then LAST occurs. In Figure 1, FIRST can be run 155 at 4 msec, the strobe can occur at point 157, and LAST would therefore be the run at 28 msec.

The Inventors have concluded that both FIRST and LAST must occur within the same Troll in orderto guarantee that the problem discussed in connection with Figure 6will be avoided, Restated, if FIRSTand LAST are not within the same Troll, then it is not certain that the numbers in the latches 48 and 51 provide data from which speed can be accurately computed. Figure 8 illustrates this problem.

Strobes 157A and 157B cause latch 51 in Figure 1 to be loaded with a number, say 3935. Then, in one case, a later strobe 1 57C in Figure 8, more than oneTroii away, causes latch 51 to be loaded with a second number, say 5986. In another case, a strobe 1 57D can load latch 51 with an identical number (5986) because counter 57 rolled over at point 159. Thus, the speed program would see the same number (5986 in both cases), butthis number represents vastly different AT's, as shown in Figure 8. The requirementthat both FIRST and LAST occu rwithin the same Troii eliminates this error which is caused by the different AT's.

One may now inquire as to the slowest propellerspeedwhich can be measured underthecircumstances just described, namely, a Troii of 32.8 msec, an asynchronous repetition of the primary program every 10 msec, and a run ofthe speed program every 4,8, and 10 msecwithin each repetition ofthe primary program.

One answerto the inquiry comes from the shifting of T,,II back and forth between the four positionsT,,acl)- Trol,l4) shown in Figure 7, in search of the position of Troii which gives the smallest number of speed program runs between FIRST and LAST. For example, if the run at 0 msec is considered to be within Troii(1), and this run is FIRST, then LAST occurs at30 msec. The intervening speed program runs are at4, 8,10,14,18,20, 24,and 28 msec, atotal of8 intervening runs. Applying a similar analysis to the test ofthe Trniis, one derives the data in Table 2.

TABLE 2 No. of FIRST LAST Intervening Intervening occurs occurs speed program Speed Pro- Min.

at at runs at gram runs AT 1)0msec 30msec 4,8,10,14,18, 8 24 20,24,28,msec 2)4msec 34msec 8,10,14,18, 8 22 20,24,28,30 msec 3)8msec 40msec 10,14,18,20,24 9 28 28,30,34,38 msec 4)10msec 40msec 14,18,20,24 8 24 28,30,34,38 msec 3) (modified): 8msec 38msec 10,14,18,20,24 8 24 28,30,34 msec Table 2 indicates that the smallest number of intervening runs ofthe speed program is 8, in the far right column.Therefore, ifthe data in latch 51, which is read during a speed program run, changes within eightor fewer runs of the speed program, then it is assumed that FIRST and LAST both occur during the same Troll.If the data in latch 21 remains unchanged for more than eight consecutive speed program runs 155, then it is asscmed that FIRST and LAST occur outside the same Trnii, and, therefore, the two strobes may have occurred outside the same Trol.

The readerwill note that the limit of eight unchanged latch readings has the effect of modifying line 3 in Tabie 2. if the actual Trcii Occurring is Troii(3) in Figure 7, then LAST, in effect, occurs at 38 msec, not 40 msec as in line 3, because a latch change occurring after 38 msec, even though otherwise qualifying as a LAST, under the eight-run rule of Step 3 in Table 1, it is not used. This modification of line 3 is a consequence ofthe asynchronicity. Even though LAST occurs at 40 msec with Troll(3), one does not know thatTrnii3 is actuallythe Troll occuring. Troiiiii could be. Thus, any speed run following eight runs of unchanged latch data is, in effect, ignored.

The minirnum speed which can always be measured underthe eight-run rule is easily computed, once the rule has been derived. This speed is related to the smallest ATthat could occur between two strobes separ ated by eight intervening program runs. This AT is the difference between the first intervening program run and the last,i.e.,28-4'=24 msecforcase 1 in Table 2. From Table 2, the minimum is 22 milliseconds (case2).

Computing in the same manner as in equation 2, for an eight-tooth wheel, the speed is 340.9 rpm.

1000 = 1000 msec/sec x 60 sec/min 8 teeth/revoiution x 22 msec If an underspeed condition does not exist, as determined by the eight-run rule, Step 4then calculates the present speed. As the parenthetical expression shows,the speed is adjusted by ERROR(m)to accommodate any errors in tooth positioning shown in Figures 5 and 6. For example, let it be assumed thatthe full time of one revolution is 160,000 counts (i.e., 1/8 revolution per 20,000 counts), but that the time intervals AT(1 ) and AT(2) in Figure 5A are 15,000 and 25,000 counts, respectively. The error coefficients in Step 2 for teeth 1 and 2 will be 3/4and 1-1/4, respectively.Thus, in Step 4the actual speed computed based on AT(1)will be 314 60 x 2,000,000 750rpm= x 8 That is, even though the time interval actually measured was 15,000 counts instead of 20,000 counts, the error coefficients allowthe actual propeller speed at steady-state to be computed.

During accelerations and decelerations, however, the speed computed in Step 4will be slightly different than the actual speed. The difference will be a function ofthe relative difference between the rate of propeller acceleration and the computational speed of microprocessor 66, or, in simplerterms, of how many times per second Step 4is performed with respect to the rate of acceleration ofthe propellers. The Inventors have performed a simulation in which Step 4was executed at the rate of 300 per second and the propellers were accelerated ata maximum rate of 393 rpm per second. Figure 9 is a plot of measured propellerspeed (CSPD) and measurement error(TERR) both in rpm. The measurement error is small, never exceeding 1 rpm.For comparison, Figure 10 shows the same simulation with all errorcoefficientsfixed at 1 (i.e., omitting calculation Step 2). The errors exceed 10 rpm. This is taken to demonstrate the effectiveness of the error coefficients.

The preceding discussion has assumed that single pickup coils 33 and 36 in Figure 1 are usedforeach toothed wheel 27 and 30. However, it may be desirable to provide second, backup coils 110 and 113,together with backup latches 115. The Inventors here point out that, using the backup sensors 110 and 1 four speeds are now computed. Step (5) is executed for each offour sensors. The sensors (i.e., coils 36 and 113) for propeller 15will be termed sensors Al and A2 in Table 1, and, similarly, for propeller 18, sensors B1 and B2.

Step5 checks the sensors for properfunctioning. The phrase "sensorAl" is an abbreviation for "the speed computed based on sensorAl." 5.1 inquires whether the speeds indicated by both sensors for a given propeller are sufficiently similar; in this case, whether within 40 rpm of each other. If so, the speed is taken as the average of the two speeds and a flag for each sensor is reset indicating that both the sensors are good. A flag can be anytype of memory device, such as a memory location in RAM.

lithe difference in speeds fall outside the 40 rpm range, then Step 5.2 is executed. Step 5.2 first inquires whether both the speeds ofthe other propeller (the aft propeller in this example) are "good" based on Step 5.1: that is, within 40 rpm of each other. Steps 5.2.1 and 5.2.2 state in more detailed form thefollowing inquiry: of sensors A1 and A2 (forfore propeller 1 which deviates more from the speed (e.g., "aft speed") indicated by the other propeller's sensing system? (Aft speed is the speed computed for the aft propeller in Step 5.1.) The sensor with the smallest deviation is taken as the good one.If Step 5.2 indicates that both sensors 81 and B2 are not "good" (that is, the "aft speed" is not a reliable judge), then Step 5.3 is executed.

5.3 asks which sensor is indicating the lower speed? The sensor indicating the lower speed is chosen because the Inventors consider it preferable to overspeed the propellers 15 and 18 in case of sensorfailure ratherthan to underspeed them. Choosing the lower speed sensor causes the propeller speed control equipment (not discussed herein)to believe thatlhe propellers are going slowerthan proper, and the equipmentthustriesto accelerate the propellers, thus overspeeding them.

Step 6 is a double check. A common failure of all four sensors, such as an electrical failure of theexcitation circuit (not shown), can cause Step to set good flags for all four. Step 6 prevents this. The "IF" statement at the beginning has three conditions. (1) Core speed must exceed 10,000 rpm. (Core speed refers to the speed ofthe high speed turbine of a gas turbine engine which may powerthe propellers.) (2) The deviation ofthe actual pitch of propeller 15 from the scheduled pitch must be lessthan 3 and, similarly, (3) the pitch deviation of propeller 18 must be less than 3 . The existence ofthese conditions indicates that the propeller system is operating under power conditions.Under these engine and pitch conditions, it is assumed highly unlikely t'i'at either propellerwould be operating at less than 350 rpm. Therefore, if a reading of 350 rpm or less is obtained, the sensor providing that reading is considered to be faulty and a flag is set accordingly.

Thus far, only speed sensing has been considered. However, in a counterrotating propeller system, sensing of the phase angle between propellers may also be desired. Phase is defined with reference to Figure 2.

Figure 2 schematicaliy shows an end on view of two coaxial propellers. One propeller's blades is indicated by squares 120,the other propeller's blades is indicated by circles 123. Phase angle as defined as the angle 125 between a blade on one propeller and the nearest blade on the other propeller in the clockwise direction, but measured atthe instant when blade 123 is ata predetermined position, such as the 12 o'clock position shown. The actual angle 125 will, of course, be constantly changing because the counterrotating blades are moving toward each other. However, when measured at the predetermined time just described, if the prop ellers are operating at identical, constant speeds, the phase angle will be a measurable constant.

The phase angle, in effect, describes the crossing points in space of the propellers blades. For example, blades 1 23A and 1 20A, iftraveling at identical speeds, will cross approximately at region 130. For acoustical and other reasons, it is sometimes desirable to control this crossing point, as by moving the region 130to region 133 in Figure 2.

The present invention measures phase angle in Step 7. Step 7 is believed to be self-explanatory. In effect, Step 7 is the ratio oftwo time intervals. The intervals can be illustrated by arcs 125 and 135 in Figure 3. Arc 125 represents the length oftimetaken by blade 120A in Figure 2to travel from point 137 to point 139. Similarly, arc 135 (also shown in Figure 2) represents the time interval for blade 1 23A to travel from point 141 to point 144. The ratio ofthetwo arcs (or angles) is the phase.

The reasonthatthis ratio indicates phase angle is that itgives the relative position of blade 120 in Figure2 with respect to blade 1 23A when blade 1 23A is at a predetermined position, such as at the 1 2:00 position shown. When blade 120A is closerto blade 123A (angle 125 is smaller),then the phase in Step 7will be aller. The converse is also true.

The phase angle measured in Step 7 actually calculates what percentage angle 125 is of angle 135 in Figure 3. The largerthe percentage, the closer blade 120A in Figure 2 is to point 144when blade 123A is at the 12:00 position. Therefore, the phase angle indicates the relative position of blade 1 20A when blade 1 23A is at the 12:00 position.

An invention has been described wherein the rotational speed of an aircraft propeller is computed many times per revolution. The invention includes atachometer. Forexample, at 1200 rpm, one revolution takes 50 msec (.050 sec). Underthetiming of Figure 7, 16speed program runs occur between 0 and 50 msec,inclus- ive: the speed is computed 16 times per revolution. The invention can compute this speed in a counterrotating pair of aircraft propellers. Further, the invention also computes phase angle ofthe counterrotating propellers at the same rate of 16 times per revolution. The invention thus provides the aircraft computer and the pilot with performance data which is nearly instantaneous with measurement of the AT's.

It was stated in the Background of the Invention that it may be desirable to know when the blade No. 1 was located at the 1:30 o'clock position. This can be accomplished by adding a counter (not shown) which counts the rollovers of counter 57. A second set of memory locations in addition to RAM 70 can be used to store the data taken from the second counter. The microprocessor 66 would then read the second counter at each reading of latches 48 and 51, and store both countervalues atthe pairofmemory locations forthe tooth 45 in question.For example, a reading of 5 on the rollovercounterwhen a latch holds a value of 12,000 would indicate that the tooth causing the latch to load 12,000 crossed the coil 36 at real-time of 164,006 msec (164.006 = 5 x 32.8 msec + 12,000/2,000,000 x 32.8 msec).

An invention has been described which measures AT's by using a magnetic pickup coil 36 in Figure 1. An alternate form would use an optical pickup, known in the art, to sense the flag passings.

One important aspect ofthe invention can be explained with reference to Figures 5 and 5A. As stated above, at constant speed, a deviation of a tooth from its intended position will cause the measured ATto deviate from the idealized AT. This deviation is used to compute an error coefficient in Step 2 above. Then, latei when a time interval is again measured based on the deviant tooth, the actual speed can be computed from both the measured (i.e., not idealized) time interval and the error coefficient. In a sense, the idealized AT is regenerated from the measured AT.

The invention Operates as in the following example. Let it be assumed thatthe squares 170 in Figure l2are, in fact,theteeth (i.e., flags) 45 in Figure 1.Let it be further assumed that the angle between all neighboring flag pairs is45 degrees, butthatflag 170B is displaced such thatthe angle between flags 170B and C is20 degrees, while the angle between flags 170A and B is 70 degrees. The total angle between flag 170A and Cis thus 90 degrees. Let itfurther be assumed that the time of one revolution at constant speed is eight seconds.

With these assumptions, AT(ID) is one second and the measured AT's will each be one second, with two exceptions. Flags 170B and C have a AT of 20/45 x 1 second, while flags 170A and B have a ATof70/45 x 1 second. The error coefficients compare the measures AT's with the idealized AT. The error coefficients allow one to derive the actual angles between the flags, in at leasttwo ways: (1) the error coefficient for the 20 degree angle is 20/45. It is known that the idealized AT represents an angle of 45 degrees, and thus the actual angle isthe errar coefficienttimes the idealized angle: 20/45 x 45 = 20. (2) the entire set of flags describe a circle, vjl'ich contains 360 degrees. The ATforthe 20 degree interval was 20/45 sec.Since the entire circle represents eight seconds, 20/45 x 360/8, or 20 degrees is obtained as the angle.

Knowledge ofthe angle betweenflagsallows one to compute the speed based on the AT's. In the example above, the AT between flags 1 70B and C of 20/45 sec allows the speed to be computed: 20/45 sec for 20 dtgrees of travel corresponds to 1/45for one degree, or360/45 (namely, eightseconds),forthe entire setof flags over 360 degrees, consistent with the assumed speed.

Therefore, once the error coefficients are established, the actual angular positions of the flags becomes ",nown. Then, from a single AT, the speed can be inferred. This is true, in principle, even with a grossly : < s-ed distribution of flags as shown in Figure 12. Having the error coefficients, which contain data as to the angularseparation oftheflags, one can compute the angularspeed ofthe propeller based solely on the AT shown. Afull revolution is not needed, and, in fact, the speed can be computed several times during one revolution, providing highly up-to-date information.

The invention can be viewed as developing and storing a collection of data, including the AT's and the time elapsed for a total revolution, from which the flag positions can be computed. A model of the target wheel is generated in RAM, so to speak.

This discussion has considered onlythe effect of the positions of flags upon the AT's. However, as stated earlier in connection with Figures 5 and 5A, not only position, but also the geometry and composition ofthe flag are involved in the AT's. Thus, the AT's which the invention measures do not necessarily have a clear relationship with the teeth or the edges 90, as shown in Figure 5. However, the AT's do, in fact, have a consistent relationship with the edges 90. For example, AT(l) in Figure 5 may be measured for teeth 9A and B. AT(1) does not end with edge 93, buton phantom edge 90. This causes no problem because this ATwill, in general,always end on phantom edge 90.Thus, the AT's do not establish the actual geometricangular spacing between adjacent teeth, but instead, establish what will be called the angles between the "effective" locations of adjacent teeth. Phantom edge 90 is one such "effective" location.

It was stated earlier in connection with the explanation of Step 2 thatthe error coefficients are computed for the blade opposite to the one which just loaded a latch. One reason for this will be explained by an example.

Assume thatthe last tooth to pass was tooth No.8, and AT(8) has just been computed. Therefore, n = 8.

Also assume a deceleration is occuring, indicated by continually increasing AT's, as shown in Table 3.

TABLE 3 aT(1)=12,000 1 = (n+1) modulo8 AT(2)= 12,200 2=(n+2)modulo8 AT(3) = 12,300 3 = (n+3) modulo 8 AT(4) = 12,400 4 = (n+4) modulo 8 AT(53 = 12,500 5 = (n+5) modulo 8 AT(6) = 12,600 6 = (n+ 6) modulo8 AT(7) = 12,700 7 = (n+7) modulo 8 AT(8) = 12,800 8 = n Therefore, in this example, Time(full revolution)/8 = 99,600/8 = 12,450 This is the estimated AT(lD).

As stated above, the error coefficients are, in effect, a ratio of actual ATto AT(ID). It was also stated that AT(ID) is estimated from eight consecutive AT's. Further, speed is computed, in Step 4, from individual AT's, perhaps several times per revolution. The Inventors have found that, during a constant acceleration orcon stantdeceleration, adding 4 or5to the index in Step 2will give a more accurate speed computation from a single AT, by giving a truer erro r coefficient throug h a better estimated AT(ID). This is shown in Table 3, in the rig ht column.

Numerous substitutions and modifications can be undertaken without departing from the true spirit and scope of the present invention. What is desired to be secured by Letters Patent isthe Invention as defined in the following claims.

Claims (6)

1. Asystemformeasuring the rotational speed of an aircraft propeller having atargetwheel bearing several flags which pass a reference point during rotation, comprising: (a) meansformeasuring time intervals between successive passings ofthe reference points bytheflags; (b) means for deriving the angular spacing between adjacentflagsfrom the time intervals; and (c) meansforcomputing the speed based on onetime interval.

2. Asystem for computing rotational speed of an aircraft propeller, comprising: (a) a targetwheel connected to the propeller and bearing several flags; (b) a sensorwhich produces strobe signals in response to the passage of flags; and (c) error correction means for modifying the measured time interval between strobe signals, wherein the effects of nonuniformities in placement, composition, or geometry, or any combination ofthese, in the flags are reduced.

3. In atachometerfor measuring the rotational speed of an aircraft propeller and having a targetwheel (6) bearing flags (9) which trigger strobe signals in a sensor (12), a system for compensating for nununiformities in the timing of the strobe signals, comprising: (a) means for measuring the time intervals between strobe signals; (b) means for developing an idealized time interval; (c) means for developing error coefficients by comparing the actual time intervals with the idealized time interval; and (d) means for computing the rotational speed ofthe propeller based on one time interval and one ofthe error coefficients.

4. A system for measuring the rotational speeds of an counterrotating aircraft propellers, comprising: (a) first and second targetwheels fastened to respective first and second propellers; each targetwheel bearing a numberofflags; (b) first and second sensors for producing first and second strobe signals in response to passage offlags by the sensors; (c) clock means for providing a time signal; (d) first and second latch means for receiving the time signal and storing the time signal in response to the respective strobe signals; (e) memory meansforstoring data obtained from the latch means and from which the time intervals between consecutive flags on one of the target wheels can be computed;; (f) computation means for (i) computing the total time elapsedforthe passage oftwo or more flags; (ii) computing error coefficients for modifying the time intervals of (e) to compensate for irregularities in the distribution oftheflags on awheel; (iii) computing the nearly instantaneous speed ofthe propellers based on the time intervals of (e) and the error coefficients of (f)(ii).

5. A system for measuring the rotational speeds of the propellers in a pair of counterrotating propellers, comprising: (a) a firsttargetwheel connected tothefirst propellerand bearing a firstset of flags, and a secondtarget wheel connected to the second propeller and bearing a second set of flags, the distribution of each set of flags being not necessarily uniform; (b) afirstsensorfor producing a first strobe signal in responseto the passage of flags in thefirstsetofflags, and a second sensor for producing a second strobe signal in response to the passage of a flag in the second set of flags; (c) a clockfor providing atimesignal;; (d) a first latch means for storing the time signal existing at the occurrence of one ofthefirst strobe signals, and a second latch meansforstoring thetimesignal existing atthe occurrence of one ofthe second strobe signals; (e) processor means for reading the latches of(d) and storing the latched time signals; in a memoryarray; (f) computation means for (i) computing the total elapsed timeforthe passage of a selected numberofflags on each targetwheel; (ii) computing errorcoefficientswhich indicate nonuniformities, if any, intheflag distribution; and (iii) computing the speeds ofthe propellers based on the stored signals of (d) and the error coefficients of (f)(ii).

6. A system for measuring the rotational speed of an aircraft propeller substantially as hereinafter descri- bed with reference to and as illustrated in the drawings.