IEEE Signal Processing - March 2018 - 114

Mud
Pulse
Signal

Compared to MPT, EM telemetry
offers several advantages: it has no
moving part that makes the maintenance of the tool easier and it does not
depend on the circulation of a drilling
fluid and, hence, supports air drilling.
However, the operating range of EM
telemetry is highly dependent on the
resistivity profile of the formation to
be drilled: just one single layer with
sufficiently large conductivity contrast
may divert the electrical current from
the surface. Further, as the signal emanates in three dimensions, its range
is limited.

Siren Position

Drill
Mud
Flow

Drill Bit

Pressure

Stator
Rotor

Time

Figure 2. Signal generation in MPT. A pressure wave is generated by creating a physical obstruction

Noise cancellation

Noise is a major issue in telemetry
while drilling because its energy can
be orders of magnitude stronger than
the signal of interest. To reduce the
impact of noise on the reliability of the telecommunication
link, different noise cancellation approaches are commonly
used. Two broad classes of methods exist: model based and
diversity based.

on the drilling mud flow; in this example, using a fixed stator and a moving rotor. The angle between
the two determines how much flow energy is converted into mud-pulse signal.

controlling the flow area. The principle of signal generation is
illustrated in Figure 2. The stator is made of a metallic plate
perforated at several locations to allow the flow to circulate. The
rotor is moved by an electrical motor whose angular position is
precisely controlled. The combination of a rotor and a stator is
called a siren. In this setup, an electromechanical device is used
to manipulate a hydraulic system such that the amount of flow
energy that is converted into pressure signal can be controlled.
Many risks must be considered in such a design: erosion due
to moving particles in the drilling fluid, shocks and vibrations
in excess of hundreds of g-force, high downhole temperatures
beyond 150 cC, and limited power. The required mechanical
power can vary significantly depending on the geometry of the
siren and properties of the signal in consideration, in particular,
carrier frequency, symbol rate, and effective pulse shape. It is
desirable to choose a waveform that gives good telemetry performance and minimizes the required mechanical power. As
an example, a continuous-phase modulation with raised cosine
phase transition can be used [9]. Increasing the transition period
makes the signal smoother and decreases mechanical power.
The tradeoff for the smoothing is a slight increase of intersymbol
interference. We show one such simulation using a quadrature
PSK (QPSK) telemetry signal in Figure 3. In this example, the
peak mechanical power can be reduced by a factor of 50 and current divided by a factor of ten when the transition between symbols is smoother.

Signal generation in EM telemetry
In EM telemetry, a downhole tool equipped with an insulating
gap acts as a dipole antenna. A signal source is used to create an
electric potential across the gap, which generates an electric field
that propagates to the surface. The surface receiver measures the
electric current between the rig and one or several metal rods
electrically coupled with the ground or using deep electrodes [5],
[6], [10], [11]. This system is shown in Figure 4.
114

Model-based noise cancellation for MPT
In our model-based approach, a dynamic model is associated
with each noise source. Such a model can be formulated as a
dynamic system with hidden parameters. An example is the
so-called mud pump noise created by the surface equipment
generating the mud pressure and creating a specific noise
signature in the form of a narrowband harmonic waveform.
This specific noise signature can be associated with the number of pistons-two pistons (duplex pump) or three pistons
(triplex pumps)-creating cyclic pressure imbalance on the
drill string.
A general model for the mud pump noise recorded on a
pressure sensor is given by [12]
y (t) =

/ / a m,k (t) sin ^k $ i m (t) + i m,kh + n (t),

(1)

m= 1 k= 1

where M is the number of active pumps, K m is the number of
harmonics associated with the mth pump, a m, k (t) is the amplitude of the kth harmonic of the mth pump, i m, k is the initial
phase of the kth harmonic of the mth pump and i m (t) the
angular phase of the mth fundamental, and n (t) is the observation noise.
The estimation of a signal from a set of data measurements
is generally referred to as a filtering or estimation problem [13].
The general approach consists of gathering the unknown parameters linked to the estimation of y (t) under a state vector x (t) .
These quantities are not directly accessible, and they are said
to be unobservable. In contrast, the measurement vector y (t)
gathers available data that can be directly observed from the
system. They are generally obtained by sensor measurements

IEEE Signal Processing Magazine

March 2018

Table of Contents for the Digital Edition of IEEE Signal Processing - March 2018