METHOD OF OPERATING A TELECOMMUNICATIONS NETWORK AND
MOBILE STATION
BACKGROUND TO THE INVENTION
The present invention relates to a method of operating a hybrid CDMA/TDMA-TDD (code division multiple access/ time division multiple access - time division duplex) cellular telecommunications network, and in particular to a method of TS (time slot) assignment in such a network.
One of the advantages of CDMA is that frequency planning is not necessary if the codes used in the adjacent cells are orthogonal to those used in the cell in question. See A Viterbi, Principles of Spread Spectrum Communication, Addison- esley, 1995.
In a hybrid CDMA/TDMA-FDD (frequency division duplex) system a further degree of freedom is added in the time division of channels. However, this does not alter the basic interference mechanisms. These are that mobile stations (MS's) cause interference at the neighbouring base stations (BS's), and the base stations, in turn, interfere with the mobile stations of the adjacent cells.
When considering a CDMA/TDMA-TDD system the complexity in terms of interference is increased since uplink and downlink are time multiplexed on the same carrier frequency. The combination of a frequency/time slot/code has previously been defined as a Radio Resource Unit (RU), by C Mihailescu, X Lagrange, and Ph Godlewski, "Dynamic Resource Allocation For Packet Transmission in TDD TD-CDMA Systems," in Proceedings of the 1999 IEEE Vehicular Technology Conference, Houston, Texas, USA, May 16-19 1999, IEEE, pp. 1737-1741. If the frames and TS's of two adjacent cells are not synchronised additional interference scenarios can occur, as described by Harald Haas and Gordon J R Povey, "The Effect of Adjacent Channel Interference on Capacity in a Hybrid TDMA/CDMA-TDD System Using UTRA-TDD Parameters," in Proceedings of the 1999 50th IEEE Vehicular Technology Conference, Amsterdam, The Netherlands, September 19 - 22 1999, IEEE, vol. 2, pp. 1086 - 1090. These are that mobile stations can
interfere with each other as can the base stations. This scenario is depicted in Figure 1. Figure 1 shows two RU's one of which is used for the uplink and the second for the downlink. This is shown for four entities (BSa, MSa, BSb and MSb) where BSa and MSa form a communication link, and BSb and MSb form another communication link. Therefore, if BSa transmits MSa will receive and vice versa. The misalignment of two frames is modelled by a time offset (t0jj) which is normalised to the slot duration (tst0i) resulting in: toff
It can be seen that due to the frame misalignment BSa and BSb interfere with each other, the interference being designated IBB. In the same way as there is BSa and BSb interference, MSa and MSb will cause mutual interference (IMM). This means that these interference types are coupled (if there is IBB then IMM also exists). Since these interference scenarios are between same entities we define it as same entity interference. Same entity interference is inherent in a TDD system and as t0ff increases so, too, does same entity interference. In the following we assume that α can only take values of 0 or 1 which means that the slots are synchronised, but the direction of transmission of the same RU (RU's which are used at the same time) between adjacent cells can be opposite if α = 1 or synchronous if α = 0. It can be shown that with α = 1 only same entity interference exists.
If α = 0 the constellation is similar to an FDD scenario where base stations interfere with mobile stations (IBM) and vice versa (IMB). Therefore, using the same logic as with same entity interference (see above) the interference scenario for α = 0 can be defined as other entity interference.
These findings already highlight the fundamental difference between a TDD system and an FDD system, assuming that the RU's can be handled flexibly. It is possible in the TDD mode to choose between two principal interference mechanisms (same entity interference or other entity interference). This additional degree of freedom in a TDD system can be exploited to minimise interference. This results in better system performance in particular if interference limited multiple access methods such as DS-CDMA (direct sequence CDMA) are employed. It has been shown by the inventors that for a simple two cell
scenario it is possible to achieve greater capacity in the TDD mode compared to an equivalent FDD mode by using a simple DCA (dynamic channel allocation) strategy. We now consider a network of TDD cells to account for full spatial coverage and show that it is still feasible to achieve better system performance compared to an equivalent FDD air interface. The present invention relates to a decentralised DCA.
Interference characterisation in a CDMA/TDMA— TDD network
With the aid of Figure 2, which schematically shows a cellular network, the total other- cell interference at the receive TS of any MS m can be found as
K
= y ^ +(i -«.) ^ , (2) 'MM IBM
where K is the number of surrounding cells. K is confined to cells which significantly contribute to interference in the cell of interest (usually the first and second tier of cells). is the total number of active users in the neighbour cell j, a,,k is the path loss between the user of interest, m, and the interfering user i, similarly a m is the path loss between the user of interest, m, and the BS of cell j, T g is the transmission power of user i in cell j and Tβ$ is the total power which the BS in celly transmits in the current TS. Similarly the interference observed at any BS u can be calculated as
From (2) it can be seen that the interference at any MS is composed of MS-»MS interference scaled by a synchronisation factor α,. Due to the synchronisation factor the
MS-»MS interference is modified which is indicated by (IMM) instead of (IMM). The second type of interference experienced at any MS is BS- MS interference (IBM) for
which a similar scaling applies as for (IMM). Hence, the magnitude of each type of interference can be manipulated by ,- withy = 1, 2, ..., K.
It can be shown that the interference is composed of IMM and IBM, with each having a
5 different magnitude. Therefore, any location within the network can be characterised by an interference vector with one component being the accumulation of the interference resulting from all mobiles. The second component describes the sum of the total interference caused by the BS entities.
° SUMMARY OF THE INVENTION
It is an aim of the present invention to provide a method of dynamic channel allocation involving little interference and less blocking and outage.
5 The present invention provides a method of allocating a channel to a mobile station in a telecommunications network comprising contiguous cells each containing a base station, the method comprising measuring at said mobile station the sum of the interference resulting from all mobile stations, measuring at said mobile station the sum of the total interference caused by base stations, and using the measurements to allocate a channel ° comprising a pair of time slots for which the minimum interference component applies.
The minimum interference on average can be lower than the other cell interference which is the only interference mechanism in an equivalent CDMA/TDMA-FDD network. This is the reason why this method can achieve better perfonnance than an equivalent FDD system. 5
However, the problem of measuring the total interference ratios is that α,- for each celly = 1, 2, ..., K can be different and therefore IMM or IBM does not reflect the total interference power of the mobile stations and base stations respectively. The real interference contribution from the other mobiles (IMM) can be determined only if 0 α,= 1 V / (4)
Similarly, IBM can be found only if
O V (5)
order to always fulfil these requirements, the present invention preferably uses a fixed TS assignment scheme which maintains the properties as stated by (4) and (5), but gives the flexibility to exploit the fact that the interference at any location in a TDD network is composed of two components. Consequently it will be possible to assign a new mobile to TS's for which the minimum interference component applies.
Preferably, then, the inventive method comprises defining fixed subsequent and repetitive time slots in each cell, the time slots in each cell comprising pairs each including a transmit time slot and a receive time slot, such that at least one of the time slot pairs in each cell has its transmit time slot simultaneous with respective receive time slots in every cell adjacent to said each cell, and its receive time slot simultaneous with respective transmit time slots in every cell adjacent to said each cell.
Preferably, the determination of IMM and IBM, in a case when a mobile station is to be allocated to a channel which is not defined by said at least one time slot pair (i.e. a channel other than the opposed channel), comprises measuring, at said mobile station, the modified interference between mobile stations IMM scaled by a synchronisation factor α,- for each adjacent surrounding cell j = 1, 2, ..., K, as well as measuring, at said mobile station, the modified interference of the surrounding base stations with said mobile station IBM scaled by said synchronisation factor, and calculating IMM and IBM from the respective modified interferences. In order to find IMM and IBM it is preferable that an idle frame for TS n in cell n (i.e. the time slots in the opposed channel) be introduced. This idle frame enables the opposed pair of TS's to track the measurements of IMM and IBM after the RU has been allocated. In particular, IMM and IBM can be found respectively from the sum of IMM in the idle frame, and of IMM in cell n, and the sum of IBM in the idle frame, and of IBM in cell n.
This novel scheme of fixed TS assignments is related to the fast DCA as defined in the 3rd Generation Partnership Project (3GPP), Technical Specification Group Radio Access Network, "Radio Resource Management Strategies," 3G IS 25.922 V3.0.0 (2000-01), January 2000. (Note that this TS assignment scheme does not mean that some TS's can only by used in certain cells. The frequency and TS re-use of one is still maintained. The TS assignment is only with respect to RX (receive) and TX (transmit) transmission direction of TS's in a TDD system.) It can be shown that the fixed TS assignment scheme of the invention in addition to a simple fast DCA improves user blocking and outage significantly. With the combination of both methods the results outperform those of an equivalent FDD system.
In an embodiment of the inventive method, when a new mobile station enters the network, IBM and IMM are determined for the opposed channel in the new mobile's cell, and if IMM < IBM then the opposed channel is allocated to the new mobile station. If IBM ≤ IMM, IBM and IMM are determined for the "quasi" synchronous channel n and if both IBM < IMM and IBM < IMM are satisfied then channel n is allocated to the new mobile station. If these two inequalities are not satisfied then this previous step is repeated continually with n incremented by 1 each time. Preferably, if a channel has still not been allocated when n has reached the maximum number of channels then the channel in which the interference is the minimum base-mobile interference min(i ) is allocated to the mobile station.
The algorithm prevents destructive MS«-»MS interference which is a problem in the TDD mode.
The present invention also provides a mobile station for use in a telecommunications network comprising contiguous cells each containing a base station, the mobile station comprising means for measuring the sum of the interference resulting from all mobile stations, means for measuring the sum of the total interference caused by base stations, and means using the measurements to allocate a channel comprising a pair of time slots for which the minimum interference component applies.
Preferably, the mobile station comprises means for measuring the modified interference between mobile stations IMM scaled by a synchronisation factor α,- for each surrounding cell j = 1, 2, ..., K, adjacent to a cell in which the mobile station is located, as well as means for measuring the modified interference of the surrounding base stations with said 5 mobile station IBM scaled by said synchronisation factor, and calculating IMM and IBM from the respective modified interferences, i order to find IMM and IBM it is preferable that an idle frame for a time slot n in cell n (i.e. the tune slots in the opposed channel) be introduced. Preferably, the mobile station comprises means for finding IMM and IBM respectively from the sum of IMM in the idle frame, and of IMM in cell n, and the sum of ° IBM in the idle frame, and of IBM in cell n.
BRIEF DESCRIPTION OF THE DRAWINGS
Particular embodiments of the present invention will now be described in more detail, by ^ way of example only, with reference to the accompanying drawings, in which:
Figure 1 is a diagram showing interference in a known network as discussed above;
Figure 2 is a diagram showing other-cell interference in a known network as discussed 0 above;
Figure 3 is a graph of possible interferences in a network in which the method of the invention can be used;
5 Figure 4 is a schematic network map showing the TS assignment scheme of the invention;
Figure 5 is a diagram showing the TS assignment scheme of the invention;
Figure 6 is a flow chart showing the DCA algorithm of the invention; 0
Figure 7 is an example of a cell topology and interference plot;
Figure 8 is an example of a 3D interference plot;
Figure 9 is a graph of relative interference probabilities in a cross section of the plot of Figure 8;
Figure 10 is a graph of outage;
Figure 11 is a graph of downlink (DL) and uplink (UL) blocking;
Figure 12 is a graph showing the percentage reduction in blocking achieved by the invention as compared to an equivalent FDD case;
Figures 13 and 14 are a 3D plot and a cross section thereof respectively showing interference when two networks operate with no mutual x and >> offset;
Figures 15 and 16 are a 3D plot and a cross section thereof respectively showing interference when two networks operate with a first mutual x and y offset;
Figures 17 and 18 are a 3D plot and a cross section thereof respectively showing interference when two networks operate with a second mutual x and y offset;
Figure 19 is a graph of outage for the three two-network cases of Figures 13 to 18;
Figure 20 is a graph of DL and UL blocking for the three two-network cases of Figures 13 to 18; and
Figure 21 is a graph of percentage reduction in blocking achieved by the invention as compared to an equivalent FDD case for the three two-network cases of Figures 13 to 18.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
In deriving a fixed TS assignment scheme, we assume that the most significant interference results from the first tier of cells around the cell of interest (COI) with the underlying assumption of hexagonal cells. A pair of TS's forms a set of communication channels the number of which is dependent on the number of available spreading codes. In Figure 3 a possible interference scenario in a TDD system is depicted. One channel consists of at least two TS's, one for the uplink and one for the downlink. It is assumed that one frame is composed of 7 channels. Each channel can be characterised by an IMM and an IBM component. According to the invention, an intelligent DCA algorithm is able to find the lowest interference component, min(7), out of all possible channels. If min(i) is associated with an Zwtype of interference, the process of assigning a new mobile to TS's for which the minimum interference applies will now mean that this user will be allocated to a channel with opposed TS's.
Here a fixed TS scheme is developed where at least one pair of TS's is opposed to all 6 neighbouring cells to enable the measurement of IMM and IBM- Figure 4 depicts the novel TS assignment plan and shows a multiple of 7-cell clusters. Each cell is supposed to occupy 7 subsequent pairs of TS's. In cell 1, for example, the first pair of TS's is opposed to all 6 neighbouring cells (cell 2 to cell 7), illustrated by the capital "X", and hence a} — 1 V = 2, ..., 7 from the viewpoint of cell 1. It automatically follows that the TS pair 1 is synchronous in all 6 neighbouring cells (cell 2 to cell 7), i.e. cell 2 to cell 7 transmit and receive at the same time with respect to the first two TS's. Similarly, in cell 2, the second pair of TS's (TS 3 and TS 4) are opposed to all their neighbouring cells. This is repeated in cell 3, and so on. In Figure 5 the actual TS assignment is depicted. This plan illustrates how the transmission and receive directions are distributed throughout a 7-cell cluster. This cluster can be repeated as often as desired to achieve full spatial coverage. It can be seen that in cell 1 only TS\ and TS are opposed to all other adjacent cells. In cell 2 it is TS2ι and TS2 2, in cell 3 TS3ι and TS3 2, etc. This scheme enables the easy measurement of I M and IBM- When it is an opposed channel, as in TS\ and TS 2 in cell 1, the MS will be able to measure IMM at TS*2 and IBM at TSS respectively. This is because , = 1 Vj = 2, 3, ..., 7. This measurement is slightly more complicated for the "quasi" synchronised TS's as,
for example, TS 2 TS ι and TS 2 in cell 1, the reason being that α2 = 1, but ,- = 0 for j = {1,3,4,5,6,7}. Therefore, for any TS n with n ≠ cell n, the interference components are measured as:
BM = ∑ l MX uMn (6) ι,ιψn
In order to find IMM and IBM it is suggested that an idle frame for TS n in cell n be introduced. This idle frame enables the opposed pair of TS's to track the measurements of IMM and IBM after the RU has been allocated. The duration of a multi-frame having one idle frame will depend on the maximum specified time for "service establishment" and the maximum speed of a mobile in the TDD deployment enviromrient. Let the maximum speed of a mobile be 200 km/li and assuming that the slow-fading channel will not change its characteristic within a range of 5 m, the time interval within which an idle frame is required will be 90 rήs. With a frame duration of 10 s, the idle frame would need to occur every ninth frame. In the event of an idle frame, (6) and (7) become
feMi ie — ∑ IBM,- i,i≠n (8) MMIΛ ZJ IMMI- > w* (9)
and subtracting (8) from (6) and (9) from (7) gives
?MMπ = ΪBM ~ΪBMidle
(10) BM„ = lMM -ΪMMi e , (H)
and consequently,
MM = tΛ + uMn (12) l M - MuιXhMn ■ (13)
Again, IMM in, for example, cell 1 for TS 2 can be directly measured at TS ι and IBM at
TS2 2.
The DCA algorithm
The concepts presented in the previous section can be used to build the foundation for the DCA depicted in Figure 6 which is operated as a slow DCA according to "Radio Resource Management Strategies" (supra). The algorithm in Figure 6 ensures that severe MS— MS interference is avoided. This can be proven with the aid of (6) and (7).
Let a = Σu≠nlBMi, c = IMM„, b = IBMn and d = ∑]>i≠nlMM
Then
and it follows that
IBM = a+ b (16) IM = c + d . (17)
The "if statement in Figure 6 can now be expressed as,
a+b < c+d (18) a+ c < d+b , (19)
and solving the second equation for a and substituting into the first equation gives
b < c (20)
which
tne sulgl
e ceft with opposed TS's, n, out of six adjacent cells, and this is exactly what was intended to be demonstrated. Note that five out of six cells are synchronous as can be seen from Figure 4.
Simulation platform
Simulations have been conducted under the following conditions:
- Non-optimal power control applied.
• - Mobiles are assigned to a BS based on the minimum path loss.
- Handover margin used.
- Correlated shadowing used.
- 16 kbps service assumed.
- UTRA (UMTS Terrestrial Radio Access)-TDD parameters applied.
- 7-cell cluster with the COI in the centre.
The parameters used are summarised in Table 1 below:
Interference characterisation for a single operator network
Different user population have been applied for the first tier of interfering cells. Subsequently the interference was evaluated using a quadratic mesh. At each grid point it is assessed whether IMM is greater than IBM. This is done using the following assignment:
-1 if IMM > IB , + 1 otherwise, (21)
Then N= 1000 Monte Carlo runs were carried out and for each grid point the interference is characterised according to:
N
(22)
Using m it is possible to calculate the probability that IBM>IMM
z = Pr(IBM > IMM = 0.5 (l + ^) (23)
A contour plot and the cell topology is shown in Figure 7. Positive contours indicate a low value for z whereas negative numbers point to a greater value of z, i.e. IMM interference is dominating. The same picture is revealed by the 3d plot in Figure 8. It can be found that within the centre cell IMM is slightly dominating with the probability of 55% - 70 % depending on the load in the interfering ring of cells. This can be seen in Figure 9 where PT(IBM > IMM) is depicted for a cross section in Figure 8. The cross section chosen is the line determined by x = y, thus the line in the x-y plane with the slope of gradient being 1. In addition the load in the ring of interfering cells is varied between 2 and 7 users per TS. It can be seen that at the BS of the centre cell the probability, PT(IBM > IMM), is between 30% and 45%. This probability increases to about 55% for a mobile location at approximately 180 m distance from the BS.
System performance for a single operator network
Figure 10 shows outage compared to an equivalent FDD system. It can be seen that for all populations the decentralised DCA scheme outperforms the equivalent FDD scenario. The greatest benefits are for 6 active users per set of channels. In Figure 11 user blocking measured in "average attempts for call establishment" is depicted. Also in terms of blocking the proposed scheme DCA algorithm results in better performance. This is underpinned by Figure 12 where the relative total blocking is shown. The results for 2 and 3 users are less significant since in this state the absolute rate of blocking is very small. However, for 6 users the improvement is still about 22%.
Interference characterisation for a two-operator network
In the previous section the scenario with only a single carrier has been investigated. In a second scenario we assume that in addition to this a further carrier adjacent to the previous one is employed. According to the UMTS (Universal Mobile Telecommunications System) specification: 3rd Generation Partnership Project (3 GPP),
Technical Specification Group Radio Access Network, "Physical channels and mapping of transport channels onto physical channels (IDD)" 3G IS 25.22 1 V3.1.1 (1999-12), December 1999, the carrier spacing is 5 MHz. The licensed TDD frequency band in UMTS is 20 MHz, and thus it is very likely that one operator only gets one TDD carrier. As a consequence, the networks are not co-ordinated and may overlap in a random manner. Hence, in the simulation we generated an overlaid cluster of 7 cells modelling the network of a second operator. The adjacent channel protection factor is assumed to be 30 dB. Again, users in both networks have been distributed equally in space and the interference calculated at grid points of a quadratic mesh. Three scenarios have been investigated, hi the first scenario the x and v offset of the overlaid cell cluster equals zero, i.e. the second network sits on top of the first network. In a second scenario the x andy offset is set to 32.48 m and in the third scenario the x andy offset is 64.95 m. In all cases the load in each cell for each TS is 4 users.
Figures 13 and 14 illustrate the resulting interference for the "on-top" case. Figure 14 highlights the interference for the cross section determined byx =y.
Figures 15 and 16 illustrate the resulting interference for the case where the x and j shift is 32.48 m.
Finally, Figure 17 and Figure 18 demonstrate the resulting interference for the case where the x andy shift is 64.95 m.
All scenarios reveal that the most significant difference compared to the single operator case occurs at locations around the BS of the co-existing network. At these locations the probability that IBM is greater than IMM is about 1. However, the most interesting result is that this state does only apply for a very limited area around the BS of the overlaid network, i.e. the impact of the interference generated by the base stations of the coexisting network is spatially limited indicated by a sharp negative gradient of the function P (IBM > IMM) at a distance of 5 - 10 m from the BS. This is indicated by the spikes occurring in the interference plots.
System performance for a two-operator network
In Figure 19 outage is plotted for the three scenarios described above. In Figure 20 uplink versus downlink blocking within the centre cell of the victim network is depicted. In Figure 21, blocking using the decentralised DCA of the invention compared to the case where all TS's are aligned is illustrated. It can be summarised that for all scenarios except for the co-location of the BS's (x and y offset equals zero) the new techniques result in better system performance compared with the "ideally" synchronised case. This case has always been assumed to be the optimum state in a TDD network.
It is anticipated that in the case of co-locating the BS's the decentralised opposing algorithm cannot achieve better results than using synchronous transmission and transmission and reception because the BS<->BS interference is too high. Therefore, any decision of the MS to use an opposed channel is very likely to result in outage.
Conclusions
The fixed TS assignment scheme of the invention, in combination with the decentralised DCA, is able to achieve less blocking and outage than an ideally synchronised TDD network (TX and RX is at the same time throughout the entire network). Since such an ideally synchronised network can be compared to an equivalent FDD system, the results reveal that with the TDD mode it is possible to obtain better system performance which can be attributed to the greater flexibility in TDD. This result, in particular, is quite significant because it implies that the spectrum efficiency in the TDD mode can be better than in the FDD mode — also when the TDD mode is used for full spatial coverage.
It has also been demonstrated that the methodology proposed results in an improved system performance in an environment with adjacent channel interference.
The decentralised DCA algorithm of the invention can also be applied to the TD-SCDMA
(time division - spatial code division multiple access) system which is related to the UTRA-TDD system. The basic difference is that the 5 MHz bandwidth of a UTRA-TDD
carrier is subdivided into three carriers, each with a bandwidth of 1.6 MHz. In addition, the number of time slots in the TD-SCDMA system is reduced to 7 as compared to 15 in the UTRA-TDD system. The first two time slots are fixed in terms of the link direction. To be precise, one time slot is used for the uplink and the other for the downlink. The remaining 5 time slots can be used for either the uplink or the downlink. Due to these similarities, the invention can also be applied to the TD-SCDMA system.