Draft copy (presented at IFAC MMM Workshop on Automation in Mining, Mineral and Metal Industry, 2009) 1/6

Wireless SISO Channel Propagation Model for Underground Mines

Walter Grote

Electronic Engineering Department, Universidad Técnica Federico Santa María,

Valparaíso, Chile (e-mail: [email protected])

Abstract: Modern mining procedures aim to replace human resources directly involved in the mineral
extraction process by automatic or remote control systems. Data may be collected from sensors, or data
and/or video may be transmitted by telecommunication equipment. The actual technology trend is to do
these transmissions using wireless technology (WSN, Wi-Fi, among others). This work summarizes the
experience that has been collected from characterizing a Single-Input-Single-Output (SISO) large scale
870 MHz and 2.45 GHz wireless propagation channel in an underground copper mine. Also the Rician K
factor and channel coherence time are calculated for a 2.45 GHz mobile link.
Keywords: telecommunication, wireless propagation, underground copper mines

at almost grazing angles on the walls and hence suffer low

1. INTRODUCTION
loss. At lower frequencies (than 500 MHz) the losses in the
Open or underground mine working conditions provide walls obviously increase while at frequencies higher than 2
hazardous working conditions and therefore there has been an GHz, the excess loss due to the wall roughness becomes
increasing trend to replace humans by locally or remotely increasingly important as has been observed by Emslie et. al.
controlled intelligent machines. In both cases remote data (1975). Thus the working range of radio frequencies for free
collection and control are essential. Wireless technology is an propagation is considered to lie roughly between 500 MHz –
attractive solution to perform this task, since it provides 2 GHz”. Underground mine tunnels are imperfect
mobility, communications and reduces installation costs. In waveguides, since its walls are not perfect conductors, the
spite of the increasing trend of using wireless electrical parameters are not known and vary according to
communications, electromagnetic (EM) wave propagation in rock composition within the mine. Values of the cutoff
tunnels is still not well understood. In part, this can be traced frequencies depend on the shape and the transverse
to the fact that wireless technologies are aimed to achieve dimensions of the tunnel and the given propagation mode,
higher transmission bandwidths, which calls for higher which vary for different carrier frequency values, as it was
carrier frequencies. EM propagation, radiated by antennas or experimentally confirmed, Deryck (1978).
leaky conductors can be traced to the 40’s, Shanklin (1947).
Significant contributions were made by James Wait and have 2.1 Reported Attenuation Values
been very well summarized by Tripp (2000). In this
publication, we present the wireless propagation channel Dudley and Mahmoud (2006) found that the attenuation for
characteristics derived from measurements performed in the the lower order modes for the TE0m modes are much less
Chilean copper mine El Teniente, said to be the largest
radius tunnel. For εr = 12, when excited by a linear current
severe than for the TM0m case at a 1GHz frequency in a 2m
underground excavation of its kind in the world, located 50
kms east of Rancagua, 3269 m above sea level, in the Andean source, it is 1.098 dB/100m for the TE01 mode. Changsen
mountain range. This mine, which began activities in 1904, (2006) summarized EM propagation characteristics of UHF
has 2400 kilometers of underground drifts. El Teniente radio wave in circular, rectangular, arched, ellipse, trapezium
produces 404,738 metric tons of fine copper per year using
α can be predicted by (1), when the transmission frequency
and semi-circle tunnels. In his publication the attenuation rate
panel caving methods. Different layers of tunnels are named
Teniente N, being N=4 at a higher level than N=8. exceeds the cutoff frequency.

xλ 2 ⎡εr 1 ⎤
α = + 3⎥
ε r − 1 ⎢⎣ w
2. MINE TUNNEL LARGE SCALE PATH LOSS MODELS
h ⎦
(1)
3

Tunnels can be treated as waveguides and therefore

transverse electric and magnetic modes may propagate above λ is the carrier’s wavelength, the internal free space
a certain cutoff frequency. This effect was described by electricity parameter is (ε0, μ0), the exterior tunnel magnetic
conductivity rate is μ0, εr is relative dielectric constant, w is
Mahmoud (1974), who stated that “frequencies above about
500 MHz propagate through the tunnel via successive
the maximum width and h is the height of the tunnel. In
reflections from the tunnel walls. Due to the large dimensions
general, for rectangular, oval, circular, trapezoid semicircle
of the cross section relative to the wavelength, the waves fall
arch tunnels 4.34 < x < 5.13. In the case of a circular tunnel, x
h =2a, a = 2m radius tunnel, εr = 12, α = 4.7 dB/100m. This
= 5.09 and if (1) is evaluated for a 1 GHz frequency in a w = Another model that is frequently used to calculate received
power from a transmitter is the 2 Ray model, where it is
result is similar to the one published by Dudley and assumed that there is one reflection point of the signal
Mahmoud (2006). It is of further interest to our study to between the transmit and receive antenna, Rappaport, (2000).
reproduce Fig. 1, which shows the relation between the It can be shown to be approximated by equation (3), where
predicted attenuation and w/λ. hTx and hRx are the relative heights of the transmitting and
receiving antennas, above ground.

P G G λ2 ⎛ ⎡ 2πh h ⎤ ⎞
PRx (dBm) = 10 log10 Tx Tx Rx2 ⎜⎜ 2 sin⎢ Tx Rx ⎥ ⎟⎟
(4π ⋅ d ) ⎝ ⎣ λ ⋅ d ⎦ ⎠
2

(3)

2.2 Experimental Setup for El Teniente Copper Mine

In the literature reviewed by this author there is no evidence

of work to establish large scale path loss of electromagnetic
wave propagation in copper mines. This work summarizes
the experience that has been collected from characterizing the
SISO (Single Input Single Output) wireless propagation
Fig. 1. Relation between attenuation and w/λ, while the channel in the Chilean underground copper mine of El
diameter is a constant, Changsen (2006). Teniente.

Fig. 1 confirms Mahmoud’s (1974) statement that for Due to the fact that the El Teniente mine facilities are still in

(w/λ < 1). Communications in underground mines will

frequencies below 500 MHz, attenuations are extremely large full production, special measuring equipment had to be
designed and implemented to perform these measurements

frequencies happen to be such that w/λ < 15, because

greatly benefit from the use of leaky coaxial cables if carrier during regular working hours. A portable transmitter box was
built using two oscillators, the ZOS-1025 VCO (voltage
attenuations will be large at those frequencies and these controlled oscillator) from Minicircuits®, delivering 10 mW

section dimensions of the tunnel w > 15λ, the attenuation rate

cables will provide good signal coverage. When the cross in the 685 – 1025 MHz range and a NS2-2000104 M2
NovaSource® 10 dBm signal generator operating in the 2 –
depends less on the area of the cross section and the wave 2.5 GHz range, (Novasource (2008)). The transmitters were
transmits almost the same as in free space. Therefore, housed in a plastic toolbox, to prevent damage caused by
technologies like WSN and Wi-Fi benefit from the water and dirt, as shown in Fig. 2. The box included
propagation characteristics in underground mines since at rechargeable batteries and voltage regulators to provide stable
2.45 GHz, w/λ > 32.6 for tunnels 4m wide or above, resulting power supplies for the electronics.
in attenuations that are less than 2 dB/km. Benzakour et. al.
(2004), Boutin et.al. (2006), Li et.al. (2007), Kennedy and
Foster (2006) report that at distances typically less than 40 m Measurements at 870 MHz were carried out using the
from the transmitter, attenuation can be modelled best using portable ZOS-1025 transmitter connected to a 9 dBi Yagi
the LOS (Line of sight), free-space (Friis) path loss model antenna, as seen in Fig. 2, and the Anritsu MS2711D
represented by equation (2), Spectrum Analyzer, provided with an Anritsu 2000-1032
antenna.

PTx ⋅ GTx ⋅ GRx ⋅ λ2

PRx (dBm) = 10 log10
(4π ⋅ d )2
(2)

where PTx and PRx are the transmission and receive powers,
GTx and GRx the transmission and receive antenna gains and d
the separation between the 2 antennas. The before mentioned
authors also state that at distances larger than 40 m the tunnel
starts behaving like a waveguide, resulting in attenuation
much lower than the one that may be predicted by equation
(1) (however, none of the 4 before mentioned publications
invoke explicitly equation (1)).
Fig. 2. 870 MHz portable Transmitter with antenna
 -40
XC-30
-50 XC-40
XC-50
-60

Received power, dBm

LOS
-70 2 Ray

-80

-90

-100

-110

-120

-130
4 44 84 124 164 204 244 284 324 364 404

distance, m

Fig. 4. Received power levels as a function of distance

Fig. 3. Path Loss measurement locations in El Teniente 8. between antennas, compared to analytical models, 870
MHz
2.3 Experiment Outcomes, Discussion and 870 MHz Model
Therefore, the LOS model seems appropriate to represent the
Fig. 3 shows the 3 tunnels: XC-30, XC-40 and XC-50, at large scale propagation loss model at 870 MHz and can be
Teniente 8 level, where the experiments were conducted; the largely attributed to the waveguide property of the tunnel.
transmitter was located at positions Tx 1, Tx2 and Tx3, at the
centre of each drift respectively, while moving with the 2.4 Experiment Outcomes, Discussion and 2.45 GHz Model
Spectrum Analyzer along the tunnel to record received power
levels. Each tunnel cross section was 4 m wide and high.
These laneways had railways and cabling, but at the time the Another set of measurements was performed using the M2
data was collected, no mining activity was taking place. Novasource to transmit at 2.45 GHz, replacing the 9 dBi Yagi
antenna for a 2.4 GHz 8 dBi Hyperlink Technologies patch
antenna.
The experimental large-scale fading model was obtained
following the Saleh-Valenzuela (1987) procedure, recording However, since large and small-scale fading characteristics of
received power measurements every 8 m along the tunnel mobile broadband communications at this frequency were to
pathway – when close to the transmitter – and at more distant be determined, a portable sampling receiver was
spots at remote places. At each data collection spot, 10 implemented to perform measurements. The 2.45 GHz
measurements were performed with the spectrum analyzer receiver shown in Fig. 5 is provided with a ZFSC 2-2500
moving on 3 m diameter circle. The data was then averaged front end amplifier connected to the 2.45 GHz 8 dBi patch
and the standard deviation was calculated for each spot. antenna, followed by a ZFM-4212 Mixer, which produces a
Averages for each spot are shown in Fig. 4. In tunnel XC-30, 70 MHz intermediate frequency by means of a Novasource-
the largest standard deviation registered was 5.1 dB, in XC- M2 local oscillator, set to 2,443 MHz.
40 it was 8.1 dB and in XC-50, 6.7 dB.

Fig. 4 shows the average values obtained at each location, as

compared to the free-space LOS equation (2) and the 2 ray
model equation (3) as well. The measurement outcomes
clearly show that propagation characteristics within the
tunnel are not following the 2 ray model. Instead, because of
the wave guiding properties of the tunnels, they fit the LOS,
free-space equation quite well. Measurements outcomes
performed in tunnel XC-40 show a significant 18 dB drop in
the 90-110 m range, but thereafter follow the LOS pattern
quite closely. The 3.3 dB/100m attenuation value at the far
end of the tunnel from Fig. 4 is similar to the 3.51 dB/100m

assuming εr = 5 and h = w = 4m.

that can be obtained from (1) for a rectangular waveguide

Fig.5. 2.45 GHz sampling receiver

There was no need for an input filter, since no interference distance, m
signals in the neighbourhood of the carrier frequency 2.45 4 24 44 64 84 104 124
0
GHz were recorded in the measuring areas with the Spectrum
Analyzer. -20

Received Power, dBm

Experimental LOS 2 Ray
The intermediate frequency signal was fed to a ZFL-500 -40
amplifier followed by 70 MHz Precision Devices FTR 4030
-60
bandpass filter, having a bandwidth of 7.5 kHz. The output of
this filter was once more amplified to feed a Frankonia® -80
PMS 1080 RF power data acquisition system, which samples
the signal at a rate of 100 ms for the large-scale propagation -100

channel measurements. The receiver amplifier chain, in -120

conjunction with the antenna gains provided the necessary
sensitivity of the sampling data acquisition system, whose
linear input-output power response is shown in Fig. 6.
Fig. 8. Received power levels as a function of distance
The receiver was housed in a plastic toolbox provided with between antennas, compared to analytical models,
rechargeable batteries and voltage regulators, for portability 2.45 GHz
and protection. The output of the data acquisition system was
fed into a notebook running Labview. This system was
checked against the readings obtained with the Spectrum To perform the measurements, the transmitter was kept at a
Analyzer, to make sure that it recorded the correct power fixed location in the tunnel, while the receiver was moved
levels, considering the amplification factor. along the laneway. This was a tunnel used for rock removal
and therefore its shape was not as regular as the once chosen
to perform the previous measurements at 870 MHz. At each
The actual portable transmitter – receiver arrangement is measuring location, 1000 samples were taken with this
shown in Fig. 7. As it can be seen, it is a compact portable arrangement. Since the receiver took a longer time to set up,
measuring system that protects the electronics and battery less measuring locations were selected, to be able to cover the
power supplies, while making it possible to move the full extent of the tunnel in the assigned time span.
antennas to perform the measurements.

Experiment results are shown in Fig. 8, where they are

compared to the free space LOS (equation (2)) and the 2 Ray
(equation (3)) models. As it was pointed out beforehand in
the literature discussion, experimental measurements follow
the LOS free space model quite well at distances less than 30-
40 m, while at distances greater than these, the tunnel guides
the waves, resulting in path losses that are less than predicted
by these 2 models. However, beyond the 80 m distance, the
slope of attenuation is of the order of 10 dB/100m, much

rectangular waveguide h = w = 4m assuming εr = 5, which is

more than the values predicted by equation (1) for a

1.25 dB/100m.

3. MINE TUNNEL SMALL-SCALE FADING MODEL

Fig. 6. Input-output response of the sampling receiver
operating at 2.45 GHz Random multipath components arriving at the receiver from
reflections superimpose to the dominant straight line LOS
signal, if it exists. These components sometimes reinforce the
received signal, while at times they weaken it significantly,
producing signal fades. The received signal strength envelope
distribution is known to be Rician, if a strong dominant
signal is present, superimposed with multipath components
and Rayleigh if no such strong component is present, as it
typically happens in non-LOS (NLOS) conditions.

⎛ r 2 + A2 ⎞ ⎛ Ar ⎞
p (r ) = exp⎜⎜ − ⎟I 0 ⎜ ⎟ A ≥ 0, r ≥ 0
σ 2σ 2 ⎟⎠ ⎝ σ 2 ⎠
r
⎝
p (r ) = 0
2

r<0
(4)

⎛ A2 ⎞
K (dB ) = 10 log⎜⎜ 2 ⎟⎟
Fig. 7. 2.45 GHz portable transmitter and sampling receiver ⎝ 2σ ⎠
Equation (4) shows a Rician distribution, where A is the
amplitude of the dominant signal r and I0 is the modified
type-one Bessel function of order 0. When A = 0, only
multipath components are present, the envelope of the
received signal distributes according to Rayleigh. Due to
many reflections at the tunnel walls it is not obvious that the
predominant signal will actually be the LOS signal. Abdi
et.al. (2001) proposed a method to estimate the K parameter
based on statistical performance of two moment-based
estimators, as a less complex alternative to the maximum
likelihood estimator. In our analysis of the collected
empirical data we use this estimator. This result was verified
by comparing the empirical data with a representation of (4), Fig. 10. Sample o received power level recordings from
adjusting its K value to fit the empirical data. moving loader.

Empirical data was obtained transmitting from a moving

underground mine loader, while at work, with the receiver of
Fig. 5 at a fixed location, using a sampling rate of 10 ms. The
portable transmitter was fastened on top of the machine hood
of the front loader as shown in Fig. 9, with the antenna
pointing towards the entry of the mine pit, where the receiver
was located. Data was recorded and split into time windows
of 2 minutes. One of these time windows is shown in Fig. 10.
It can be seen from Fig. 10 that significant fading takes place
as the loader moves within the tunnel to perform its job. A
note of caution is needed to interpret the received power
readings of Fig. 10: the actual received signal strength of the
data acquisition system can be obtained by applying th linear
conversion scale of Fig. 6. The actual reading value does not
affect the calculation of the K factor. The average K factor Fig.11. Autocovariance function of received signal from
that was obtained was of 0.17 dB, showing, in fact, that moving loader.
multipath signals are dominant. Li et.al. (2007) reported
Fig. 11 shows the result of this calculation for the data
measurements in an empty mine laneway, with values of K
collected. It can be seen that the channel coherence time is of
ranging 4<K<8 dB. Differences can be attributed to the
the order of 1.7 seconds, if we apply the criteria mentioned in
characteristics of the mine and the possible effects of antenna
the previous paragraph. The implications of this are relevant.
directivity on the loader.
This time exceeds by almost two orders of magnitude
Channel coherence time is a statistical measure of the sending a 1500 Byte maximum Ethernet size using the IEEE
minimum time required for the magnitude change of the 802.11b protocol at its slowest transmission rate (1 Mbps)
channel to become decorrelated from its previous value. It is and by far more a communication between WSN devices.
related to the Doppler spread of the received signal due to Thus it is possible, for example, to safely send video images
multipath propagation. This is a relevant parameter if channel for the remote control of underground loaders using Wi-Fi
equalization is used at the receiver, because it tells the technology, as it has been demonstrated in El Teniente. It
amount of time the training of the equalizer will be valid also shows that WSN can be mounted on mobile
while receiving a data frame. It may be computed by underground machinery to transmit relevant data.
evaluating the autocovariance function gradually increasing
the computed time window. It is normally accepted that the 4. CONCLUSIONS
channel coherence time is less or equal to the value when the
Large and small scale propagation models have been
autocovariance function is 0.7, Rappaport (2000).
analyzed for an underground copper mine, comparing it to
published results for different kind of mining environments,
mainly coal and gold mines.
Due to the wave guiding characteristics of the tunnel it can be
seen that in LOS conditions the free-space equation provides
a good large scale propagation model fit for UHF frequencies
applications in the range of 870 MHz and 2.45 GHz, a result
which is slightly different from published results.
Small scale propagation characteristics at 2.45 GHz show that
Fig.9. K factor measuring system, 2.45 GHz multipath transmission is dominant when the transmitter is
mobile. However, the channel coherence time is of the order Novasource M2 Specifications, (2008) http://www.nova-
of 1.7 seconds at 2.45 GHz, given the fact that underground eng.com/Inside.asp?n=Products&p=NovasourceM2
loaders are not able to make very fast movements in a tunnel, Rappaport, T. (2001), Wireless Communications: Principles
which in turn is quite convenient for broadband and Practice, 2nd Edition, pp. 138, Prentice Hall
transmissions. Saleh, A. M., Valenzuela, R. A., (1987), "A Statistical Model
for Indoor Multipath Propagation", IEEE J. Sel. Ar.
5. ACKNOWLEDGMENTS Com., Vol. SAC-5, No. 2, pp. 128-137.
This publication was made possible as part of a consulting Shanklin, J. P. (1947). VHF Railroad communications in
project issued by César Ortega, Head of the Superintendencia tunnels, Communications, volume 27, pp. 16-19.
de Telecomunicaciones El Teniente, Codelco. I would also Tripp, A. C., McNearny, R. and Furse, C.. Prof. James R.
like to thank David Guzman, who helped perform the (2000) Wait and Mining Production Technology—An
measurements and the anonymous reviewer, whose Appreciation, IEEE Trans. On Antennas and
comments are greatly appreciated and made it possible to Propagation, Vol. 48, No. 9, pp. 1438-1441
improve this publication. Partial funding was provided by the
PBCT ACT 11/2004 project.

REFERENCES
Abdi, A., Tepedelenlioglu, C., Kaveh, M., Giannakis, G.,
(2001) On the Estimation of the K Parameter for the Rice
Fading Distribution, IEEE Communications Letters,
Vol.5, No3, pp. 92-94.
Benzakour, A. et. al. (2004), Wideband Measurements of
Channel Characteristics at 2.4 and 5.8 GHz in
Underground Mining Environments, Proc. of IEEE
VTC’04-Fall, Los Angeles , California, USA, Vol. 5. pp
3595-3599.
Boutin, M., Benzakour, A., Despins, C., Affes, S., (2006),
Characterization and Modeling of a Wireless Channel at
2.4 and 5.8 GHz in Underground Tunnels, 3rd Internat.
Symp. on Wireless Communication Systems, ISWCS '06.
pp.: 517 – 521.
Changsen, Z. Yan, M., (2006), Effects of cross section of
mine tunnel on the propagation characteristics of UHF
radio wave, ISAPE 06, 7th Internat. Symp. on Antennas,
Propagation & EM Theory, 26-29 Oct., pp.:1 – 5
Deryck, L. (1978), Natural Propagation of Electromagnetic
Waves in Tunnels, IEEE Transactions on Vehicular
Technology, Vol. VT-27, N° 3, pp.: 145-150
Dudley, D. and Mahmoud, S., (2006), Linear Source in a
Circular Tunnel, IEEE Transactions on Antennas and
Propagation, Vol 54. N°7, pp. 2034-2047
Emslie, A., Lagace, R. and Strong, P., (1975), Theory of
propagation of UHF radio waves in coal mine tunnels,”
IEEE Transactions on Antennas and Propagation,
Volume 23, Issue 2, pp.:192 - 205.
Kennedy, G. and Foster, P., (2006), High Resilience
Networks and Microwave Propagation in Underground
Mines, Proc. of the 9th Conf. on Wireless Technology, pp.
193-196.
Li, N., Wei, Y., Ping Z., (2007), Analysis and Comparison of
the Performance about Wireless Modulation
Technologies in the Empty and Straight Mine Laneway,
Internat. Symp. on Microwave, Antenna, Propagation
and EMC Technologies for Wireless Communications,
pp.: 756 – 760
Mahmoud, S., (1974), Characteristics of Electromagnetic
Guided Waves for Communication in Coal Mine
Tunnels, IEEE Transactions on Communications,
Volume COM-22, N° 10, pp. 1547-1554