Note: Descriptions are shown in the official language in which they were submitted.
13DV 8584
~ . ., _.
PRO~ELLER SPEED AND-PHASE SENSOR
The inven~ion relates to rpm sensors for
aircraft propellers and, more specifically, to a system
which measures the rpms of both propellers in a
counterrotating propeller pair. The system also
measures the phase relationship between the propellers.
BACKGROUND OF THE INVENTION
It is common to measure the rotational speed
o an aircraft propeller 3 in Figure 4 by attaching a
toothed wheel 6 (calIed a "target wheel") to the
propeller shaft. Each tooth (or flag) 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.
If the toothed 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, if the frequency is 160 Hz (i.e.,
160 teeth per second), and there are eight *eeth per
revolution, then the rpm inferred is 1200 rpm (i.e.,
lZ00 = 60 x 160/8).
One disadvantage to this approach is that it
only provides an average speed over the time interval
.
- - ,
' . ~ -
.~
~58~ 13DV-8584
--2--
taken by several teeth to pass the coil. Accelerations
and ~ecelerations 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
know the precise instant in time when blade No. 1 on
the propeller was located at the 1:30 o'clock position.
OBJECTS OF T~E INVENTION
It is an object of the present invention to
provide a new and improved speed measurement system or
aircra~t 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 OF THE DRAWING
FIGURE 1 illustrates one ~orm o the invention
associated with a pair of counterrotating propellers.
FIGURE 2 is a schematic of a pair of
counterrotating propellers.
FIGURE 3 illustrates two arcs 125 and 135,
which are also shown in Figure 2.
~ IGURE ~ illustrates a prior art sensor for
propeller speed measurement.
FIGURES 5 and SA illustrate variations in
positioning of the teeth 45 of the wheel 30 in Figure lo
30 FIGURE 6 illustrates a sequence of pulses
produced by the counter 57 in Figure 19 together with
the strobe pulses identified in Figure 1.
FIGURES 7 and 8 are timing diagrams used to
exp~ain the asynchronous operation of the counter 57
and the microprocessor 66 in Figure 1.
FIGURES 9 and 10 are plots of simulations~
which compare the operation-of one form of the
.
': .
125891~ 13DV-8584
--3--
invention which uses error coefficients in computing
speed, with another form which does not.
FIGUR~S 11 and 12 show two schematic
arrangements of the teeth 45 in ~igure 1.
SU~MARY OF THE INVENTION
In one form of the invention, a clock stores
the real-times at which propeller blades cross a
reference point. From these real-times, the rpm of the
propeller can be computed nearly instantaneously.
DETAILED DESCRIPTION OF TH~ 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 signals
thus cause the latches 48 and 51 to load the data
~resent 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) at the
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 0 to decimal 65,535, and then
starts at zero ("rolls over") and continues.
The outputs 60 and 63 of latches 48 and 51 are
fed to a microprocessor 66 indicated by the
symbol ~qP. The data bus 54 also feeds the
microprocessor 66. Thus, both the latches 48 and 51,
~, . .
-~; "
...
.1~5891~ 13DV-8584
--4--
as well as the microprocessor 66, have inputs from
counter 57, and thus have access to a real-time
signal. The microprocessor 66 is programmed according
to the flowchart described by ~he eight steps listed in
the following Table 1. A detailed description of each
step follows the listing. The reader is invited to now
jump to this description, which refers step-by-step to
Table 1.
TABLE 1
1. Calculate time it takes for full revolution.
Time (full revolution) = ~T(l)+ ~T(2)~ dT(3)+ aT(4)~
AT(5)+ ~T(6)+ ~T(7)~ ~T(8)
2. Calculate error coefficient for the tooth opposite
the current tooth. m and n are indices.
m = tn~4) modulo 8
error(m) = ~T(m)
Time (full revolution)/8
3. In an underspeed condition (less than 340.9 rpm), the
coefficients are reinitialized to one. They will
gradually converge to their correct values 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.
25 speed = ERROR(n) x (60 sec/min) x (2,000,000 counts/sec)
aT(n) 8 teeth/revolution
,. . .
. .
., : .
: ~ . . .", ~ :
`:
12S8914 13DV-858~
--5--
5, Selec~ good sensors. (Al and A2 refer to two
sensors on one propeller, Bl and B2 refer to two
sensors on the other.)
5.1 IF absolute value (sensor Al - sensor A2)
40.0 rpm then
speed = (sensor Al ~ sensor A2)
2 and
reset the flags indicating tha~ bo~h fore
sensors are good
IF NOT, then do this:
5.2 IF sensor Bl and sensor B2 are both good
then
5,2,1 IF absolute value (sensor Al - aft
speed)~absolute value tsensor A2 - aft
speed) then fore speed = Al and then
set flags indicating that fore sensor
Al is good and fore sensor A2 is bad,
5,2.2 IF NOT then fore speed = sensor A2 and
se~ flags indicating that sensor Al is
bad and sensor A2 is good,
5.3 IF sensor Bl and sensor B2 are not both good
then pick lower.
5.4 Repeat S.l - 5.3 for other propeller,
replacing Al by Bl, A2 by B2, B1 by Al, and B2
by A2.
,
. . .
:- .
--
,
5 ~9 ~ 4 13DV-8584
--6--
6. Check for sensors not reading when the engine is
running.
IF core speed ~lO,000 rpm AND ABS (fore pitch
- scheduled fore pitch) ~3.0 degrees AND ABS
(aft pitch - scheduled aft pitch) ~3.0 degrees
THEN
IF fore sensor A ~350 rpm THEN
set flag indicating fore sensor A is
bad.
IF fore sensor B C350 rpm THEN
set flag indicating fore sensor B is
bad.
IF aft sensor A ~350 rpm THEN
set flag indicating that aft sensor A
is bad.
IF aft sensor B ~350 rpm THEN
set flag 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 x 45
The flowchart is written based on the
assumption that each propeller 15 and 18 in Figure l
has eight blades, and, correspondingly, eight teeth on
each target wheel 27 and 30. However, for ease of
illustration, each propeller is shown as having only
four blades.
,...................................... .
- . .
.
7 13DV-858~
Step 1 is a summation in which the ~otal time
for one revolution of a propeller is computed. This
computation is done for each propeller. The
computation is executed as follows. As stated above,
when a tooth 45 passes the pickup 36, the signal
produced on line 42 causes latch 51 to load the number
presently existing on bus 5~. In effect, latch 51 is
loaded with the exact time of day at which tooth ~5
passed pickup 36.
The fact that counter 57 counts from zero to
64k and then starts over at zero again does not
significan~ly affect this 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 ~. -
The microprocessor 66, on a continuing basis,reads each latch ~8 and 51, and places the real time
data into a random access memory (RAM) array 70. One
subarray of the RAM is indicated by four boxes 73 for
blades 1-4 on propeller 15, and a similar subarray 75
for 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.
The microprocessor 66 reads latch 51 and stores the
number just 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, thereby storing the
real-time occurrences o~ the strobe signals. This is
tantamount to storing the real-times of tooth passings,
which is tantamount to storing the real--time
occurrences when blades cross a predetermined point,
.
~5~ 13DV-8584
--8--
such as a point 82. The latter is 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 an example, for the clock rate of 2 Mhz
described above, for an eight-toothed wheel and for a
constan-t propeller speed of 1200 rpm, at a given
instant the numbers contained in the RAM for propeller
15 might be those, such as "t=9,000", as shown. The
reader will note that all numbers differ by 12,500,
which is the number of counts occurring during the
0.00625 second interval between tooth crossings.
The microprocessor 66 also stores data in
subarray 75 for the 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 per second
for an 8-toothed wheel at 1200 rpm), that no problem
exists for the microprocessor to 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 66
to compute the time intervals ( ~T's) between
successive blade crossings. The interval is the
difference between the stored real-time Eor two
successive blades, as shown by symbol ~T in Figure 1
near boxes 77 and 79. aT is, in this example, lZ,500.
~T~1) refers to the time interval (~T) between
the crossing of tooth No. 8 and tooth No. 1. ~T(2)
refers to the time interval between the crossings of
tooth No. 1 and tooth No. 2, and so on. Thus, Step No.
1 computes the total time interval for a single
revolution of each blade~
~5~ 13DV-8584
g
Step No. 2 computes an error coefficient. One
reason for the error coefficient is explained with
reference to Figure 5. During the manufacture of
toothed wheels 6 in Figures 1 and 5, it is almost
inevitable that a tooth 9~ in Figure 5 will 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 of the
three cases, edges 90 can be displaced from intended
positions 93, and by up to 0.1 degrees, indicated by
angles ~. Thus, the signals produced by pickup 12 will
in fact occur at different times than if edges 90 were
in their intended positions. As a consequence, the
time intervals measured between the leeth bearing the
edges 90 [shown as ~T(l) and ~T(2)1 will be dif~erent
than the time intervals measured between teeth 9C and
9D, [i.e., ~T(3) and ~T(~)~, even if the toothed wheel
6 is rotating at a constant speed.
Unless corrected, the data in latch 51 in
Figure 1 would indicate that the wheel 30, and thus the
propeller to which it is attached, is undergoing an
acceleration followed by a deceleration because time
interval ~T(l) is less than time interval ~(4).
Further, even if the toothed wheel 30 were
perfectly manufactured, nonuniformities in the
reluctance of the wheel material can induce
nonuniformities in the the strobe signals. One reason
is that the coil 12 is triggered 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.
.
,.
.
~5~ 13DV-8584
-10-
As indicated in Step 2, m and n are indices.
("Modulo 8" means that the highest number used is 8, so
that if n = 6, m is not 10, but 2. 9 becomes 1, 10
becomes 2, 11 becomes 3, and 12 becomes 4. A kitchen
clock could be viewed as "modulo 12." The highest
number used is 12. Adding 4 hours to 11 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, and thus, with 8 teeth, an error
coefficient for the tooth opposite the tooth currently
loading latches 48 and 51 in Figure 1 is being
computed. This ha; significance during accelerations
and will be discussed in greater detail at the 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 the time
interval for a full revolution For example, if the
toothed wheel were perfectly manufactured, and o a
perfect material, and if the propeller speed were
constant, all ~T's in Step 1 would be identical. If
the time for a full revolution were 8 units, then
each ~T would be 1 unit, and the error coefficient in
Step 2 would be unity.
~lowever, if time interval ~T(l) in Figure 5A
were 3/4 units, and time interval ~T(2) were 1-1/4
unit, then the error coefficient for tooth 1 would be
3/4 according to the equation in Step 2 (3/4 = ~ .
(The ~T's are referenced to lines 99 running through
the centers of the teeth rather than through the edges
for ease of illustration.) The error coefficient is a
ratio of the actual time interval ~T(l) in Figure 5A to
an ide~lized time interval T(ID) at constant speed.
~T(ID) would result from ~erfect geometry and perfect
composition. ~T(ID) is estimated by dividing TIME~full
revolution) by 8, as Figure 5A indicates.
. .
~ 13DV-8584
The error coef~icients are used in Step 4, but
first the microprocessor 66 inquires whether an
underspeed condition exists in Step 3. One such
undersp~ed 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 coefficients to unity is
that at such low speed there is no requirement for high
accuracy of propeller speed measurements. Also,
start-up seems a logical time to set variables9 such as
error coefficients, to nominal values3 such as unity.
Further, the 64k 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
reerence to Figure 6.
Figure 6 shows 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 for this explanation as equivalent to the
pulse train 101 in Figure 6, with each pulse separated
by 1/2,000,000 sec as shown. If the pulse 103 produced
by strobe 42, corresponding to the passage o~ a tooth
45 in Figure 1, is separated from the following pulse
105 by a distance which is equal to or greater than 64k
x 1/2,000,000 secs, the microprocessor 66 cannot
distinguish pulse 105 from a pulse 107 occurring
exactly one Troll earlier Both pulses 105 and 107
present the same real-time data to latches 48 and 51 on
bus 54. The speed computed based on pulses 103 and 107
would be the same as that computed based on pulses 103
and 105, yet pulse 105 represents a slower actual speed.
5 ~9 ~ ~ 13DV-858
-12-
Another way to state this is that strobe
pulses 103 and 105 must be closer than 64k counter
pulses in train 101 in order to correctly compute the
speed. In the case of 8 teeth, a 64k counter, and a
clock rate of 2 Mh~, the lowest rotational speed
measurable is 228.7 rpm, computed as follows.
ro~ in Fig, 6 - _ ' ro~c~ 0~0328 rollover~ (1
~ sec
T is t~e ma~ mu~ timeinter~albe~weentwoteeth. For
ro~
an eight-tooth wheel, Tro~ correspondsto 228.7 rp~
1 rollo~er 1 revolution sec (2)
0 0328 second x ~8 rol`lover x 60 ~in
This limitation could be eliminated by using a counter
larger than 16-bits, such as a 32-bit or larger counter
which rolls over less often, thus increasing the time
interval Trol1 in Figure 6, but such would impose
lncreased cost, as well as impose possible hardware
availability problems.
This limit on speed measurement just dlscussed
assumes that the data in R~M 70 is continuously
updated. However, if the updating is not continuous,
but periodic, a different limit is obtained, The
different limit results chiefly from the fact that the
2Mhz clock running tlle counter 57 can be asynchronous
with respect to the clock running the microprocessor
66, as will now be explained.
There exists a larger control sys~em (the
"primary control system," not shown), for ~he engine
and aircraft, with which the propellers 15 and 18
operate. The reader need not be concerned with the
primary control system except to know that a larger
- 30 computer program (the "primary program"~ for the
.. ..
- , , : . : - : . .
,.
..
: ' ~
~S~9~ 13DV-8584
-13-
p~imary control sys~em must run, start to finish, every
10 milliseconds (msec). That is, the primary program
repeats every l~ msec, as shown by arrows such as 150
in Figure 7. The arrows 150 indicate the startups of
the primary program. The 10 msec requirement is
imposed by factors unrelated to the present invention.
The program of Table 1 herein (the "speed
program") is run within the primary program every 4, 8,
and 10 msec during each run of the primary program.
The speed program can be viewed as a subroutine of the
primary program. Tne runs of the speed program are
illustrated as lines 155 in Figure 7. The time
required for one run o~ the speed program is short,
say, 50 microseconds, a microsecond being 1/1,000,000
of a second. This time is so much faster than the 10
msec (i.e., 10/1,000 second~ intervals between the
startups 150 of the primary program, that the running
time of the speed program cannot be drawn to scale on
Figure 7. The running time is too short. The running
time would occur, or example, in the 50 microsecond
interval between the times 49.975 and 50.025 shown in
the Figure. Such a length of time would probably be
invisible to the naked eye under the scale shown.
Therefore, the speed program runs every 4, 8,
and 10 msec during each run of the primary program.
The runs of the speed program are so fast that they can
be viewed as instantaneous on the scale of Figure 7.
They can also be viewed as instantaneous with respect
to Troll, which is 32.8 msec. Each run of the 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 at 0, 4, 8,
and 10 msec. Counter 57 in Figure 1 can start a~ zero
(i.e., rollover) at any of these points, or at any
point in-between. Thus, in a sense, counter 57 and
-- .
- --
- ~
4 13DV-8584
-l4-
microprocessor 66 are asynchronous: the startup time
150 for the primary program does not necessarily
coincide with the start of Troll~ nor does the
startup time 150 have any fixed, known, relationship
with the start of Troll~ In this asynchronous
situation, the Inventors' analysis has led to this
conclusion: subject to an exception identified later,
the following relationship between runs 155 of the
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 order to
guarantee that the problem discussed in connection with
Figure 6 will be avoided. Restated, if FIRST and 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 s~robe 157C in Figure 8, more than one
Troll away, causes latch 51 to be loaded with a
second number, say 5986. In another case, a strobe
157D can load latch 51 with an identical number (5986)
because counter 57 rolled over at point 15~. Thus, the
speed program would see the same number (5986 in both
cases), but this number represents vas~ly different
.
~ 25 ~ 13DV-8584
-15-
~T's, as shown in Figure 8. The re~uirement that both
FIRST and LAST occur within the same Troll eliminates
this error which is caused by the different ~T's.
One may now inquire as to the slowest
propeller speed which can be measured under the
circumstances just described, namely, a Troll of 3Z.8
msec, an asynchronous repetitiorl o the primary program
every 10 msec, and a run of the speed program every 4,
8, and 10 msec within each repetition of the primary
program.
One answer to the inquiry comes from the
shifting of TrOll oack and forth between the four
P Troll(l)~TrOll(4) sho~n in Figure 7 in
search of the position of Troll which gives the
smallest number of speed program runs between FIRST and
LAST. For example, if the run at 0 msec is considered
in Troll(l)~ and this run is FIRST, then
LAST occurs at 30 msec. The intervening speed program
runs are at 4, 8, 10, 14, 18, 20, 24, and 28 msec, a
total of 8 intervening runs. Appl~ing a similar
anslysis to the rest of the Trolls~ one derives the
data in Table Z
~5~4 13DV-8584
-16-
TABLE 2
No. of
FIRST LAST Intervening Intervening
occurs occurs speed program Speed ~ro- Min.
at at runs at gram runs aT
1)0 msec30 msec 4,8,10,14,18, 8 24
20,24,Z8 msec
2~4 msec34 msec 8,10,14,18, 8 22
20~24,28,30 msec
3)8 msec40 msec 10,14,18,20,24 9 28
28,30,34,3$ msec
4) 10 msec 40 msec 14,18,20,24 8 ~24
~8,30,34,38 msec
3) (modified):
8 msec 38 msec 10,14,18,20,24 8 24
lS 28,30,34 msec
Table 2 indicates that the smallest number of
intervening runs of the speed program is 8, in the far
right column. ThereEore, if the data in latch 51, which
is read during a speed program run, changes within eight
or Eewer 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
assumed that FIRST and LAST occur outside the same
Troll~ and, therefore, the two strobes may have occurred
outside the same Troll~
The reader will note that the limit of eight
unchanged latch readings has the effect of modifying line
3 in Table 2. If the actual Troll occurring is
Troll(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
,
; .~
~5~ 13DV-85~4
-17-
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 of the asynchronicity. Even though LAST
occurs at 40 msec Witil Troll(3~, one does not know that
Troll(3) is actu~lly the Troll occurring. Troll(l)
could be. Thus, any speed run following eigh~ runs of
unchanged latch data is, in effect7 ignored.
The minimum speed which can always be measured
under the eight-run rule is easily computed, once the rule
has been derived. This speed is related to the
smallest a T tha~ could occur between two strobes
separated by eight intervening program runs. This ~T is
the di~ference between the first intervening program run
and the last, i.e., 28 - 4 = 24 msec for case 1 in Table
2. From Table 2, the minimum is 22 milliseconds (case
2). Computing in the same manner as in equation 2, for an
eight-tooth wheel, the speed is 340.9 rpm.
340 9 ~ 1000 msec/sec x 60 sectmin
8 teeth/revolution x 22 msec
~f an underspeed condition does not exist, as
determined by the eight-run rule, Step 4 then calculates
the present speed. As the parenthetical expression shows,
the speed is adjusted by ERRORtm) to accommodate any
errors in tooth positioning shown in Figures 5 and 6. ~or
example, let it be assumed that the full time of one
revolution is 160,000 counts (i.e., 1/8 revolution per
20,000 counts), but that the time intervals aT(l) and
~T(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 3/4 and 1-1/4, respectively. Thus, in
- 30 Step 4 the actual speed compute~ based on ~T(l) will be
750 rpm = 314 x 60 x 27000,000
15,000 8
.
, .
13D~-8584
~S~
-18-
That is, even though the time interval actually measured
was 15,000 counts instead of 20,000 counts, the error
coefficients allow the actual propeller speed ~t
steady-state to be computed.
During acceleratiolls and decelerations, however,
the speed computed in Step 4 will be slightly different
than the actual speed. The difference will be a function
of the relative difference between the rate of propeller
acceleration and the computational speed of microprocessor
66, or, in simpler terms, of how many times per second
Step 4 is performed with respect to the rate of
acceleration of the propellers. The Inventors have
performed a simulation in which Step 4 was executed at ~he
rate of 300 per second and the propellers were accelerated
at a maximum rate of 393 rpm per second. Figure 9 is a
plot of measured propeller speed (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 error coefficients fixed 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 used for each
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 113, four speeds are now
computed: Step (5) is executed for each of four sensors.
The sensors (i.e., coils 36 and 113) for propeller 15 will
be termed sensors Al and A2 in Table 1, and, similarly,
for propeller 18, sensors Bl and B2. Step 5 checks the
sensors for proper functioning. The phrase "sensor Al'~ is
an abbreviation for "the speed computed based on sensor
Al." 5.1 inquires whether the speeds indicated by both
.
.
, ~ ' .
.
~8~ 13DV-85~4
-19-
sensors for a given propeller are sufficiently similar; in
this case, whether within ~0 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 any type of memory
device, such as a memory location in RAM.
If the difference in speeds fall outside the ~0
rpm range, then Step 5.2 is executed. Step 5.2 first
inquires whether both the speeds of the 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 the following
inquiry: of sensors Al and A2 (for fore propeller 15),
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 Bl
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 preerable to overspeed the propellers 15 and
18 in case of sensor failure rather than to underspeed
them. Choosing the lower speed sensor causes the
propeller speed control equipment (not discussed herein)
to believe that the propellers are going slower than
proper, and the equipment thus tries to 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 the
excitation circuit (not shown), can cause Step S to set
good flags for all four. Step 6 prevents this. The "IF"
statement at the beginning has three conditions. (l) Core
speed must exceed lO,000 rpm. (Core speed refers to the
,~
, . ..
.,, ~ . . ..
.
;. , ::
- , ,
~ 13DV-8584
-ZO-
speed of the high speed turbine o~ a gas turbine engine
which may power the propellers.) (2) The deviation of the
actual pitch of propeller lS from the scheduled pitch must
be less than 3 and, similarly, (3) the pitch deviation of
propeller 18 must be less than 3. The existence of these
conditions indicates that the propeller system is
operating under power conditions. Under these engine and
pitch conditions, it is assumed highly unlikely that
either propeller would 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 ~s 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
~igure 2. Figure 2 schematically 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 at the instant when blade lZ3 is at a
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. ~lowever, when measured at the
predetermined time just described, if the propellers 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 123A and 120A, if traveling at identical
speeds, will cross approximately at region 130. For
acoustical and other reasons, it is sometimes desirable to
. , .
:
,.
13 DV - 8 5 8
-21 -
control this crossing point, as by moving the region 130
to region 133 in Figure 2.
The present invention measures phase angle in
Step 7. Step 7 ls believed to be self-explanatory. In
ef~ect, Step 7 is the ratio of two time intervals. The
intervals can be illustrated by arcs 125 and 135 in Figure
3. Arc 125 represents the length of time taken by blade
120A in Figure 2 to travel from point 137 to point 139.
Similarly, arc 135 (also shown in Figure 2) represents the
10 time interval for blade 123A to travel from point 141 to
point 144. The ratio of the two arcs (or angles) is the
phase.
The reason that this ratio indicates phase angle
is that it gives the relative position o~ blade 120 in
15 Figure 2 with respect to blade 123A when blade 123A is at
a predetermined position, such as at the 12:00 position
shown. When blade 120A is closer to blade 123A (angle 125
is smaller), then the phase in Step 7 will be smaller.
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 larger the percentage, the closer blade
120A in Figure 2 is to point 144 when blade 123A is at the
12:00 position. Therefore, the phase angle indicates the
25 relative position of blade 120A when blade 123A 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 a
30 tachometer. For example 9 at 1200 rpm, one revolution
takes 50 msec (.050 sec). Under the timing of Figure 7 a
16 speed program runs occur be~ween 0 and 50 msec,
inclusive: the speed is computed 16 times per
revolution. The invention can compute this speed in a
counterrotating pair of aircraft propellers. Further, the
.. . .
13DV-85~4
-2~-
invention also computes phase angle of the 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 4T'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 coun~er values at the
pair of memory locations for the tooth 45 in question.
For example, a reading of 5 on the rollover counter when 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
20 12,000/2,000,000 x 32.8 msec).
An invention llas been described which measures
~T's by using a magnetic pickup coil 36 in Figure 1. An
alternate form would use an optical pickup, known in the
~rt, to sense the flag passings.
One important aspect of the invention can be
explained with reference to Figures 5 and 5A. As stated
above, at constant speed, a deYiation of a ~ooth from its
intended position will cause the measured ~T to deviate
from the idealized ~T. This deviation is used to compute
an error coefficient in Step 2 above. Then, later, 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 dT is regenerated
from the measured ~T.
:,, .
~S~9~ 13DV-8584
-23-
The invention operates as in the following
example. Let it be assumed that the squares 170 in Figure
12 are, in ~act, the teeth ~i.e., ~lags) 45 in Figure 1.
Let it be further assumed that the angle between all
neighboring flag pairs is 45 degrees, but that 1ag 170B
is displaced such that the angle between flags 170B and C
is 20 degrees, while the angle between flags 170A and B is
70 degrees. The total angle between flag 170A and C is
thus 90 degrees. Let it further be assumed that the time
of one revolution at constant speed is eight seconds.
With these assumptions, ~T(ID) is one second and the
measured aT's will each be one second, with two
exceptions. ~lags 170B and C have a ~T of 20/45 x 1
second, while flags 170A and B have a ~T of 70/45 x 1
second. The error coefficients compare the measured ~T's
with the idealized dT. The error coefficients allow one
to derive the actual angles between the flags, in at least
two ways: tl) 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 is the error coefficient times the ideali~ed angle:
20/45 x 45 ~ 20. ~2) the 0ntire set of flags describe a
circle, which contains 360 degrees. The ~T for the 20
degree interval ~as 20/45 sec. Since the entire circle
represents eight seconds, 20/45 x 360/8 9 or 20 degrees is
obtained as the angle.
Knowledge of the angle between flags allows one
to compute the speed based on the aT's. In the example
above, the aT between flags 170B and C of 20/45 sec allows
the speed to be computed: 20/45 sec for 20 degrees of
travel corresponds to 1/45 sec for one degree, or 360/45
(namely, eight seconds), for the entire set of flags over
360 degrees, consistent with the assumed speed.
Therefore, once the error coefficients are
established, the actual angular positions of the flags
., ~ ~ ' '` ..
,
~ 9~ 13DV-8584
-24-
beco~es known. Then, from a single ~T, the speed can be
inferred. This is true, in principle, even with a grossly
skewed distribution o~ flags as shown in Figure 12.
Having the error coefficients, which contain data as to
the angular separation of the flags, one can compute the
angular speed of the propeller based solely on the aT
shown. A full 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 ~T'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 RA~I, so to speak.
This discussion has considered only the effect of
the positions of flags upon the ~T's. However, as stated
earlier ln connection with Figures 5 and 5A, not only
position, but also the geometry and composition of the
flag are involved in the ~T's. Thus, the aT's which the
in~ention measures do not necessarily have a clear
relationship with the teeth or the edges 90, as shown in
Figure 5. ~lowever, the ~r~S do, in fact, have a
consistent relationship with the edges 90. For example,
~T(l) in Figure 5 Inay be measured for teeth 9A and B.
25 ~T(l) does not end with edge 93, but on phantom edge 90.
This causes no problem because this ~T will, in general,
always end on phantom edge 90. Thus, the ~T's do not
establish the actual geometric angular 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 "efective" location.
It was stated earlier in connection wi~h the
explanation of Step 2 that the 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.
~2~9~ 13DV-8584
-25-
Assulne that the last tooth to pass was tooth No.
8~ and ~T(8) has just been computed. Therefore, n = 8.
Also assume a deceleration is occurring, indicated by
continually increasing aT's, as shown in Table 3.
TABLE 3
~T~l) = 12,100 1 = (n~l) modulo 8
~T(2) = 129200 2 = (n+2) modulo 8
~T(3) = 12,300 3 = (n~3) modulo 8
~T(4) = 12,400 4 = (n+4) modulo 8
~T(5) = 12,500 5 = (nl5) modulo 8
aT(6) = 12,600 6 = (n~6) modulo 8
~T(7) = 12,700 7 = (n~7) modulo 8
~T(8) = 12,800 8 = n
Therefore, in this example,
Time(full revolution)/8 = 99,60~/8 = 12,450
This is the estimated QT(ID).
As stated above, the error coefficients are,
in effect, a ratio of actual dT to aT(ID). It was also
stated that ~T(ID) is estimated from eight consecutive
aT~s. ~urther, speed is computed, in Step 4, from
individual ~T's, perhaps several times per revalution.
The Inventors have found that, during a constant
acceleration or constant deceleration, adding 4 or 5 to
the index in Step 2 will give a more accurate speed
computation from a single aT, by giYing a truer error
coefficient through a better estimated ~T(ID). This is
shown in Table 3, in the right 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 is the Invention as
defined in the following claims.
.
.
