Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
\
~39L~
Docket No. DW-483
SPECIFICATION
BOREHOLE MEASUREME~T ~HILE DRILLING SYSTEMS and METHODS
This application is a continuation in part of a
co-pending application, Serial No. 857,677 filed by
Serge A. Scherbatskoy on December 5, 1977 on "Improved
Systems, Apparatus and Methods for Logging While Drilling".
Field of the Invention
This invention generally pertains to measurements
while drilling a bore hole in the earth and more part-
icularly pertains to systems, apparatus, and methods
utilizing hydraulic shock waves in the drilling mud
column for transmission of signals representing one or
more downhole parameters to the earth's surface. It
also pertains to systems and methods for de~ecting these
signals in the presence of interfering noise.
-1-
2~
Description of the Prior Art
This invention relates to data transmission systems
for use in transmitting data from the bottom of a well
bore to the surface while drilling the well.
It has been long recogniæed in the oil industry
that the obtaining of data from downhole during the
drilling of a well would provide valuable informa~ion
which would be of interest to the drilling operator.
Such information as the true weight on the bit, the
inclination and bearing of the borehole, the tool face,
fluid pressure, and temperature at the bottom of the
hole and the radioactivity of substances surrounding or
being encountered by the drill bit would all be expressed
by quantities of interest to the drilling operator. A
lS number of prior art proposals to measure these quantities
while drilling and to transmit these quantities to the
surface of the earth have been made. Various transmission
schemes have been proposed in the prior art ~or so doing.
For a d~scription of prior art see fQr instance U.S.
~ Patent No. 2,787,795 issued to J.J. Arps, U.S. Patent
No. 2,887,298 issued to H.D. Hampton, U.S. Patent No.
4,078,620 issued to J.H. Westlake et al, U.S. Patent
No. 4,001,773 issued to A.E. Lamel et al, U.S. Patent No.
3,964,556 issued to Marvin Gearhart et al, U.~. Patent
No. 3,983,948 issued to J.D. Jeter, and U.S. Patent No.
3,791,043 issued to M.K. Russell. All of the above listed
patents are incorporated in this specification by
reference.
Perhaps the most promising of these prior art
proposals in a practical sense has been that of signalling
by pressure pulses in the drilling fluid. Various
methods have been suggested in the prior art to produce
such mud pulsations either by a controlled restrîction
of the mud flow circuit by a flow restricting valve
appropriately positioned in the main mud stream or by
means of a bypass valve interposed between the inside
-2-
of the drill string (high pressure side) and the annulus
around the drill string (low pressure side).
It has been suggested in the prior art to produce
mud pressure pulses by means of valves that would either
restrict the mud flow inside the drill string or bypass
some flow to the low pressure zone in the annulus around
the drill string. Such valves are of necessity-~low
because when used inside the drill string the valve
must control very large mud volumes, and when used to
control a by-pass, because of the very high pressure
differences, the valve was of necessity also a slow
motorized valve. For example, such a motorized valve,
interposed between the inside of the drill string and
the annulus produced in response to a subsurface measurement
slow descreases and slow increases of mud pressure. These
were subsequently detected at the surface of the earth.
In order to understand more fully the operation
of a slowly acting motorized valve as suggested in the
prior art, reference is made to Fig. lA which shows the
opening and the closing of such a valve as a function
of time. Referring now specifically to Fig. lA, the
abscissas in Fig. lA represents time, t, whereas the
ordinates represent the degree of opening of the valve, R.
R S(t3 (1)
O
where S0 is the total area of the opening and S(t) is
the area which is open at the instant t during the process
of opening or closing of the valve. Thus when R = 0 the
valve was closed and when R = 1 the valve was fully opened.
The times involved in the operation of the valve were
as follows:
taV) = OAl was the time at which the valve started
to open;
t(V) = OBl was the time at which the valve was fully
open;
-3-
~a4~
t(V) ~- OCl was the time at which the valve started
to close;
t(V) = ODl was the time at which the valve was
fully closed.
The time interval:
T(v~ = t(v) _ taY) = tdV~ ~ ~cV) (~)
TaV) will be referred to as the "time of opening or closing
of the valve". The time interval
T(V) = t(V) t~v) (3)
T(V) will be referred to as the "time of open flow". Thus,
the total period of the actuation of the valve was
T(v) = 2T(v) ~ T(v)
In the above attempts one had TaV) = 1 second,
T(V) = 2 seconds and consequently the total time of the
actuation of the valve was TtV) = 4 seconds. These re-
latively slow openings and closings of the valve pro-
duced correspondingly slow decreases and increases of
mud pressure at the surface of the ear~h (see Fig. lB).
It can be seen that the mud pressure decreased from
its normal value of for example, 1000 psi (when the valve
was closed) to its lowes~ value of 750 psi (when the
valve was open). The times involved in these observed
pressure variations were as follows:
t(a) = OEl was the time at which ~he mud pressure starts
to decrease from its normal level at 1000 psi;
t(b) = OFl was the time at which the mud pressure
attained its lowest level at 750 psi and
was maintained at this level until time
t(c) = ~1;
-4-
t(c) = OGl was the time at which the mud pressure
starts to increase;
t(d) = OHl was the time at which the mud pressure
attained its normal level at 1000 psi.
Thus, the pressure decreased during the time interval
T(S) - t~bS) - t(a), then it remained constant during the
interval T~S) = t(c) - t~b)~ and then it rose from its
depressed value to the normal level during the time
interval T3S) = t~d) - t(c). Thus, the total time of
the mud flow through the bypass valve for a single
actuation of the valve was
T(S) = T(S) ~ T2S) ~ T3S~ (5)
I have designated quantities in Fig. lA (such as
taV), tbV), tcv)l tdV~, TaV), T(V) and TtV) with super-
script "v" to indicate that these quantities relate to theoperation of the valve which is below the surface of the
earth. On the other hand the quantities t~a), t(b)~ t(c),
t(d~ T(S), T2S). T3S) and T(S) in Fig. lB are designated
with superscript "s" to indicate that these quantities
relate to measurements at the surface o~ the earth. This
distinction between the quantities provided with super-
script "v" and those with s'uperscript "s" is essential
in order to fully understand some of the novel features
of my invention. It is essential in this connection to
distinguish between the cause and the effect, or in other
words, between the phenomena occurring downhole, in the
proximity of the valve and those at the detector at the
surface of the earth.
An essential feature of the previously proposed
arrangement is based on the relationships:
-5-
~ ~4~
T(S~ = TaV) ~5
T(S) = TbV) (7~
T(s) = ~(v) ~8)
These relationships show that the period of decrease
or increase of the pressure at the earth's surface was the
same as the corresponding period of opening and closing
of the valve, and the period at which the pressure was
substantially constant (at a decreased level) was the
same as the period during which the valve was fully
open. In other words, the decrease and subsequent in-
crease of the mud pressure at the earth's surface was
in exact correspondence with the opening and closing
of the valve. This condition as expressed by the re-
lationships (6), (7), and (8) will be referred to in
this specification as relating to a "regime of slow
variations of pressure".
The regime of slow pressure variation as suggested
in the prior art was not suitable for telemetering in
measurement while drilling operations, particularly when
several down hole parameters are being measured. By
the time a first parameter has been measured, encoded,
transmitted to the surface and then decoded, the well
bore can have been deepened and the second parameter may
no longer be available for measurement. Relatively
long time intervals were required for the conversion of
the measured data into a form suitable for detection and
recording. The entire logging process was lengthy and
time consuming. Furthermore various interfering effects
such as pulsations due to the mud pump and noise associated
with various drilling operations produced additional
difficulty. A slow acting motorized valve, such as
that suggested in the prior art, is believed to be in-
adequate to satisfy current com~ercial requirements.
--6--
~. ~ 3 ~ 5
Summar~ of the Invention
According to a first broad aspect of the present
invention, there is provided an apparatus for making measure-
ments in a borehole while drilling lncluding a valve electri-
cally energizable to produce mud pressure variations in a mud
column by recurrent motion of said valve, a sensor for sensing
a value of a parameter in said borehole, a first means to sup-
ply electrical power by generating an elec-trical current and
for recurrently moving said valve in response to said sensor
thereby producing said pressure vari.a-tions in said mud column
to indicate the magnitude o~ said parameter and a second means
responsive to the motion of said valve for producing a signal
representing said motion.
According to a second broad aspect of the present
invention, there is provided for use in performing measure-
ments in a borehole in conjunction with operations of drilling
said hole a telemetering apparatus for transmitting information
through a transmission channel from an appropriate depth in
said hole to the top of said hole, said information being in
the form of succession of elementary signals, said succession
beinq arranged in accordance with a pattern representing the
magnitude of a parameter representing said in~ormation, said
telemetering apparatus including a first means for positioning
at the top of said hole and operatively connected to said
transmission channel for detecting a superposition of said suc-
cession of elementary signals and of interfering signals which
results from at least one of said drilling operations, and a
selective means dependent upon the shape of said elementary
signals for receiving said detected signal and selectively
transmitting said succession by minimizing said interfering
signals and a means responsive to said selective means for in-
--7--
~ ~ !
dicating the value of said parameter.
Some of the ob~ectives of my invention are accompli-
shed by using hydraulic shock waves -Eor telemetering logging
information while drilling is in process. These shock waves
are produced by a very rapidly acting (for all practical pur-
poses almost instantaneously acting) bypass valve interposed
between the inside of the drill string and the annulus around
the drill string. When the bypass valve suddenly opens, the
pressure in the immediate vicinity of the valve drops and then
returns to normal almost instantaneously and a sharp negative
pulse is generated, and conversely, when the bypass valve sudden-
ly closes, a sharp positive pulse is generated. Elasticity of
mud column is employed to assist in the generation and trans-
mission of such shock waves. The phenomenon is analogous to the
well known water hammer effect previously encountered in hyd-
raulic transmission systemsO (See for instance John Parmakian
on "Water Hammer Analysis", Prentice Hall, Inc., New York, N.Y.
1955 or V.L. Streeter and E.B. Wylie on 'IHydraulic Transients"
McGraw-Hill Book Co., New York, N.Y.)
Significant features of my invention such as the
generation and detection of hydrauIic shock waves are shown
schematically in Figs. 2A and 2B. The graph in Fig. 2A shows
the openings and closings of a fast acting shock wave produ-
cing valve, whereas the graph of Fig. 2B shows pressure varia-
tions detected at the earth's surface and resulting from the
operation of the valve as in Fig. 2A. Symbols such as Al, Bl,
(v) (v) t (v) t (v) T (v) T ( ) d T (v) n
Cl, Dl, ta ~ tb ' c ' d ' a ' b an t
Fig. 2A have a similar meaning as the corresponding syl~ols in
Fig. lA. However, the time scales in Figs. lA, lE, 2A and 2B
have been considerably distorted in order to facilitate descrip-
tion, and in the interest of clarity of explanation.
-7a-
s
The first thing which should be noted in examining
Fig. 2A is that the ti~es of opening and closing o~ the
valve in accordance with my invention are by several
orders of magnitudes shorter than the corresponding
times obtained by means of the motorized valve as re-
ported in connection with Fig. lA. In the arrangement
previously sugges~ed taS in Fig.lA) one had TaV) = 1
second whereas in accordance with my inven~ion as in
Fig. 2A one has TaV) = 5 milliseconds. A similar si~-
uation applies to the time interval during which a valveremains open. In the arrangement previously suggested
~as in Fig. lA) one had T(V) = 2 seconds whereas in
Fig. 2A one has T(V) = 100 milliseconds. Thus, for all
practical purposes, the openings and closings of the valve
in Fig. 2A may be considered as instantaneous or almost
instantaneous.
Rapid or almost instantaneous openings and closings
of the valve have an important and far reaching influence
on the performance of a telemetering system in a measuring
~hile drilling operation. The pressure variations detected
at the earth's surface in accordance with my invention
(Fig. ~B) show no similarity whatever to the pressure
variations obtained by means of a slow acting valve (Fig.
lB). I have previously pointed out the existence of
equations (6), (7), and (8) which show relationships between
the events illustrated in Fig. lA and those illustrated in
Fig. lB. Analagous relationships do not exist between the
events in Fig. 2A and 2B.
As shown in Figs. lA and lB, the opening of the valve
produced a corresponding decrease in the mud pressure
at the surface of the earth, and conversely, the closing
of the valve produced a corresponding increase in pressure.
--8--
4~
For the sake oi emphasis I wish to repeat that in
the prior art the opening of the valve produced a single
event namely a decrease in pressure and the subsequent
closing of the valve produced another single event --
an increase in pressure. On the other hand in my in-
vention the fast opening of the valve as in Fig. 2A
produces two events: a rapid decrease and subsequent
increase in pressure (negative pulse "M" as in Fig. 2B).
This is in contrast to the case shown in Fig. lA and Fig.
lB where an opening and a subsequent closing of the valve
is required in order to produce a decrease and a sub-
sequent increase in pressure. Furthermore, the fast
closing of the valve as in Fig. 2A produces an increase
and a subsequent decrease of the mud pressure (positive
pulse "N" as in Fig. 2B). Such an increase and sub-
sequent decrease in pressure does not occur in the ar-
rangements suggested in the prior art. In my inven-
tion, there are two shock waves produced by a single
operation of the valve. A wave form such as shown in
Fig. 2B, which comprises both a negative and a positive
pulse, will be referred to in this specification as à
"valve wavelet". Pressure pulses associated with a valve
wavelet have an onset rate of several thousand psi/sec.
and are of short duration.
It is of interest to point out the rapidity of the
phenomena associated with the observed valve wavelets.
The times involved in Fig. 2B are as follows:
t(S) = OK is the time of appearance of the negative
pulse "M";
t2S) = OL is the time at which the negative pulse
"M" decayed;
t(S) = OM is the time oE appearance of the positive
pulse "N";
_g_
t(S) = ON is -the time at which the positive pulse
"N" decayed.
The time interval Tns) representing the "len~th" of the negative
pulse "M" ~or the positive pulse "N") is 100 milliseconds,
whereas the time interval TmS) from the appearance of the
negative pulse "M" to the appearance of the positive pulse "N"
is 105 milliseconds. Thus, the total period of ~low as shown
in Figure 2B; i.e.,
T(S) =, T(s) + T(S)
u n m
iS 205 milliseconds whereas the -total period of flow as shown
in Figure lB (see equation 5) was TtS) = 4 seconds.
The graphs in Figures lA, lB, 2A, and 2B have been
simplified and idealized by eliminating ripples and other
extraneous effects. It should also be noted (see Figure 2B)
that the bypass valve is at least partially open during the
time interval from t( ) to t( ). During this time interval,
there is a slow pressure decline which is eliminated at the
detection point by an appropriate filter. Such a pressure
decline is not shown in the graph of Figure 2B.
It should also be pointed out that the numerical
values attached to Figures 2A and 2B are gi.ven merely as an
example. These values should not be interpreted as restricting
my invention to any particular example given.
The process as explained in connection with Figures
2A and 2B will be referred to as relating to a "regime of
hydraulic shock waves". Thus, a distinction is made between
the regime of hydraulic shock waves as in Figures 2A and 2B
and the regime of slow variations of pressure as in Figures lA
and lB.
-- 10 --
~ ~9~2~i
By providing a regime of hydraulic shock waves, I
obtained a telemetering sys~em by me~ns of which large
amounts of information can be transmitted per unit of time.
Such a system is considerably better adapted to satisfy
current commercial requirements than the one which is
based on the regime of slow variations of pressure.
The valve, in accordance with my invention, is
operated by the output of one or more sensors
for sensing one or more downhole parameters in the
earth's subsurface near the drill bit. One single
measurement of each parameter is represented, by a
succession of valve wavelets. Each valve wavelet
corresponds to a single opening and closing of the valve.
The succession of valve wavelets (which represen~s
lS the useful signal) when detected at the earth's surface
is usually mixed with various interfering signals such
as those produced by the operation of the pump and by other
drilling operations. In a typical drilling arrangement,
a large pump located at the surface is used to pump
drilling mud down the drill stem ~hrough the bit and
back to the surface by way of annulus between the drill
pipe and the well bore. The interfering effects due to
the pump are eliminated in this invention by a process
which takes into account the periodicity of these effects.
Other effects associated with drilling operations usually
appear as noise signal comprising a relatively wide
frequency spectrum. This noise signal is in some instances
white noise and in other instances it departs considerably
from white noise. A digital filtering system which may
be a matched filter or a pulse shaping filter or a spiking
filter is employed to remove the noise signal. The
matched filter m~;m;zes the signal to noise ratio at the
reception point, a pulse shaping filter minimizes the
-11-
~94~5
mean square difference between a desired output and the
actual output, whereas a spiking filter transforms the
useful signal by contracting it into one which is suf-
ficiently sharp so that it can be distinguished against
a background noise. A special technique is applied to
adapting these filters to the objectives of this invention.
Such a technique requires storage and subse~uent re-
production of two reference signals. The first re-
ference signal is a wavelet produced by the opening
and closing of the valve and the second reference signal
represents noise due to the drilling operations. Detection
and storage of the first reference signal is obtained
by removing the weight on the bit and stopping the
actual drilling (but maintaining the mud pumps in normal
action). Thus, a signal is obtained which is free from
the ambient noise. Detection and storage of the second
reference signal is obtained when drilling is in progress
during a period of time when the valve is closedO An
appropriate digital computing system is arranged to receive
the data representing one or both of these reference
signals, and derives from the data a memory function
for the matched filter, for the pulse shaping filter,
or for the spiking filter.
One aspect of my invention pertains to improvements
involving the bi-stable action of valve assembly 40 of
the special telemetry tool 50. Another aspect o my
invention concerns the provision of a special hydraulically
operated mechanical arrangement that will periodically
positively move the valve 40 to the closed positicn.
In addition there is provided an electric system that
will inhibit operation of the valve 40 in case of an
electrical failure in the downhole apparatus.
-12-
~:~9~5
Further aspects of my invention concern improve-
ments involving the power supply 95 and the power drive
104 of the special telemetry tool 50. Such improvements
serve to greatly increase the number of satisfactory
valve actuations attainable without downhole battery re-
charge or replacement.
Another aspec~ of my invention concerns improve-
ments in pulse time codes wherein only short pulses of
substantially constant duration are transmitted, and the
time intervals between successive pulses are the measures
of ~he magnitude of the relevant parame~er. In addition
there is disclosed a system for improving the precision
and accuracy in the transmiss;.on and detection of mud
pressure pulses generated at the downhole equipment,
which system involves the generation at the downhole
equipment and the transmission of a group of at least
3 unequally spaced mud pressure pulses for each in-
formation carrying single pulse, and the provision of
appropriate equipment at the surface for detecting
and translating the transmitted pulse groups.
The novel features of my invention are set forth
with particularity in the appended claims. The in-
vention both as to its organization and manner of
operation with further objectives and advantages
thereof, may best be presented by way of illustration
and examples of embodiments thereof when taken in con-
junction with the accompanying drawings.
.,
-13-
11 94Z~5
Brief Description of Drawings
Figs. lA, lB~ 2A and 2B are graphs which relate
in part to the portions of the specification entitled
"Field of the Invention" and IlDescription of the Prior
Art'l. The rPm~;n;ng figures, as well as Figs. lA, lB,
2A, and 2B relate to the portions of the specification
entitled "Summary of the Invention" and "Description
of the Preferred Embodiments".
Fig. lA shows schematically the operation of a
slow acting valve as was suggested in the prior art.
Fig. lB shows schematically pressure variations detected
at the earth's surface and resulting from the operation
of the valve as shown in Fig. lA. Both Figs. lA and
lB describe a condition referred to in this specification
as a "regime of slow variations of pressure";
Fig. 2A shows schematically the operation of a
fast acting valve in accordance with my invention.
Fig. 2B shows schematically presssure variations
detected at the earth's surface and resulting from
the operation of the valve as shown in Fig. 2A. Both
Figs. 2A and 2B describe a condition referred to in
this specification as a "regime of hydraulic shock waves".
Fig. 3 schematically and generally illustrates a
well drilling system equipped to simultaneously drill
and to make measurements in accordance with some aspeets
my invention.
Fig. 4A shows schematically a portion of the sub-
surface equipment including a special telemetry tool
in accordance with my invention;
Fig. 4B shows schema~ically a portion of the
arrangement of Fig. 4A.
Fig. 5A shows, schematically and more in detail, the
electronic processing assembly comprised within the
dotted rectangle in Fig. 4A.
Fig. 5B shows schematically a power supply including
a capacitor charging and discharging arrangement ~or
providing the required power and energy for actuating
. -14-
2~3S
the valve of the special telemetry tool.
Fig. 5C shows schematically electronic circuitry
whieh may be utili~ed to accomplish automatic cut-off
o~ the power drive for the val~e of the special tele-
metry tool. Fig. SD and 5E are graphs to aid in the
explanation of the automatic cut-off for the signalling
valve power drive.
Figs. 6A, 6B and 6C show diagrammatically the
operation of hydraulic "auto close" of the signalling
valve.
Fig. 6D is an engineering drawing of the arrange~
ment shown in Figs. 6A, 6B, and 6C.
Fig. 6E shows schematically an electronic "fail
safe" arrangement applicable to the si~nalling valve.
Fig. 7A shows schema~ically the "sub" and housing
structure for the special telemetry tool.
Fig. 7B shows schematically the cross-sectîon
shape of centralizers that may be utilized with the
structure of Fig. 7A.
~20 Fig. 7C shows schematically special connector means
that may be utilized for joining the sub-sections of
housing portion 250b of Fig. 7A.
Figs. 8A to 8E show graphs representing variations
of pressure as measured at the earth's surface and
corresponding to various values of TaV) (~imes of open-
ing or closing of a valve) and of T(V) (time of open flow).
Graphs in these figures show results of certain tests
which I performed in order to obtain the optimum condition
for a regime of hydraulic shock waves~ More specifically
Figs. 8A to 8E can be described as follows:
Fig. 8A corresponds to TaV) = 1 second and T(V) -
2 seconds,
Fig. 8B corresponds to T(V) = 200 milliseconds and
T(V) = 1 second.
-15-
Fig. 8C corresponds to T~V) = 60 milliseconds and
T(V) = 0.5 seconds.
Fig. 8D corresponds to TaV) = 20 milliseconds and
TbV = 0.25 seconds.
Fig. 8E corresponds to TaV) = 5 milliseconds and
T(V) = 10 1 seconds;
Fig. 8F shows an exact reproduction of the pressure
signal showing a valve wavelet as received at the surface
from the depth of 9,800 feet at an actual oil well being
drilled in East Texas.
Fig. 9 is a schematic illustration showing ~ypical above
ground equipment to be used in conjunction with a downhole
pressure pulse signalling device in accordance with my
invention and comprising a matched filter for elimin~ting
random noise when the random noise is white. Figs. lOA to
lOG provide a graphic illustration of certain wave forms
and pulses as they vary with time, which are shown in
order to aid in explanation of the operation of the
equipment in Fig. 9. The time axes in Figs. lOA to
Fig. lOC and the time axes of Fig. lOD to Fig. lOG are
posi~ioned one below the other so that one can compare
these signals and wave forms in their time relationship
one to another. More specifically, Figs. lOA to lOG
can be described as follows:
Fig. lOA contains three graphs showing three com-
ponents of a signal detected at the top of the drill
hole. These components represent, respec~ively, an
information carrying signal, pump noise, or in the case
of use of several pumps in tandem, the noise from the
group of pumps and random noise.
Fig. 10~ contains three graph showing respectively
the delayed information carrying signal, the delayed
pump noise and the delayed random noise. The delay
is by an amount Tp representing the period of the operation
-16- -
~9~2~35
of the pump (when several pumps are used the pressure
variations, although not sinusoidall are still perio~ic
because the tandem pumps are maintained relatively close
to being "in phase").
Fig. 10C contains two graphs showing, respectively,
differences of the corresponding graphs in Fig. 10A
and Fig. 10B. One of these graphs represents random
noise while the other graph represents an information
carrying signal.
Fig. 10D shows a function representing the output
of a digital filter or of a cross-correlator in the
embodiments of my invention. This function is sub-
stantially simil~r to that represent;ng the information
carrying signal in Fig. 10C. The digi~al filter used
herein may be a matched filter, a pulse shaping filter
or a spiking filter.
Fig. 10E shows a function similar to that of Fig.
10D but delayed in time by an appropriate time interval.
Fig. 10F shows a function as that in Fig. 10E but
reversed in time.
Fig. 10G results from a comparison of graphs of
Fig. 10D and 10F and repres~nts instants corresponding
to pulses that occur in coincidence in these graphs.
Fig. 11 shows schematically certain operations
performed by a digital filter.
Fig. 12 shows schematically an arrangement for
storing an information carrying signal or for storing
a noise signal.
Fig. 13 shows, schematically a portion of the above
ground equipment comprising a correlator for noise elimination.
Fig, 14 shows schematically a portion of the above
ground equipment comprising a matched filter for noise
elimination when the noise is not white.
Fig. 15 shows schematically a portion of the above
3~ ground equipment which contains a pulse shaping filter.
-17-
Fig 16 illustrates schematically certain operations
performed by a pulse shaping filter
Fig. 17 shows schematically a portion of the above
ground equipment comprising a spiking ~ilter wherein
a spiking filter is used to transform a double wavelet
into a corresponding pair of spikes.
Figs. 18A to 18F show six possible choices for
a spike lag for a pair of spikes, as produced by means
of the arrangement of Fig. 17.
Fig. 19 shows schematically a portion of the above
ground equipment comprising a spiking filter wherein
a spiking filter is used to transform a single valve
wavelet into a corresponding single spike.
Fig. 20A to Fig. 20F show six possible choices
for a spike lag for a single spike as produced by means
of ~he arrangement of Fig. 19.
Fig. 21A to Fig. 21C show schematically certain
operations associated with a spiking filter for various
time delays. More specifically: Fig. 21A corresponds
to a desired spike at time index 0; Fig. 21B corresponds
to a desired spike at tim~ index l; Fig. 21C corresponds
to a desired spike at time index 2.
Fig. 22 shows schematically an arrangement for
determining the perfor~ance parameter P of a spiking filter.
Fig. 23 is a graph showing how the performance of the
Eilter P may vary with spike lag for a filter of fixed
duration.
Fig. 24 is a graph showing how the performance
parameter P of a spiking filter may vary with filter
length (or memory duration) for a fixed spike lag.
Fig. 25 contains graphs showing how the performance
parameter P of a spiking filter may vary with filter
length and filter time lag.
Fig. 26A shows schematically a pulse time code system
in accordance with the prior art.
-18-
s
Fig. 26B shows schematically a pulse tîme code
system of my invention, wherein the magnitude of the
parameter being transmitted is represented by the time
in~erval between successive single short pulses of sub~
stantially constant time duration.
Fig. 26C further illustrates schematically the pulse
time code system of Fig. 26B.
Fig. 26D shows schematically a pulse time code system
of the type shown by Figs. 26B and 26C but wherein
"triple group" pulses are utilized.
Fig. 27 is a schematic block diagram showing a
"Code Translator" utilized to enable a system to receive
signals that are in the "triple group" pulse time
code form.
Fig. 28A is a schematic block diagram showing cir-
cuitry of the selector 316 of Fig. 27 in further detail.
Figs. 28B, 28C, 28D and 28E are graphs to aid in
explanation and understanding of the operation of the
circuitry of Fig. 28A.
Fig. 29 is a schema~ic block diagram of downhole
circuitry for generating the "triple group" pulses shown
in Fig. 26D.
Fig. 30 is a schematie diagram showing the principles
of eircuitry that can accomplish the pulse time code
of my invention.
It should be noted that identical reference numerals
have been applied to similar elements shown in some of
the above figures. In such cases the description and
functions of these elements will not be restated in so
far as it is unneccessary to expl in the operation of
these embodiments.
-19`-
2~
Description of the Preferred ~mbodiments
GENERAI. DESCRIPTION OF APPARATUS FOR DATA
TRAN SMI S S 10~ WHILE DRILLING
Figure 3 illustra~es a typical layout of a system
embodying the principles of this invention. Numeral
20 indicates a standard oil well drilling derrick with
a rotary table 21, a kelly 22, hose 23, and standpipe
24, drill pipe 25, and drill collar 26. A mud pump or pumps
27 and mud pit 28 are connected in a conventional manner
and provide drilling mud under pressure to the stand-
pipe. The high pressure mud is pumped down the drill
string through the drill pipe 25 and the standard drill
collars 26 and then through the special telemetry tool
50 and to the drill bit 31. The drill bit 31 is provided
with the usual drilling jet devices shown diagra~atically
by 33. The diameters of the collars 26 and the telemetry
tool 50 have been shown large and out of proportion to
those of the drill pipe 25 in order to more clearly
illustrate the mechanisms. The drilling mud circulates
downwardly through the drill string as shown by the
arrows and then upwardly through the annulus between
the drill pipe and the wall of the well bore. Upon
reaching the surface, the mud is discharged back into
the mud pit (by pipes not shown) where cu~tings of rock
and other well debris are allowed to set~le and to be
further filtered before the mud is again picked up and
recirculated by the mud pump.
Interposed between the bit 33 and the drill collar
26 is the special telemetering transmitter assembly or
telemetry tool designated by numeral 50. This special
telemetering transmitter assembly 50 includes a hous-
ing 250 which contains a valve assembly, or simply a
valve 40, an electronic processing assembly 96, and
sensors 101. The valve 40 is designed to momentarily
by-pass some of the mud from the inside of the drill
collar into the annulus 60. Normally (when the valve
40 is closed) the drilling mud must all be driven
-20-
~4~
through the jets 33, and consequently conslderable
mud pressure (of the order of 2000 to 3000 psi) is present
at the standpipe 24. When the valve 40 is opened at the
command of a sensor lOl and electronic processing assembly
96, some mud is bypassed, the total resistance to flow is
momentarily decreased, and a pressure change can be
detected at the standpipe 24. The electronic processing
assembly 96 generates a coded sequence of electric
pulses representative of the parame~er being measured
by a selected sensor 101, and corresponding openings
and closings of the valve 40 are produced with the
consequent corresponding pressure pulses at the stand-
pipe 24.
N~eral 51 designates a pressure transducer that
generates electric voltage representative of the pressure
changes in the standpipe 24. The signal representative
of these pressure changes is processed by electronic
assembly 53, which generates signals suitable for recording
on recorder 54 or on any other display apparatus. The chart
of recorder 54 is driven by a drive representative of the
depth of the bit by means well known (not illustrated).
II. GENERAL DESCRIPTION OF SPECIA~ TELEMETERING
TRANSMITTER
Fig. 4A shows certain details of the special tele-
metering transmitter 50. Certain of these and other
details have also been described in the above referred
to co-pending application Serial No. 857,677 filed by
S.A. Scherbatskoy, of which this application is a
continuation in part. Fig. 4A is diagrammatic in nature.
In an actual tool, the housing 250, which contains the
valve 40, the electronic processing assembly 96, and
the sensors 101, is divided into two sections ~50a and
250b. The upper portion 250a (above the dotted line 249)
-21-
contains the valve assembly 40 and associated mech-
anisms and, as will be pointed out later in the
specification, is of substantially larger diameter than
250b. The lower section 250b ~below the dotted line
249) contains the electronic processing assembly 96,
sensors 101~ and associated mechanisms, and as will be
explained later in the specification, has a substantially
smaller diameter than the upper section 250a. As shown
in Fig. 4A, ~he drilling mud circulates past the special
telemetry tool 250a, 250b downwardly (as shown by the
arrows 65) through the bit nozzle 33 and then back (as
shown by the arrows 66) to the surface in annulus 60
and to the mud pit 28 by pipe means not shown. The valve
assembly 40 comprises valve stem 68 and valve seat 69.
The valve stem and seat are constructed in such manner
that the cross sectional area of the closure A is slightly
larger than the cross sectional area B of the compensating
piston 70. Thus, when the pressure in chamber 77 is
greater than that in the chamber 78, the valve stem 68
is forced downwardly; and the valve 40 tends to close
itself more tightly as increased differen~ial pressure
is applied.
The fluid (mud) pressure in chamber 77 is at all
times substantially equal to the fluid (mud) pressure
inside the drill collar, designated as 26 in Fig. 3
and 50 in Fig. 4A, because of the opening 77a in the
wall of the assembly 250. A fluid filter 77b is inter~
posed in passageway 77a in order to prevent solid part-
icles and debris from entering chamber 77. When the
valve 40 is closed, the fluid ~mud) pressure in chamber
78 is equal to the fluid (mud) pressure in the annulus
60. When the valve 40 is open and the pumps are running
mud flow occurs from chamber 77 to chamber 78 and through
orifice 81 to the annulus 60 with corresponding pressure
drops-
-22-
`~;
~1~3':~`V~
Double acting electromagnetic solenoid 79 i5 arrange~
to open or close valve 40 in response to electric current
supplied by electric wire leads 90.
Let P60 indicate the mud pressure in the annulus 60~
P77 the pressure in chamber 77, and P78 -the pressure in chamber
78. Then, when valve 40 is closed, one has P78 = P60. When the
pumps 27 are running and valve 40 is "closed'l, or nearly closed,
and P77 > P78 the valve stem 68 is urged towards the valve seat
69. ~hen valve 40 is in the "open" condition (i.e., moved
upwardly in the drawing) flow of mud from chamber 77 to the
annulus 60 results; and because of the resistance to flow of
the orifice C (Figure 4B), one has the relationship P77 ~ P78 >
P60. Chambers 83 and 269are filled with a very low viscosity
oil (such as DO~ CORNING 200 FLUID, preferably of viscosity 5
centistokes or less) and interconnected by passa~eway 86.
Floating piston 82 causes the pressure P83 in the oil filled
chamber 83 to be equal at all times to P78. Thus, at all times
P78 = P83 = P84. Therefore, when the valve 40 is "open", since
P78 P84 and P77 > P84, the valve 40 is urged towards the
"open" position by a force F = (area s) (P77 - P84)o The valve
40 can therefore be termed bi-stable; i.e., when "open" it tends
to remain "open" and when "closed" it tends to remain "closed".
Furthermore, when nearly open it tends to travel to the open
condition and when nearly closed, it tends to travel to the
closed condition. The valve 40 can therefore be "flipped" from
one state to the other wi-th relatively little energy. The valve
action can be considered the mechanical equivalant oE the
electric bi-stable flip-flop well known in the electronics art.
Fig. 4B shows the valve 40 in the open condition;
w~ereas, in Fig. 4A it is closed.
Referring again to Figure 4A, numeral 91 indicates
an electric "pressure switch" which is electrically
conductive when P77 > P78 (pump running) and electrically
non-conductive when P77 = P78 (pumps shut down - not
running). Wire 92 running from pressure switch 91 ~o
power supply 93 can, therefore, turn the power on or off.
Also, by m~ans of electronic counter 94 and electromagnetic
sequence switch 95, any one of the four sensors 101 can be
operatively connected to the electronic processing assembly
~6 by sequentially stopping and rllnn;ng the mud pumps 27
or by stopping then rllnn;ng the pumps in accordance with
a predetermined code that can be interpreted by circuitry
in element 94.
III. DESCRIPTION OF ELE~TRONIC PROCESSING
ASSEMBLY PORTION OF SPECIAL TELEMETRY TO~L
We have described the operation of the bi-stable
valve 40 and the sequence switch 95 which makes the
selective electrical connection of the various sensors
101 ~o the electronic processing assembly 96.
For further details of the electronic processing
assembly 96 reference is made to Fig. 5A, where like
numbers refer to like numbers of Fig. 4A.
Various types of sensors that generate electric
signals indicative of a downhole parameter are well
known. Examples are gamma ray sensors, temperature
sensors, pressure sensors, gas content sensors, magnetic
compasses, strain gauge inclinometers, magnetometers,
gyro compasses, and many others. For the illustrative
example of Fig. 5A, I have chosen a gamma ray sensor
such as an ionization chamber or geiger counter or
-24-
2~35
scintillation counter (with appropriate electronic
circuitry). All these can be arranged to generate
a DC voltage proportional to the gamma ray flux which
is intercepted by the sensor.
It is understood that the switching ~rom one type
sensor to another as accomplished by switch mechanism
95 of Fig. 4A is well within the state of the art,
(electronic switching rather than the mechanical switch
shown is preferable in most cases). Consequently, in
Figo 5A for reasons of clarity of description, only
a single sensor 1~1 has been shown. Also, the power
supply 93 and mud pressure actua~ed switch gl of Fig.
4A are not illustrated in Fig. 5A.
In Fig. 5A, the sensor 101 is connected in cascade
to A/D convertor 102/ processor 103, and power drive 104.
The power drive 104 is connected to windings 105 and
106 of the double acting solenoid designated as sole-
noid 79 in Fig. 4A. The power drive 104 may be similar
to that shown by Fig. 3E of the parent application.
The operation is as follows: the sensor 101 gen-
erates an output electric analog signal as represented
by the curve lOla shown on the graph immediately above
the sensor rectangle 101. The curve shows the sensor
output as a function of the depth of the telemetering
transmitter 50 in the borehole. The A/D converter con-
verts the analog signal of 101 into digital form by
measuring in succession ~he magnitude of a large number
of ordinates of curve lOla and translating each individual
ordinate into a binary number represented by a binary
word. This process is well known in the art and requires
no explanation here. It is important, however, to
realize that whereas graph lOla may represent the
variation of the signal from the transducer in a matter
of hours, the graph 102a represents one single ordinate
(for example, AB of the curve lOlb). Thus, the time
scale of the axis of absissas on Figure 102a would be
in seconds of time and the whole graph 102a represents
-25-
one ~inary 12 bi~ word, and in actuality represents
the decimal number 2649. Thus, each 12 bit word on
grap~ 102a represents a single ordinate such as the
ordinate AB on the graph lOla. The usual binary coding
involves ti~e pauses between each binary word. After
the pause a s~art up or precursor pulse is transmitted to
indîcate the beginning of the time interval assigned to the
binary word. This precursor pulse is not part of the binary
word but serves to indicate that a binary word is about
to commense. The binary word is then transmitted which
is an indication of the value of an ordinate on graph
lOla; then a pause (in time) followed by the next binary
word representing the magnitude of the next ordinate,
and so on, in quick succession. The con~inuous curve
of graph lOla is thus represented by a series of binary
numbers or words each representing a single point on
the graph lOla. It is important to understand here
that between each binary word there is always a pause
in time. This pause ~during which no signals are
transmitted) is frequently several binary words long,
and the pause will be employed for an important purpose
which will be explained later in the specification.
In order to permit decoding at the surface, the clock
No. 1 must be rigorously constant (and in synchronism with
the corresponding clock 212 or 309 located at the surface),
and it generates a series of equally timed spaced pulses
in 2 manner well known in the art of electronics.
The graph 103a represents a single bit of the binary
word 102a, and the axis of abcissas here again is quite
different from the previous graphs. The time on graph
103a is expressed in milliseconds since graph represents
only a single bit. Each single bit is translated into
two electric pulses each of time duration tx and sep-
arated by a time interval ty. &raph 104a is a replica
of 103a, which has been very much amplified by the power
drive 104. Electric impulse 104b is applied to solenoid
winding 105 (which is the valve "open" winding), and
-26-
electric impulse 104c is applied to solenoid winding
106 (which is the valve "close" winding). The valve
40 of Fig. 4A thus is opened by pulse 104b and closed
by pulse 104c and, therefore, the valve 40 remains in
the "open" condition for approximately the time ty.
The times tx are adjusted to be proper for correct act~
uation of the solenoid windings and the time ty is pro-
portioned to open the valve 40 for the correct length
of time. Both of these times are determined and con-
trolled by the clock ~2.
In telemetering information from a sensor to the
earth's surface, I provide appropriate pauses between
transmission of successive binary words. Because of
these pauses, it is possible to store in an appropriate
electronic memory at the surface equipment the noise
caused by the drilling operation alone (without the
wavelet). The necessary arrangements and procedures for
doing this will be described later in this specification.
IV. DESCRIPTION OF POWER SUPPLY FOR SPECIAL
TELEMETERING TRANSMITTER
As was pointed out pre~iously, the valve 40 of
Fig. 4A must be very fast acting, and to drive it fast
requires considerable power. (It has been determined
as a result of appropriate testing that such a valve
requires about 1/2 to 3/4 horsepower to operate at the
necessary speed).
Although this power is very substantial, it is
applied only very briefly, and consequently requires only
small energy per operation.
In actual operation during tests, it was found that
1/2 horsepower applied for about 40 milliseconds provided
the required energy to produce a satisfactory single
valve actuation. This energy can be calculated to be
about 15 Joules. A battery pack that is sufficiently
small to be contained within housing 250b of Fig. 7A
-27-
can provide approximately ~ million Joules, without
requiring recharge or replacement, The system is
therefore capable of generating 130,000 complete valve
operations (open plus close), In actuality the energy
consumption is less than 15 Joules per operation. The
inductance, the Q, and the motional impedanc~ of the
solenoid winding cause the current build up to be
relatively slow and along a curved rise as shown in
curve 272A of Fig. 5C and 300, 301, of Fig. 6E. Thus
the total energy per pulse is subs-tantially less than
15 Joules and has been measured at 9 Joules thus pro-
viding a capability of 216,000 complete valve actuations.
(A still greater capability is achieved by use of the
circuitry described later in connection with Fig. 5C.~
From the above, it can be seen that providing the nec-
essary downhole energy from batteries for a practical
telemetry tool is quite feasible. Providing the nec-
essary very large power (1/2 horsepower), however, pre-
sents difficult problems
It was clear that the solution to such a problem
would involve the storage of energy in a mechanism that
could be caused to release it suddenly (in a short time~
and thus provide the necessary short bu~sts of high
power. One such mechanism was "hammer action" which was
utilized in the tool disclosed in my co-pending appli-
cation, but which has been found to be sometimes insuf-
ficient. Other mechanisms considered ear:Ly were the use
of compressed air, compressed springs and others.
~apacitor energy storage systems required large values
of capacitance: The energy stored in a capacitor varies
as the first power of the capacitance and as the square
of the stored voltage, and since low inductance, ~ast
acting, solenoid drive windings are required, the
necessity of low voltage devices becomes apparent,
initial calculation indicated that unduly large capacitors
would be required.
-28-
After further evaluation, it appeared that an
operable system might be feasible. By mathematical
analysis and by experiments and tests it was determined
~hat a set of optimum circuit parameters would be as
follows:
l. Inductance o solenoid winding:
.1 henrys when in the actuated position and
.07 henrys when in the non-actuated position
(i.e., a tapered armature solenoid).
2. Resistance of solenoid winding: 4.5 ohms.
3. Voltage at which energy is stored: 50 volts.
4. Magnitude of storage capacitor: lO,000 mfd.
5. Current capability of drive circuit: 10 amperes.
It was determined that in order to have fast sole-
noid action, low inductance windings are desirable. T t
was also determined that current capabilities of elec-
tronic drive circuits can be increased well beyond 10
amperes. Low voltage, however, requires unduly large
values of capacitance.
Recent advances in so called molten salt batteries
have produced energy sources of very good compactness.
The same recent technology has also developed capacitors
of extraordinarily high values, 10 farads in as little
space as 1 cubic inch. These were unacceptable because
the required heating to a high temperature (500C) which
was deemed impractical; and ~he cost was prohibitive.
Consequently, still further efforts were required.
Following a thorough and lengthy investigation, finally
it was discovered that a tantalum slug capacitor made
in accordance with the latest developments would meet
the specifications if the other parameters and factors
outlined above were optimized to match the characteristics
of such capacitors.
From the above it can be seen that at least 215,000
complete valve operations can be realized from one battery
-29-
charge. Assuming that the telemetry system can provide
adequate continuous data by transmitting five pulses per
minute, the system is capable of operating continuously
in a bore hole for a period of 440 hours. It must be
pointed out however that continuous operation is often
not necessary. The tool can be used only intermittently
on command by the circuitry controlled by switch ~1 and
elements 94 and ~5 of Fig.~4A.
Furthermore, as will be explained later, when ad-
vantage is taken of the improved circuitry of Fig. 5C
an even greater number of valve operations can be
achieved. Operation at a rate of one pulse per second
is considered practical.
There is another parameter to be determined: the
proper recharging of the capacitor after discharge. The
capacitor can be charged through a resistor connected to
the battery, (or other energy source) but this sometimes
proved to be slow because as the capacitor became part-
ially charged, the current through the resistor diminished,
and at the end of the charge cycle, the charging current
approached zero. If the ohmic valve of the resis~or is
made small, the batteries would be required to carry
excessive momentary current because the initial current
surge during the charging cycle would exceed the value
for maximum battery life~ The best method is to charge
the capacitor through a constant current device. The
capacitor would then be charged at an optimum charging
current corresponding to the optimum discharge current
for the particular type of battery for m~im~lm energy
storage. By correctly determining the charging current,
a substantial increase ~sometimes a factor of ~ or 3)
in the amount of energy that is available from a given
battery type can be achieved. Constant current devices
are well known and readily available electronic integrated
circuits, and are available for a wide range of current
values.
-30-
Fig. 5B shows schematically a power supply which
may be incorporated in the power drive 104 of Fi~. 4A
including a capacitor charging and discharging arrange-
ment ~or providing the required power and energy ~or the
windings of solenoid 7g. In Fig. 5B, 450 indicates
a battery or turbo generator or other source of direct
current electric potential, 451 the constant current
device, and 452 the capacitor. The capacitor is charged
through the constant current device 451 and discharged
via lead 453. The lead 454 provides the regular steady
power required for the balance of the downhole electronics.
V. DESCRIPTION OF HYDRAULIC "AUTO-CLOSE"
SIGNALLING VALVE
I have also provided an arrangement which will
operate in case of a malfunction which could occur
when the valve is "stuck" in an open position for a
long period of time. An arrangement for automatically
closing the valve in case of such malfunction (indicated
by reference numeral 269 in Fig. 4A) is illustrated
diagramatically in connection with Fig. 6A, 6B, and 6C.
As was pointed out earlier in the specification,
the valve is designed to have a hydraulic detent or bi-
stable action; i.e., when opened by an impulse from the
solenoid winding 105 it tends to remain open and later,
when closed by an impulse from the solenoid winding 106,
it tends to remain closed. It is possible that because
of an electrical or mechanical malfunction the valve
could become "stuck" in the open position. It should
be noted that if such a malfunction occurs the drilling
operation can proceed. Some wear would occur at the
orifice 81 of Fig. 4A. However, the disturbance to
the mud system hydraulics by having the valve open
for long periods of time is not desirable; and even
though drilling can continue, it is very advantageous to
31-
%`~
have the valve closed most of the time and opened only
to produce the short pulses required to generate the
h~draulic shock wave.
In the diagrammatic drawings o~ Fig. 6A, 6B, and 6C,
the rod 100 is used to push the val~e closed by exerting
a force downward on the rod 80 of Fig. 4B (the solenoid
armature shaft).
Referring now to Figs. 6A, 5B, ~C, and 6D, the upper
end of the mechanism is exposed to ''drill pipe mud";
i.e., mud under the hydrostatic pressure plus the differ-
ential pressure across the bit; i.e., the difference in
pressure between the inside of the tool 50 and the annulus
60. When the pumps are not running, the pressure at
the zone 111 is hydrostatic only; and when the pumps
are running, the pressure is hydrostatic plus differ-
ential. Since the differential pressure is of the
order of 1000 to 2000 psi, a large pressure change occurs
at the zone 111 when the pumps are started up (i.e., an
increase of 1000 to 2000 psi). In Fig. 6A, when the
pumps are not running, zones 112, 113 are at annulus
pressure because tube 114 is connected to the chamber
84 which contains oil at annulus pressure ~see Fig.
4A) and because the orifice 115 interconnects the zones
112 and 113.
Assume now that the pumps are started up. The
pressure in zone 111 then increases substantially (i.e.,
by 1000 to 2000 psi) the piston 116 is pushed downward
compressing the spring 107 (not illustrated in Fig. 6B)
and the high pressure oil in zone 112 pushes the piston
108 downward and compresses the spring 110 (not illustrated).
Thus, when the pumps are started up, the parts of Fig.
6A change to the configuration of Fig. 6B, and both
the pistons 116 and 108 are in the downward position
and the rod 100 is extended downwardly as shown.
-32-
Now because of the orifice 115 and the action of
spring 110, the piston 68 is pushed upwardly with a vel-
ocity determined by the size of the orifice 115, the
spring constant of spring 110, and ~he viscosity of the
oil in the zones 112, 113. This velocity can be easily
controlled and made equal to any desired value; as for
example, a velocity such that the piston 108 will return
to i~s original upward location in about 1 minute. There-
fore, after one minute the arrangement assumes the con-
figuration of Fig. 6C. For the same reasons, when the
pump is stopped the action of the spring 107 and the
orifice 115 will cause the piston 116 to rise back to the
original condition of Fig. 6A.
It can be seen, therefore, that every time the
mud pump is started the rod 100 will move downwards by
the distance d as shown in Fig. 6B and then return to
the normal retracted position. Since in normal drilling
the pump is stopped every time a joint of drill pipe
is added, it follows that every tlme a joint of drill
pipe (usually 30 feet long) is added, the rod 100 will
make a single downward excursion and then return to
its original upward position.
As was pointed out previously, the rod 100 is ar-
ranged so that when it is extended downwardly it pushes
solenoid armature shaft 80 of Fig. 4A downwardly and
closes the valve. Thus, the device of Fig. 6A, 6B,
6C, and 6D i5 a "safety" device; i.e., should the
valve get stuck in the open position because of an
electrical or mechanical malfunction, the valve will be
forced shut after a maximum of 30 feet of drilling.
Fig. 6D shows the engineering drawing of the device
diagrammatically illustrated in Fig. 6A, 6B, and 6C.
In the actual instrument, the device as illustrated in
Fig. 6D is placed in the location 269 of Fig~ 4A. Like
numbers on Fig. 6D represent the elements having like
numbers on Fig. 6A, 6B, 6C, and Fig. 4A.
-33-
~ ~ ~ L~
VI. DESGRIPTION OF ELECTRONIC "FAIL SAFE"
FOR SIGNALLING VALVE
The hydraulic "auto close" system described in
connection with Fig. 6A, 6B, 6C, and Fig. 6D will auto-
matically close the ~alve every time the mud pumps are
stopped and restarted, and thus any mechanical sticking
of the valve can be remedied. TherP is a case, however,
that requires further attention: If the "close" electric
circuitry 103, 109 of Fig. SA were to fail for any reason
(e.g. a burned out solenoid winding) then the valve
would reopen electrically, shortly after the hydraulic
"auto close" device closed it.
Fig. 6E shows an electric system that will inhibi~
operation of the valve in case of an electrical failure
in the downhole apparatus. Sl designates the winding o
the solenoid that "closes" the valve and S2 the winding of
the solenoid that "opens'l the valve. The resistor Rl is
connected in series with the portion o~ the solenoid drive
104 which actuates the "close" solenoid winding Sl.
The resistor R2 is connected in series with the portion
of the solenoid drive 104 which actuates the "open"
solenoid winding S2. These resistors are of very low
ohmic value (about .05 to .2 ohms). It is understood
that the operation of the system described in detail
with respect to Fig. 5A in this specification is as follows:
The "open" electric current pulse is generated first
and is shown diagramatically in Fig. 6E as the pulse
300; the "close" electric current pulse is generated
later (after a time ty) and is sho~n diagramatically
as 301 in Fig. 6E. It must be noted that these electric
pulses 300 and 301 represent the current drawn by the
solenoid windings and not the voltage applied (the
resistors Rl and R2 generate voltage drops Rlil and
-34-
4~
R2i2, and il, i2 indicate the currents through the r~-
spective solenoid windings); consequently, if one of the
windings Sl or S2 is burned out or open circuited, no
current will flow and no corresponding pulse will be
produced (similarly, any other electrical failure will
cause no current to flow through one or both of the re-
sistors Rl, R2).
The magnitude of the time ty of Fig. 6E and the
length of the time tx has been explained and defined
previously in this specification in connection with
Fig. 5A.
The delay of the delay element 302 is equal to ty.
In other words, block 302 produces at its output a pulse,
identical to the input pulse but delayed by the Time ty.
Such delay systems are well known and need no description
here.
Since the delay of element 302 is equal to ty,
the pulse as shown by 303 will be in time coincidence
with the pulse 301.
304 is an anti-coincidence circuit (also known
as an OR gate) and produces at its output 305 an electric
signal only when one of the pulses 301, 303 is impressed
on it, but produces no output when both pulses 301 and
3Q3 are present. 306 is a relay ac~uated by the signal
on lead 305 and is arranged to disconnect the power
to the downhole tool. Thus, if only a "close" pulse
is present (without the "open" pulse~ or if only an
"open" pulse is present (without the "close" pulse),
the power to the downhole power drive is disconnected
then be closed mechanically by the "auto close" hydraulic
-35-
system described in com~ection with Fig. 6D.
~ s an alternate arrangemen~ in Fig. 6E, the relay
306 (which of course can be an electronic switch com-
prising transistors) can be arranged to interrupt the
power only to the circuitry for the "opening" solenoid.
This would have certain advan~ages because the "closing"
circuitry will continue to function, and one of the
objectives is to insure the "closing" of the valve.
Furthermore, an electronic counter 314 may be inter-
posed between the "OR" circuit and the relay circuit306 so that a single electric malfunction will not
disconnect the power. The power will then be dis-
connected o~ly after, for example, 2, 4, or 8 success-
ive malfunctions.
VII. DESCRIPTION OF AUTOMATIC CUT-OFF FOR
SIGNALLING VALVE POWER DRIVE
As has been pointed out previously in this speci-
fication, very fast operation of the valve 40 of Fig.
4A is important. The requisite shock wave will not
be produced if the valve opera~ion is slow. Since the
valve and its drive mechanism contain considerable mass,
substantial power is necessary to open or close the valve
in the time that is considered desirable. This power
is of the order of 1/2 to 3/4 horsepower and can be
provided by a power supply which has been described in
section IV hereof. As in all designs of this nature,
a margin of power is required in order to be sure`that
the valve always opens or closes upon command. The
various electronic "logic circuits" and "power drive
circuits" shown in Fig. 5A are designed to provide
rectangular voltage pulses 104b and 104c that have a
duration of about 40 to 50 milliseconds in order to
make sure that the solenoid windings 105 and 106 are
energized for a sufficient time to ensure the
-36-
opera~ion of the valve. Fi~. 5E shows the voltage
pulse 104b of Fig. 5A in greater detail. At the time
0 the voltage is suddenly applied by the power drive
104 and rises almost instantaneously to the value
shown by numeral 270, remains at this voltage value
for 50 milliseconds, and then is cut of and falls
(again almost instantaneously) to the value 0.
It is very informative to study the motion of ~he
valve by making measurements of the current flow into
the solenoid drive winding and constructing a graph
(see Fig. 5D). From such a graph, the behaviour of the
valve can be quantitatively studied. Figure 5D shows
such a graph in the form of an oscillogram of the
current versus time. (This cùrrent is measured, for
example by the voltage across resistor Rl or R2 of Fig. 6E.)
It is important to understand that it is the current
through the solenoid winding that determines the force
upon the valve stem 68 of Fig. 4A, since ampere turns
determine the electromagnetic pull. Since the windings
of the solenoid have inductance, the current will not
build up instantaneously when a sudden voltage is
applied as in Fig. 5E. If the solenoid comprised a
simple inductor, then the current would build up according
to simple exponential curve 271 of Fig. 5D as shown by
the dotted curve. In actuality something quite different
occurs: When the valve actuates (opens or closes) there
is a sudden motion of the armature of the solenoid 79
of Fig, 4B and a back e~m.f. is generated, This back
e.m.f. is caused by the velocity of the armature that
quickly changes (increases) the inductance of the per-
tinent coil of the solenoid 79. In Fig. 5D, 271 shows
the approximate current versus time curve in the
solenoid winding when the armature of solenoid 79 and
the valve stem 68 is "blocked" in the "open" or "closed"
condition. The solid curve 272 in Fig. 5D shows the
-37-
~314~S
actual current buildup when the valve is not blocked;
i.~., in actual working conditiorl ~opening or closing).
The curves 272 for opening or closings are very similar.
It is seen that curve 272, after the application of
the voltage, gradually rises (since the respective
solenoid coil 105, 106 has inductance) until it reaches,
in the example shown, the value of 4 amperes at the
time T. which is 20 milliseconds. Then there is ~hP
sudden drop of current that reaches the lower value
of 2.2 amperes at the time Tl which is 25 milliseconds.
After the Tl = 25 milliseconds, the current again increases
according to the familiar "exponential" until i~ reaches,
assymptotically, the value of about 10 amperes at-the
time of approximately 60 milliseconds ~this value is
determined by the resistance of the solenoid winding
which in the example given is about 4.7 Ohms).
From a study of the curve 272 in Fig. 5D, it will
become apparent that the valve 40 starts opening or
closing at the time To = ~0 milliseconds and completes
the motion at the time Tl = 25 milliseconds. As was
pointed out previously, an almost identical situation
occurs during the "opening" or the "closing" of the
valve; and the curve 272 would indicate that at the
time To = 20 milliseconds the valve starts its motion,
and at the time Tl = 25 milliseconds the motion is
completed.
It is important to note tha~ the tirne Tl = 25
milliseconds on Fig. 5D is given as a typical example,
and Tl depends on a number of factors. Thus, at high
differential pressures Tl will be greater than 25
milliseconds and could be 30, 35 or 40 milliseconds.
Suffice it to say that the time Tl on Fig. 5D indicates
the time when the valve actuation has been completed,
-38-
~4~S
and the curren~ between the times Tl and 50 milliseconds
is in effect "wasted" since the acutation of the valve
has already been completed. This Pxtra time is a
"safety ~actor" to ensure that, even under adverse
conditions, the valve will alwa~s be actuated when
the voltage pulse is applied.
In accordance with my invention I use the signal
at the time Tl to turn off any fur~her current to the
solenoid 79. Consequently all the current between the
time Tl and 50 milliseconds will be saved (thus reducing
very substantially the total amount of energy required
to operate the valve 40). It must be noted that the full
"safety factor" referred to above is maintained; the
current will continue to be applied until the valve has
completed its (opening or closing) operation.
The electronic circuitry that is employed to
accomplish the above objective is shown by Fig. 5C,
wherein 104 indicates the power drive of Fig. 4A. Be-
tween the power drive 104 and ground is interposed a
resistor (Rl or R2) of low value (compared to the re-
sistance of the solenoid) for example 0.2 ohms. The
voltage across this resistor is, therefore, propor-
tional to the current fed to the particular solenoid
winding 10~, 106. (Two circuits as shown in Fig. 5C are
necessary -- one for the opening solenoid power drive
and a second for the closing solenoid power drive, bu~
for simplicity of illustration, only one circuit is
shown in Fig. 5C.) 273 is a conventional amplifier and
at its output the ~oltage curve 272a of Fig. 5C will
be a replica of the curve 272 in Fig. 5D. 274 is a
derivator (well known in the electronics art) which
generates an output voltage proportional to the first
time derivative of its input voltage. Curve 275 shows
this derivative voltage. It can be seen from observing
curve 272 or 272a that the derivative (slope) of the
-3g-
curve is always positive except during ~he times be-
tween To and Tl, during which time the slope (derlvative)
is negative. On the curve 275 only the impluse 276
is negative. 277 is a conventional receifier arranged
to pass only the pulse 276, as shown on ~he graph 278.
279 is an electronic delay circuit (also well known
in the art) which generates an output pulse 276b which
is a replica of the input pulse but delayed by the
time Tl - To~ Thus, the pulse 276b as shown in graph
280 occurs slightly later than the time Tl. This pulse
276b is applied to an electronic switch 281 that is
arranged to cut off the power to the power drive 104,
thus stopping the current flow almost immediately after
the val~e 40 has completed its operation (opened or
closed). The eleçtronic switch 281 is arranged to re-
store the action of power drive 104 after an appropriate
~ime. The process repeats itself when the next impulse
104a (or 104b) occurs.
It is important to note that the saving in energy
that can be achieved by utilizing this aspect of my
invention can be very substantial. Since very large
powers are required to operate the valve 40 with the
great speed required, this saving is very significant,
and it could in the example shown increase the battery
life by as much as 5 times.
VIII. DESCRIPTION OF SUB AND HOUSING STRUCTURE
FOR SPECIAL T~LEMETRY TOOL
An important characteristic of the Measurement While
Drilling (M~D) apparatus of this invention is its practi-
cality; i.e., convenience and ease of adaptability to
existing oil well drilling hardware and tools and drill
strings. In the attempts of the prior art, large special
steel housings 30 feet or more in length and 8 inches
-40-
~l9 ~Z~5
~n diameter are required ~o house the complicated in-
strtlmentation; and their ~ransportation from location
to location requires specially constructed vehicles.
In the apparatus of this invention, because there is
no valving mechanism interposed in the main mud stream,
it is possible to eliminate the heavy, very long, ex-
pensive special housing (approximately 30 feet long)
and only a short section of drill collar ~called a "sub")
is required. In the practical embodiment of this in-
vention, this sub is only 36 inches long and 6 3/4 inches
in diameter (instead of 30 fee~ which was previously
required).
One of the important features of this invention,
therefore, is that no heavy, long special housings are
required. This is advantageous especially when downhole
magnetic measurements such as compass indications (e.g.,
steering the drilling of a deviated hole) are to be made,
which require use of non-magnetic drill colla s. Non-
magnetic drill collars are not only heavy (2-3 tons)
but also extremely expensive ($20,000. each) since they
must be manufactured of stric~ly nonmagnetic material
such as K Monel. In the construction of the apparatus
of this invention "Standard" API Drill Collars having
outside diameters of 6" to 9" (which are the most common
si~es) are utilized. All of the standard API collars
have an inside diameter of 2-13/16" ~ l/16" - 0". The
simplicity, small size and coaxial construction of the
valve system of this invention and its associated parts
allow a special feature to be accomplished: All of
the pertinent power drive and associated equipment can
be located in a pressure resisting tube sufficiently
small in diameter to permit it to be inserted into the
inside bore (2-13/16") of a standard API Drill Collar
without unduly interfering with mud flow. Some Sensors
should be placed as near to the drill bit as possible.
-41-
In par~icular, a downhole gamma ray Sensor should
be capable oE detecting the penetration of the
bit into a given lithologic ormation as soon as such
penetration occurs. Furthermore, some sensors, such as
a downhole compass-inclinometer require accurate in-
dexing with respect to the "tool face" used in direc-
ional drilling. In addition, a compass-inclinome~er
must be placed at a substantial distance from any mag-
netic or paramagnetic material. Furthermore, when a
compass-inclinometer is employed, the housings 250a
and 250b in Fig. 7A must be carefully indexed angularly
with respect to the sub 253; which in turn is indexed
with respect to the "Bent Sub" used in directional
drilling.
The "bent sub" is equipped with an indexing mark
253a and the angle of this indexing mark must have a
constant and measured angular relationship to the in-
dexing mark 254a that is placed on the telemetering
sub 254. This known angle (representing the angle be-
tween indexing marks 253a and 254a is then introduced
into the computation for the determination of the bear-
ing and angle with respect to a vertical plane of the
"Bent Sub".
Fig. 7A is a schematic showing of the special
telemetry tool 50, illustrating the arrangement wherein
the l'special long tool" is eliminated and only a
short section of drill collar sub is requi.red, as was
previously mentioned. In Fig. 7A, a housi.ng designated
by numeral 250 is made up of an upper sect:ion 250a and
a lower section 250b, as was previously described with
reference to Fig. 4A. The upper section 250a is
contained within a short sub 254 (only about 36 inches
long~. This short sub is especially bored out to
provide an inside diameter (e.g. 4 1/2") sufficient
to house the valve assembly 40 and also to permit the
-42-
unrestricted ~low of drilling mud past upper section
250a through passages 61, which are also designated
by numeral 61 in Fig. 4A. The housing 250a is of small
diameter, preferably, 2 11/16" OD or less. A drill
Collar 255 provided by the user (the oil company or
the drilling contractor) is usually 30 feet long and
of great weight and cost. The inside diameter of a
standard API Drill Collar as was pointed out previously
is 2-13/16" - O + 1/16". Centralizer members ~56 are
provided for lower housing 250b. These are slightly
smaller in diameter than the ID o the standard API
Drill Collar, for example, 2-3/4" O.D. Small clearance
is very important in order to prevent "chatter" when
the tool is vibrated during drilling. Discharge passage
85 is the same as that shown in Fig. 4A. The housing
250b is suspended within the sub 254 by securing means
not shown. The cross-section shape of the centralizers
256, as indicated in Fig. 7B, is such as to provide
slots or passages 258 to permit free flow of drilling mud.
The housing lower section 250b is actually made up
of several sub-sections which are connected, one to an-
other, by a special connector means shown in Fig. 7G.
As shown in Fig. 7C, each sub-section is provided at its
upper end slot 260 and at its lower end a protusion or
tooth 261. A protusion ~.61 of one sub-section matingly
engages a slot 260 of the adjacent sub-section. The
adjacent sub-sections are retained by a connector sleeve
262 which is matingly received by the end portions o the
sub-sections. Circular openings 263 in the sub-sections
are aligned with respective threaded openings 264 in the
connector sleeve 262, and the parts are secured by screws
265. The special connector means of Fig. 7C provides
for accurate angular indexing when sub 253 is a "Bent Sub".
-43-
~ 9~
As was pointed out previously, the angle between
indexing marks 253a and 254a must be known in order to
compute the angle with respect to vertical of the "Bent
Sub?'. It is also necessary that the angular displacement
between the axes of a magnetometer-inclinometer and the
mark 254a be known and invariable during the drilling
operation (it is preferred but not necessary that the
angle between one of the horizontal axes of the magnet-
ometer-inclinometer and the indexing mark 254a be zero).
For this purpose the tool 250b is assembled with angular
indexing teeth 261 as shown on Fig. 7C and Fig. 7A.
In order to design an efficient telemetering system
two requirements will be considered. One of these deals
with optimum conditions for obtaining the regime of hy-
draulic shock waves. The other requirement is concerned
with obtaining shock waves of sufficient intensity to
override extraneous noise effects,
IX. OPTIMUM CONDITIONS FOR DETERMINING THE REGIME
OF HYDRAULIC SHOCK WAVES (DETERMINATION OF
PARAMETERS Kl (or K2) and T(V))
I have performed a series of experiments in order
to determine the optimum conditions for the regime of
hydraulic shock waves.
The occurrence of a hydraulic shock wave is analogous
to that of the water hammer effect. By suddenly stopping
the 10w in a localized ~egion in the line of flow, we
suddenly increase pressure in that region. This initially
localized increase in pressure propagates itself along
the line of flow as "water hammer'l. It is well known that
a sudden and localized change (decrease or increase)
in velocity produces a corresponding localized change
(increase or decrease) in pressure, and conversely, a
sudden and localized change in pressure produces a sudden
and localized change in velocity. Because of elasticity
-44-
and inertia of the fluid, the change is being transmitted
further from the volume element where it origina,tes to
neighboring volume elements with a veloci~y ~hich is
the velocity of propagation of compressional waves.
The problem of propagation of shock waves is of extreme
complexity. To satisfy practical requirements, we need
to determine a parameter which will be the most representative
from the standpoint of obtaining a clearly defined shock
wave. Two parameters will be considered which we designate
as parameter Kl and parameter K2. When either of these
parameters exceeds an appropriate value, a clearly de-
fined shock wave is produced.
(a) Parameter Kl
This parameter is the mean rate of change of the
velocity of mud flow through the bypass valve during the
period of opening (or closing) of the valve:
Let V(t) be the velocity of the mud flow through
the bypass valve as it varies with time (in cm/sec. or
feet/sec.). A~ the instant t = 0 when the valve begins
to open, the velocity is zero; i.e., V(0) = 0). At the
instant t = TaV) when the valve is fully open, the vel-
ocity ~f the valve attains a certain value Vf which is
the bypass velocity during the period of full flow. Thus,
V (T(V)) = Vf ~10)
~5 Consequently the parameter Kl which is mean rate of change
of the velocity during the period TaV) is
Ta
Kl is measured in cm/sec~.
-45-
s
We assume that when Kl exceeds an appropriate
threshold value; i.e., when
1 ~ 1 (12)
we obtain a clearly defined shock wave. In the exper-
iments performed it was determined that
Ml ~ 2 x 105 cm/sec2 (13)
(b) Parameter K2
This parameter represents the mean rate of change
of the area of the opening of the valve during the
periOd T(v)
We previously defined (see equation (1) ) S(t~ as
the area of the valve which is open at the time t. Thus,
at t = 0 one has S(0) = 0 and at t = TaV) one has
S(TaV)) = S0 (14)
where S0 is the total opening of the valve. The parameter
K is
2 K2 = S0 cm2/sec. (15)
We assume that when K2 exceeds an appropriate thres-
hold value; i.e., when
K2 > M2 ` (16)
we obtain a clearly defined shock wave. In experiments
performed, it was determined that M ~ 100 cm /sec.
Roughly speaking, Kl is proportional to K2. The
parameter K2 is perhaps more useful because it tells us
directly how to design and operate the valve.
-46-
There i5 also a parameter T(V) (see BlCl, in Fig.
2A) which needs to be considered in the discussion on
Fig. 8A to 8E. Each of these fi~ures corresponds to
a set of numerical values of Kl and T(V) or K2 and T(V~.
Fig. 8A to 8E show the effect of varying Kl and
TbV3 or K2 and T(V) in effecting the transition from the
regime of slow pressure variation to the regime of hy-
draulic shock waves. More specifically each of ~hese
figures show how the pressure detected and the earth's
surfa~e (ordinate) varies with time, t (abscissa).
The size of the orifice was 0.5 cm2 in these experiments.
Experimental data were obtained at a number of wells.
These wells were selected in Oklahoma, West Texas, East
Texas and in the Netherlands. Moreover, some of the
tests were made on an "experimental well" that was ex-
plicitly drilled to perform telemetry experiments.
In performing the above experiments, account was
taken of the great variety of existing mud pump
installation and of various interfering effects. There
are many kinds of mud pumps: Single Duplex, Double Duplex,
Single Triplex, Double Triplex, and the pump pressure
variations for a given average mud pressure vary a great
deal from installation to installation. Elimination of
the large interfering mud pressure signals is complex.
The pump pressure signals from a single Duplex System can
be 10 or even 20 times larger than those from a carefully
adjusted Double Triplex. For determination of the optimum
values of K2 ~or Kl) and of T(V), the drilling opera-
tion was stopped and a very good (smooth) Tiplex pump
was used. Thus, graphs in Figs. 8A to 8E are not representative of a typical condition but represent a
condition where the various noises (from the pumps and
other sources) were minimized and then averaged out by
calculation and drafting in order to obtain optimum
values for the paramters K2 (or Kl) and T(V). The
-47-
~L~942~5
.,
corresponding va~ues of K2 (or Kl) and T(V~ for each
of Figs. 8A to 8E are given in the table which follows:
T A B L E
K2 (in cm2/sec) T(V) ( in seconds)
5 Fig. 8A .5 2
Fig. 8B 2.5
Fig. 8C 8.5 - 0.5
Fig. 8D ~ S~ 0.25
Fig. 8E 100 0.1
, .
~The graphs of Fig. 8A to 8E represent average numbers
obtained in a large number of tests. In these tests the
normal standpipe pressure was 3000 psi and variations of
pressure were in a range oE 100 psi. The above tests
were made using various types of valves: motor driven,
rotary,~poppet, etc. Fig. 8F is an exact replica of
the standpipe pressure recorder chart obtained in tests
conducted at 9800 feet with standpipe pressure ~800 psi
and perormed in an actual drilling of an oil well in
West Texas.
The graph in Fig 8A was obtained by using a slow
acting valve. The numerical val~es of the pertinent
parameters in Fig. 8A were K2 = 0-5 cm2/sec and T(V)
= 2 sec; i.e., they were similar to those suggested
in the prior art as in Figs. lA and lB. Consequently,
both Fig. 8A and Fig. lB represent the regime of slow
pressure pulsation. On the other hand, Fig. 8E was
obtained using a fast acting valve for which K2 = 100
cm2/sec and TbV~ = 10 1 seconds. Consequently, Fig.
8E represents the regime of hydraulic shock waves,
and the valve wavelet in Fig. 8E is very similar to
the valve wavelet in Fig. 2B.
-48-
~9~ 35
Figs. 8B, 8C, and 8D as defined in the table above
show the transition from the regime of slow variations
of pressure to the regime of hydraulic shock waves.
In the tests shown in Figs. 8B, 8C, and 8D the
conditions were kept as similar as possible. The in-
strument was located near the bottom of clrill holes o
depth of about 8000 feet, the mud viscosity was about
40 Funnel, and the weight 12 pounds per gallon. The valve
when "open" had an eff~ctive open area of 0.7 cm2. The
normal standpipe pressure was 3000 psi and the valve used
in these tests was similar to the valve 40 but modified
to permit slower-action (without the bi-stable action);
i.e.,`the valve was a simple pressure balanced valve,
and the flow ra~e was controlled by a restriction at
the inlet passageway. It should be noted that the valve
action which accounts for Fig. 8B was quite fast,
but it did not produce the desired regime of hydraulic
shock waves. The sharp onsets, however, indicated that
faster action is desirable. The discharge rate was
of the order of 5 gallon/sec2,
By adjusting the inlet restriction and the outlet
restriction and the electric power supplied to the drive
solenoids, various valve operation speeds werP ob~ained.
It is seen from the above that no shock waves are
produced when K2 = 0 5 cm2/sec, and an almost ideal shock
wave is produced when K2 = 100 cm2/sec.
X. OBTAINING SHOCK ~AVES TO OVERRIDE NOISE
(ALTERNATIVE VERSION)
I will introduce another paraMeter which will
express a requirement regarding the intensity o the
shock wave. Two different approaches will be considered.
One of these is based on a parameter K3 which represents
-49~
the amount oE mud ~measured in cm3 ~ in g~llons) which
passed through the ~Tal~7e during ~he period T~V~ (This
quanti~y is known as fluence). The other approach is
based on a parameter K4 which represents the average
flux of the mud stream during the period T(~ Thus,
amount mud passed during Ta (17)
Consider the period of opening of the valve; i.e.,
the period T(V). To simplify the problem we assume that
the rate of increase of the velocity of flow during the
period T(V) i5 constant and is equal to Kl. Thus,
V(t) = Kl t in cm/sec. (13)
Assume also that the rate of increase of the opening
o the valve is constant, and equal to K2. Thus,
S(t) = K2 t in cm2/sec. (19)
Consequently the volume that passes through the valve
during the time TaV) is
,T(V) ~KlK2T(V))3 C1ll3
K3 J KlK2t dt 3 a (20)
Thus, parameter K3 is the amount in ~m3 of ~luid ~hat passed
throu~h the valve during ~he period T(V). This is the
amount of fluen~e for the period of a single opening
and closing of ~he valve. Another alternative is to take
instead of the parameter K3 a parameter K4 representing
-50-
4~5
the ~lux for the period TaV~; i.e.,
K3
Ta (21)
~5~, g~
XI. GENERAE PROCEDURE FOR NOISE ELIMINATION
Consider now the general procedure for decoding
signals from the pressure transducer 51. Fig. 9 shows
the apparatus arrangement,and Figs. 10A to 10G illus-
trate certain wave forms and pulses which are in~olved
in the decoding of signals by means of the arrangement
o Fig. 9.
The signal obtained from the pressure transducer
51 comprises a useful information carrying signal to~
gether ~ith interfering signals which tend to obscure
Ol- mask the useful signal. The useful information
carrying si.gnal represents the coded message obtained
by means of the valve 40 in response to a sensor. There
are various interfering signals. One of these produced
by the pump 27 contains an intense steady component of
the mud pressure produced by ~he pump. This component
accounts for the circulation of the mud through the drill
2Q pipe and back through the annulus between the drill pipe
and the wall of the bore. Superimposed th~reon is an
alternating component produced by the recurrent motion
o~ the reciprocating pistons in the pump.
In order to improve reception, it is desirable to
remove from the output of the transducer 51 the steady
component in the pressure generated by the pump 27.
Accordingly, a frequency selective filter 150 is connected
to the transducer 51 for transmitting frequencies in a
range from 0.1 to 10 Hertz and attenuating frequencies
outside of this range. The frequencies contained within
the steady pressure component are below 0.1 Hertz.
-51-
3s
In the terminology applied to t~is specification,
a distinction will be made between the term "filterl' as
applied to the frequency selective filter 150 and the
term "digital filter" to be used later in the description
of my invention. In a "filter" such as the ilter 150,
conventional filtering is performed by means of analog-
type electronic networks, whose behavior is ordinarily
treated in the frequency domain. The term '7filter" may
be used to designate what is known to be a "wave filter",
a "Shea filter" (see for instance, "Transmission Networks
and Wave Filters" by T.E. Shea, D. VanNostrand Co., New
York, N.Y., 1929) and other filters such as Tchebyshev
and Butterworth filters. On the other hand a dlgital
filter such as a matched filter, a pulse shaping filter
or a spiking filter is more fruitfully treated in time
domain. The output of a digital filter is obtained
by convolving the digital i~put trace with the filter's
weighting coefficients. A digital filter is a computer.
The signal produced at the output terminal 151 of
the filter 150 will be expressed by a function F(t)
which is
F(t) = B(t) ~ P(t) -~ U(t) (22)
where B(t) is the useful information carrying signal,
P(t) is the interfering signal caused by the periodic
pressure variation from the pump (pump noise), and U(t)
represents random noise. The random noise is produced
by various efects such as the action of teeth of a
cutting bit (as toothed core bit) during drilling, by
the gears in the mechanical drill string9 and by other
devices involved in rotary drilling operations. In
some cases U(t) approximates white noise, although in
o~her cases the departure of U(~) from white noise
may be substantial.
-52-
The coded message expressed by the information
carrying signal B(t~ is a series of binary words and
each of these binary words contains a succession of
bits. A single bit in a binary word is produced by
a single "operation" (i.e., by a single opening and
closing) of the valve 40. Such a single operation generates
a hydraulic shock wave which manifests itself at the
surfac~ of the earth as a single valve wavelet, such
as the valve wavelet in Fig. ~B. Consequently, the
message expressed by B~t) is in the form of a coded
sequence of valve wavelets, each of said valve wavelets
being of the type shown in Fig. 2B. Figs. lOA to lOG
show various steps to be accomplished in order to
separa~e the information carrying signal B(t3 from the
interfering signals. In order to facilitate explana-
tion, I have expressed B(t) in Fig. lOA by a single
valve wavelet rather than by a coded succession
of valve wavelets. Thus, the valve wavelet in Fig. lOA
is of the same type as a single valve wavelet in Fig. 2B.
There is however, a slight change in no~ation. I eliminated
in Fig. lOA the superscript "s" which appeared in Figo
2B. Thus, in Figs. lOA to lOG various times have been
designated as tl, t2, . . tl5, tl6
script "s'1. Various graphs in Figs. lOA to lOG have
been appropriately labeled in these figures. In the
interest of clarity and to facilitate explanation, the
time scales corresponding to these graphs have been
distorted.
In order to eliminate the interfering noise signals
(pump noise and random noise) and to produce a signal
representing the coded message, three successive operational
steps are provided which may be identified as follows:
-53-
- .
2~S
S~ep 1: In this step a signal having three com-
ponents as shown in Fig. lOA is transformed into a signal
having two components as shown in Fig. lOC. The purpose
of this step is to eliminate the pump noise P(t). As a
result of this step a "valve wavelet" of the type shown
in Fig. lOA is transformed into a "double wavelet".
Such a double wavelet is shown in Fig. lOC.
Step 2: The purpose of this step is to eliminate
the random noise signal.
lO Step 3: In this step each double wavelet as shown
in Fig. lOD is transformed into a single pulse as shown
in Fig. lOG. Consequently, one obtains a coded sequence
of single pulses representing in dig~tal format the
parameter measured by the sensor 101 at an appropriate
lS depth in the borehole.
XII. ELIMINATION OF PUMP NOISE ~STEP NO.l)
Consider now Fig. lOA. This figure shows the three
components of the signal F(t) as expressed by the equation
(22). These are: the valve wavelet B(t), the pump noise
P(t), and the random noise U(t~. As previously pointed
out, the signal F(t) has been obtained by means of the
filter 150. This filter is connected to a delay element
152 which is effective in delaying the input signal F(~)
by an amoun~ Tp where Tp is one period in the oscillation
produced by the pump 27. Thus, the signal obtained at
the output lead 153 of the delay element 152 can be ex-
pressed as F(t-Tp). The three components of the signal
F(t-Tp) are shown in Fig. lOB. They are: the delayed
valve wavelet B~t-Tp), the delayed pump noise P(t-Tp)
and the delayed random noise U(t-Tp). The time interval
Tp depends on the speed of rotation of the pump, and
-54-
~ ~ 9L~L2~i5
since the speed of the pump is not constant, the delay
Tp is a variable delay. Thus, an appropriate control for
the delay element 152 must be provided in response to
the speed of rotation of the pump 27. Accordingly ~he
5 delay element 152 is arranged to receive through the lead
154 timing impulses obtained from a pulse generator 155,
which is mechanically driven by the pump to produce an
appropriate number of pulses per revolution of the pump.
A chain drive transmission 156 is provided for this
purpose. The delay element 152 may be Reticon Model
SAD-1024 Dual Analog Delay Line as marketed by Reticon
Corporation, Sunnyvale, California, U.S.A.
Assume that the pump 27 produces Nl strokes per
second. Thus, Tp = l/Nl. The pulse generator 155 pro-
duces ti~;ng pulses at a relati~ely high rate N2, which
is a multiple of Nl. Thus, N2 = KNl, where K is a constant
which has been chosen to be 512. Thus if the strokes
of the pump are one per second, this w~uld require ~he
signal generator to produce 512 pulses per second.
It is apparent that the rate of pulsation of the mud
pump 27 varies with time and, accordingly, N2 will
vary so as to insure that the delay produced by delay
element 152 will always be equal to one period of the
mud pressure oscillations produced by the mud pump 27.
The signal F(t - Tp) derived from the delay element
152 is applied to an input lead 153 of a subtractor 160.
The subtractor 160 also receives at its input lead 161
the signal F(t) derived from the filter 150, and it
produces at i~s output lead 162 a difference signal
which is
x(t) = F(t) - F(t-Tp)
= B~t)-B(t-Tp)-tP(t)-P(t-Tb)+U(t)-U(t-Tp) (23)
-55-
2~3~
Since P(t) is periodic and has a period Tp, one has
P~t) - P~t - Tp) = O (24)
Thus, because of the periodicity of pulsations produced
by the mud pump 27, the noise due to the pump has been
eliminated and the signal obtained at the output lead
162 of the subtractor 160 can now be expressed as
x(t) = b(t) ~ u(t) (25)
where
b(t) = B(t) - B(t - Tp) (26)
is the information carrying signal and
u(t) - U(t) - U(t - Tp) (27)
is a random noise signal.
Both the information carrying signal b(t) and the
noise signal u(t) are shown in Fig. 10C. It can now be
seen that by performing the step No. 1 as outlined above,
I have transformed the information carrying signal B(t)
as shown in Fig. 10A and having the form of a valve
wavelet into a different information carrying signal
b(t) as shown in Fig. 10C. The signal b(t) will be
referred to as a l'double wavelet" in contrast to the
signal B(t) which represents a l'valve waveletll. A dou-
ble wavelet comprises two valve wavelets such as the
valve wavelets 'IA'l and "B" in Fig. 10C. These valve
wavelets are separated one from the other by a time
interval Tp. The valve wavelet "A" is similar to that
of Fig. 10A whereas the valve wavelet "B" represents
the inverted form of the valve wavelet "A".
The signal x(t) (equation (25)) is further applied
to the input lead 162 of an analog to digital (A/D)
converter 163 which is controlled by a clock 178. One
obtains at the output lead 164 of the A/D converter a
signal expressed as
xt= bt + Ut ~28)
-56-
where in accordance with the symbolism used here xt,
bt, and ut are digital ~ersions of the analog signals
x(t), b(t~, and u(t) respectively. The signals xt and
Ut are in the form of time series, such
x = ( . . x 2' x_l, xO, xl~ g
and
Ut ( . U-2' u_l. u0, ul~ u ) (30)
respectively, and the signal bt is finite length wavelet
b = (bo, bl, b2, ., bn) ~31)
XIII. ELIMINATION OF RANDOM NOISE WHEN RANDOM
NOISE IS WHITE (STEP NO. 2)
The mixture of a double wavelet bt and of noise
signal ut is now applied to a (n + 1) length digital
filter 170 having a memory function
at (aO' al, a2, , an) (32)
I select in this embodiment a digital filter known as a
matched fllter, and I choose the memory function at to
optimize the operation of the filter. Optimum conditions
are achieved when the signal to noise rat:io at the output
of the filter 170 is at its maximum value. (For a des-
cription of matched filters, see for instance a publi-
cation by Sven Treitel and E.A. Robinson on "Optimum
Digital Filters for Signal-to-Noise Ratio Enhancement",
Geophysical Prospecting Vol. 17, No. 3, 1969, pp. 2h8-293,
or à publication by E.A. Robinson on '1Statistical
Communication and Detection with Special Reference ~o
Digital Data Processing of Radar and Seismic Signals",
Hafner Publishing Company, New York, N.Y., 1967, pages
250-269.)
-57-
s
I arrange the memory func~ion at of the matched
filter 170 to be controllable so as to insure at all
times during the measuring operations the optimum con-
ditions in the operation o the filter. The control
of the filter is effected by means of a computer 172
which receives appropriate data from a storage and re-
call element 173 in a manner to be described later.
A signal Yt obtained at the output lead 174 of
the matched filter 170 can be expressed as a convolu-
tion of the input function xt and the memory function
at. Thus,
Yt t at aOXt ~ alXt-l + - ~ anXt (33)
where the asterisk indicates a convolution. Substituting
in (33), xt = bt ~ Ut one obtains
Yt = Ct + Vt (34
where
Ct - bt at
is the filter response to a pure signal input and
Vt Ut at (36)
is the noise output. A schematic block diagram indicating
these relationships is shown by Fig. 11.
In order to insure the optimum conditions in the
operation of the matched filter 170, a certain ~ime,
say time t = to is chosen and it is required that the
instantaneous power in the filter output containing a
signal at time t = to be as large as possible relative
to the average power in the filtered noise at this in-
stant. Hence, in order to detect the signal ct in ~he
filtered output ut, use is made of the signal-to-noise
ratio defined as
-58-
vs
(Value of filtered signal at time to)2
= (37)
Power of filtered noise
If one convolves the tn ~ 1) length signal (bo~ bl,
..., bn) with the (n + 1) length filter, one obtains a
(2n -t 1) length output series (cO, Cl, ... cn, ....
c2n 1~ c2n~ where cn is the central value of this outpu~
series. Thus, at time to = tn, ~ becomes
c~ (aObn + albn_l + .. + anbO)~
{Evn} E{vn} (3~)
where E{v2} is the average value of the noise output
power.
-59-
I assume here that the random noise ut is white
noise. Then it can be shown (see for ins~ance a pub-
lication by Sven Treitel and E.A. Robinson "Optimum
Digital Filters for Signal to Noise Ratio Enhancement".
Geoph. Prosp., Vol. XVII, ~3, 1969, pp. 240-293) that
the maximum value of the signal to noise ratio ~ can
be obtained when
(aO ~ a~ , an) = (kbn , kbn_l, . . ., o
where k has been chosen to be one. Thus, in the presence
of a signal immersed in noise, when the noise is white
the optimum conditions can be obtained when the memory
of the filter is given by the inverted signal; namely,
by the coefficient sequence (bn, bn l~ ~ bo)
The memory of the filter 170 is determined at all
times by means of the computer 172 which is connected
to the filter by a channel 175. The term "channel" as
used herein refers to suitable conductors, connections
or transmission means, as a particular case may require.
A storage and recall element 173 is provided for storing
the function bt for a subsequent transmission of bt
through channel 176 ~o the computer 172. The function
of the computer is to reverse the input data expressed
( 0~ bl, ~ bn) so as to provide at its
output channel 175 a sequence (bn, bn 1~ ~ bo) which
in turn is applied to the matched filter through the
channel 175 and stored therein as a memory of the filter
in accordance with the equation (39).
The filtered output Yt obtained at the output lead
174 of the matched filter 170 is further applied to a
digital to analog (D/A) converter 181. Since Yt represents
a signal in digitalized form, the corresponding analog
function obtained at the output lead 182 of the D/A
-60-
:~L1 94,'2~5
converter 181 will be expressed in accordance with the
symbolism used here as ytt).
It should be noted that maximizing the signal to
noise ratio in the filtered outpu~ y~ is equivalent to
minimizing the noise signal vt (or its analog equivalent
v(t)) when compared to the information carrying signal
Ct (or its analog equivalent c(t)). Thus,
v(t) ~ c(t) (40)
and
y(t) ~ c(t) ~ b(t) (41)
Therefore the output function y(t) of the ma~ched
filter as shown in Fig. lOD closely resembles the fun-
ction b(t) shown in Fig. lOC.
An important feature of my invention involves storing
(for a subsequent reproduction) the func~ion bt by means
of the storage and recall element 173. The required
procedure for storing bt will now be explained in con-
nection with Fig. 12. The procedure consists of sev-
eral steps as follows.
Step (a). Drilling operations are stopped; i.e.,
the bit 31 is lifted a short distance above bottom,
the draw works are maintained stationary, and the
rotary 21 is stopped from turning.
Step (b). Pump 27 continues to operate as during
normal drilling procedures; i.e., at a uniform pump
rate and a pump pressure representative of that used
during the actual "measurements while drilling". All
other sources o~ interference, such as A.C. electric
pick-up from generators, operation of cranes, etc.
are stopped. "Heave" and other sources of noise are
eliminated insofar as possible in the case of offshore
operations (as by selecting a calm day).
-61-
9~2~3~
Step (c). As previously described and shown in
connection with Fig. 5A, the subsurface encoding is
determined by a "clock" which is in rigorous synchron-
ism with the "clock" at the surface equipment. Conse-
quently it is possible at the surface to make a deter-
mination o~ when a single pulse is generated at the
subsurface, for example the precursor pulse; and by
knowing the velocity of transmission through the mud
column, also to know the exact time at which the hy-
draulic impulse is recei~ed at the surface. Thus it
is possible to receive a single ~'wavelet" at the-sur~
face and to know in advance when in time the-waYelet
appears, even though it could be obscured by noise.
(In many cases the single wavelet will stand out over
the noise so that visual observation on an oscilloscope
is practicable.) Thus the hydraulic transient genera~ed
by the valve 40 is received by the transducer 51
at a known time.
Step (d). The signal obtained by ~he transducer
51 is transmitted through the filte~ 150 to selectively
transmit frequencies in ~he range from .1 Hertz to 10
Hertz. Because the drilling operations have been stopped
(as outlined in the step (a) above), the random noise
U(t) is negligible, and consequently the signal obtained
at the output of the filter 150 is of the form F(t)
B(t) + P(t).
Step (e). The signal F(t) obtained Xrom the filter
150 is passed through the delay element 152, subt~actor
160, and A/D converter 163 in a manner which has been
explained above. Usually when drilling is in progress,
the signal obtained at the output of the A/D conver~er
163 is of the form xt = bt ~ Ut where ut accounts for
the noise due to drilling operations. However, here
-62-
again, because the drilling operations have been stop-
ped, the random noise signal ut is negligible. Under
these conditions one has xt~, b~. The signal xt represents,
as nearly as practicable, a "noiseless signal" which
would correspond to a wavelet representing the information
carrying signal.
S~ep (f). The function xt~ bt is stored.
The operational steps (a), (b), ~c), (d), (e), and
(f) as outlined above are performed by means of the
operational arrangement as shown in Fig. 1~. In this
arrangement~ the output of the A/D converter 163 is
applied to the input of the storage and recall element
173 for recording of Xtf~ bt.
It should be noted that in frequency domain the
lS memory function at of the matched filter 170 can be
expressed as
A(f) = e~2~vfn B*(f) (42)
where f is frequency and B*(f) is the complex conjugate
of the Fourier transform of the signal bt.
Elimination of random noise when the noise is
white can, in some instances, be accomplished by means
of an autocorrelator rather than by means of a matched
filter, such as the matched filter 170 in Fig. 9. To
do this the schematic diagram of Fig. 9 should be modi-
fied by eliminating the matched filter 174, computer
172, and storage and recall element 173. Instead, an
autocorrelator would be used. The input terminals of
the autocorrelator would be connecLed to the output
leads 164 of the A/D converter 163. At the same time
-63-
\
the output leads of the autocorrelator would be connected
to the input leads 174 of the D/A converter 181. The
output of the D/A converter may be processed by means
of the delay element 190, polarity reversal element 192,
AND gate 193, etc., as shown in Fig. 9. However, in some
instances the output of the D/A converter may be applied
directly to a recorder.
XlV. T~ANSFORMATION OF A CpDED SEQUE~CE OF DOUBLE
WAVELETS I~TO A CODED SEQU~:NCE OF SHORT
PULSE5 (STEP NO. 3)
Consider again the operational arrangement of Fig.
9. I have now obtained at the output lead 182 o the
D/A converter 181 a signal represented by a function
y(t) which is shown in Fig. lOD and which has a shape
similar to that of the double wavelet b(t); i.e.,
Y(t)~V b(t).
The function y(t) r~ b(t) represents a single bit
in the digitalized signal which operates the valve 40.
It is apparent that such a function is not very con-
venient for representing a very short time intervalcorresponding to a single opening and closing of the
valve 40. It is therefore necessary as outlined in the
Step No. 3 to transform a double wavelet into a single
short pulse coincident with the operation of the valve.
I accomplished this objective by means of a delay
element 190 controlled by a clock 191 in combination
with polarity reversal element 192 and an AND gate 193;
(coincidence network) arranged as shown in Fig. 9.
The delay element 190 receives through lead 182 the
3`0 signal y(t) from the D/A converter 181. This delay
element is controlled by a clock 191 so as to obtain
-64-
- ~ \
~9~
at ~he output lead 195 a delay e~ual to time interval
Tm. The delayed function b~t - Tm) as shown in Fig.
lOE is further applied through lead 195 to a polarity
reversal element 192 to produce at ~he output lead 197
of the element 192 an inverted delayed double wavelet
expressed as -b(t - Tm) and shown in Fig. lOF.
The signal -b(t - Tm) is applied through ~he lead
197 to the AND gate 193. At the same time, the signal
b(t) deri~ed from the D/A converter 181 is applied
through the leads 18~ and 200 to the AND gate 193. Each
of the signals b(t) and -~(t - Tm~ includes pulses which
have positive and negative polarity. By comparing the
signal b(t) as in Fig. lOD with the signal -b(~ ~ Tm)
as in Fig. lOF, it can be seen that the~e is ~nly one
pulse in Fig. lOD which is in time coincidence wi~h the
pulse at Fig. lOF. This pulse occurs at the time interval
from t3 to t4 in Fig. lOD and from tg to tlo in Fig. lOF.
We note that the instants t3 and tg are coincident since
t3 = tl + Tm and tg = tl ~ Tm. Similarly the instants
t4 and tlo are coincident since t4 = tl + Tn + Tm and
tlo = tl ~ Tn + Tm Consequently a single coincidence
pulse is derived from the double wavelet b(t) and is
shown in Fig. lOG. Accordingly the AND gate 193 which
received at its input leads 200 and 197 signals repre~
senting the function b(t) and -b(t - Tm), respectively,
generates at its output lead 210 a single pulse as
shown in Fig. lOG.
It should be recalled that in the interest of sim-
plicity I have shown in this embodiment a single pulse
which is produced and is substantially coincident with
-65-
~L~9~
a single opening and closing of the valve. One should
keep in mind that in actual drilling and simultaneous
measuring operation one obtains at the output lead 210
a coded succession of single pulses which represent a
measurement performed by a selected sensor of a selected
parameter.
The coded succession of single pulses obtained at
the output lead 210 of the AND gate 193 is applied to
a D/A converter 211 which is controlled by a clock 212.
At the outpu~ lead 214~of the DlA converter 211, one
obtains in analog form a signal representing the measure-
ment of the selected parameter. This signal is recorded
by means of the recorder 54.
XV. USE OF CROSS CORRELATION
In another embodiment of my invention, a cross
correla~or can be used instead of a matched filter for
noise elimination. There is a close analogy between
convolution of two functions, as sho~n by means of the
equation (20a), and cross correlation. Cross correla-
ation of one function with another produces the same
result as would be produced by passing the first function
through a filter (matched filter) whose memory function
is the reverse of the second function. (See for in-
stance a publication by N.A. Anstey, "Correlation
Techniques--A ReviPw," Geoph. Prosp, Vol. 12, 1964,
pp. 355-382, or a publication by Y.W. Lee on
"Statistical Theory of Communication," John Wiley and
Sons., New York, N.Y., 1960, p. 45.)
I show in Fig. 13 how the same operations which
can be performed by a matched filter may also be per-
formed by a cross correlator 200. The cross correlator
-66-
200 is provided with two input terminals 201 and 202
and an output terminal 203. The signal xt derived
from the A/D converter 163 is applied to the input terminal
201, whereas the signal bt derived from the storage
and recall element 173 is applied to the input terminal
202. A signal representing the cross correlation of
Xt and bt is thus obtained ak the output lead 203.
It is easily seen that the cross correlation signal
obtained at the output lead 203 is identical to the
convolution signal Yt as expressed by equation ~33)
and produced in Fig. 9 by the matched filter 170. The
cross correlation signal is futher processed as shown
in Fig. 13 in the same manner as the signal obtained
by means of the matched filter 170 was processed in the
arrangement of Fig. 9. The cross correlator 200 may
be of the type known as Model 3721A as manufactured
by Hewlett Packard Company of Palo Alto, California.
XVI. ELIMINATION OF RANDOM NOISE I~HEN RANDOM
NOISE IS NOT WHITE NOISE
When random noise is white noise, the auto cor-
relation qt of the noise function is zero for t ~ 0.
Consider now the case when the unwanted no-ise ut has
a known auto correlaticn function qt, where the co-
efficients qt are not necessarily zero for t ~ 0.
This is the case of "auto correlated noise," in
contradistinction to pure white noise, whose only non-
vanishing auto correlation coefficient is q0. An appro
priate form of a matched filter and related components
is shown in Fig. 14. In this case it is required to
-67-
store not only the information carrying signal bt
(as by means of the element 173) but also the noise
signal ut. Accordingly, the arrangement of Fi~. 14
includes two storage and recall elements which are 173
and 224. The storage and recall element 173 performs
a function which is identical to that of the element
designated by the same numeral in Figs 9.and 12. It
serves to store and subsequently to generate the fun-
ction bt. On the other hand the function of the storage
and recall element 224 is to store and subsequently
reproduce the nbise function ut. The data representing
the functions bt and ut obtained from s~orage elements
173 and 224 re~pectively, are applied through channels
225 and 225, respectively, to a computer 228. The
function of the computer 228 is to transform the input
received from the înput channels 225 and 226 into data
required to determine the memory func~ion of the matched
filter 220. The latter data are applied to the matched
filter 220 through the channel 230.
The notation now will be the same as beore,
except that one must now bear in mind that the noise
Ut is no longer white noise. The matched filters to
be discussed here are indeterminate in the sense of an
arbitrary amplification factor k, which is set equal
to unity for convenience~
The same definition of the signal to noise ratio
will be used~ Thus
c2 (43)
E{Vn}
It is desired to maximize ~ subject to the assump-
-68-
tion that the input noise ut is of the auto correlated
kind. It will be convenient to introduce matrix.nota-
tion at this point. Let
a = ~aO, al. --, an)
5 designa~e the (n ~ 1~ row vector which characterizes
the memory of the matched filter 220. Furthermore, let
b = (bn~ bn_l~ .--, bo) (45)
be the (n ~ l) row vector that defines the time reverse
of the signal bt and let
~= ' .. .
_qn q O .
be the (n ~1) by (n + 1) auto correlation matrix o the
noise. Then one can write
~ = (ab ) (ba') (46
where the prime (') denotes the matrix transpose.
In order to m~;m; ze ~ the quantity (46) will be
differentiated with respect to the filter vector a and
20 the result will be set as equal to zero.
A relationship is obtained
qa' = b' (47)
which can be written out in the orm
qO -- qn aO bn
al bn_l
= (48)
qn _ - - bo
-69-
~19~5
This is the matrix formulation of a set of (n + 1)
linear simultaneous equations in the (n ~ 1) unknown
filter coefficients (aO, al, ..., an). It's solution
yields the desired optimum matched filter in the pre-
sence of auto correlated noise. The equation (48) may
be solved by the Wiener-Levinson recursion technique
(See N. Le~inson "The Wiener RMS error criterion in
Filter Design and Prediction," Jour of Mat. and Phys.
1947, v. 25, pp. 261-278 and S. Treitel and E.A.
Robinson, 'ISeismic Wave Propagation in Terms of Commun-
ication Theory", Geophysics, 1966, Vol. 31, pp. 17-32~.
This recursive method is very efficient, and it is thus
- possible to calculate matched fil~ers of great length
by means of the computer 228. The known quantities in
this calculation are the noise auto correlation matrix
q and the time reverse of the signal wavelet bn t while
the unknown quantities are the filter coefficients at.
These filter coefficients represent the memory function
of the matched filter 220.
The computations which are required to determine
the memory function of the matched filter 220 are per-
formed by the computer 228. The computer receives from
the storage and recall elements 173 and 224 data regarding
the functions bt and ut respectively. Upon the reception
of qt the noise auto correlation matrix is calculated,
and upon the reception of bt the time reverse of this
signal is determined. Subsequently the unknown filter
coefficients at are calculated and then transmitted
through the output channel 230 to the matched filter 220.
The output of the matched filter 220 is applied
to a D/A converter 181 and further processed in the
szme manner as the output of the matched filter 170
was processed in the arrangement of Fig. 9.
-70~
s
In ~requency domain, the memory function of the
matched filter 220 can be expressed as
A~f) = e~~fn ~ ~49)
where B*(f) is the Fourier transform of the time reverse
of the signal b = (bo~ bl, ..., bn) and Q(f) is the power
spectrum of the noise in the in~erval (f + df). The
physical meaning of the expression (49) is simple.
The larger the amplitude spectrum ¦B(f)¦ of ~he signal
and the smaller the power density spectrum Q(f) o~ the
noise in the interval (f, f + df), the more the matched
filter transmits frequencies in that interval. Thus,
if the power spectral density Q(f) of the noise is small
in some interval of the frequency band occupied with the
signal, the matched ~ilter is essentially transparent
(attenuates very little) in this inter~al.
Consider now the signal storage and recall elements
173 and 224. The procedure required to store the signal
bt by means of the element 173 has been previously de-
scribed in connection with steps (a) to (f) as performed
by means of the arrangement of Fig. 12.
A different approach is required in order to storethe noise signal ut by means of the element 224. As was
pointed out previously in connection with Fig. 12, it
is possible to receive and store a "noiseless signal".
Similarly, because of the synchronism bet~een the
subsurface and surface l'clocks", it is possible to
receive and store "signalless noise"; i.e., the signal
received by the transducer 51 during its normal drilling
operation (containing all the noises incidental to this
drilling operation but no information carrying signal).
In this case the arrangement of Fig. 12 can also be
-71-
used to illustrate the required procedures. The
steps for obtaining a record of the function u(t)
can be stated as follows:
Step ~). Weight on the bit is applied and normal
drilling operations are conducted.
Step (~). A time is chosen when no information
carrying signal is present; i.e., a pause between
binary words.
Step (y). A signal is obtained which represents
lQ pressure variation of the drilling fluid at the trans-
-72-
ducer 51. This signal is transmi~ed through the filter
150. Because of the time chosen in step ~) abo~e,
the signal b(t~ is nonexistent, and consequently, the
signal obtained from the output of the filter 150 has
the form F~t~ = P(t) ~ U(t).
Step (~). Pump nois~s signal P(t3 is elimina~ed.
This is accomplished by means of the delay element 152
and subtractor 160. Then the resultant signal is ap-
plied to the A/D converter 163. Since no information
carrying signal is present, bt = . and consequently,
the signal obtained from the output of the A/D
converter 163 has the form xt = ut.
Step (~). A record of the function xt = ut is
obtained by utilizing the storage and recall element
224 at the output of the A/D converter 163, as shown
in Fig. 12.
Summarizing what has been said above, it can be
seen that if the noise is white, then ~he matched filter
170 and related components as shown in Fig. 9 guarantee
the optimum value o signal to noise ratio ~. If the
noise is not white but has a known auto correlation
function, then the matched filter 220 and related com-
ponents as shown in Fig. 14 guarantee the optimum
value of ~. ~
Z5 XVII. PULSE S~PING FILTER OF WIENER TYPE
Fig. 15 shows a portion of the surface equipment
comprising a filter which operates on a principle which
is different from that of the matched filter in Fig. 9
-73-
or in Fig. 14. The matched filter in Fig. 9 or Fig.
14 is optimum in the sense that it is a linear ilter
that optimizes the signal to noise ratio. On the other
hand the filter 240 in Fig. 15, designated as a pulse
shaping filter or Wiener filter, is optimum in the
sense that it is a linear filter that minimizes ~he
mean square difference between a desired output and
an actual outpu~. (For a description of such a filter
see3 for instance, the publication by E.A. Robinson
and Sven Treitel on "Principles of Digital Wiener
Filtering," Geophysical Prospecting 15, 1967, pp. 312-
333 or a publication by Sven Treitel and E.A. Robinson
on "The Design of High-Resolution Digital Filters,"
IEEE Transactions on Geoscience Electronics, Vol. GE-
4, No. 1, 1966, pp. 25-38.~
The pulse shaping filter 240 in Fig. 15 receives
via its input channel data regarding the functio~
Xt = bt + Ut derived from the A/D converter 163. The
pulse shaping filter is an (m ~ 1) length filter
having a memory
ft (oJ fl~ ~ f~) (50)
which converts in the least error energy sense the
(n + 1) length input x~ = (x0, xl, ..., xn~ into an
(m + n + 1~ length output Zt = (Z0~ Zl' ~ m+n
A model for such a filter is shown in Fig. 16. In this
model there are three signals, namely: ~1) the input
signal xt, (2) the actual output signal zt,and (3) the
desired output signal bt. The signal bt is a double
wavelet as in Fig. 10C.
-74-.
os
The output signal Zt is a convolution of the filter
memory function ft with the input function xt, that is,
Zt ft Xt (51)
The problem is to determine the memory function ft
so that the actual output ~t is as close as possible (in
the least error energy sense) to the desired output bt.
To select the memory function, the following quantity
lS ml nl ml zed:
I = ~Sum of squared errors between desired output
and filtered signal wavelet) + v ~Power of the
filtered noise).
where v is a preassigned weigh~ing parameter.
The computations which are required for minimizing
I are performed by the computer 245 ~hich is provided
with three input channels 246, 247, and 248, respectively.
A storage and recall element 173 transmits to the computer
245 through the channel 246 data regarding the function
bt. Similarly, storage and recall element 224 transmits
to the computer 245 through the channel 247 data regarding
the function ut. The channel 248 is used to transmit
to the computer 245 data regarding the functions xt
which are also applied to the input lead 241 of the
pulse shaping filter 240.
Upon reception of the input signals bt, ut and xt
applied through the channels 246, 247, and 248 respectively,
the computer 245 is arranged to perform certain calcula-
tions to be described later and to transmit through the
output channel 251 to the pulse shaping filter 240 the
required data concerning the memory function of the
filter 240. Thus, the actual filter output Zt is in the
-75-
-,
least error energy sense as close as possible to the
desired output bt. In other words
Ztr~ bt ~.52)
as shown in Fig. lOD~
Consider now more in detail the opera~ions which
are performed by means of the computer 245. In symbols,
the quantity I to be minimized is
m + n
I = ~ (bt ~ Zt) + v E {vt} ~53)
t - o
where the notation E {} indicates the ensemble average
and where
m
t ~ s Ut-s
~
represents the filtered noise. Simplifying the expression
for I, one obtains
m~n ~ m ~2 m
I = ~ bt ~ ~ fs Xt-s + v ~ L fsqt-sft (54
~ ~ s=0 / ~ t=0
Here
qt-s E {UT-s ~i-t} (55)
where T iS a dummy time index, and where qt s is the auto
correlation of the received noise. Differentiating the
expression for I with respect to each of the filter co-
efficients, and setting the derivatives equal to zero,
a set of simultaneous equations is obtained which is
-76-
4~
~ s rt-s + v qt-s = gt ~56)
for t = 1, 2, ..., m. In the above equations the quant-
ities rt s~ qt s and gt are `known, whereas the quantities
fs are unknown.
Calculations performed by the computer 245 serve to
determine the parameters rt s' q~ ~ and gt from the in-
put functions applied to channels 248, 247, and 246
respectively, and then to solve the equations (56) ~or
the unknown quantities fs. The parameters rt s~ qt s~
v, and gt are defined as follows:
The parameter rt s is the auto correla~ion of the ~nput
signal xt, which is applied to the computer 245 through
the channel 248. The para~eter q~ is the auto correla-
tion of the noise signal ut, whi~h is applied to the
computer 245 through the channel 247. The parameters
gt are defined as the cross product coefficients between
~he desired outpu~ bt and input xt. Thus,
m
gt ~ bs Xs-t (57)
s=O
for t = 0, 1, 2, ..., m. In the expression for gt the
desired output bs is applied to the computer 245 through
the channel 246 and the input xt is applied through the
channel 248. The parameter v is a weighting parameter
to which an appropriate value is assigned as discussed
later in this specification.
-77-
\
Thus the parameterS rt_s~ qt-s and gt
by the computer 245 and then the computer solves the
equations, thus producing at the output channel 251 the
quantities fs These quantities are applied to the
memory of the pulse shaping filter 240. The actual
output Zt of the filter 240 is in the least energy sense
as elose as possible to the desired output bt.
Since the matrix of the equation, namely the
matrix [rt s ~ vqt s] is in the form of an au~o corre-
lation matrix, these equations can be solved efficientlyby the recursive method. This recursive method has been
described in the following two publications: N. Levinson,
"The Wiener RMS (root mean square criterion) in Filter
Design and Prediction," Appendix B of N. Wiener, "Extra-
polation, Interpolation, and Smoothing of Stationary
Time Series", John Wiley, Ne~ York, N.Y. 1949, and
Enders A. Robinson on "Statistical Communication and
Detection with Special Reference on Digital Data
Processing of Radar and Seismic Signals," pages 274-
279, Hafner Publishing Company, New York, N.Y. 1967.
It should be noted that the machine time required
to solve the above simultaneous equations for a filter
with m coefficients is proportional to m2 for the re-
cursive method, as compared to m3 for the conventional
method of solving simultaneous equations. Another ad-
vantage of using this recursive method is that it only
requires computer storage space porportional to m
rather than m as in the case of conventional methods.
-78-
9fl~2~
In designing a pulse shaping filter two require-
ments may be considered; namely,
(a) to shape as close as possible the function
Zt into the desired function bt.
(b) to produce as little output power as possible
when the unwanted stationary noise is its only input.
In many practical situations a filter is needed
for per~orming both of the above requirements simul-
taneously, and so one is faced with the problem of
finding some suitable compromise between the two re-
quirements. Hence one selects an appropriate value
for the parameter v which assigns the relative weighting
between these two requirements.
There are situations where the value zero is
assigned to v. In such case the expression (53) takes
a simpler form, namely
m~n
I = ~ (bt ~ æt) (58)
t=0
and the computer 245 does not require the data repre-
senting ut. In such case the storage and recall element
224 shown in Fig. 15 is removed and, thus, the computer
245 is provided with two input channels only; namely,
the channel 246 and the channel 248.
It should now be observed that the performance of
a pulse shaping filter and that of a matching filter are
not exactly alike; i.e., for a given input, the outputs
of these filters are not identical. The expression
Yt ~ bt, which is applicable to a matched filter, has
been used above in order to indicate that the signal
expressed by Yt (which represents the output of a matched
filter) closely approximates the double wavelet bt.
-79-
Accordingly, it was pointed out that the same graph
in Fig. lOD represen~s the functions y(t) as well as
the functions b(t). It should also be noted that the
expression-zt~ bt which is applicable to a pulse
shaping filter has been used above in order to indicate
that the signal expressed by Zt (which represents the
output of a pulse shaping filter) closely approximates
the double wavelet bt. Accordingly, it was pointed out
that the sa~e graph in Fig. lOD represents the function
z(t) as well as the function b(t). Strictly speaking,
the same graph in Fig. lOD should not be used to represent
the functions, b(t), y(t) and z(t). However, since both
y(~) and z(t) closely approximate b(~), it is believed
to be appropriate for the purpose of explanation to use
the same graph oE Fig. 10D to discuss the performance
of a matched filter and that o~ a pulse shaping ilter.
XVIII. ~AVELET SPIKING
Consider now the operational arrangement shown by
Fig. 17. I have obtained at the output lead 162 o~ the
subtractor 160 both the information carrying signal b(t)
and the noise signal u(t). The signal b(t) is the double
wavelet as shown in Fig. lOC. The mixture of the signals
b(t) and u(t) is applied to the A/D conver~er 163, ~here-
by producing at the output lead 164 of the converter
digitalized signals bt and ut; these signals corresponding
to the analog signals b(t) and u(t), respectively. The
mixture of these two signals bt and ut is in turn applied
to the input lead 300 of a spiking filter 351. The
-80-
spiking ilter is a particular case of a Wiener type
pulse shaping filter in which the desired shape is
simply a spike. (For a description of a spiking filter
see, for instance, the publication by S. Treitel and
E.A. Robinson on "The Design of High-Resolution Digital
Filter`' IEEE Transactions on Geoscience Electronics,
Vol. GE-4, No. 1, 1966 pp. 25-38.)
It should now be recalled that a double wavelet
b(t) such as shown in Fig. lOC consists of two valve
wavelets; i.e., a valve wavelet "A" and a valve wave-
let "Bl'. The valve wavel~t "B" follows the valve wave-
let "A" after a time Tp. The function of the spiking
filter to be used in the embodiment of Fig. 17 is to
transfonn the valve wavelet "A" as well as the valve
wavelet l'B'r into a respective clearly resolved spike.
Thus, a double valve wavelet bt is converted by means
of the spiking filter 351 nto a pair of spikes.
Figs. 18A to 18F show respectively six possible
positions of a pair of spikes (such as Ml and Nl ) with
respect to the double wavelet applied to the input
terminal 300 of the spiking filter 351. Let Tk be the
time interval between the spikes Ml and Nl, which is the
same in all of Figs. 18A to 18F. Let Hl be the point
of intersection of the spike Ml with the axis of abscissas
texpressed in milliseconds). Thus, the distance OHl
(in milliseconds) will represent the time lag of the
spikes with respect to the double wavelet. Accordingly,
in Fig. 18A the time lag OHl =0); i.e., the initial
spike Ml is placed at the very beginning of the double
wavelet. The five cases shown in Figs. 18B to 18F cor-
-81-
`\
z~
respond to increasing values of the time lag OHl.
One of these figures represents the optimum value o~ the
time lag for which the resolution of the spiking filter
is the highest. For such an optimum time lag the out-
put signal derived from the spiked filter is noticeablysharper than for any other time lag. A procedure for
obtaining the optimum value of the time lag, the optimum
length of the memory of the filter, and the opti~um
value of the time interval Tk will be described later
in this specification.
A double spike obtained at the outpu~ terminal of
the spiking filter 351 represents a single bit in the
digitalized signal which operates the valve 40~ It is
desirable as pointed out in connection with Fig. 9 to
transform a double spike into a single spike or pulse.
I apply here a processing system similar to that in Fig.
9. (See Section XIV of the specification entitled "Trans-
formation Of A Coded Sequence Of Double ~avelets Into A
~oded Sequence Of Short Pulses - Step #3"). Accordingly,
I provide a delay element 303 controlled by a clock 304
in combina~ion with polarity reversal elernent 306 and
an AND gate (coincidence network) 307. These are ar~
ranged in a manner similar to that shown in Fig. 9.
However, the amount of delay in Fig. 17 is different
from that in Fig. 9. Namely, in Fig. 17 the delay
element 303 should produce an output signal which is
delayed with respect to the input signal by an amount
Tk, whereas in Fig. 9 the delay produced by the corre-
sponding delay element 193 amounts to Tm.
A coded succession of single pulses obtained at the
output lead of the AND gate 307 is applied to a D/A
converter 308 controlled by a clock 309. At the output
-82-
~-~g~3S
lead of the D/A converter 308, one obtains in analog
form a signal representing the measurement of a selected
parameter in the borehole. This signal is recorded
by means of recorder 54.
It should be noted that in some instances, depending
on the specific electronic circuitry chosen for the
various blocks shown in Fig. 17, the polarity reversal
element 306 may not be required, since its function may
be performed by the appropriate design of the AND gate.
Fig. 19 shows an alternate arrangement for noise
elimination by means of a spiking filter. In Fig. 17 a
special means was provided for eliminating the pump
noise (i.e., delay element 152 in combination with the
subtractor 160). On the other hand, in Fig. l9 the
procedure for noise elimination has been simplified.
Thus, the signal processing using a delay element 152
and a subtractor 160 has been eliminated. Accordingly,
in Fig. 19 the signal F(t) obtained at the output ter-
minal 151 of the filter 150 is digitali~ed by means of
an A/D converter 350 and then applied to a spiking filter
351a. The spiking filter 351a is designed differently
from the spiking filter 351 of Fig. 17. In Fig. 17 the
spiking filter 351 was designed to convert a double
wavelet such as shown in Figs. 18A to 18F into a pair
of spikes separated one from the other by a time interval
Tk. On the other hand the spiking filter 351a in Fig.
19 has been designed to convert a single valve wavelet
into a single spike. Various positions of a single spike
with respect to a single valve wavelet are shown in
Figs. 20A to 20F.
It should now be recalled that each single valve
wavelet applied to the input terminal of the filter 351a
and each single spike obtained rrom the output terminal
-83-
of the ~ilter 351a represents a single bit in the digit-
alized signals which operate the valve 40. The coded
succession of the spikes in the digitalized format ob-
tained from the output terminal of the spiking ~ilter
351a is then applied to a D/A converter 352 where it is
transformed into a coded succession of spikes, each spike
representing a single bit of the information encoded at
the subsurface instrumentation. The succession of these
bits represen~s in a digital format the measurement of
the selected parameter. It is, however, necessary for
recording and/or display purposes to represent this
measurement in an analog form. Accordingly, the signal
obtained from the D/A converter 352 is applied to a D/A
converter 362 to produce at the output of the converter
362 a signal having magnitude representing the measure-
ment of the selected parameter. This signal is recorded
by means of the recorder 54.
It should be noted that the conversion of a double
wavelet into two spikes by means of the spiking filter
351 as in Fig. 17 or the conversion of a single valve
wavelet into a single spike by means of the spiking
filter 351a as in Fig. 1~ can only be approximated. A
pure spike; i.e., a delta function, cannot be obtained.
~lowever, the objective of this invention is to increase
resolution; i.e., to obtain an output signal which isnoticeably sharper than the input signal.
Consider now the manner in which a spiking filter
should be designed. In theory~ this purpose may be
achieved exactly i~ one could use a filter whose memory
function is allowed to become infinitely long. For
exact filter performance one also needs, in general, to
delay the desired spikes by an infinite amount of time
relative to the input wavelet. tSee publication by
-84-
~9~S
! J.C, Claerbout and E.A. Robinson on "The Error in Least
Sguares Inverse Filtering", Geophysics, Vol. 29, 1964
pp. 118-120,). In practice, it is necessary to design
a digital filter whose memory function has finite duration
and hence, at best, one can attain ~he objective only
approximately.
Suppose that for practical reasons one wan~s to
consider a filter which has memory function of the order
of duration of an input wavelet. Assume that one is at
liberty to place the desired spike at any selected loc-
ation. For instance, Figs. 18A to 18F show six possible
positions or locations of spikes for a spiking filter
301 of Fig. 17. Similarly Figs. 20A to 20F show six
possible positions of spikes for spiking filter 351 of
Fig. 19. The optimum position of spikes was determined
for each of these cases. It should be noted that ~he
position of the spike is an important factor governing
the fidelity with which the actual output resembles the
desired spike.
A spiking filter is a particular case of a Wiener
type pulse shaping filter previously herein described.
Consequently, the required procedures to design such a
filter are analogous to those which have been previously
herein described. We are concerned with the determination
of the elast error energy for a filter w~hose output is
a spike.
In order to determine the optimum value o the time
lag and the optimum length of the memory function for
the spiking filter 301 in Fig. 17, it is necessary to
obtain a record of a double wavelet bt (which is a dig-
ital version of b(t)~. The necessary steps for obtaining
such a record; i.e., steps (a), (b~, (c~, ~d), (e) and
(f) have been previously described with reference to
Fig. 12. Thus, the record of bt is stored in the element
173 in the arrangement of Fig. 12. Similarly, in order
-85-
~L19~2l)~i
to determine the optim~m value of the time lag and the
optimum length of the memory function for the spiking
filter 351a, it is necessary to obtain a record of a
single valve wavelet Bt (which is a digital version of
B(t).
Consider the spiking filter 351a in Fig. 19.
Various posi~ions of a spike corresponding to various
delays ~Figs. 20A to 20F) can be expressed as
~1,0,0,..... 0,0): Spike at time index ) or zero delay
spiking filter.
(0,1,0,..... 1,0): Spike at time index m + n - 1 or
(m ~ m - 1 - delay spiking filter.
(0,0 ....... 0,1): Spike at time index m ~ n; (m + m)
- delay spiking filter.
Performance of a spiking filter corresponding to various
delays is illustrated diagrammatically in Figs. 21A,
21B and 21C. In all these figures the input wavelet is
the same; i.e., valve wavelet B~ which has been recorded
and stored as explained hereinabove. The desired out-
put in Fig. 21A is a spike (1,0~0); i.e., a spike having
zero delay. The corresponding memory function for a
zero delay spiking filter is F = (Fl, F2, F3, ..... F
and the actual output is Wt = (Wl, W2, ..... ~.
Similar notation applicable to Figs. 21B and 21C is shown
in these figures. To each position of the spike there
corresponds an energy error E. The normali~ed minimum
energy error E represents a very convenient way to measure
the performance of a pulse shaping Wiener type filter,
and more particularly, of a spiking filter. When the
filter performs perfectly, E = 0, which means that the
desired and actual ilter outputs agree for all values of
-86-
\
of time. On the o~her hand, the case E = l corresponds
to the worst possible case, that is, there is no agree-
at all between desired and actual outputs. Instead
of the quantity E, it is desirable to consider the one ' 5
complement of E which shall be called the filters
performance parameter P.
P = l - E (46)
Perfect filter performance then occurs when P = 1, while
the worst possible situation arises when P = 0.
Fig. 22 illustrates schematically the process of
measuring the performance parameter P. A computer 400
is provided with three input channels 401, 402 and 404.
The input channel 401 receives from the storage and recall
element 403 data representing a valve wavelet Bt; input
channel 402 receives from t~e lag control 405 data re- X
regarding spikes for various time delays; and input
channel 404 receives data from memory duration control
406 regarding spikes for various memory durations. The
output at 410 of the computer 400 provides by means of
a meter 411, a measure of the performance parameter P.
For a constant filter duration, one might suppose
that there must exist at least one value of lag time
'at which P is as large as possible. In Fig. 23 there is
shown a plot of P versus the lag time of output spikes
for a family of filters of a fixed duration. It is
observed that the highest point of curve (point Ml)
corresponds to a time lag ONl and the choice of thls time
leads to the optimum time lag filter. It should be re-
called that the curve in Fig. 23 relates to a filter of
a fixed duration.
One can also see what happens as one increases the
filter memory duration at a constant time lag. Fig, 24
shows a plot of P vs. filter length for a desired and
fixed spike time lag. It can be observed that this curve
is monotonic, and that it approaches asymptotically the
-87-
largest value of P as the filter length becomes larger
and larger. The graphs as shown in Figs. 23 and 24 are
obtained by means of the arrangement schematically shown
in Fig. 22.
The two important design criteria that have been
discussed here are the filter time lag and the filter
memory duration. One can always improve performance
by increasing the memory function duration, but physical
considerations prevent one from making this duration
indefinitely long. On the other hand, one may search
for that desired output time lag which leads to the
highest P value for a given selected filter duration.
This time lag in filter output harms in no way and can
improve filter output drastically.
The filter performance parameter P as a function
of time lag and a constant duration (Fig.`23~, or the
parameter P as a function of ilter's memory duration
for a constant time lag ~Fig. 24) are helpful, but do
not tell the whole story. Ideally one would like to
investigate the dependence of P on ~ime lag and memory
duration for all physically reasonable values of these
variables. One way this could be done is to plot P
by taking filter time lag as ordinates and filter mem-
ory duration as abscissas. The array of P values can
then be contoured so that one may see at a glance which
combination of time lag and memory duration yields
optimum filter performance. Such a contour map is
given in Fig. 25 The map shows only the contours for
Pl, P2 and P3. Obviousl~, one is most interested in
the larger P values for it is there that the best filter
perfornlance is obtained. This display enables one to
select the best combination of fil~er time lag and
memory duration by inspection.
88-
~ ~42~:95
XIX. PULSE TIME CODE
Although in the examples shown I described telemetering
systems as employing a Binary code system, other codes
are sometimes more suitable. Tor example, for a gamma ray
sensor or an electronic compass-inclinometer a Pulse-
Time code may be preferable. In some cases, especially
where the sequential transmission of several numbers is
required, a Pulse-Time code has advantages. For some
arrangements of an electronic compass, it is necessary
to sequentially transmit 5 numbers in order to measure
the magnetic bearing. By using a telemetering system
based on Pulse-Time code, a considerable saving both in
~he energy required from the battery and the time re-
quired for the transmission of the data can be achieved.
A conventional Pulse T;me code transmission system
is illustrated in ~ig. 26A (for example for transmission
of the values of 3 parameters~. A series of voltage
pulses is transmitted and the time duration (tl, t2, t3)
of each pulse a, b, c, is representati~e (e.g., propor-
tional or inversely proportional) of the magnitude of the
parameter being transmitted. It is noted that between
each pulse a pause in time is required in order to sep-
arate the end of one pulse from the beginning of the next.
Thus in Fig. 26A the pulses a, b, c are somewhat analogous
to ~hree binary "words", each being separàted from the next by
a time internal Tw, These pauses are, of course, a detriment
to rapid data transmission since the pause itself carries
no information. Purthermore, long time duration pulses
are repugnant to the telemetering system of this invention.
I propose a type of Pulse Time code as shown in Fig. 26B.
In this system it is not the duration of the pulse that
is a measure of the parameter but the time between successive
very short pulses. Instead of transmitting long pulses of
variable duration, only short pulses of substantially con-
stant duration (in the telemetering system of this invention,
-89-
~9~
pulses lasting several milliseconds) are transmitted,
and the time between the pulses is ~he measure of the
magnitude of the parameter. Thus, no time is required
to separate one time interval (representing a parame~er)
from the next. In Fig. 26B parameter No. 1 is represented
by the time (tl) between pulse P0 and pulse Pl. Parameter
No. 2 is represented by the time (t~) between pulse Pl
and pulse P2, and parameter No. 3 is represented by the
time (t3) between pulse P2 and pulse P3. It is seen that
in the example above, pulse Pl represents the end of time
interval tl and also the beg;nn;n~ of time interval t2,
and pulse P2 represents the end of time interval t2 and
also the beg;nnin~ of time interval t3 etc. Thus, there
is no lost time between each significant time interval
(i.e., Tw of Fig. 26A is zero).
Thus it can be seen that by utilizing the pulses
Pl, P2, P3, both to indicate the end of one time interval
and also the beginnin~ of the next time interval, the lost
time (unused time) is zero; and all the time used for ~he
transmission of data (i.e., identification of the time
intervals tl, t2, t3,) is useful time. In terms of binary
coding, each "word" (identifying a number) is followed
in~ediately by the next "word", and so on. Only at the
end of a sequence of transmission is there a pause Tp,
and then the sequence repeats. In the next sequence,
of course, the time intervals between ~0, Pl, P2, p3
will be usually somewhat different since the data re-
presented by the times tl, t?~ t3 usually vary with time;
and each new transmitted ba~ch of data represents, for
example, a new condition in the borehole.
-90-
o~
Fig 30 shows the principles of the circuitry that
can accomplish the Pulse Time Code of this invention.
In the practical borehole instrumentation, of course,
modern electronic integrated circuits ~as for example,
a type CD4066 Bilateral Switch) would be used Csee also,
Fig. 29). For ease of explanation, I will show a simple
mechanical stepping switch and a simple mechanical relay
so the principles of the logic of the system can be
simply illustrated.
In Fig. 30, sensors are connected to the stepping
switch terminals 1, 2, 3 of the stepping switch 285 which
has an electromagnetic drive winding 286 as shown. Assume
that we start the sequence with the stepping switch at
position "O" as shown in Fig. 30. The battery 288 gen-
erates a reference DC voltage. This DC voltage will appear
across the resistor 289 and will charge the capacitor 290
at a predetermined rate determined by the value of
resistor 289, the size of the capacitor 290, and the
voltage of the battery 288. 291 is a Trigger Circuit
which generates a single sharp electric pulse w~en the
voltage applied to its input exceeds a predetermined
value (Trigger voltage2, The output of the Trigger
Circuit 291 activates the winding 286, and the arm 287
of stepping switch 285 moves over to the next contact
(No. 1 in this case2. Simultaneously, the Trigger
Circuit 291 momentarily operates the relay 292J which
discharges the capacitor 290 to ground.
When the arm 287 is moved from position "0" to
position "1", the process repeats itself, except that,
instead o the reference voltage of the ~attery 288, the
voltage output of sensor No. 1 is connected to the circuit,
and the pulse Pl is generated at the time the capacitor
again reaches the trigger voltage of the Trigger Circuit
291. This time is proportional to the value ( RC
-q,l-
~314~
where R is the ohmic value o the resis~or 289, C the
capacitance of the capacitor 290 and ~s the output voltage
o~ the sensor. Thus the time tl is inversely proportional
to the output voltage of the sensor.
After the activation of the Trigger Circuit 291 by
the voltage ~rom the sensor No. l, the process repeats
itself and again, and when the voltage on the capacitor
290 reaches the trigger voltage, the Trigger Circuit 291
generates a sharp pulse which operates the relay 292,
discharges the capacitor 290 and energizes the stepping
switch 285 and moves the arm 287 to the next contact.
Thus, the stepping switch 285 steps in sequence and
connects the Sensors l, 2, 3, one after another to the
resistor 289. The pulse generated by the Trigger Circuit
291 when the arm 287 is at position "0" corresponds to
pulse P0 (of Fig. 26B); and the pulses generated by the
Trigger Circuit when the arm 287 is at respective positions
"l", "2", "3" correspond to respective pulses Pl, P2, and
P3 (of Fig. 26~ he corresponding time intervals tl,
t2, t3 are representative ~inversely proportional~ of the
voltage outputs of the sensors No. l, No. 2, No. 3.
The above paragraphs describe the principle of the
Pulse Time encoder which may be employed in the down-
hole equipment in place of the A/D Converter 102 in
Fig. 4A. The decoding at the surface may be accomplished
by conventional Pulse Time decoding circuitry and need
not be discussed in detail here.
g , TPo, TPl, TP2, TP3, etc. represent suc-
cessive pulses received at the detecting point at the
surface. These pulses occur at times To~ Tl, T2, T3
etc. respectively. In the Pulse Time code as described
in relation to Fig. 26B, the time between successive
pulses is used to indicate the magnitude of a parameter.
_9~_
Thus, if 3 parameters are to be telemetered, the code can
be as shown in Fig. 26C, in which
Tl - To is a time interval representing parameter No.l.
T2 ~ Tl is a time interval representing parameter No.2.
T3 - T2 is a time interval representing parameter No.3.
In mud pulse Measurement ~ile Drilling operations
it is necessary in some cases that the measurements be
made with great precision. Since the sound velocities
in the mud column are not always constant, and noise and
attenuation conditions vary, the time interval between
pulses received at the surface will not be in precise
agreement with the time interval between the corresponding
pulses generated at the downhole equipment. In other words,
there is frequently an uncertainty at the sur:Eace as to the
exact time at which a specific pulse arrives.
Assume that thP absolute uncertainty of each
pulse arrival time is plus or minus 0.2 seconds, or a total
of 0.4 secGnds. In order to achieve an accuracy of ~ 1%
for Tl - To with a total absolute error of 0.4 seconds,
the time between pulses must be at least Q.4 x 100 or
40 seconds. Furthermore, since the apparatus could
sometimes fail to develop a clear sharp pulse, at least
two detection 'Isweeps" are necessary. If both sweeps give
the same answer, the data then would have been "~erified".
Consequently, to achieve the desired accuracy and certainty
(for a practical case, an accuracy of ~ 1%) about 80 to 120
seconds will be required per parameter measured (i.e., about
2 minutes per parameter2-
In my improved Pulse Time Code I propose an addition
which in many cases can result in much greater accuracy.I propose to use for each transmitted pulse P0, Pl, P2,
P3 not one single mud pressure pulse but a group of at
least 3 unequally spaced mud pressure pulses as shown in
Fig. 26D (hereinafter called a triple group).
-93-
?5
Let the time spacings in each triple group be
The time from the first pulse to the second = tl.
ThP time from the second pulse to the third = t2.
The time from the first pulse to the third = t3.
In this situation, again To represents the time of
arrival of triple group TPo; Tl the time of arrival of triple
group TPl; T2 the time of arrival of triple group TP2 and
T3 the time of arrival of triple group TP3, and;
~1 ~ To is a time interval representing parameter
No. 1.
T2 ~ Tl is a time interval representing parameter
No. 2.
T3 - T2 is a time interval representing parameter
~o. 3.
The advantage of this system is that in case of a
momentary failure causing one pulse in the group not to
be received, the failure can immediately be recognized --
because one triple group will con~ain two pulses instead
of three. ~urthermore, since the times tl, t2, t3 are
unequal and known, one can determine which pulse in the
group is missing; and still furthermore, since again tl,
t2, t3 are known, one can apply the proper correction and
determine the times Tl - To~ Tl - T2, T2 3
same accuracy as would be in the case where all pulses
were preser,t in t~e triple group. The triple group sys-
tem has one further advantage: Since it is difficult to
determine the exact arrival time of a particular pulse,
the triple group permits substantially closer determina-
tion of the arrival time. One could, for example, take
the arithmetic average of the arrival times of each pulse
in the triple group or by use of modern computer techniques
-94-
obtain even greater accuracy o~ the arrival time.
Fig. 29 shows a block diagram of a downhole electronics
logic system that will generate the triple group pulses
shown i~ Fig. 26D.
Numeral~101 is a sensor (see Fig. 4A) that generates
an electric ~oltage indicative o~ the magnitude of a down-
hole parameter. 601, 602, and 603 are respectively a
voltage controlled oscillator, a scaler, and a trigger
circuit that generate in a well known manner a series of
electric pulses separated by time intervals that are
indicative of the magnitude of the voltage output of the
sensor 101 and, therefore, of the downhole parametar
being measured. The time interval between the pulses
PO and Pl as shown in Pi~. 26B is, therefore, a measure
of one para~eter measured by one of the sensors 101 of
Fig. 4A.
The portion of ~ig. 29 that is enclosed within
the dashed rectangle shows the detai~s of the circuitry
for generating the triple group pulses referred to previously.
607, 608 and 60~ are electronic "one shot" univibrators
that in response to the impulse from the trigger circuit
603 each produces a single output pulse of durations
Dl, D2, D3 respectively as shown in Fig. 29. Blocks
610 are electronic derivators; i.e~, each produces
at its output a signal that is proportional to the
1st first time derivative of the input signal (such
electronic devices are well known in the art). Their
outputs, therefore, will be as shown in ~ig. 29 as G,
H and I; i.e., two opposite polarity pulses separated
by respective ~ime intervals Dl, D2 and D3. Blocks
611 are rectifiers and transmit only the positive pulses
appearing at the outputs o~ derivators 610. The out-
puts of rectifiers 611 are connected in parallel at
612 and produce the signal J, which is the signal
-95-
desired (also shown in Fig. 26D). Each single pulse
generated by the trigger circuit 603, therefore, will
generate three pulses separated by known and unequal
time intervals (triple group~ as shown by J. In practice
the înterval Dl is made very short compared to D~ and D3
and would be only a few microseconds, whereas D2 and D3
are intervals of a few milliseconds to several hundred
mil`liseconds. In the analysis of the operation we can
therefore assume that Dl = 0.
Therefore in the output shown as J in Tig. 29,
the pulse Pl indicates the end of ~he output pulse from
607 Cwhich as pointed out above is for all practical
purposes also the beginning of ~he output pulse since
its lengLh is assumed to be zero); the pulse p~ is the
end of the output pulse from 608; and the pulse p3 is the
end of the output pulse from 60~. Thererore, (since
Dl was assumed to be zero2 the time interval tl = D2;
the time interval t3 = D3; and the tIme interval t2 =
D3 - D2. Thus the portion of Fig. 29 within the dashed
rectangle creates the triple group at its output at 612
(shown as J in ~ig. 22) in response to a single pulse
impressed upon its input.
The circuitry shown in Fig. 2~ may be interposed in
Fig. 4A between a selected $ensor 101 and the power drive
104. In other words, when the Pulse Time Code system of
Fig. 29 is employed, the ~tD converter 102 and processor
103 are eliminated (since they are adapted for binary en-
coding~ and the Power Drive 104 is driven directly by the
output of amplifier 613 of ~ig, 29.
When the triple group Pulse Time Code is used instead
of the Binary Code, it will be necessary to decode the
triple group signals at the surface. In Figs. 9, 12, 13,
17 and 19 the signals that represent the downhole parameter
are assumed to be in binary code form. To change the sys-
tem to receive signals in the triple group Pulse Time Code
-96-
form as described in connection with ~igs. 29 and 26D
it is necessary to interpose between the filter 150 and
the subsequent apparatus at the sur~ace a special "Code
Translator" such as that shown in Fig. ~7. For this pur-
pose the wire 151 in Figs. 9, 12, 13, 17 and 19 will be
interrupted and the Code Translator interposed. ~n some
cases it is more desirable to interpose the "Code Trans-
lator" between the subtractor 160 and the A/D Converter
163 at lead 162 of Figs. 9, 12, 13, 17, and the preferred
location for insertion will be clear to those skilled
in the art.
With reference to ~i~g. 27, 316 is a "Selector" which
is designed to generate a single output pulse in response
to the triple group described ~n connection with Tig. 29.
Numeral 317 designates a Time-to-Amplitude converter;
i.e., an electronic c~rcuit that generates an output DC
voltage at the wire 312 that is a predetermined function
of the time between two input pulses impressed upon its
input by wire 318. ~uch devices are well known in the
electronics art and need no detailed descrip~ion here.
320 is an Analog to Digital converter and is also well
known in the art.
Fig. 28~ shows the Selector 316 in more detail. 321,
322, 323 are one shot univibrators designed to generate
a single output pulse o selected predetermined time dur-
ation in response to an input pulse. 321 generates a
long pulse of time duration 13, 322 a shorter pulse of time
duration 12, and 323 a still shorter output o~ pulse of
time duration 11 as shown above each ~lock 321, 322, and
323. Blocks 324 are Deri~ators; i.e., electronic circuits
that generate an output proportional to the first time
derivative of a signal impressed upon the input. Such
units are also well known and will generate output $ignals
as shown on the curve above each Derivator block. Blocks
325 are "inverters"; i.e., they generate an output signal
which is a replica of the input signal but inverted in
sign as shown on the curve immediately above each inverter.
~97-
Blocks 326 each contain a rectifier and generate at the
output a single positive electrical pulse, 326a, 326b,
326c as shown. Blocks 327 are "coincidence circuits" or
"AND GATES" well known in the art. Each block 327 pro-
duces an output pulse at its terminal "c" only when there
is a pulse at input "a" and a pulse at input "b" that are
in time coincidence. The outputs of all three coincidence
circuits 327 are connected in parallel at wire 329 and ap-
plied to the input of the Time to ~mplitude converter 317.
The Time to ~mplitude converter 317 produces an output DC
voltage that is a predetermined function of the time be-
tween two successive input pulses. The output of the Time
to Amp`litude converter 317 is connected to the AID conver-
ter 320 which then translates the input DC voltage into
1~ binary coded pulses ~n a ~anner well known in the art.
The circuitry of Figs. 27 and 28A is made up of
conventional electronic integrated circuit components
that are well known in the art. The over-all operation
of the "selector" requires some more detailed description.
P Pl, P2, P3 generated by amplifier 613 of
Fig. 29 are impressed on power drive 104 of ~ig. 4A and
are transmitted to the surface as mud pressure pulses by
valve 40. At the surface these mud pressure pulses are
picked up for example by the elements of Fig. 9 comprising
transducer 51, filter 150, delay element 152 and subtractor
160. The pulses that appear on lead 162 of ~ig. 9 (or
Fig. 12, ~Fig. 13 or ~ig. 17~ we shall designate as TPl,
TP2 and TP3 Cand correspond to pulses P~. ~2 and P3 gen-
erated by the downhole electronics of ~ig. 29).
Figs. 28B, 28C, 28D, 28E show the response of the
circuit of ~ig. 28 to the pulses TPl, TP2, TP3. When
pulse TPl arrives and is impressed on wire 151 (or wire
162) of Fig. 28A, all three one shot univibrators 321,
322 and 323 are triggered and each generates a respective
output pulse having its own characteristic and predeter-
-98-
mined and fixed length 13, 1~ and 11, respectively. Thus,
when pulse TPl triggers the univibrators they generate
the output voltages (pulses) Al, Bl, Cl as shown in Fig.
28B.
When pulse TP2 arrives it cannot trigger univibrator
321 because it is already in the "ON" state. Pulse TP2
does, however, trigger univibrators 322 and 323, since
they have returned to the "OF~" state, and they generate
output pulses B2 and C2 as shown in Fig. 28B. When pulse
P3 arrives it cannot trigger one shot univibrator 321 or
322 because they are already in the 'ION'' condition.
Pulse TP3 does, however, trigger univibrator 323, since
it has returned to the "OFF" state, and it generates out-
put pulse C3 as shown in Fig. 28B.
The time intervals 13, 12, and 11 of the univibra-
tors 321, 322 and 323 in Fig 28A are so proportioned
that they correspond to the time delays caused by the
action of the univibrators 609, 608, 607 of Fig. 29,
and consequently, the ends of the group of univîbrator
pulses of Fig. 28B are in "coincidence" and activa~e
the ~ND gates of Fig. 28~.
Fig. 28B shows the conditions of operation when
all pulses TP , TP2, TP3 are present,
Fig. 28C shows the same conditions as in ~ig. 28B;
but with one of the pulses Ce.g. TPl~ missing.
Fig. 28D s~ows the same conditions but with pulse
TP2 missing, and Fig. 28E w~en pulse TP3 is missing. It
must be noted that no matter which pulse (TPl, TP2, or
TP3) is missing, two univibrator output pulses always
end at the time T. This characteristic of the circuit
of Fig. 28A is used to always generate at least two time
coincident pulses at the time T no matter which of the
_99_
pulses TPl, TP2, and TP3 is missing. So ~ong as at
least two of the group of pulses are detected, the
time of occurrence of the output pulse at ~29 of Fig.
28A will be the ~ame. The single pulse 328a in ~ig.
28A is generated whenever a "group" of pulses is
received by the uphole apparatus and the pulse 328a
will be present whenever any two pulses in the "group"
are detected at the surface.
Referring again to Fig, 28~l block 317 is a conven-
tional tlme to amplitude converter and generates a DC
output vol~age that has a predetermined functional rela~ion-
ship to the time between successi.ve pulses 32~a. 320 is
a conventional AtD converter and translates the magnitude
of this DC output voltage into a binary word. The binary
words follow one another in ~uick success~on as deter-
mined by the characteri~stics of 320 and its associated
clock.
~t can be seen, therefore, that the apparatus of
Fig. 28A translates the Pulse-Time code using the triple
group into a Binary code; and the appara~us succeeding
wire 151 or 162 of Figs. 9., 12, 13, and 17 will operate
in exactly the same manner as if the data were initially
transmitted in binary code form from the subsurface.
XX. ADDITlONAL NOTE~
C12 In order to obtain the shock waves described
earlier in this specification, there are certain limits
imposed on K2 (mean rate of change of the opening of the
valve~ and on TbV) (the time of open flow2. Experiments
have shown that K2 should be at least 5 cm2/sec. and
preferably within the range from 20 cm2/sec. to 150 cm2/sec.
TbV) should be at most 500 milliseconds and preferably
be in the range from 50 milliseconds to 150 milliseconds.
-100-
(2). It must be noted that although the synchronization
pulses (clock 155~ in the examples shown are generated
either by the generator connected to the pump shaft, or
by the phase loc~ed loop described in the parent case,
other means for providing the clock frequency that is
synchronous with the pump action can be provided. ~or
example, the well known "pump stroke counter" usually
used on the connecting rod of the pump can be used to
generate one electric pulse per pump stroke, The period
between each such successive pulse can ~e divided into
an appropriate number (e.g. 512 or 1024) of equal ti~e
intervals by a microprocessor or by a phase locked loop
or by other means well known in the computer and electronics
art. In such arrangement no access to the pump jack shaft
is necessary, and the clock frequency equal to that of
generator 155 can be generated bY the microprocessor
and the pump stroke counter switch.
(3) I have described in the first part of this
specification the conditions for occurrence of hydraulic
shock waves and related "valve wavelets" in some detail.
At some depths, for example shallow depth, conditions
can arise when the valve wavelet as described above is
not well formed. For such a valve wavelet it is necessary
to have a sufficient ~olume of mud flowing in the drill
pipe and suEficient hydrostatic pressure at the transmitter
e~d. It should be clearly understood that my invention
is not limited to the particular wavelet shown and is
applicable to other forms of pressure pulses which can
be detected at the earth's surface as a result of the
actuation of the valve 40.
(4~ Various digital ~ilters including matched
filters, pulse shaping filters and spiking filters ~ave
been described above in a considerable detail. In part-
icular, the performance of each digi~al filter has been
clearly explained by providing a detailed sequence of
operations to be performed These operations have been
101
e~plained and specified by appropriate mathematical
formulations. It is clearly understood that by using
modern computing technigues a person skilled in the art
can provide the necessary programs on the basis of the
descriptiQns provided in this specification, and -the
operations described in connection with Figs. 9, 12, 13,
14, 16, 17, l9, 21 can be performed by suitable software.
~ 52 Various digital filters which I described can
be also applied to other forms of transmitting measurement
by mud pulsations by means other than the by-pass valve
of the t~pe described herein. These other forms may in-
clude the method based on controlled restriction of the
mud flow circuit by a flo~ restricting valve appropriately
positioned in the main mud stream as described in the U.S.
Patent 2,787,795 issued to J. J. ~rps. Broadly speaking,
the digital filterIng systems wh~ch I have described can
be applied to any forms of telemetering system in logging
while drilling and other forms of logging in which the
drilling equipment is removed in order to lower the measuring
equipment into the drill hole. It can be applied to
telemetering systems utilizing pulses representing any
forms of energy, such as electrical electromagnetic,
acoustical pulses and other pulses.
C6~ The pulse time code employing the triple pulse
group as described above can also ~e applied in acoustic
well logging in order to facilitate the processing of
acoustic well logging signals and to obtain a highly
effective method allowing automatic correction of errors
due to pulse skipping in the measurement of the transit
time of acoustic waves. Acoustic well logging methods
and apparatus are usually designed to measure translt
time of an acoustic wave between a first and a second
pulse. In U.S. Patent 3,900,824 issued on August 19, 1975
to J. C. Trouiller et al~, it was proposed to prevent
pulse s~ipping by the measurement made during a sequence N-l
-102-
stored in an auxiliary memory and comparing this measurement
with the next measu~ement (sequence N~. Patent 3,900,824
is included in this application by reference. The
alternate method which I propose and which is based on
the pulse time code is effective in correcting measurement
errors due to cycle skipping in a more efficient and
reliable manner.
(7) The triple pulse group time code has a very
broad field of applications outside of logging while
drilling. It can be used in any communication system for
transmitting messages from a transmitting station to a
receiving station as well as in various types of well
logging (not necessarily in logging while drilling) such
as acoustical logging Csee Note ~6).
(8~ It is understood that in order to capture and
store a wavelet for later use in the digital filters
described herein, certain steps must be taken at the site
as have been described in this specification. It is
desirable sometimes to capture a single wavelet (rather
than a double wavelet2, as is necessary in the embodiment
of Fig. l9~comprising the spiking filter 351A. To capture
a single wavelet it is convenient to synchronize the
generation of the signal produced by the s~bsurface equipment
with detection equipment on the surface. This can be
done by replacing one of the sensors 1, 2, 3 ar.d 4 at the
subsurface equipment in ~ig. 4~ by a device such as a
"clock" or constant time controlled signal generator that
will cause uniformly time spaced operations of the valve
40 of ~g. 4A. The operation would be 2S follows:
Ca) By stopping and starting the mud pumps at the
surface in the appropriate sequence, the switch 91 of
Fig. 4A can be made to connect the modified sensor ~i.e.,
generator of uniformly spaced pulses~. Thus a sequence
of pulses will be generated by the valve at known times.
(Correction of course must be made for the travel time
-103-
of the pulse from the subsurface to the surface which
has been determined previously by ~ethods well known.
(b) The surface equipment is controlled by its
own clock that is in synchronism in time and phase with
the subsurface signal transmitter.
~c~ By suitable switching at the surface the cap-
turing and storing of the double wavelet can be interrupted
so that the storing circuit is connected only for ~he
time of one wavelet and is automatically disconnected
during the occurrence of the second wavelet.
Of course the same operation can ~e done by ~and
~by the operator~ This is easily done when the wavelet
is distinct and clearly overrides the noise. When the
wavelet is submerged in noise, the automatic system ~such
as described herein is used.
(9) There are two interfering noise signals which
tend to obscure the reception of the useful signal B~t)
(see equation 22). One of these represents the pump
noise P(t) and the other represents the noise U(t) which
is associated with various drilling operations other than
the action of the pump. In order to eliminate these
interfering signals I provide three filtering systems
identified as filtering systems #1, #2, and #3.
Filtering system #1 is the analog filter 150. The
purpose of this filter is to suppress the steady component
of the transducer output representing the pressure gen-
erated by the pump 27 and other frequencies outside of
the range of interest~
Filtering system #2 comprises a delay element 15~
and a subtractor 16~. The purpose of this system is to
suppress or eliminate the pump noise P(t).
Filtering system #3 comprises a correlator, or a
digital filter which may be a matched filter, a pulse
shaping filter or a spiking filter and also comprises
various associa~ed elements such as storage and recall
-104-
elements, and computers for determining the optimum
values of the memory elements for the corresponding
digital filters (see Figs. 9, 12, 13, 14, and 15). The
purpose of the system #3 is to elimina~e or suppress
the noise U~t).
The filtering systems ~ 2, and #3 are connected
in cascade. In the embodiments of my invention described
above, the filtering system #l is connected to the pressure
transducer 51, the system #2 is connected to the output
lead 151 and the system #3 is connected to the output
lead 164 of the system #2.
Each of the above filtering systems is a linear
system. Therefore, the function of these systems may
be interchanged or reversed. I can proceed at first
with the filtering system #1 and then interchange the
order of the filtering systems #2 and #3. Also, in some
cases it may not be necessary to use all three filtering
systems. Any two may be sufficient and in some cases
only one. Also, the system between wire 182 and wire
210 can sometimes be eliminated and the D/A converter
211 arranged to accept double wavelets.
(10) When the signal produced by the process de~
scribed in Section XIII (steps a to f) is captured and
stored, it can be cross-correlated with the raw signal
as generated by the transducer 51 or with the precon-
ditioned signal at wire 162 of Figs. 9-19. In the case
of cross-correlation with the raw signal at the trans-
ducer 51 the second wavelet in the "double wavelet" will
have to be eliminated by suitable means well known in
the art, so as to be able to cross-correlate with the
single wavelet at the output of transducer 51.
-105-
