sensors

Article
The Compass Error Comparison of an Onboard
Standard Gyrocompass, Fiber-Optic Gyrocompass
(FOG) and Satellite Compass
Krzysztof Jaskólski 1, * , Andrzej Felski 1 and Paweł Piskur 2
1 Polish Naval Academy, Institute of Navigation and Maritime Hydrography, Smidowicza St, 69,
81127 Gdynia, Poland; a.felski@amw.gdynia.pl
2 Polish Naval Academy, Institute of Electrical Engineering and Automatics, Smidowicza St, 69, 81127 Gdynia,
Poland; p.piskur@amw.gdynia.pl
* Correspondence: k.jaskolski@amw.gdynia.pl; Tel.: +48-261-262-837




Received: 18 March 2019; Accepted: 22 April 2019; Published: 25 April 2019


Abstract: The aim of the presented research was to analyze the accuracy indications of three types of
compass systems for the purposes of meeting warship modernization requirements. The authors
of this paper have made an attempt to compare the accuracy of an onboard standard gyrocompass,
a fiber-optic gyrocompass (FOG) and a satellite compass in real shipping circumstances. The research
was carried out in the Gulf of Gdansk area, during the preparation of hydrographic surveys on stable
courses. Three heading recordings have been taken into consideration. The helmsman’s operation and
vessel inertia were analyzed and removed according to a spectrum analysis. Transient characteristics
and the spectrum analysis (based on the Fourier transform theory and headings descriptions in the
frequency domain) are presented. Data, processed using a band-stop finite impulse response (FIR)
filter to reduce low-frequency heading distortions, are presented for further analyses. The statistics
of errors of the compasses investigated, as well as the spectrum of these errors, are also presented.
Based on accuracy measurements, the possibility of using the most accurate heading data as the input
signal to the automatic ship control system was considered.

Keywords: gyrocompass; accuracy; fiber-optic gyrocompass; satellite compass; band-stop;

finite impulse response (FIR) filter; spectrum analysis; Fourier Transform

1. Introduction
Despite being equipped with receivers of the Global Navigation Satellite System (GNSS),
Automatic Radar Plotting Aid (ARPA), and Electronic Chart Display and Information System (ECDIS),
contemporary ships still cannot sail without a compass. According to the International Convention
for the Safety of Life at Sea (SOLAS) convention and International Maritime Organization (IMO)
regulations, a gyrocompass is obligatory. A magnetic compass (or simply compass, according to IMO
terminology) is based on a 19th-century design, but nowadays is equipped with elements assuring
transmission to contemporary digital devices. Interestingly, gyrocompasses possess a number of
similar limitations to their predecessor, except for the many advances made in the form of digital
elements. However, more advanced solutions have been implemented over the last 100 years [1]. In the
1980s, laser gyrocompasses were introduced on warships, followed by fiber-optic modifications in the
1990s. The latter have become increasingly popular, even on small, unmanned vehicles [2,3], and have
also been accepted by off-shore suppliers and hydrographic ships.
Satellite compasses trace in a completely new way [4,5]. It is a specific version of a GNSS receiver
having, inter alia, an option to determine a ship’s heading. These devices were accepted by many
users, but current regulations do not allow the use of them as a substitution for a ship’s compass.

Sensors 2019, 19, 1942; doi:10.3390/s19081942 www.mdpi.com/journal/sensors

Sensors 2019, 19, 1942 2 of 20

It is evident that different construction solutions possess different metrological characteristics.

The gyrosphere of the classical gyrocompass behaves as the pendulum, meaning it possesses its own
oscillations. These oscillations appear after maneuvers of the ship, especially after rapid alterations
of the heading. However, present compasses, except for other modifications, are characterized by
miniaturization, so one can expect less influence of the gyroscopic sphere inertia on the components of
the instrument. Authors did not find such information in the literature.
Alternatively, fiber-optic gyrocompasses (FOGs) and satellite compasses do not possess dynamic
elements, so one can expect a different nature of system response. It is especially difficult to find
information on the behavior of these new constructions, which are installed on typical ships, in the
literature. Additionally, it should be taken into account that the dynamics of the reaction of a ship is
dependent on her maneuverability and potential to maintain her on a stable course, which can only
be shown by the onboard installed compass. Thus, the dynamics of the indications of compasses are
relative to their properties, but also to real reactions of the ship, and develop according to the value of
the heading shown by the compass in use.
In literature, this problem is treated superficially; usually, the accuracy of a compass is given along
with its possible deviation. The most common approach to the problem is based on the foundation
of the random character of errors and is defined by a standard deviation. It is worth mentioning
that we have no evidence that the process has a normal distribution. However, many uses in the
area of sensor integration demand familiarity with the spectrum of errors. Over the last 100 years,
many research papers [4,5] concerned with the nature of errors of the gyroscopic compass have been
published and it is known that these errors had a character of the ergodic process with low frequencies.
For certain, we know less about the character of errors of more advanced solutions. Some investigations
in this field have been executed by the Polish Naval Academy and have been published [6], but these
investigations have been devoted to installations with mechanical gyroscopes, which are currently
not in use. More complex analysis has been conducted and documented in Reference [1], with the
results of similar studies on satellite compasses being published in Reference [4] and Reference [5].
Interesting research outcomes of preliminary numerical simulation studies and the preliminary static
experimental evaluation of a true-North gyrocompass system—employing a commercially available
low-cost inertial measurement unit (IMU) comprising a 3-axis FOG with micro-electro-mechanical
systems (MEMS) accelerometers—were presented in [7]. The article presents the results of a numerical
simulation study to evaluate the potential accuracy of the proposed system, and a numerical sensitivity
study to evaluate how this accuracy will change with variation in the sensor measurement noise of
the gyroscopes and accelerometers. Interesting results were achieved in Reference [8] with the use
of a MEMS gyroscope, suitable for northfinding in pointing and targeting applications with the bias
stability of 0.03 deg./h.
Similar accuracy analyses have already been presented in recent years, however, they were limited
to single devices [9–11] and have been focused on small autonomous vehicles [2]. Some aspects of
this problem have been discussed in books concerning the integration of global positioning systems
(GPSs) and inertial techniques, for example Reference [12] or Reference [13]. Many ship accidents have
arisen from an error in course indication. Considering that actual errors in gyrocompasses and satellite
compasses are very minor, such devices may be regarded as valid inputs for autopilot functions,
provided that any failure in their operations is controlled by means of a secondary heading source [14].
Meanwhile, real information regarding both the character of the oscillations and the dynamics of a
compass in real circumstances encounter the problem of separating the natural oscillations of the ship
(yaw) from the oscillations of the compass, resulting from its design. That is why the results of an
investigation of any devices devoted to robotics cannot be translated to a ship’s devices, as the latter
are designed with different tasks and operate in different circumstances.
Research material collected during such experiments as those discussed above can be used to
analyze the effect of a ship’s motion on the accuracy of its compass readings. It was assumed that the
Sensors 2019, 19, 1942 3 of 20

main disturbance—having an essential significance with regard to the accuracy of the readings of a
compass—would be the ship’s motion.
It should also be remembered that the real ship’s heading deviations would result from the
movement of the ship. In other words, the motion will determine both the real ship’s heading deviation
and the errors in measuring it.
It might seem that the comparison between the compass headings with the reference heading,
determined by Equation (1), will enable a full analysis of changes caused by the ship’s motion.
Unfortunately, it appears that this kind of comparison will only provide information on absolute values
of heading deviations for specific moments of observation, giving no information on their character.
The analysis conducted
Sensors 2019, 19, xled to the
FOR PEER conclusion that it is necessary to analyze the distributions
REVIEW 3 of 20 not only in
the time domain but also in the frequency domain. One of the basic tools used in the theory of signal
It should also be remembered that the real ship’s heading deviations would result from the
processing ismovement
Fourier transformation,
of the ship. In otherwhich words, was used will
the motion for determine
frequency bothdomain
the real analysis.
ship’s heading
This paper is organized as follows:
deviation and the errors in measuring it. Section 2 discusses the measuring principles and the design of
experiments. Based It might
on seem that the comparisonobtained
the measurements between the atcompass headings kwith
time interval = the
5 s,reference heading,recordings of
the heading
determined by Equation (1), will enable a full analysis of changes caused by the ship’s motion.
three types of compass systems
Unfortunately, are this
it appears that presented. Sectionwill
kind of comparison 3 describes
only provide the transient
information on characteristics
absolute and
spectrum analyses
values ofthatheading were carried
deviations for out after
specific heading
moments reduction
of observation, with
giving no the use ofonMATLAB
information their R.2015
character. The
software. Harmonic analysis conducted
components led to the
resulting fromconclusion
long that it is necessary
periods to analyze the distributions
of disturbances by the helmsman and
not only in the time domain but also in the frequency domain. One of the basic tools used in the
ship inertial theory
possibilities were attenuated.
of signal processing Section 4which
is Fourier transformation, provides
was useddiscussion
for frequency ofdomain
the method
analysis. and results.
Finally, conclusions are presented
This paper is organizedin SectionSection
as follows: 5. 2 discusses the measuring principles and the design
of experiments.
The research questions Based on the
this measurements
paper attempts obtained are as 𝑘follows:
at time interval
to answer = 5 s, the heading recordings
of three types of compass systems are presented. Section 3 describes the transient characteristics and
Does the modelanalyses
spectrum of thethat band-stop
were carried finite impulse
out after headingresponse
reduction with(FIR) filter
the use eliminate
of MATLAB low frequency
R.2015
heading distortions?
software. Harmonic components resulting from long periods of disturbances by the helmsman and
ship inertial should
What frequencies possibilitiesbewere
usedattenuated.
to filter Section
out the4 harmonic
provides discussion
componentof the method and results.
resulting from long periods
Finally, conclusions are presented in Section 5.
of disturbances by Thethe helmsman?
research questions this paper attempts to answer are as follows:
The output Does
signal theof which
model compass
of the band-stopisfinite
the impulse
most accurate and filter
response (FIR) has eliminate
the mostlow potential
frequencyfor use as an
heading
autopilot input distortions?
signal?
What frequencies should be used to filter out the harmonic component resulting from long
periods of disturbances by the helmsman?
2. Materials andTheMethods
output signal of which compass is the most accurate and has the most potential for use as
an autopilot input signal?
2.1. Background
2. Materials and Methods
According to their technical specifications, the satellite compass, gyrocompass and FOG compass
2.1. Background
transmit National Marine Electronics Association Interface Standard for communication between
marine electronic devices to
According their technical
(NMEA) specifications,
0183 messages the the
with satellite
HDG compass, gyrocompass
(heading gyro) orand
HDT FOG
(heading true)
compass transmit National Marine Electronics Association Interface Standard for communication
sentence. Thebetween
station for signals recording consisted of a FOG NAVIGAT 3000 and
marine electronic devices (NMEA) 0183 messages with the HDG (heading gyro) or HDT a satellite compass
FURUNO SC50, shown
(heading true) in Figure
sentence. The1,station
which forwere
signalsdeveloped at theofPolish
recording consisted Naval Academy.
a FOG NAVIGAT 3000 and a These were
satellite
installed on the compass
survey FURUNO
vessel ORPSC50, shown in Figure 1, which were developed at the Polish Naval
“HEWELIUSZ”.
Academy. These were installed on the survey vessel ORP “HEWELIUSZ”.

Figure 1. NAVIGAT FOG 3000 and satellite compass FURUNO SC50 recording stations (source:
K. Jaskólski).
Sensors 2019, 19, x FOR PEER REVIEW 4 of 20

Figure 1. NAVIGAT FOG 3000 and satellite compass FURUNO SC50 recording stations (source: K.
Jaskólski).
Sensors 2019, 19, 1942 4 of 20
The investigation was carried out in the area of the Gulf of Gdansk, during hydrographic surveys
from the 8th to 9th of February, 2018, with the course over ground 091.5 deg. and 271.5 deg. according
The investigation was carried out in the area of the Gulf of Gdansk, during hydrographic surveys
to the directions of the hydrographic profiles.
from the 8th to 9th of February, 2018, with the course over ground 091.5 deg. and 271.5 deg. according
At least ten data recordings were carried out. For the first four data recordings, the time of the
to the directions of the hydrographic profiles.
measurements was originally 200 s, but the first calculation indicated the dominant frequencies equal
At least ten data recordings were carried out. For the first four data recordings, the time of the
to 0.01 Hz. Thus, the recording time was extended to no less than 1000 s and the sampling frequency
measurements was originally 200 s, but the first calculation indicated the dominant frequencies equal
of recorded data was changed from 𝑓(𝑠) = 1 Hz to𝑓(𝑠) = 0.2 Hz. The weather conditions during the
to 0.01 Hz. Thus, the recording time was extended to no less than 1000 s and the sampling frequency
test numbers 4, 6 and 7 were: Wind SWbW-2B (Beaufort) and state of the sea 1.
of recorded data was changed from f (s) = 1 Hz to f (s) = 0.2 Hz. The weather conditions during the
According to the technical specifications of the navigational instruments, the accuracy of
test numbers 4, 6 and 7 were: Wind SWbW-2B (Beaufort) and state of the sea 1.
NAVIGAT FOG 3000 was [15]:
According to the technical specifications of the navigational instruments, the accuracy of NAVIGAT
FOG 3000 was [15]: 𝛿 = 0.8° ∙ sec(𝜑 ),
δFOG = 0.8◦ · sec (ϕm ),
while the accuracy of the satellite compass FURUNO SC50 was [16]:
while the accuracy of the satellite compass FURUNO SC50 was [16]:
𝛿 = 0.5° ∙ sec(𝜑 ).
δSC50on
The reference gyrocompass, installed 0.5◦ · secand
= board (ϕmpresented
). in Figure 2, was NAVIGAT X
MK1, with an accuracy of [17]:
The reference gyrocompass, installed on board and presented in Figure 2, was NAVIGAT X MK1,
with an accuracy of [17]: 𝛿𝑀𝐾1 = 0.4° ∙ sec(𝜑 ).
δMK1 = 0.4◦ · sec (ϕm ).
where 𝜑 is the mean geographical latitude.
whereAsϕmthe
is mean geographical
the mean latitude
geographical of the area was 54.6 deg. N, the mean error of each device
latitude.
could be determined. The results of such calculations are shown in Table 1.

Figure 2.
Figure Shipboard gyrocompass
2. Shipboard gyrocompass NAVIGAT
NAVIGAT XX MK1
MK1 as
as aa recording
recordingstation
station(source:
(source: K.
K. Jaskólski).
Jaskólski).

As the mean geographical latitude of the area was 54.6 deg. N, the mean error of each device
Table 1. Mean errors of investigated compass systems.
could be determined. The results of such calculations are shown in Table 1.
The Type of Compass Mean Error [deg.]
Table 1. Mean errors
NAVIGAT FOGof3000
investigated compass
1.4 systems.
TheFURUNO SC50
Type of Compass 0.9 [deg.]
Mean Error
NAVIGAT X MK1 0.7
NAVIGAT FOG 3000 1.4
FURUNO SC50 0.9
2.2. The Scope of Research NAVIGAT X MK1 0.7
Sensors 2019, 19, 1942 5 of 20

2.2. The Scope of Research

To fully and explicitly determine the error values, as well as the degree of their dependence on
the vessel’s motion, it was necessary to compare the directions measured with a reference direction.
Discussion concerning the variation of these errors cannot be described in a deterministic way. One of
the fundamental problems related to the conduct of investigations of compass errors is the selection of
a reference heading. A reference heading can be used for comparison against the heading defined by
the compasses under investigation. In the current investigation, the reference heading was calculated
as the mean of all the devices, according to the formula:

( ni=1 CHFOG )  ( ni=1 CHSC50 ) 

  P   P  
Pn  Pn 

 i=1 
CH ( i ) FOG − n  i=1 
CH ( i ) SC50 − n  

1  n + n

RH =  , (1)

( ni=1 CHMK1 ) 
 P
m  Pn 
CH ( i ) −

 i=1 
 MK1 n  
+

n

where RH is the reference heading, CD(i) is the course distortion for the iteration (i), CH (i)FOG is the
compass heading per FOG compass for the iteration (i), CH (i)SC50 is the compass heading per satellite
compass for the iteration (i), CH (i)MK1 is the compass heading per standard gyrocompass for the
iteration (i), n is the number of iterations, and m is the number of compass systems. Note that in this
study m = 3.
In References [9,18,19], comparisons between the direction defined and the direction of the
reference system for accuracy determination were presented. In these works, the satellite systems
were used to determine the heading and an estimation of the course system accuracy. The methods
performed consisted of comparing the defined direction with the direction of the survey profile, which
may be used to check the accuracy of a shipboard compass for constant errors. Excluding the fact
that the authors reduced the accuracy and were dependent on many factors including visibility, wind
speed, currents, state of the sea, and helmsman’s ability to keep the ship accurately in line with the
survey profile, their research provides a comparison for short periods of observation. Consequently,
the methods described could only be used to determine constant correction, called compass error (CE),
but could not investigate compass accuracy in dynamic conditions.
In the current experiment, recordings were made based on constant heading readings from a
classical gyrocompass with internal correction, a FOG compass and a satellite compass. Exordial, short
data registrations lasting up to 200–350 s were discarded, as they contained too little information. For
further discussion in this paper, heading recording numbers 4, 6, and 7 were taken into account as the
most valuable. The heading recordings from the three types of compasses as a function of time were
presented in Figures 3–5.
 Sensors 2019, 19, x FOR PEER REVIEW 6 of 20

Sensors 2019, 19, x FOR PEER REVIEW 6 of 20

Sensors 2019, 19, 1942 6 of 20

Figure 3. Heading recording No 4: Measured data before data processing.

Figure 3. Heading recording No. 4: Measured data before data processing.

Figure 3. Heading recording No 4: Measured data before data processing.

Figure4.4.Heading
Figure Headingrecording
recordingNo.
No.6:6:Measured
Measureddata
databefore
beforedata
dataprocessing.
processing.

Figure 4. Heading recording No. 6: Measured data before data processing.

 Sensors 2019, 19, x FOR PEER REVIEW 7 of 20
Sensors 2019, 19, 1942 7 of 20

Figure 5. Heading
Figure recording
5. Heading No.7:7:Measured
recording. Measureddata
databefore
beforedata
dataprocessing.
processing.

Analyzing
Analyzing the theheading
headingdistributions,
distributions, presented
presented in inFigures
Figures3–5,
3–5,periodical
periodicaldeviations
deviationsmainlymainly
caused
causedby byyawing
yawingcan canbebeidentified.
identified.The Thedifference
differenceininamplitudes
amplitudesand andthethephases
phasesofofthese
thesedeviations
deviations
shows
showsthe thevarying
varyingcharacter
characterofofboth bothdistributions
distributionsresulting
resultingfrom
fromthe thegyro
gyroelement
elementsuppression
suppressionofof
oscillation
oscillation for the classical gyrocompass. A satellite compass FURUNO SC 50 equippedwith
for the classical gyrocompass. A satellite compass FURUNO SC 50 equipped withMEMSMEMS
(micro-electro-mechanical
(micro-electro-mechanical systems) technology is more resistant to short-period swinging thanaa
systems) technology is more resistant to short-period swinging than
gyrocompass
gyrocompass NAVIGAT NAVIGATXXMK1. MK1. TheThecourse
coursedistortions
distortions depends
depends on on the
the potential
potential to to set
setaatime
timeofof
smoothing function based on the Kalman filtering
smoothing function based on the Kalman filtering methodology [3]. methodology [3].
These
Theseexamples
examplesshow showperceptible
perceptibleresemblances
resemblances of ofthe
theoscillation
oscillationin inall
allindications
indicationsofofthe thethree
three
compasses.
compasses. Certainly, this is the result of the ship’s yaw, which is especially visible in Figures44and
Certainly, this is the result of the ship’s yaw, which is especially visible in Figures and5.5.
However,
However,in in Figure
Figure 5, 5,the
thegradual
gradualapproach
approach of values
of values recorded
recorded fromfrom NAVIGAT
NAVIGAT X MK1 X MK1 to values
to values shown
shown by the remaining systems can be observed. This is the characteristic
by the remaining systems can be observed. This is the characteristic behavior of the classical behavior of the classical
gyrocompass
gyrocompassafter afteralteration
alterationin inthe
the ship’s
ship’s course.
course.
InInFigures 4 and 5 the satellite compass
Figures 4 and 5 the satellite compass and andFOG
FOGindications
indicationsare areanalogous,
analogous, even
even in in terms
terms of
of the
the phase, while in Figure 3 the phase shift in SC 50 indications—in relation
phase, while in Figure 3 the phase shift in SC 50 indications—in relation to the reference system to the reference system
(NAVIGAT
(NAVIGATXXMK MK1)—can
1)—canbe beobserved. Differences are
observed. Differences are even
even upuptoto33degrees,
degrees,while
whilethe theaccuracy
accuracyof
ofeach
each instruments
instruments amounts
amounts approximately
approximately 1 degree.
1 degree. During
During the the experiment
experiment the ship
the ship navigated
navigated with
with
meridional (E and W) courses, so speed deviation, as a possible additional source of error,ofdid
meridional (E and W) courses, so speed deviation, as a possible additional source error,
not
did not appear.
appear.
3. Results
3. Results
3.1. Transient Characteristics and Spectrum Analysis
3.1. Transient Characteristics and Spectrum Analysis
Three data recordings with the use of post-processing methodology were taken into consideration.
Three data recordings with the use of post-processing methodology were taken into
The effect of ship motion and the accuracy of the three compass systems were analyzed. In addition,
consideration. The effect of ship motion and the accuracy of the three compass systems were
the effects of the helmsman’s operation and vessel inertia were analyzed and removed with the use of
analyzed. In addition, the effects of the helmsman’s operation and vessel inertia were analyzed and
spectrum analysis.
removed with the use of spectrum analysis.
Firstly, course distortions CD(i, j) were calculated according to the formula:
Firstly, course distortions 𝐶𝐷(𝑖, 𝑗) were calculated according to the formula:
P  
n (∑𝑛 𝐶𝐻 )

 = [(𝐶𝐻(𝑖, 𝑗) −i=1 CH
𝑖=1 j 𝑗
CD(i,𝐶𝐷(𝑖,
j) = 𝑗)
)],

CH (i, j) − (2)
 n
𝑛  , (2)
Sensors 2019, 19, x FOR PEER REVIEW 8 of 20

where 𝐶𝐷(𝑖,
Sensors 2019, 19,𝑗) is the
x FOR
Sensors 2019, 19, 1942
course
PEER REVIEW for the iteration (𝑖), and the compass type (𝑗) (FURUNO
distortion 8 ofSC
20
8 of 20
50, or NAVIGAT X MK1, or NAVIGAT FOG 3000), 𝐶𝐻𝑗 is the compass heading according to
where 𝐶𝐷(𝑖,type,
instrument 𝑗) isand
the course
𝑛 is thedistortion
number of the iteration (𝑖), and the compass type (𝑗) (FURUNO SC
foriterations.
50, or NAVIGAT X MK1,
(i, j) isdistortions
CDcourse
whereThe or NAVIGAT
the course distortion
results arefor theFOG
iteration
presented in (i),𝐶𝐻
3000),
Figuresand is
thethe
𝑗6–8. compass
compass typeheading according
( j) (FURUNO SC 50,to
instrument
or NAVIGAT type,
X MK1, 𝑛 NAVIGAT
and or is the numberFOGof3000),
iterations.
CH j is the compass heading according to instrument
The course distortions results
type, and n is the number of iterations.are presented in Figures 6–8.
The course distortions results are presented in Figures 6–8.

Figure 6. Heading recording No. 4: Course distortions.

Figure6.6. Heading
Figure Heading recording
recordingNo.
No. 4:
4: Course
Course distortions.
distortions.

Figure 7. Heading
Figure 7. Heading recording
recording No.
No. 6: Course distortions.
6: Course distortions.

Figure 7. Heading recording No. 6: Course distortions.

Sensors 2019,
Sensors 2019, 19,
19, 1942
x FOR PEER REVIEW 99 of 20
of 20

Figure 8.
Figure 8. Heading
Heading recording
recording No.
No. 7: Course distortions.

Secondly,
Secondly, absolute deviation outcomes
outcomes for
for heading
headingrecordings
recordingsNo.
No.4,4,6, 6,
andand 7 are
7 are presented
presented in
in Table
Table 2. 2.

Table 2. Absolute deviation outcomes of compass heading for test No. 4, 6, and 7.
Table 2. Absolute deviation outcomes of compass heading for test No. 4, 6, and 7.
Test
TestNumber
Number NAVIGAT FOG
NAVIGAT FOG 3000
3000 [deg.]
[deg.] FURUNO SC
FURUNO SC 50
50 [deg.]
[deg.] NAVIGATXXMK1
NAVIGAT MK1[deg.]
[deg.]
Test No.44
Test No. 1.53
1.53 1.23
1.23 1.30
1.30
Test
TestNo.
No.66 1.34
1.34 1.30
1.30 1.00
1.00
Test
TestNo.
No.77 1.02
1.02 0.88
0.88 1.18
1.18

The main
The main disturbances
disturbancesin inheading
headingare arenot
notonly
onlythe
theresult
result
ofof the
the ship’s
ship’s unstable
unstable motion,
motion, (changes
(changes in
in heading). Therefore, the ship’s motion was determined by the heading deviation
heading). Therefore, the ship’s motion was determined by the heading deviation (dynamic properties (dynamic
properties
of of the
the ship) and ship)
at the same and
timeat measurement
the same time measurement
errors errors of were
of heading deviations heading deviations
generated by thewere
ship
regardless of whether it was being manually controlled or in automatic mode. The comparison mode.
generated by the ship regardless of whether it was being manually controlled or in automatic of the
The comparison
compass headingof the the
with compass heading
reference with
heading wasthedetermined
reference heading was determined
by the onboard by theaccording
gyrocompass, onboard
gyrocompass,
to the referenceaccording to the reference
heading calculations and heading calculations
hydrographic profile and hydrographic
direction. profile direction.
The comparison delivers
The comparison delivers information about heading deviations for
information about heading deviations for the moments of observations according to the moments ofgyrocompass
observations
according
signal to gyrocompass
recordings signal recordings
without information about without information
their character about The
of changes. theirrecorded
characterassumptions
of changes.
The recorded assumptions
were presented as follows: were presented as follows:
If:
If :∆t = 5 s,then : f (s) = 0.2 Hz
𝛥𝑡 = 5 s,
where f (s) is the sampling frequency and ∆t is the sample period.
then:
The recorded assumptions depended on the errors which appeared in the classical and FOG
𝑓(𝑠) = 0.2 Hz
gyrocompasses.
where 𝑓(𝑠) is the Errors caus a frequency
sampling long periodanddeviation, which
𝛥𝑡 is the in itself
sample depended on the dynamic deviation.
period.
In
Theconclusion, the analysis ofdepended
recorded assumptions the distribution
on the in the time
errors whichdomain was in
appeared insufficient. Thus,
the classical anditFOG
was
necessary to analyze the distribution in the frequency domain. For this purpose,
gyrocompasses. Errors caus a long period deviation, which in itself depended on the dynamic a basic tool in signal
processing
deviation. theory—the Fourier transform theory—was applied. This tool ensured a signal description
determined in the frequency
In conclusion, of thef distribution
domain
the analysis or pulsationinω the
= 2π f . domain was insufficient. Thus, it was
time
x ( t )
necessary to analyze the distribution in the frequency domain. For thisformula
Signal can be presented according to the Fourier integration [6,20,21]:
purpose, a basic tool in signal
processing theory—the Fourier transform Ztheory—was
+∞ applied. This tool ensured a signal
1
description determined in the frequency or(ω
x(t) =domain 𝑓 X )·e jωt dω, 𝜔 = 2𝜋𝑓.
pulsation (3)
2π
Signal 𝑥(𝑡) can be presented according to −∞
the Fourier integration formula [6,20,21]:
where X(ω) is the Fourier transform of signal
1 x (t).
+∞
𝑥(𝑡) = ∫−∞ 𝑋()𝑒 𝑗𝑡 𝑑, (3)
2𝜋
Sensors 2019, 19, 1942 10 of 20

The Fourier transform of signal x(t) is as follows [6,20,21]:

Z +∞
X (ω) = x(t)·e−jωt dt, (4)
−∞

Contrary to the continuous signal, x(t), the distribution of heading is determined by the discrete
signal of the periodical character x(i). The period of sampling for such a signal is given by:

x(i) = x(i∆t), (5)

where i = 0 . . . N−1.
Mathematical notation of the signal is [6,20,21]:
X N−1
x(i) = A(0) + 2
r=1
A(r)· cos [ω0 ri·t + ϕ(r)], (6)

where A(0) is the initial signal amplitude, A(r) is the amplitude of the harmonic signal components,
ω0 is the pulsation of the main components for the harmonic signal, where r = 1, and ϕ(r) is the phase
of the specific harmonic signal component.
The pulsation of the main components for the harmonic signal, where r = 1, is [6,20,21]:

2π
ω0 = , (7)
N∆t

The notation (6) of the signal x(i) gives information about what part of the distribution x(i) has
the particular components in the form r for the harmonic signals of amplitudes A(r), phases ϕ(r)
and initial amplitude A(0). The notation is true for odd N. For even N, r has to be in the range of
1 . . . N2 [6,20,21]. Note that the relationship was found to exist between the evenness of the samples,
and the mathematical notations were presented in odd N elements mathematical apparatus.
Thus, Fourier analysis of the determined signal (6) enabled the distribution of the frequency
domain. The amplitude frequency of signal A(s) is constrained in the lower range of frequency f 0
determined with the observation period T in the upper range of Nyquist frequency f (n), where [6,20,21]:

f (s)
f (n) = , (8)
2

where f (s) is the sampling frequency.

The main task of the mathematical model for the ship’s heading distribution was finding an
amplitude of heading errors. It is more convenient to study the complex spectrum. This is the heading
distribution and the heading disturbances distribution.
According to equation:
CE(i∆t) = CH (i∆t) − RH, (9)

where CE(i∆t) is the heading disturbances for the sample i = 0 . . . N − 1, RH is the reference heading
according to Equation (1), and CH (i∆t) is the compass heading for the sample, i.
The compass heading was maintained according to the gyrocompass readings, with long periods
of disturbances by the helmsman being implemented, as presented in Figures 3–5. Before frequency
analysis, the calculated RH (1) had to be reduced by subtracting the recorded compass headings
according to formula CE(i∆t) (9).
The post-processing data for the FURUNO SC50, NAVIGAT FOG 3000, and NAVIGAT X MK1
after heading reduction (000 deg.) are depicted in Figures 9–11.
Sensors 2019, 19, x FOR PEER REVIEW 11 of 20
Sensors 2019, 19, 1942 11 of 20
Sensors 2019, 19, x FOR PEER REVIEW 11 of 20

Figure 9. Heading recording No 4: Post-processing data with the heading reduction (000 deg.).

Figure9.9.Heading
Figure Headingrecording
recordingNo.
No 4:
4: Post-processing data with the heading
heading reduction
reduction (000
(000 deg.).

Figure 10.Heading
Figure10. Headingrecording
recordingNo.
No. 6:
6: Post-processing
Post-processing data
data with
with the
the heading
heading reduction
reduction (000
(000deg.).
deg.).

Figure 10. Heading recording No. 6: Post-processing data with the heading reduction (000 deg.).
Sensors 2019, 19, x FOR PEER REVIEW 12 of 20

Sensors 2019, 19, x FOR PEER REVIEW 12 of 20

Sensors 2019, 19, 1942 12 of 20

Figure 11. Heading recording No. 7: Post-processing data with the heading reduction (000 deg.).
Figure 11. Heading
Figure11. Heading recording
recording No.
No. 7:
7: Post-processing data with the heading reduction
reduction (000
(000 deg.).
deg.).
Frequency analysis for compass course distortions depends on long-term disturbances. The
distributions
Frequency
Frequency of compass
analysis heading
analysis for distortions
for compass
compass were
course
course equallydepends
distortions
distortions susceptible
dependsontoondisturbances
long-term
long-term for the satellite
disturbances.
disturbances. The
compass,
The FOG
distributions
distributions compass and
of compass
of compass classical
heading
heading gyrocompass,
distortions
distortions and such
were equally
were equally disturbances
susceptible
susceptible depend on ship
to disturbances
to disturbances motion
for the
for the satellite
noises.
satellite The frequency
compass, FOG analyses
compass for compass
and course
classical distortions
gyrocompass, are
and presented
such in Figures
disturbances
compass, FOG compass and classical gyrocompass, and such disturbances depend on ship motion 12–14.
depend on ship
motion
noises. noises. The frequency
The frequency analysesanalyses for compass
for compass course distortions
course distortions are presented
are presented in Figures
in Figures 12–14. 12–14.

Figure 12. Heading recording No. 4: Single-sided amplitude spectrum of NAVIGAT FOG 3000,
Figure 12. Heading recording No. 4: Single-sided amplitude spectrum of NAVIGAT FOG 3000,
NAVIGAT X MK1, and FURUNO SC50 course distortions, determined using the fast Fourier
NAVIGAT X MK1, and FURUNO SC50 course distortions, determined using the fast Fourier
Figure 12.algorithm.
transform Heading recording No. 4: Single-sided amplitude spectrum of NAVIGAT FOG 3000,
transform algorithm.
NAVIGAT X MK1, and FURUNO SC50 course distortions, determined using the fast Fourier
transform algorithm.
The single-sided amplitude spectrum of three types of compasses for the heading recording No.
4, depicted in Figure amplitude
The single-sided 12, presents a harmonic
spectrum partial
of three types 0.015
of of Hz for for
compasses thethecourse distortions,
heading with
recording No.a
maximum amplitude spectrum of 1.3 deg. These course distortions depend on the implementation
4, depicted in Figure 12, presents a harmonic partial of 0.015 Hz for the course distortions, with a
of long periods
maximum of helmsman/autopilot
amplitude disturbances
spectrum of 1.3 deg. and the
These course ship’s inertial
distortions depend abilities.
on the implementation
of long periods of helmsman/autopilot disturbances and the ship’s inertial abilities.
Sensors 2019, 19, 1942 13 of 20
Sensors 2019, 19, x FOR PEER REVIEW 13 of 20

Figure 13. Heading recording No. 6: Single-sided amplitude spectrum of NAVIGAT FOG 3000,
NAVIGAT X MK1, and FURUNO SC50 course distortions, determined using the fast Fourier
transform algorithm.

Similar investigations were carried out with the use of heading recording No. 6. The single-sided
amplitude spectrum, presented in Figure 13, depicts a harmonic partial of 0.01 Hz for the course
distortions, with maximum amplitude spectrum of 0.9 deg. As in the previous case, the greatest
Figure 13. Heading
Figure 13. Heading recording
recording No. 6: Single-sided
No. 6: Single-sided amplitude
amplitude spectrum
spectrum of
of NAVIGAT
NAVIGAT FOGFOG 3000,
3000,
course distortions
NAVIGAT
depended on the disturbances created by the helmsman/autopilot and the ship’s
NAVIGAT X MK1, and FURUNO SC50 course distortions, determined using the fast Fourier
X MK1, and FURUNO SC50 course distortions, determined using the fast Fourier
inertial abilities.
transform algorithm.
transform algorithm.

Similar investigations were carried out with the use of heading recording No. 6. The single-sided
amplitude spectrum, presented in Figure 13, depicts a harmonic partial of 0.01 Hz for the course
distortions, with maximum amplitude spectrum of 0.9 deg. As in the previous case, the greatest
course distortions depended on the disturbances created by the helmsman/autopilot and the ship’s
inertial abilities.

Figure 14. Heading

Figure 14. Heading recording
recording No. 7: Single-sided
No. 7: Single-sided amplitude
amplitude spectrum
spectrum of
of NAVIGAT
NAVIGAT FOGFOG 3000,
3000,
NAVIGAT
NAVIGAT X MK1, and FURUNO SC50 course distortions, determined using the fast Fourier
X MK1, and FURUNO SC50 course distortions, determined using the fast Fourier
transform
transformalgorithm.
algorithm.

The single-sided amplitude spectrum of three types of compasses for the heading recording
The same research results on the basis of heading recording No. 7 registered data, presented in
No. 4, depicted in Figure 12, presents a harmonic partial of 0.015 Hz for the course distortions, with a
Figure 14, exhibit a harmonic partial of 0.01 Hz and an amplitude spectrum of 0.68 deg.
maximum amplitude spectrum of 1.3 deg. These course distortions depend on the implementation of
The low-frequency distortions below 𝑓 = 0.02 Hz are a phenomenon of undetermined
long periods
Figure 14.of Heading
helmsman/autopilot
recording No.disturbances
7: Single-sidedandamplitude
the ship’sspectrum
inertial abilities.
of NAVIGAT FOG 3000,
fluctuations. This phenomenon is related to Schuler oscillations for the frequency distortions
Similar
NAVIGAT investigations
X MK1, and were carried SC50
FURUNO out with the distortions,
course use of heading recording
determined No. the
using 6. The
fast single-sided
Fourier
𝑓 = 0.0018 Hz, presented in Figures 12–14. This oscillations make a standard compass NAVIGAT X
transform
amplitude algorithm.
spectrum, presented in Figure 13, depicts a harmonic partial of 0.01 Hz for the course
MK1 heading deviation from 0.5 deg. to 0.65 deg., which was a source of heading for the helmsman.
distortions, with maximum amplitude spectrum of 0.9 deg. As in the previous case, the greatest
If the helmsman had diverted the course according to the standard gyrocompass, the course
course The same research
distortions results
depended ononthethe basis of heading
disturbances recording
created No. 7 registered data,
by the helmsman/autopilot andpresented in
the ship’s
distortions would have been duplicated by the FOG and satellite compasses.
Figure 14,
inertial exhibit a harmonic partial of 0.01 Hz and an amplitude spectrum of 0.68 deg.
abilities.
The low-frequency
TheModel
same of research distortions
results on the below
basis 𝑓 = 0.02recording
Hz are aNo.phenomenon of undetermined
3.2. The the Band-Stop Filter with theof heading
Finite Impulse Response 7 registered data,
(FIR) presented in
fluctuations. This phenomenon is related to Schuler oscillations for the
Figure 14, exhibit a harmonic partial of 0.01 Hz and an amplitude spectrum of 0.68 deg. frequency distortions
𝑓 = 0.0018 Hz, presented
The low-frequency in Figures
distortions 12–14.
below f =This
0.02oscillations make a standard
Hz are a phenomenon compass NAVIGAT
of undetermined X
fluctuations.
MK1 heading deviation from 0.5 deg. to 0.65 deg., which was a source of heading for
This phenomenon is related to Schuler oscillations for the frequency distortions f = 0.0018 Hz, presented the helmsman.
If Figures
in the helmsman
12–14. This hadoscillations
diverted makethe course according
a standard compass toNAVIGAT
the standard
X MK1 gyrocompass, the course
heading deviation from
distortions would have been duplicated by the FOG and satellite compasses.
0.5 deg. to 0.65 deg., which was a source of heading for the helmsman. If the helmsman had diverted

3.2. The Model of the Band-Stop Filter with the Finite Impulse Response (FIR)
Sensors 2019, 19, 1942 14 of 20

the course according to the standard gyrocompass, the course distortions would have been duplicated
by the FOG and satellite compasses.

3.2. The
Sensors Model
2019, 19, x of
FORthePEER
Band-Stop
REVIEWFilter with the Finite Impulse Response (FIR) 14 of 20

One solution for disabling low frequency heading distortions is to employ the FIR model presented
One solution for disabling low frequency heading distortions is to employ the FIR model
in Figure 15.
presented in Figure 15.

Figure 15. Band-stop, finite impulse response (FIR) filter to reduce low-frequency heading distortions
Figure 15. Band-stop, finite impulse response (FIR) filter to reduce low-frequency heading distortions
(source: MATLAB R.2015).
(source: MATLAB R.2015).
where:
where:
f (s) = 0.2 Hz, the sampling frequency,
𝑓(𝑠) = )0.2
f (pass1 = Hz, theHz,
0.005 sampling
the firstfrequency,
passband frequency,
𝑓(𝑝𝑎𝑠𝑠1) = 0.005 Hz,
f (stop1) = 0.0075 Hz, the the first
first passband
stopband frequency,
frequency,
𝑓(𝑠𝑡𝑜𝑝1) = 0.0075 Hz, the
f (stop2) = 0.0125 Hz, the second first stopband
stopband frequency,
frequency,
𝑓(𝑠𝑡𝑜𝑝2) = 0.0125 Hz, the second stopband
f (pass2) = 0.015 Hz, the second passband frequency. frequency,
𝑓(𝑝𝑎𝑠𝑠2) = 0.015 Hz, the second passband frequency.
According to the filter model assumptions, and the results of the single-sided amplitude spectrum
According
(presented to the12–14),
in Figures filter model assumptions,
the bandwidth from fand the
(pass1 ) =results
5 mHzoftothe single-sided
f (pass2 ) = 15 mHzamplitude
will be
spectrum (presented in Figures 12–14), the bandwidth from 𝑓(𝑝𝑎𝑠𝑠1) =
attenuated, and the other frequencies will be passed. Band-stop FIR was proposed to eliminate the 5 mHz to 𝑓(𝑝𝑎𝑠𝑠2) =
15 mHz will be attenuated, and the other frequencies will be passed. Band-stop
helmsman’s distortions. The helmsman’s disturbances frequency could not be measured directly, FIR was proposed to
eliminate
and the helmsman’s
hence filter parameters were distortions.
estimated The helmsman’s
according to the disturbances frequency
system’s accuracy could and
investigation not the
be
measured directly, and hence filter parameters were estimated according to the
analyzed spectrum results. With the use of a band-stop finite impulse response filter, it is possible tosystem’s accuracy
investigation
filter out harmonicand the analyzed spectrum
components resulting results.
from longWith the use
period of a band-stop
helmsman’s finite impulse
disturbances response
and ship inertia
filter, it is possible to filter out harmonic components resulting from long
possibilities. The recorded signal and the signal after filtration are presented in Figures 16–18. period helmsman’s
disturbances and ship inertia possibilities. The recorded signal and the signal after filtration are
presented in Figures 16–18.
 eliminate the helmsman’s distortions. The helmsman’s disturbances frequency could not be
measured directly, and hence filter parameters were estimated according to the system’s accuracy
investigation and the analyzed spectrum results. With the use of a band-stop finite impulse response
filter, it is possible to filter out harmonic components resulting from long period helmsman’s
disturbances and ship inertia possibilities. The recorded signal and the signal after filtration are
Sensors 2019, 19, 1942
presented in Figures 16–18. 15 of 20

Sensors 2019, 19, x FOR PEER REVIEW 15 of 20

Figure 16. Heading recording No. 7. Heading amplitude for recorded and after filtration signal for (1)
NAVIGAT
Figure FOG 3000;
16. Heading (2) FURUNO
recording No. 7. SC 50; andamplitude
Heading (3) NAVIGAT X MK1. and after filtration signal for
for recorded
(1) NAVIGAT FOG 3000; (2) FURUNO SC 50; and (3) NAVIGAT X MK1.

Figure 17. Heading recording No. 4. Heading amplitude for recorded and after filtration signal for
(1)Figure 17. Heading
NAVIGAT recording
FOG 3000; No. 4. Heading
(2) FURUNO amplitude
SC 50; and for recorded
(3) NAVIGAT X MK1.and after filtration signal for (1)
NAVIGAT FOG 3000; (2) FURUNO SC 50; and (3) NAVIGAT X MK1.
 Figure 17. Heading recording No. 4. Heading amplitude for recorded and after filtration signal for (1)
Sensors 2019, 19, 1942 FOG 3000; (2) FURUNO SC 50; and (3) NAVIGAT X MK1.
NAVIGAT 16 of 20

Figure 18. Heading recording No. 6. Heading amplitude for recorded and after filtration signal for
Figure 18. Heading recording No. 6. Heading amplitude for recorded and after filtration signal for (1)
(1) NAVIGAT FOG 3000; (2) FURUNO SC 50; and (3) NAVIGAT X MK1.
NAVIGAT FOG 3000; (2) FURUNO SC 50; and (3) NAVIGAT X MK1.
According to the research outcomes presented in Figures 16–18, the heading amplitude for the
According to the research outcomes presented in Figures 16–18, the heading amplitude for the
filtered data was reduced. Additionally, the harmonic component resulting from long periods of
filtered data was reduced. Additionally, the harmonic component resulting from long periods of
disturbances by the helmsman and the ship’s inertial possibilities were attenuated.
disturbances by the helmsman and the ship’s inertial possibilities were attenuated.
According to the research outcomes for the filtered data presented in Figures 19–21, the heading
According
Sensors 2019, 19, toPEER
x FOR the research
REVIEW outcomes for the filtered data presented in Figures 19–21, the 16
heading
of 20
amplitude (calculated according to Equation (2)) was halved. This is easily observed by comparing
amplitude (calculated according to Equation (2)) was halved. This is easily observed by comparing
Figures 6–8 with Figures 19–21, respectively.
Figures 6–8 with Figures 19–21, respectively.

Figure 19. Heading

Figure19. Heading recording
recording No.
No. 7:
7: Heading
Heading distortions
distortions of
of the
the filtered
filtered signal.
signal.
 Figure 19. Heading recording No. 7: Heading distortions of the filtered signal.
Sensors 2019, 19, 1942 17 of 20

Sensors 2019, 19, x FOR PEER REVIEW 17 of 20

Figure 20.Heading
Figure20. Headingrecording
recordingNo.
No.4:4:Heading
Headingdistortions
distortionsof
ofthe
thefiltered
filteredsignal.
signal.

Figure 21. Heading

Figure 21. Heading recording
recording No.
No. 6:
6: Heading
Heading distortions
distortions of
of the
the filtered
filtered signal.
signal.

The band-stop finite impulse response filter halved the amplitude of the heading distortions and
The band-stop finite impulse response filter halved the amplitude of the heading distortions and
reduced the mean errors for the three heading recordings. The research outcomes are presented in
reduced the mean errors for the three heading recordings. The research outcomes are presented in
Tables 3 and 4.
Tables 3 and 4.
Table 3. Compass heading mean distortion outcomes with the use of a filtered signal for heading
Table 3. Compass
recording heading
No. 4, 6, and 7. mean distortion outcomes with the use of a filtered signal for heading
recording No. 4, 6, and 7.
Test Number NAVIGAT FOG 3000 [deg.] FURUNO SC 50 [deg.] NAVIGAT X MK1 [deg.]
Test Number NAVIGAT FOG 3000 [deg.] FURUNO SC 50 [deg.] NAVIGAT X MK1 [deg.]
Test No.
Test 4 4
No. −0.10
−0.10 0.01
0.01 −0.15
−0.15
Test No. 6
Test No. 6 0.13
0.13 −0.15
−0.15 0.02
0.02
Test No. 7 0.14 0.01 −0.15
Test No. 7 0.14 0.01 −0.15

Table 4. Compass heading mean error outcomes with the use of a filtered signal for heading recording
No. 4, 6, and 7.

Test Number NAVIGAT FOG 3000 [deg.] FURUNO SC 50 [deg.] NAVIGAT X MK1 [deg.]
Test No. 4 0.54 0.52 0.52
Sensors 2019, 19, 1942 18 of 20

Table 4. Compass heading mean error outcomes with the use of a filtered signal for heading recording
No. 4, 6, and 7.

Test Number NAVIGAT FOG 3000 [deg.] FURUNO SC 50 [deg.] NAVIGAT X MK1 [deg.]
Test No. 4 0.54 0.52 0.52
Test No. 6 0.71 0.70 0.68
Test No. 7 0.48 0.51 0.53

Mean errors of the investigated compass systems (illustrated in Tables 1 and 4) indicate a double
reduction of potential errors after FIR usage, which results in the three compass systems having
similar accuracy. Similar conclusions can be drawn after analyzing the compass heading mean
distortion outcomes.

4. Discussion
The comparison of three compass systems, FURUNO SC 50, NAVIGAT X MK1, and NAVIGAT
FOG 3000, was elaborated in both the time and frequency domains. Investigations were conducted
during bathymetric works executed by the survey ship ORP “HEWELIUSZ”. Measurements for
data analysis were taken when the ship navigated in the directions of nearly E and W, which was
beneficial for the diminution of errors for all instruments. Examples of the registration (Figures 3–5)
indicated that the ship’s direction of movement had no impact. The registrations also appeared to
suggest that there were no essential differences in the accuracy of the examined compasses; although
the uncertainties (errors) were different between consecutive tests (around 0.3 deg.), the differences
between the heading recordings from the compasses in the same test were similar.
The low-frequency distortions were in the band of 0.01 Hz < f < 0.02 Hz This could be related to
the Schuler oscillations of the NAVIGAT X MK1 compass, which was the source of information for the
helmsman and—as consequence—was duplicated by the remaining compasses as the ship, in fact,
navigated along such courses.
Analysis of the courses taken in the time domain gave the opportunity to diagnose the problem
deeply and to eliminate the helmsman’s impact on the gyrocompass’ errors. From this experiment,
we established that the uncertainty (errors) of the three examined instruments were similar. It was
found to be approximately 1 to 1.5 deg., but it must be stressed that the experiment took place in good
weather conditions, which was a requirement for the bathymetric measurements to be taken. Owing
to the ship’s movement on a stable direction for the bathymetric profiles, a specific methodology for
calculating the reference direction was proposed.
The Fourier transform was used for more accurate analysis. The frequency domain analysis of
the ship’s motion distortions creates a new research opportunity with respect to the ship’s inertia,
and weather conditions impact, etc. The frequency analysis of the compass errors constitutes a useful
tool for investigating a simultaneous data error comparison. Frequency analysis enables the isolation of
errors according to the ship’s yawing. This errors appear in data recordings regardless of the compass
type. Frequency analysis creates an opportunity for analysis of the outer factors of the measurement
process and provides a better methodology for compass error estimation. This method could be very
important in estimating errors of navigational devices. With the use of a band-stop finite impulse
response filter, it is possible to filter out the harmonic components resulting from long periods of
disturbances by the helmsman and a ship’s inertia possibilities.

5. Conclusions
This paper discusses the accuracy of three different ships compasses. On the basis of parallel
recordings made by the different devices installed on the same ship, the authors were able to investigate
the uncertainty of measurements in the context of both amplitude and frequency. In the real experiment,
it was checked that the price indications of different compasses differed insignificantly.
Sensors 2019, 19, 1942 19 of 20

The frequency analysis of compass errors constitutes a useful tool for investigating simultaneous
data error comparison. It was demonstrated that frequency analysis enables the isolation of the ship’s
yawing errors. The model of the band-stop FIR is a good tool for eliminating low-frequency heading
distortions. According to the filter model assumptions, and the results of a single-sided amplitude
spectrum, the bandwidth from f (pass1) = 5 mHz to f (pass2) = 15 mHz will be attenuated and
other frequencies will be passed. Band-stop FIR was proposed to eliminate helmsman distortions.
After band-stop FIR usage, the accuracy of the three types of compass systems appeared to be
comparable. During the experiment, the ship navigated according to information from a NAVIGAT X
MK1 classical gyrocompass. Consequently, a low-frequency component existed in each registration,
probably being descended from the dynamic properties of this onboard compass. The association
between the dynamic properties of the ship and the properties of the gyrocompass is interesting.
If a different compass—without this low frequency oscillation—was the source of information for
the heading, how would the remaining systems of the ship behave? This is a potential direction for
future investigations.

Author Contributions: The statements: conceptualization, F.A.; methodology, J.K.; P.P.; software, P.P.; J.K.;
validation, J.K.; P.P.; formal analysis, F.A.; J.K.; P.P.; investigation, F.A.; J.K.; P.P.; data curation, J.K.; P.P.;
writing—original draft preparation, J.K.; P.P.; F.A.
Funding: This research received no external funding.
Acknowledgments: The authors would like to thank two anonymous reviewers for their time devoted to improve
the paper and constructive criticism. We also express our gratitude to Marcin Jaskólski, from Gdańsk University
of Technology for his expertise concerning the manuscript editing.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Felski, A. Gyrocompasses—Their Condition and Direction of Development. TransNav 2008, 2, 55–59.
2. Szymak, P.; Malec, M.; Morawski, M. Directions of development of underwater vehicle with undulating
propulsion. Pol. J. Environ. Stud. 2010, 19, 107–110.
3. Praczyk, T. Correction of navigational information supplied to biomimetic autonomous underwater vehicle.
Pol. Marit. Res. 2018, 1, 13–23. [CrossRef]
4. Felski, A. Exploitative properties of different types of satellite compasses. Annu. Navig. 2010, 16, 33–40.
5. Felski, A.; Nowak, A. Practical experiences from the use of the satellite compass. In Proceedings of the ENC
Conference, Parthenope University, Naples, Italy, 4–6 May 2009.
6. Felski, A.; Mi˛esikowski, M. Some Method of Determining the Characteristic Frequencies of Ship’s Yawing
and Errors of Ship’s Compasses During the Sea-Trials. Annu. Navig. 2000, 2, 17–24.
7. Spielvogel, A.R.; Whitcomb, L.L. Preliminary Results with a Low-Cost Fiber-Optic Gyrocompass System.
In Proceedings of the OCEANS MTS/IEEE Conference, Washington, DC, USA, 19–22 October 2015.
8. Johnson, B.R.; Cabuz, E.; French, H.B.; Supino, R. Development of a MEMS Gyroscope for Northfinding
Applications. In Proceedings of the IEEE-ION Position Location and Navigation Symposium, Palm Springs,
CA, USA, 4–6 May 2010.
9. Ueno, M.; Santerre, R.; Babineau, S. Impact of the Antenna Configuration on GPS Attitude Determination.
In Proceedings of the 9th World Congress of the International Association of the Institutes of Navigation,
Amsterdam, The Netherlands, 18–21 November 1997.
10. Goodall, C.; Scannel, B. The Battle between MEMS and FOGs for Precision Guidance. Available
online: https://www.analog.com/en/technical-articles/the-battle-between-mems-and-fogs-for-precision-
guidance.html (accessed on 25 December 2018).
11. Niu, X.; Nassar, S.; Syed, Z.; Goodall, C.; El-Sheimy, N. The Development of A MEMS-Based Inertial/GPS
System for Land-Vehicle Navigation Applications. In Proceedings of the 19th International Technical
Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2006), Fort Worth, TX, USA,
26–29 September 2006; pp. 1516–1525.
12. Groves, P.D. Principles of GPS, Inertial, and Multisensor Integrated Navigation Systems; Artech House: Boston,
MA, USA, 2008; pp. 58–521.
Sensors 2019, 19, 1942 20 of 20

13. Britting, K.R. Inertial Navigation Systems Analysis; Artech House: Boston, MA, USA, 2010; pp. 22–249.
14. Basterretxea-Iribar, I.; Sotes, I.; Uriarte, J.I. Towards an Improvement of Magnetic Compass Accuracy and
Adjustment. J. Navig. 2015, 69, 1325–1340. [CrossRef]
15. Sperry Marine. Operation, Installation and Service Manual NAVIGAT 3000 Fiber-Optic Gyrocompass and Altitude
Reference System; Sperry Marine: Hamburg, Germany, 2014; pp. 9–98.
16. Furuno Electric CO. LTD: Operator’s Manual, The Satellite Compass SC-50; Furuno: Nishinomiya, Japan, 2011;
pp. 7–101.
17. Sperry Marine. Operation, Installation and Service Manual NAVIGAT X MK1 Digital Gyrocompass Systems;
Sperry Marine: Hamburg, Germany, 2013; pp. 16–180.
18. Comp, C. Optimal Antenna Configuration for GPS Based Attitude Determination. In Proceedings of the
ION-GPS, Salt Lake City, UT, USA, 22–24 September 1993.
19. Mi˛esikowski, M.; Felski, A. Some Aspects of DGPS Based Heading Determination. Geodezja Kartografia 1999,
XLVIII, 97–104.
20. Oppenheim, A.V.; Schafer, R.W. Discrete-Time Signal Processing, 2nd ed.; Prentice-Hall: New Jersey, NY, USA,
1989; pp. 541–755.
21. Otnes, R.K.; Enochson, L. Analiza Numeryczna Szeregów Czasowych, 1st ed.; Wydawnictwo
Naukowo-Techniczne: Warsaw, Poland, 1978; pp. 122–258.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).