Interference Analysis of a Total Frequency Hopping GSM Cordless Telephony System1

Jürgen Deißner, André Noll Barreto, Ulrich Barth*, and Gerhard Fettweis
Endowed Chair for Mobile Communications Systems
Dresden University of Technology, D-01062 Dresden, Germany
Tel. +49-351-463-4660, Fax +49-351-463-7255
e-mail: {deissner | noll | fettweis}@ifn.et.tu-dresden.de
* Alcatel Corporate Research Center, D-70499 Stuttgart, Germany
e-mail: [email protected]
ABSTRACT II. DISCUSSION OF PREVIOUS WORK
ON TOTAL FREQUENCY HOPPING1
In this paper we present an interference analysis which reveals
capacity reserves for a hierarchical GSM radio network. In the In [1] a new multiple access method, called Total Frequency
investigated concept a low-power Cordless Telephony System Hopping (T-FH), was presented for such a CTS system. It is a
reuses the frequencies of a cellular GSM system based on Total slow FH technique, which uses as much of the cellular GSM
Frequency Hopping (T-FH) multiple access. Based on a novel frequencies for FH in the CTS systems as agreed with the GSM
simulation method, which considers the dynamics of a T-FH operators and the regulator, ideally the whole GSM frequency
cellular network, we present results for different cordless user range (e.g. 124 frequencies in GSM900). The more frequencies
densities and numbers of T-FH frequencies which exhibit the are used the more obvious is the effect of averaging the CTS-
benefits of our CTS concept. Moreover, we show two possible caused interference over a lot of cellular GSM channels. In that
ways to further improve the CTS downlink performance in case paper it was concluded that cellular GSM and CTS systems can
of difficult scenarios. co-exist, if the GSM-to-CTS power ratio is well over or around
20 dB for 25 or 75 FH frequencies, respectively. However, the
I. INTRODUCTION assumed simulation environment does not represent a very dense
Solutions for hierarchical radio networks where different layers urban residential area, which we expect as a worst case as well
share the same channels were already discussed earlier, e.g. for as one of the first deployment scenarios. Furthermore, the
hot-spots. An application scenario for GSM radio networks applied Monte-Carlo simulation method with SIR values for
providing low-power cordless indoor coverage was introduced GSM bursts calculated for totally independent snapshots of the
as a home base station [1]. This culminated within ETSI into the system model does not allow for an analysis of the time
current standardization efforts on CTS. correlation of SIR values according to the dynamic environment.
Especially, neither the ability of a FH algorithm to avoid sequen-
A GSM Cordless Telephony System (CTS) [2,3,4] consists of a ces or accumulations of hits to the same frequency can be shown
low-cost GSM base station, the CTS Fixed Part (CTS-FP) and a nor the ability of a GSM receiver to recover from poor SIR
modified GSM mobile station, the CTS-MS which can work values in few bursts due to the deinterleaving and decoding gain.
either in the conventional cellular or in the new cordless mode.
The CTS-FP is connected to the fixed network like for example In [5] a FH algorithm is presented that is especially designed for
a DECT Fixed Part, however it uses GSM frequencies. the requirements of a T-FH GSM CTS system. Together with the
GSM CTS concept proposed by Alcatel [4], which incorporates
In order to convince a GSM network operator to share his the T-FH technique, this is the basis for our simulations.
spectrum for GSM cellular operation with GSM cordless
systems, our simulations have to reveal under which III. SIMULATION ENVIRONMENT
circumstances a GSM CTS system is able to perform well Our model of the cellular GSM radio network represents a dense
including: urban environment with a high traffic load, because this is likely
• a dense GSM macrocell structure and a tight reuse scheme, to be also an area of high GSM cordless user penetration. Our
• different CTS user densities, worst case assumptions are listed in the tables 1-6. As the result
• different numbers of reused frequencies or a small number of of the discussions within ETSI these assumptions were mainly
separate frequencies for CTS use only (e.g. for an agreed for their CTS simulation studies except for the BS-to-BS
introduction and test scenario), and distance (ETSI: 500m), the portion of GSM-MSs being indoor
• different GSM-to-CTS transmit power ratios.
1
This work was supported by Alcatel.
(50%), the BCCH timeslot consideration (not modelled), and the Table 5. Cellular GSM transmission characteristics
CTS transmit power (17dBm, not varied) [6]. GSM-BS GSM-MS
• transmit power: 33 dBm • transmit power: 33 dBm
• sector antenna with a horizontal pattern • omnidirectional antenna
A. Radio network configuration o
having a 3dB-beamwidth of 60 and a front-
to-back ratio of 40 dB
Table 1. System layout
• 54 dBm EIRP due to 21 dB max. gain
cellular GSM CTS (vertical pattern not considered)
• regular hexagonal cells • CTSs are generated within round buildings Furthermore, for cellular GSM and CTS we assume:
• 3/9 reuse structure with a radius of 20 m, which are uniformly • Neither power control nor DTX is applied.
distributed over the investigation area.
• BS-to-BS distance D = 750 m • The noise floor, including thermal noise and a noise figure, is assumed with -114 dBm.
• The number of occupied buildings
• cell radius R = 250 m
corresponds to the CTS density.
• all GSM-BS and GSM-MS are outdoor
• The CTS density [1/ km ] is a simulation
2
B. Propagation models2
parameter.
• The position of a CTS-FP and a CTS-MS All propagation models are two-dimensional, statistical path loss
in “its” building upon generation is
R uniformly distributed over the building models and based on the assumption of a heterogeneous
area.
• all CTS-MS and CTS-FP are indoor
environment. The shadowing S(σ) follows a log-normal
D distribution with standard deviation σ and is uncorrelated
between locations, but does not change as long as the position
Table 2. Frequency usage and traffic load remains the same3 (Table 6). In particular, the shadowing value
GSM CTS does not change during an interleaved block of 8 consecutive
• 3 transceivers per cell result in 27 • Total Frequency Hopping is applied
2
available frequencies ; no frequency • Every CTS user is provided with a unique
bursts in a certain channel (refer to IV.A).
hopping T-FH sequence.
• BCCH modelled by the continuous • The number of frequencies that are re- Table 6. Path loss equations
occupation of the downlink timeslot 0 at used by the CTS systems for T-FH is a GSM-BS - GSM-MS
one of the frequencies in every cell simulation parameter.
Walfish-Ikegami [03.30] with f = 900 MHz, hBS = 25 m, hMS = 1.5 m, hroof = 15 m, w = 20 m, b =
• SDCCH not considered, but handled like a • Each CTS-FP supports only 1 CTS-MS. 50 m, ϕ = 90
o
further TCH
• 0.1 Erlang per CTS-MS
• 23 available TCHs per cell 315. + 20 ⋅ lg(d / m) + S(8dB) for d ≤ 21m
• exponentially distributed call interarrival LGG = 
• 15.3 Erlang offered traffic (corresponds to time; mean call arrival rate per cell: 7.7 + 38 ⋅ lg(d / m ) + S(8dB) for d > 21m
1.6% blocking probability and 65% mean 0.00083/s
load) GSM-BS - CTS-MS
• exponentially distributed call hold time;
• New calls that encounter 23 occupied mean value: 120 s LGC = LGG + Lindoor with Lindoor = 12dB
TCHs in their best-server-cell are blocked.
CTS-FP - GSM-MS
• exponentially distributed call interarrival
time; mean call arrival rate per cell: 0.17/s Walfish-Ikegami [03.30] with hBS = 4 m and Lindoor = 12 dB
• exponentially distributed call hold time; 315. + 20 ⋅ lg(d / m) + S(8dB) + Lindoor for d ≤ 23.5m
mean value: 90 s LCG = 
−7.54 + 49 ⋅ lg(d / m ) + S(8dB) + Lindoor for d > 23.5m

Table 3. MS Movement CTS-FP - CTS-MS

linear attenuation indoor model [COST231] with attenuation coefficient α = 0.9 dB/m up to the
GSM-MS CTS-MS
CTS building diameter of 40 m
• MS‘s are generated for the duration of a call.
315. + 20 ⋅ lg(d / m) + S(6dB) + 0.9 ⋅ d / m for d ≤ 40m
• The initial positions at the beginning of a • The initial positions at the beginning of a LCC = 
−7.54 + 49 ⋅ lg(d / m) + S(8dB) + 2 ⋅ Lindoor for d > 40m
call are uniformly distributed within the call are uniformly distributed within the
investigation area. building of its CTS-FP.
• The speeds are uniformly distributed within • The speeds are uniformly distributed within
[0, vmax = 20m/s). The speed remains [0, vmax = 1.5m/s). The speed remains IV. PRINCIPLES OF SIMULATION AND
constant throughout the call. constant throughout the call.
INTERFERENCE ANALYSIS
• The directions are uniformly distributed • The directions are uniformly distributed
within [0,2π). The direction remains also within [0,2π). When reaching the update
constant throughout the call. However, an distance, a new direction is taken from that A. Novel Simulation Method
MS will be reflected at the border of the distribution.
investigation area. • The update distance for the position and Our C++ simulation program combines the discrete event (DE)
• Initial assignment and handover (without a best- server-decision is 1 m.
hysteresis) is carried out due to the best-
and random sampling simulation methods. The DE simulation
server-criterion including the provides signal-to-interference ratios (SIR) that are correlated in
6 neighbouring cells.
• The update distance for the position and
time due to the dynamics of the modelled cellular network
best- server-decision is 10 m. caused by the user traffic, the movement of the users during
calls, and the applied total FH multiple access technique in the
Table 4. CTS transmission characteristics
CTS. The time basis for the DE simulation is the TDMA frame
CTS-FP CTS-MS
• The CTS Fixed Part (CTS-FP) and CTS Mobile Station (CTS-MS) transmit power [dBm
period TTDMA = 4.615 ms. Through random sampling we select
EIRP] is a simulation parameter. (Our reference scenario for comparisons uses 17 dBm.) sequences of SIR values in one and the same GSM radio channel
• Both, CTS-FP and CTS-MS, have an omnidirectional antenna.
• The CTS systems are modelled to be timeslot-asynchronous. 2
The allocation of these 27 frequencies within the cluster is fixed. However, the order is not
important because adjacent channel interference is not considered.
3
These and all further assumptions for the 4 propagation models comply with the definitions
for the studies within ETSI [6].
(defined by frequency and timeslot) at random instants during an We apply the two following performance criteria to the SIR
active call for post-processing in a Matlab program. We chose distributions that we get for each of the 4 network element types
the TCH interleaving depth of 8·TTDMA as the length of such a from our DE simulation:
sequence in order to evaluate the influence of the dynamics in • the outage probability Pout = P[ SIR < 9 dB ] as a model for
the cellular radio network on the TCH performance with consi- signal detection, i.e. an outage appears if the SIR of a burst is
deration of a decoding and deinterleaving gain in the receiver. smaller than the threshold of 9dB, and
In order to start the DE simulation in the stationary state, a • the frame erasure rate (FER), which we define as the
random number of active GSM-MS and CTS-MS is already probability that within a sequence of 8 bursts of consecutive
generated in the simulation setup according to the traffic values. TDMA frames 4 or more encounter an outage, as a model for
the deinterleaving and decoding gain [9].
B. Investigation area
V. SIMULATION RESULTS
Three tiers of interfering clusters are taken into consideration
(Fig. 1). The GSM BS and MS in all clusters cause interference, For comparison we defined a reference scenario with 27
but only in the central cluster the GSM BS and MS interference frequencies for Total Frequency Hopping (T-FH) in the CTS 4,
is evaluated. 17 dBm CTS transmit power, and a density of 1500 CTS/km 2.
Based on that (section V.A), we performed simulations with the
CTSs which cause interference are generated within a radius of parameter variations as follows:
1100 m, but the CTS-FP and CTS-MS interference is only
• CTS density between 750 and 5000 CTS/km2 in shared band
evaluated within the inner radius of 700 m.
operation with 27 T-FH frequencies as well as in separate
band operation with 3 T-FH frequencies (section V.B),
• number of T-FH frequencies between 13 and 124 (sect. V.C),
7

evaluated cluster • CTS-FP transmit power between 13 and 38 dBm (sect. V.D),
7 1 4 7
• a reduced frequency list with 24 and 6 shared T-FH
7 1 4 7 1 4 7 1st tier of
interfering clusters frequencies (section V.E).
1 4 7 1 4 7 1 4
2nd tier

7 1 4 7 1 4 7
A. The reference scenario for comparison
3rd tier
1 4 7 1 4 7 1 4 Table 7 summarizes the results for our reference scenario. We
1 4 7 1 4 CTS evaluation area
observe the strongest influence in the outage probability of the
7 7

CTS interference area

GSM uplink: a CTS- caused increase from 0.9% to 1.8%. The
1 4 7 1 4 7 1 4
GSM uplink is the more sensitive GSM link due to the lower
investigation area
1 4 7 1 4 transmit power of a GSM-MS in comparison with the GSM-BS.
1 4
The corresponding FER values indicate that the T-FH algorithm
distributes the CTS interference in time and in frequency across
Fig. 1. Investigation area. all GSM connections in a way that the GSM receivers mostly are
C. Interference Analysis able to recover from single hits within interleaved blocks
according to our rough model for the GSM receiver's
Only co-channel interference is evaluated. Signal and
deinterleaving and decoding gain.
interference power levels are mean values for the whole timeslot
duration of TTS = 577 µs. The GSM BS and MS are timeslot- The CTS uplink also shows a reasonable performance, but the
synchronous in our model. However, asynchronous CTS are CTS downlink has a poor outage probability. In spite of an
modelled by a constant timeslot offset [ 0, TTS ) that is randomly assumed average outdoor-to-indoor penetration loss of 12 dB,
chosen for each CTS-FP. Overlapping power portions from the CTS-MS suffers seriously from GSM-BS-caused interference
asynchronous interferers are averaged over the whole timeslot levels, which are considerably high in comparison with the
duration of the interfered receiver. signal level received from the CTS-FP.

The interference performance at all 4 network element types, This is due to our model of the GSM-BS, which always transmits
GSM-BS, -MS, CTS-FP, -MS, can be studied in one and the with its maximum power without power control and DTX for
same simulation. We are always able to distinguish the interference reduction. The CTS downlink is effected so strongly
interference that is caused by the GSM cellular system from the
total interference, which includes the CTS- caused interference. 4
In the case of a cellular GSM 3/9 cluster with 3 frequencies per cell, we consider the reuse
of all those 27 GSM frequencies in the co-existing CTS as a worst case because the mutual
interference between the cellular and the cordless systems is the most.
Table 7. Performance results for 17 dBm CTS transmit power, 1500 CTS/km2, small set of frequencies. However, it can be a solution for the
and 27 shared T-FH frequencies
CTS introduction and for medium CTS densities.
Outage probability Frame Erasure Rate (FER)
total, with CTS GSM-caused part total, with CTS GSM-caused part
C. Different numbers of frequencies for T-FH in the CTS
GSM uplink 1.8 % 0.9 % 0.9 % 0.9 %
GSM downlink 0.9 % 0.7 % 0.7 % 0.7 % Fig. 2 illustrates the relation between the GSM performance and
CTS uplink 2.9 % 2.6 % 0.2 % 0.1 %
the number of T-FH frequencies for 17 dBm CTS transmit
CTS downlink 25.5 % 25.3 % 22.8 % 22.7 %
power and 1500 CTS/km2. As expected, the results reveal the
increase in the effect of interference averaging, when
because of the different power ratios of the base stations and of
considerably more than 27 GSM frequencies are used for T-FH
the handsets of both systems; the GSM-BS-to-CTS-FP ratio is
in the CTS.
25 dB depending on the angle to the main lobe of a GSM-BS
antenna, whereas the GSM-MS-to-CTS-MS ratio is only 16 dB.
0.02

B. Different CTS densities 0.018

0.016

Table 8 shows the dependence of the performance on the CTS 0.014

density in the case of 17 dBm CTS transmit power and shared

Outage Probability
0.012

band operation with 27 T-FH frequencies. 0.01

0.008
The results for zero density represent the GSM-caused part of
interference. In the range of 750 to 5000 CTS/km2, which
0.006

0.004 GSM uplink, total interference

already covers a very optimistic maximum CTS user penetration, GSM
GSM
uplink, GSM−caused interf.
downlink, total interference
0.002 GSM downlink, GSM−caused interf.
the CTS influence is not strong and only visible in the GSM
0
uplink. The GSM uplink outage probability only varies in the 0 20 40 60 80
No. of frequencies for T−FH
100 120

range of 0.9% to 4.4%, the FER only in the range of 0.9% to Fig. 2. GSM outage probability for 13..124 T-FH frequencies.
1.2%.
D. Different CTS-FP transmit powers
Table 8. Outage probability (Pout) and frame erasure rate (FER) without CTS
and for densities between 750 and 5000 CTS/km2 In this scenario with increased CTS-FP transmit power we ex-
2
CTS density [1/km ] 0 750 1500 3000 5000 pect similar results like in a scenario with interference reduction
GSM uplink Pout 0.9 % 1.4 % 1.8 % 2.8 % 4.4 % features at the GSM-BS. Fig. 3 shows an exchange of perfor-
GSM uplink FER 0.9 % 0.9 % 0.9 % 0.9 % 1.2 %
mance between GSM-MS and CTS-MS for increasing transmit
GSM downlink Pout 0.8 % 0.9 % 1.0 % 1.2 % 1.5 %
powers, where the reduction of outage probability and FER at
GSM downlink FER 0.7 % 0.8 % 0.7 % 0.7 % 0.8 %
CTS uplink Pout 2.5 % 2.6 % 2.9 % 3.2 % 3.6 %
the CTS-MS is considerably stronger than the increase at the
CTS uplink FER 0.1 % 0.1 % 0.2 % 0.2 % 0.4 % GSM-MS. This indicates that we can reach a reasonable perfor-
CTS downlink Pout 25.3 % 25.4 % 25.5 % 26.0 % 26.2 % mance in the CTS downlink by means that reduce the GSM-BS
CTS downlink FER 22.7 % 23.0 % 22.8 % 23.4 % 23.7 % transmit power in relation to the CTS-FP transmit power.

The CTS results also show the poor CTS downlink performance. 10
0

However, we can still learn from the CTS uplink performance, GSM downlink outage probability
GSM downlink FER
CTS downlink outage probability
which shows only a CTS-caused change in the total outage CTS downlink FER
Outage probability or Frame Erasure Rate

probability from 2.6% to 3.6%, that the reserves in the usage of

the GSM frequencies for CTS are not exhausted yet - also for the 10
−1

reason that we only modelled two-dimensional propagation and

did not consider floor losses. For the same reasons, this principle
may also be applied to GSM picocells with the advantage of
avoiding frequency planning. 10
−2

We also investigated the influence of the CTS density in the case

of 3 T-FH frequencies solely dedicated to CTS operation. The 10 15 20 25
CTS transmit power [dBm]
30 35 40

results exhibit a linear increase of the outage probability in both, Fig. 3. GSM downlink performance for CTS-FP transmit powers of 13..38 dBm.
CTS uplink and downlink, in the range of 1.3 % to 8.6 % at
750 .. 5000/km2. Using only 3 frequencies shows that the main
T-FH advantage of interference averaging is not gained at this
E. A reduced frequency list for T-FH Facing the poor CTS downlink performance in our reference
scenario, we showed that the GSM-BS-to-CTS-FP power ratio
A further way to improve the CTS performance is to exclude the
should be taken into consideration. As a way for improvement, a
frequencies of the strongest interferers from the list for T-FH and
reduced frequency list was proposed. First simulations with such
thus to prevent both, the generation and the reception of
a list indicate a considerable decrease in interference.
interference on these frequencies. However, the less frequencies
are available for T-FH the smaller is the interference averaging Therefore we will further investigate the gain that can be
effect. achieved by a reduced frequency list. Results for the
consideration of a GSM microcell layer can be found in [10].
Possible strategies for a reduced frequency list are to exclude:
• all frequencies of the potential serving cell, VII. CONCLUSIONS
• all BCCH frequencies of the potential serving and their We presented an interference analysis of a hierarchical GSM
neighbor cells (expecting that the other frequencies are not radio network which reveals capacity reserves in the GSM for
occupied with the maximum BS transmit power due to CTS or GSM picocells. The investigated CTS concept benefits
downlink power control), from the Total Frequency Hopping multiple access technique,
• all BCCH frequencies (as before) and all further frequencies which allows for low-cost CTS Fixed Parts that make use of
of the potential serving cell, GSM handset chipsets with only modified software and
• all frequencies of the potential serving and their neighbor protocols. The simulation results revealed poor CTS downlink
cells, performance for the assumptions of our reference scenario due to
• the frequencies of the strongest interferers due to not modelled interference reduction features. However, for
measurements. improvement we proposed a separate consideration of the
transmit power ratios in uplink and downlink and a reduced
For the latter two cases (18 from 27 available GSM frequencies
frequency list.
excluded due to measurements and 21 frequencies excluded due
to the frequency plan) we already carried out simulations for the It is reasonable to assume that these observations are not only
scenario with 17 dBm CTS transmit power and 1500 CTS/km2. GSM-specific. Therefore, a similar hierarchical concept should
The results presented in Table 9 indicate a considerable early be considered in the UMTS standardization efforts.
performance improvement already for reduced frequency lists.
But as expected, the interference portion that is caused by other REFERENCES
CTS is visible now, because the CTS have to share in only 6 or 9 [1] M.I. Silventoinen, M. Kuusela, and P.A. Ranta, “Analysis of a New
frequencies within a certain area. Channel Access Method for Home Base Station,” in Proceedings of
the ICUPC’96, 1996, pp. 930-935.
Table 9. First results for a reduced frequency list (gray) in comparison with the
respective results for our reference scenario and for separate band operation with [2] G. Zimmermann, W. Stahl, and R. Toy, “GSM Based Cordless
3 T-FH frequencies [GSM-caused part of interference in square brackets] Telephone System in Cellular Environment,” in Proceedings of the
Number of T-FH 27 shared 6 shared 9 shared 3 separate EPMCC’97, 1997, pp. 419-426.
frequencies (ideal AFA) (measurements) [3] “Digital cellular telecommunications system (Phase 2+); GSM
CTS-FP Pout 2.9 % [2.6 %] 1.3 % [0.1 %] 0.8 % [0.3 %] 2.8 % Cordless Telephony System (CTS), Phase 1; Service Description,”
CTS-MS Pout 25.5 % [25.3 %] 18.5 % [17.2 %] 18.4 % [18.0 %] 2.7 % TDoc SMG1 509/97, ETSI 1997.
[4] “Low interference GSM-Cordless Telephony System description,”
VI. SUMMARY TDoc SMG2 WPB 121/97, ETSI 1997.
Our simulator applies a novel method that allows for the evalu- [5] A. Noll Barreto, J. Deißner, G.P. Fettweis, “A Frequency Hopping
ation of the dynamics of the cellular environment and especially Algorithm for Cordless Telephone Systems,” to be presented at
ICUPC’98.
of the algorithm for Total Frequency Hopping in relation to the
deinterleaving and decoding gain of a GSM receiver. [6] “Definition of the Interference Environment for the Evaluation of
the Interference Performance of GSM-CTS Radio Interface
Our simulations of the CTS density (up to 5000/km2) indicate, Concepts,” TDoc SMG2 WPB 69/98, ETSI 1998.
that the increase of that parameter in the shared band scenario [7] “Digital cellular telecommunications system; Radio network
with 27 frequencies does not significantly deteriorate the planning aspects (GSM 03.30 v. 5.0.0),“ ETR 364, ETSI 1996.
interference performance neither in the GSM nor in the CTS [8] “Draft COST 231 Final Report,” COST231 TD(96)042-D, 1996.
system. The main advantage of the T-FH multiple access [9] M. Mouly and M.B. Pautet, “The GSM system for mobile
method, the effect of interference averaging, could be shown in communications”, 1992.
the GSM results for the varying number of T-FH frequencies. [10]“Interference Performance of GSM-CTS: Simulation Results,”
TDoc SMG2 WPB 95/98, ETSI 1998.