See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3755188

Slow frequency hopping solutions for GSM networks of small bandwidth

Conference Paper · June 1998

DOI: 10.1109/VETEC.1998.686454 · Source: IEEE Xplore

CITATIONS READS

7 367

4 authors, including:

J. Wigard Per Henrik Michaelsen

Nokia GomSpace A/S
80 PUBLICATIONS   944 CITATIONS    29 PUBLICATIONS   673 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Delay and Disruption Tolerant Networking applied in Low Earth Orbit satellite systems. View project

All content following this page was uploaded by Per Henrik Michaelsen on 13 July 2016.

The user has requested enhancement of the downloaded file.

 1

SLOW FREQUENCY
HOPPINGSOLUTIONS FOR GSM NETWORKS
OF SMALL BANDWIDTH

Thomas Toftegard Nielsen, Jeroen Wigard, Per Henrik Michaelsen and Preben Mogensen
Center for PersonKommunikation (CPK), Aalborg University
Fredrik Bajers Vej 7A, DK-9220 Aalborg @st,Denmark
E-mail: [email protected].

-
Abstract Slow frequency hopping is a quality and capacity border areas and another with frequencies that can be used.
enhancement feature for the GSM system. However, in These latter frequencies are denoted preferred frequencies.
order to achieve a significant gain from frequency Depending on the amount of country neighbours, the number
hopping, it has been shown that at least 3 to 4 frequencies of preferred frequencies is limited to a fraction of the total
are required in the hopping sequence. In band limited spectrum of the operator. In several Nordic countries the
networks that many frequencies per cell are not always number of preferred frequencies is limited to 18 or even less.
available. Two network features, MAIO-management and With only 18 frequencies it can be difficult to apply FH
using a lower reuse with admission load control, successfully.
compensate for the lack of spectrum even with very
Another situation in which it can difficult to exploit the gain
narrow bandwidths. The performance of these two
features have been investigated and compared to a non- from FH, is in traffic dense urban areas where it is necessary to
hopping network. A network with a bandwidth of 21 exploit both micro and macro cells. In order to accommodate
frequencies has been analysed. Simulation results show the required traffic density (Erlang/km2)micro cells are used
that MAIO-management gives the best quality. The and a splitting of the available frequencies into two groups is
comparison is based on individually reported frame necessary. Such a splitting of the spectrum is shown in Figure
erasure rates of each mobile station. 1, using a spectrum handwidth corresponding to a typical
GSM 900 operator of 7.5 MHz (equivalent to 37 frequencies).

I. INTRODUCTION 'Qpical Total Spectrum of a

European GSM Operator:
The enormous growth in digital cellular phones sets very high
16 firequencies 21 frequencies
capacity demands for the GSM operators. Partly due to these
demands extensive studies of how to exploit slow frequency
hopping (FH) to enhance network quality as well as capacity MICRO cellular MACRO cellular
Splitting between MACRO
have been performed over the last couple of years [2,3,6]. The and MICRO cellular layer :
potential gain from FH is two fold. A frequency diversity gain
is achieved by spreading the bursts over a wide bandwidth and Figure 1. The available xpectrum afer having divided the total
hence avoiding Rayleigh fading. Furthermore an interference spectrum into ,a micro and a macro cellular layer.
diversity or interference averaging is achieved due the fact
that we only collide with each interferer during random bursts In the case of a two layer network, typically a minimum of 16
instead of continuously [3]. It has been shown that by hopping channels has to be used for the micro cellular layer
over as little as 3 to 4 frequencies a large part of the frequency (corresponds to having 2 TRX's per microcell-base station),
diversity gain is achieved, whereas hopping over only 2 leaving 21 frequencies for the macro cellular layer. With as
frequencies can result in even a degradation when compared to little as 21 frequencies it is hard to achieve a satisfactorily gain
the non-hopping case. However, in some situations it can be from FH, i.e. to hop over a minimum of 3 frequencies.
difficult to apply FH successfully, since 3 to 4 frequencies per
This paper investigates two methods, MAIO-management' and
cell may not always be available.
using a lower reuse with load admission control,* which can
For example at country borders each network operator has to exploit the gain from FH in situations like the two described
co-ordinate the frequency allocation with every influencing above, i.e. in narrow band limited network situations.
operator in the region, in order to avoid network quality
degradation due to adjacent and co-channel interference. This
includes national as well as international operators. Therefore
the spectrum available to each operator is divided into MA10 is an abbreviation of the Mobile Allocation Index Offset
subgroups. One group of frequencies not to be used in the from GSM.
This method will be calljed sof capacity throughout the article.

0-7803-4320-4/98/$5.00 0 1998 IEEE 1321 VTC '98

The study is carried out using a GSM-900 type of network, but set. Secondly, the HSN (Hopping Sequence Number) which
the achieved results are reliable for the GSM-1800 and 1900 may take 64 different values [8]. Two channels bearing the
cases as well. In section I1 the related information of the GSM same HSN but different MAIO’s never use the same frequency
system is given, followed by a description of MAIO- on the same burst, while two channels using the same
management in section 111. In section IV the 2 methods, frequency list, but bearing different MAIO’s only interfere
MAIO-management and soft capacity, are compared. Section with each other ( l / r ~ ) ‘of~ the time. Usually, channels in the
V briefly introduces the simulation tool used for the same cell use the same HSN, but have different MAIO’s: that
evaluation. The paper ends with a description of the results way interference between channels in the same cell is
achieved and a conclusion on the performance of the features impossible.
in section VI and VI1 respectively.
111. THEMA10 - MANAGEMENT
CONCEPT
11. THEGSM SYSTEM In the following, a network with 21 frequencies is used for
Each radio frequency channel in GSM is time divided into illustration. Only the macro cellular layer is considered. With a
TDMA frames of 4.615 ms, where each frame is subdivided total of 21 frequencies, if considering baseband FH, no more
into 8 full slots. Each of these slots can be given a full rate than 2 TRXs can be allocated per cell in the macro cellular
traffic channel, two half rate traffic channels or a control layer,3 see Figure 3. Hence, only 2 frequencies can be used in
channel. Besides the full rate traffic channel, this paper only the FH hopping sequence.
considers the broadcast control channel (BCCH) [8]. The
The MACRO
BCCH is for several reasons of vital importance in GSM. It is Cellular Spectrum :
a unidirectional downlink channel e.g. used by each mobile to
identify where it is located in the network. Even in idle mode,
9 TCH
a fair amount of information is exchanged for the network to Seguencies 12 BCCH fiquenciea
work efficiently. Correspondingly, every mobile station (MS) Splitting between
always has to be able to decode the BCCH channel. This also WoppinglBCCH Frequencies :
means that no downlink power control can be used on the
BCCH channel. Furthermore, no hopping can take place in Figure 3. The micro/macro cellular splitting of the spectrum
timeslot 0 on the BCCH frequency, since the BCCH located in
this timeslot is not allowed to hop. This is shown in Figure 2,
illustrated using a site with 2 TRXs and baseband FH. Some GSM network field trials have indicated that baseband
FH using only 2 frequencies may not be a lucrative network
1 TDMA Frame
feature. It has to be ensured that both of the two frequencies
Timeslot: 0 1 2 3 4 5 6 7
always have a good quality, due to the fact that the frequency
and interference diversity gain from FH is relatively small.
TRX 1: Furthermore, from a practical point of view, there is an
(BCCH) information loss from not being able to re-allocate channels
based on uplink measurements, since the measurements are
TRX 2:
averaged. Therefore, instead of using baseband FH over two
frequencies, two alternative approaches both applying
Figure 2. An example of the allocation of the BCCH channel in synthesised FH, can be taken. They are:
timeslot 0 on the BCCH frequency on a site with 2
TRXs using baseband FH. 1) Besides the required BCCH frequency (with a reuse of
4/12) allocate the remaining hopping frequencies using
In a non hopping GSM network, the macro cell frequencies a tight frequency reuse pattern, as e.g. 1/3 and limit the
typically have to be planned with a reuse pattern in the order co-channel interference in the network using a load
of at least 4/12 (3-sector sites), while the frequency reuse limiting technique.
pattern for FH channels using downlink discontinuous
transmission (DTX) and power control can be reduced to 3/9 2) Along with the required BCCH frequency, use the one
[3]. In reality this means that the BCCH frequency has to be TCH frequency channel (using a 3/9 reuse pattern)
planned with at least a frequency reuse pattern of at least 4/12, allocated to each of the three sectors orthogonal in all
while if using DTX and power control the remaining traffic three cells. This could quite easily be implemented by
(TCH) frequencies only requires a minimum frequency reuse giving channels in the different sectors of a site the
pattern of 319.

For a set of n frequencies, GSM allows 64*n different hopping

sequences to be built. They are described by two parameters.
Firstly, the MA10 (Mobile Allocation Index Offset) which One frequency has to be used as BCCH (reuse 4/12) and since nine
may take as many values as the number of frequencies in the channels remains and the minimum frequency reuse of the TCH
channels is 3/9, only one more frequency can be allocated per cell.

0-7803-4320-4/98/$5.000 1993’IEEE 1322 VTC ‘98

 same HSN, but a different MA10 [8]. This technique
is denoted MAIO-management.4
In solution 1) a frequency reuse pattern of 1/3 is applied to
allow 3 frequencies to be included in each hopping sequence
t =o t= 1 t=2
in each cell (the BCCH is not included in the hopping
sequence). Network simulations have shown that for a 1/3 Figure 4. Example of the,frequency allocation when using the
reused network configuration, the average load on each MAIO-manage,mentfeature and sequential FH.
frequency channel may not exceed 30-35% [4]. This load
limitation can be achieved using two different approaches: Despite now having 3 frequencies to hop across (f4,h,fs), and
a) Allocate 1 TRX hopping over 3 frequencies, i.e. a thereby getting the desired frequency and interference
hard limitation of 33 %. diversity gain, the frequency planning has hardly changed.

b) Allocate 2 or 3 TRXs in every cell, and exploit a The penalty of applying MAIO-management shall be found in
software algorithm to decide whether a free channel the co-channel interference introduced, since the reuse has
can be used or not. Such an algorithm is based on an become worse. By turning the frequencies over the 3 sectors of
estimation of the quality and is often denoted as a a site, the reuse corresponds to the reuse of a omnidirectional
so@ capacity load limiting technique [7].5 site. However since onty one frequency can be used at a time
in one of the three sectors, the maximum load is 33%.
Simulations have shown that the use of a soft capacity Therefore, from an interference point of view, the 319 reuse
limitation technique may improve trunking efficiency as it become a 313 reuse haded up to 33% by using MAIO-
better serves peak traffic in a cell. However in practice this management. When using a 90% area outage, the CA of a 319
should be compared to implementing alternative existing GSM reused network is 7.0 dl3 , while it is 2.3 dB for a 313 network
features such as directed retry and traffic reason handover [9], [9]. Limitation the max. load to 33% gives 4.8 dB gain and
which even requires less TRX’s in the network. compensates completely for the loss of having a worse reuse.
Solution 2) has not previously been investigated.
If using synthesised FH and a common controller for all IV. MAIO-MANAGEMENT
VERSUSSOFTCAPACITY
sectors of a base station, there is no GSM system limitations In the following only thle hopping frequencies are considered,
on allocating the same frequency group to TRX’s in different i.e. the BCCH frequencies are left out. Assuming an ideal
sectors. The MAIO-management will prevent a frequency from cluster size, interference from only the six neighbouring cells
being used in more than one of the sectors at the same time. By in the first tier and the path loss model in equation ( 1 ), the C/I
using MAIO-management, it is thus possible to hop over 3 at the cell border when using 3-sector sites, are found to 5.4
times as many frequencies (for 3 sector sites). In case of only dB for the 113 reuse (which is used for soft capacity) and 13.7
one available frequency channel per cell (apart from the dB for the 3/9 reuse (which is used for MAIO) [9].
BCCH frequency) the situation changes from not being able to
apply FH at all, to being able to hop over 3 frequencies in the L(r)=k.r-Y (1)
cell, with each frequency channel loaded no more than 33% in In equation (1) r is equal to the distance between the MS and
each sector. From the basic functionality it is obvious that the base station, y the path loss slope and k a constant
MAIO-management can only be applied to sectorized sites. compensating for the effect of base station height, frequency,
To clarify the functionality, an example is shown in Figure 4, coverage etc. If includirig Log-normal fading and using a 90%
where 6 frequencies are used on each site, i.e. each sector area outage, the CA for the 1/3 reuse is -1.3 dB while 7.0 dB
having 2 TRX’s of which one is the BCCH frequency for the 3/9 case [9]. That is, if assuming 100 % load in both
(b4,b&). For each TDMA frame the frequency of the TRX situations there is a 8.3 dB C/Z difference at the cell border
not carrying the BCCH channel is shifted in a cyclic (for between the two. This means that the 1/3 reused network can
simplicity) manner. In practice random FH should be used. In not be loaded as much as the 3/9 network, if the quality is
time instant t = 0, sector 1 uses the frequencies bl andf4. In the assumed to be the same.
next TDMA frame (t = 1) the frequencies have switched to bl However, such an evaluation of MAIO-management versus the
andf5, and to bl and f6 at t = 2. soft capacity concept, based on simple C/I distributions
assuming ideal clusters may not be accurate. Using MAIO-
management or soft capacity in combination with DTX, the
max. load on each frequency in a sector will be approximately
16%. Such a low traffic load causes additional interference
diversity from FH due to the codinglinterleaving and soft
The MAIO-management feature has originally been proposed by detection in GSM. This gain may be different for the two
Nokia Telecommunications. solutions because of the difference in C/Z distribution.
The soji capacity technique is sometimes also referred to as
admission load control or fractional loading.

0-780343!2&4/98/$5.000 1998 IEEE 1323 VTC ‘98

Therefore network simulations are carried out to get a more traffic load used in all simulations is equivalent to 30% of the
reliable performance estimation. From a practical point of available traffic channels. This corresponds to loading a
view, a large advantage lies in the frequency planning network using DTX to 60% (the hard blocking limit for a
procedure when using soft capacity. It is quite simple, since network with 2 TRX per cell). Furthermore, for simplicity, the
the frequencies are simply divided into 3 groups and then (all BCCH frequencies are not included. The results are presented
of them) allocated to each site. using the cumulative density function (CDF) of the frame
erasure rates (FER) of each MS reported once per multiframe
(0.48 sec. in GSM), see [SI.The FER has been chosen in order
V. THESIMULATION
TOOL’CAPACITY’
to present a quality measure which is correlated as close as
The performance of MAIO-management, as well as the 113 possible to the actual experience of the user.
reused synthesised FH configuration, will be evaluated using
Two sets of simulations have been carried out. At first,
the GSM network simulation tool CAPACITY, which has been
simulations with a mobile speed of 3 km/h, where the
presented on several occasions [1,41.
frequency diversity gain from FH is expected to be relatively
CAPACITY is capable of simulating mobiles dynamically large (compared to the non hopping case) are carried out.
moving around in a GSM network. Both shadow and short These results are shown in Figure 5 . With both soft capacity
term (Rayleigh) fading is included. FER and call drops (i.e. the and MAIO-management 10% of the measurements show a
quality) can be evaluated taking the coding and interleaving FER worse than 0.06, while for the non hopping case 10% of
gain into account. The general network configuration the measurements are worse than 0.08. On the other hand if
parameters used throughout all the evaluations are shown in comparing the three networks using a 1% FER outage MAIO-
Table 1. In all cases the network consists of 48 3-sector sites, management is superior to the others.
i.e. 144 cells. In every simulation at least 6000 mobiles have
been simulated.

Available radio spectrum 4.2 MHz (21 frequencies)

Path loss model Lp= 35 log d
Shadow fading standard deviation 6dB
Shadow fading correlation distance I/e at 110 m
Call mean hold time 80 s (exponential distribution)
MS velocity 50 k d h and 3 km/h
Cell radius 3km
Handover algorithm Based on interference, quality,
level and power budget
Power control Both level and quality
Dropped call RX-QUAL threshold 5
Dropped call threshold 19

Table 1. The general simulation parameters used in CAPACITY.

Figure 5. Cumulative densityfunction (CDF) of the FER on trafSic
channels for a mobile speed of 3 k d h .
VI. RESULTS
Secondly simulations with a mobile speed of 50 km/h are
The two cases described in the introduction have been carried out. The results are shown in Figure 6. At the 10%
simulated for a MS speed of 3 km/h and 50 kmk.Therefore, FER outage it is difficult to se any difference between the
the potential capacity gain using an available spectrum of 4.2 three networks, The FER is around 0.04 in all cases. If, instead
MHz is considered for both of the two scenarios: looking at the 1% FER outage, non hopping is clearly
1. Synthesised FH with a frequency reuse pattern of 113 and performing the worst. With MAIO-management, 1% of the
33 % load limitation using a soft capacity load limitation FER measurements are worse than 0.17, while for the soft
technique, while hopping over 3 frequencies. The capacity network 1% of the measurements show a FER of 0.24
frequency reuse pattern on the BCCH frequency is kept at or worse. In the non-hopping case the FER is 0.36 or worse.
4/12.
2 . Synthesised FH using MAIO-management with a reuse
pattern of 319 on the hopping frequencies and a 4/12
frequency reuse pattern on the BCCH frequency.
Furthermore, for comparison, the quality of a non-hopping
network is simulated in all cases as well. Discontinuous
transmission (DTX) is excluded in any of the simulations. The

0 1998 IEEE
0-78034320-4/98/$5.00 1324 VTC ‘98
 The simulation results show that the concept of 1/3 soft-
capacity with 33% relative load gives worse network quality
than a non-hopping network configuration using a reuse of 3/9.
However, the soft-capiicity feature may have the potential of
improved trunking efficiency if installing more TRX’s .
0.1
In contrast to the lr‘3 reuse the 319 MAIO-management
k0 concept gives much improved performance over the non-
1- - - - I- - - - hopping 319 network configurations,
0,Ol The MAIO-m feature seems to be an attractive solution for
achieving the benefits of frequency hopping at low
bandwidths, without the drawbacks of increased mean
interference level which is the case for soft capacity with a 1/3
0.001
reuse.
0 0.1 0.2 0,3 0,4 0,5 0,6 0.7 0,8 0.0 1
FER
ACKNOWLEDGEMENT
Figure 6. Cumulative density function (CDF) of the FER on the
traflc channelsfor a mobile speed of 50 M. We would like to thank Nokia Telecommunications Finland
and Sonofon Denmark for co-sponsoring the project.
Therefore, also with the mobile speed moving at a speed of 50
km/h, the MAIO-management seems to have a better quality
LITERATURE
than soft capacity. When comparing the simulations of slow
and fast moving mobiles, the quality is degraded with Wigard J., T.T. Nielsen, P.H. Michaelsen and P. Mogensen,
increased speed. This is due to the handover and power control Improved Intelligent Underlay-Overlay combined with
algorithms in GSM. Frequency Hopping in GSM, In IEEE Proc. of PIMRC 1997,
Soft capacity seems to perform about the same as non hopping Helsinki, pp. 376-380.
in the case of slow moving mobiles, while it seems to be worse Zako B. et al., The GSM radio link per$ormance with space
than non hopping when the mobile speed is relatively high. diversity and slow frequency hopping, In IEEE Proc. of VTC,
However soft capacity has the advantage that it may be 1993, pp. 480-482.
capable, as mentioned before, of solving local trunking Johansen J. and B. Vejlgaard Capacity analysis of a
problems. frequency hoppint: GSM system. M.Sc.E.E. Thesis Report,
Aalborg University, June 1995.
VII. CONCLUSION Wigard J., P. Mogensen, J. Johansen and B. Vejlgaard,
The network quality in terms of Frame Erasure Rate (FER) has Capacity of a GSM Network with Fractional Loading and
been evaluated for two frequency reuse concepts, which allow Random Frequency Hopping, In IEEE Proc. of PIMRC 1996,
utilisation of slow frequency hopping at very limited Taipei, pp. 123-72’1.
bandwidths. In the evaluation the case of a three sectorized Wigard J., T.T. Nielsen, P.H. Michaelsen and P. Mogensen,
base station configuration with 2 TRX’s per sector, where one BER and FER Prediction of Control and Trafsic Channelsfor a
GSM Type of Interjace, Accepted for IEEE Proc. of VTC 1998.
TRX carries the BCCH and only the second TRX is hopping,
Carenheim C., S. Jonsson, M. Ljungberg, M. Madsfors and
is considered.
J. Naslund, FH-GSM, Frequency Hopping GSM, In IEEE
The concept of soft capacity by using a dense frequency reuse Roc. of VTC 1994, Stockholm, pp. 1155-1159.
of 1/3 is compared to the concept of MAIO-management Beming P. and M. Frodigh, Admission Control in Frequency
(Mobile Allocation Index Offset) using a nominal frequency Hopping GSM Systems, In IEEE Proc. of VTC 1997, Phoenix,
reuse of 3/9. Both concepts allow 3 frequencies to hop over. pp. 1282-1286.
The penalty of a 113 soft capacity concept is the very dense GSM Recommendlations.
frequency reuse, which potentially worsens the mean co- European Radialcommunications Office (ERO), Final
channel interference level. A mean load factor of 33% has Document for the ER0 Workshop on Traffic Loading and
been used to compensates for this problem. Efficient Spectrum Use, 1998.

The potential advantage is the gain from both frequency and

interference diversity by applying FH hopping.
From interference and frequency planning point of view the
MAIO-management changes the nominal frequency reuse of
3/9 for a sectorized cell configuration into an omni-directional
cell configuration of 3/3 with 33% relative load.

0-7803-4320-4/98/.$5.000 1998 IEEE 1325 VTC ‘98

View publication stats