GB2356774A - Synchronisation of base stations - Google Patents

Description

1 2356774 IMPROVEMENTS IN OR RELATING TO MOBILE TELECOMMUNICATIONS SYSTEMS

The present invention relates to improvements in or relating to mobile telecommunications systems, and is more particularly concerned with improving the synchronisation of base stations within such systems.

The UMTS terrestrial radio access time division duplex (UTRA TDD) mode is based on a combination of code division multiple access (CDMA) and hybrid time division multiple access (TDMA). UMTS is an acronym for universal mobile telecommunication system as will be understood by persons skilled in the art.

Reliable operation in the UTRA TDD mode, incorporating the combined TD-CDMA multiple access scheme, requires synchronisation between base stations within a compliant telecommunications system.

Moreover, the mode also requires the provision of position information for the mobile stations affiliated to each base station. Synchronisation between base stations is also desirable in order to maximise system capacity. To these ends, the synchronisation of base stations must be achieved at the levels of time slots, frames and multi-frames, where a multi-frame is a repeating cycle of a number of frames.

Co-pending British patent application no. 9919973.9 (Ref1999PO4839) describes a method of providing synchronisation between a plurality of base stations in a telecommunications system which comprises providing a random access channel in each cell. A mobile station receives a resource unit via the random access channel and a base station uses the random access channel in one cell to transmit a synchronisation signal to other base stations within the system.

The base stations are synchronised together by means of transmissions between base stations. This involves measuring timing differences between the base stations in both directions to remove the effect of propagation delays. The timing differences are reported back to a central radio network controller (RNC) which computes the necessary timing updates and signals these back to the base stations.

In theory, this method could achieve perfect synchronisation provided the measurements are accurate and provided the clocks are all running at the same rate. However,, in practice every clock will have a rate inaccuracy such that the timings of the base stations will tend to drift apart between updates. This results in a trade-off between the required accuracy of the clocks, the timing update ratio and the achievable synchronisation accuracy.

It is therefore an object of the present invention to improve the synchronisation accuracy between base stations.

In accordance with one aspect of the present invention, there is provided a method of adjusting frequency errors in a telecommunications system comprising a radio network controller and a plurality of base stations each having its own clock, the method comprising the steps of.

a) providing a timing signal from the radio network controller for the plurality of base stations; b) determining a timing difference at each base station relative to the timing signal; c) determining a timing update for each of the base stations based on the timing difference; d) using the timing update to determine the effective frequency error for each of the base stations; and e) using the effective frequency error to effect frequency error correction for each of the clocks at the base stations.

Preferably, the method further comprises the step of substantially synchronising the base stations prior to carrying out step d).

Advantageously, the effective frequency error calculated for each clock is within tolerance for that clock.

Ideally, step e) is an iterative process.

In one embodiment of the invention,, step c) is carried out by each base station. In another embodiment of the invention, step c) is carried out by the radio network controller.

For a better understanding of the present invention, reference will now be made, by way of example only, to the accompanying drawings in which:

Figure I is a schematic diagram of a telecommunications system comprising a plurality of cells; Figure 2 is a schematic diagram of time differences for signals between base stations; and Figure 3 illustrates time differences of Figure 2 in relation to TDMA frames.

Figure I illustrates a telecommunications system 10 which comprises a plurality of cells 12, 14, 16, 18, 20, 22, 24. It will be appreciated that although the cells 12, 14, 16, 18, 20, 22, 24 are shown as being hexagonal in shape, they can be of any convenient shape. Each cell 12, 14, 16, 18, 20, 22, 24 has a base station 32, 34, 36, 38, 40, 42, 44 and some of the cells are shown having one or more mobile terminals 26.

As shown, the range between neighbouring base stations 32, 34, 36, 38, 405 42, 44 is roughly double the range from any base station to a mobile terminal 26 within its cell 12, 1411 16, 18, 20, 22, 24. This typically leads to a path loss for one base station to a neighbouring base station of the order of l2dB greater than that to a mobile terminal located at the edge of the cell.

In the UTRA TDD mode, information is transmitted in bursts at a certain combination of frequency, time and coding. Frames are divided into time slots and each time slot is just long enough for a single burst of information. Transmission of information is multiplexed through the use of orthogonal codes, that is, code division multiple access (CDMA). The information transmitted within a particular time slot is divided according to these codes, and as a result, each burst contains a plurality of independent time slot and code combinations called resource units.

Moreover, a random access channel (RACH) is provided which corresponds to a single time slot per frame. The RACH is allocated to transmissions from mobile terminals to initiate communications - usually by requesting a resource unit for uplink usage. The RACH can be used for both inter base station synchronisation and for mobile tern-linal position location.

In accordance with the invention, a base station conforming to the UTRA TDD mode is provided which uses the RACH to synchronise with other base stations which are within transmission range. The base station is arranged to steal the RACH time slot for transmissions to other base stations at suitable times. For ease of explanation, it will be assumed that the same time slot is used for RACH operation in all cells, but it will be appreciated that this assumption, whilst advantageous, is not essential.

The times at which a base station should steal a RACH time slot can be determined according to the following criteria:

1) neighbouring base stations must not steal the RACH time slot in the same frame; and 2) RACH time slots must be stolen frequently enough to maintain overall base station network synchronisation to the required accuracy.

3) schedules for RACH time slot stealing may be determined either centrally by a radio network controller (RNC) or according to sequence generators resident in the base stations. In the latter case, the sequence generators are arranged in such a way that RACH stealing schedules do not coincide in neighbouring cells. If the RNC is used, it can establish schedules according to this criterion. The schedules may be at regular, pseudo random 5 or constrained random intervals.

When the base station has a schedule assigned for RACH stealing in the near future, at a suitable time it makes a broadcast transmission (preferably on its broadcast control channel (BCCH)) to all mobile terminals affiliated to the base station, to instruct these mobile terminals that the RACH will be unavailable for mobile terminal transmissions in the forthcoming scheduled stolen RACH time slot. This will clear the stolen RACH time slot for inter cell synchronisation usage.

Arranging for the stealing base station to silence mobile terminals affiliated to the stealing base station when the RACH is stolen, will prevent unnecessary collisions on the RACH channel. However, as described so far, the neighbouring base stations will not silence their respective affiliated mobile terminals from making RACH transmissions. These RACH transmissions will be power controlled and it should be possible for the neighbouring base stations to receive the transmission from the base station stealing the RACH timeslot and to receive any RACH transmissions from their own affiliated mobile terminals. However,, in the case where stolen RACH timeslots are scheduled by the RNC, it is optionally possible to arrange for the neighbouring base stations to silence RACH transmissions from their mobile terminals using the same procedure as described for the RACH time slot stealing.

In this way, the interference to the synchronisation transmission can be substantially removed, except from distant stations. If this option is not employed then interference to the reception of synchronisation transmission in the RACH timeslot may prevent its reception. However, given the statistics of RACH traffic, a high proportion of such measurements should be received.

An alternative approach to 'stealing' RACH slots for synchronisation can be used in which RACH slots are arranged throughout the network of base stations to be allocated to synchronisation at regular fixed intervals. During these allocated RACH slots, none of the mobile terminals make RACH transmissions, and it is unnecessary to instruct the mobile terminals not to make the RACH transmissions since they are capable of determining such times for themselves. However, the base stations do transmit a simple binary signal periodically to indicate that this mode of operation applies. Such a transmission would not be necessary in a network where all base stations had associated GPS receivers. During the allocated RACH time slots all base stations are either listening for synchronisation transmissions or making them. The subset of base stations making synchronisation transmissions changes from one selected RACH time slot to the next. It is necessary to ensure that the spread of transmissions is such that only one dominant synchronisation signal is received at any given base station in any given selected RACH time slot. The planning of these subsets can be performed either manually or automatically according to scheme similar to dynamic channel assignment (DCA).

Within UTRA TDD, bursts are transmitted within time slots and each burst is sub-divided into 2560 chips which are zoned into two data fields, one midamble field and a guard period. The midamble field contains training sequences. Because the base stations are static and have accurate frequency references, it is possible to perform correlation across the entire time slot. Correlation makes use of training sequences so the synchronisation burst, with the exception of the guard period, isarranged to have no data fields and effectively becomes all midamble. Whole time slot correlation affords a processing gain of about 34dB. This high processing gain serves to compensate for the increased path loss to the neighbouring cells.

Assuming that every base station sends and receives synchronisation bursts to and from its neighbouring base stations, all of the information necessary for the network wide synchronisation can be aggregated. This can be used in one of two distinct ways, either distributed or centralised.

Either method of using the RACH timeslots described above can be implemented according to a distributed or centralised approach.

In the distributed approach, every base station acts autonomously on the basis of the information it has received to adjust its clock timing in such a way that, given that all other base stations operate similarly, they will come into synchronisation.

In the centralised approach, all base stations report their results to the RNC which then computes a set of adjustments and signals these adjustments individually to the relevant base stations. Essentially, each base station measures the timing of each received synchronisation burst relative to its own timing. This can be viewed as the timing of the received burst relative to the time at which it would make its transmission.

Referring now to Figure 2, the relative positioning of one base station A and two of its neighbouring base stations B, C is shown. The time differences dij between any two base stations can be derived from the synchronisation signals. In Figure 2, the time difference between base stations A and base stations B, C are shown as d,,,b and da,c respectively.

These time differences are also shown in Figure 3.

Figure 3 shows a TDMA frame 50 which comprises 15 time slots including a time slot 52 for the RACH transmission which is shaded. Time slot 52 carries the synchronisation signal. Boxed area 60 represents the time differences at base station A for base stations B and C as shown in Figure 2. Line A relates to base station A's own signal, line B relates to the delayed signal from base station B giving a time difference of da,b, and line C relates to the delayed signal from base station C giving a time difference of d,,,,.

Boxed area 70 represents the time differences at base station B. Line B relates to base station B's own signal and line A relates to the delayed signal from base station B giving a time difference of db,a. There is no line corresponding to base station C as it is out of range of the signal from base station B. As can be seen from line B in box 60, d,b comprises a five time slot delay. Similarly, line C shows da,v to be just over one time slot delay. In box 70, line A shows a two time slot delay for db,aSuppose we have a deployment of N base stations. Let the variable L(ij) = L(ij) indicate those base stations which are able to hear each other's synchronisation transmissions. If base station i can hear base stationj's transmission and base stationj can hear base station i's transmission then L(ij) = L(i,i) = 1. Otherwise L(ij) = L(ij) = 0. Note that L(ij) = 0 for all i. All relative timings are aggregated at the RNC. If base station i hears base stationi's transmission with delay dij and base stationj hears base station i's transmission with delay dji, then the RNC computes the time differences as Jij dij - dj,i 2 and Sili dj,i - dij 2 Referring once more to Figure 4, it is plain that L(a,b) = L(b,q) = 1 and L(a,c) = L(cq) = I but L(bc) = L(cb) = 0. Figure 4 also illustrates how the time differences, dij, are derived. Thus i5ij is the time by which base station i's time is advanced with respect to the time of base stationj and excludes any time delay due to intervening distance.

Suppose base station i will be retarded by a compensation amount Ci which is to be computed. Following such compensation, the new timing error 5 between base stations i andj will be given by.5i"i = 9i,j - Ci + Ci If all measurements were completely accurate and consistent, we could simply solve the equations to make 5,',j =- 0 for all i and all j.

However, given measurement errors it is better to solve for a 10 minimum sum square error, that is, N N Z Z L (i, j) 5i"2j i=1 j=1 should be niinimised. Expanding this gives:- N N j)1,52 Y Z L(i, + C2 + C2 + 2((5i,j -C ij i i i - 8ij -Ci - Ci Cj i=1 j=1 N Let A)= ZL(ij) be the number of base stations whose j=1 synchronisation transmissions base station i can hear and who can also hear base station i's synchronisation transmission. We can then express the sum square error as N N N N N 2 N 2 M(i)Ci2 + ZZL(ij).5j2 - 4Y Ci I L(i, jJj,j - 2Z E L(i, j)Cj Z j i=1 i=1 j=1 i=1 j=1 i=1 j=1 Now differentiate with respect to Ci and equate to zero. We obtain N N 4M(iCj - 4Y L(i,j).(5i,j - 4Y L(i,j)Cj = 0 j=1 j=1 Thus N N M (iCj - Z L(i, jCj = I L(i, j),5i,j j=1 j=1 We can express this in matrix notation as (diag(M)- L)C = D where diag(M) is the diagonal matrix with elements M(i i c- 11 NJ along the diagonal, L is the matrix with elements L(ij), C is the vector with elements Ci, and D is a vector with elements M(i) Di L(i, j)gi,j I 9j's, W j=] j=1 where sjij i c= f1..M(i)J is the set of indices of base stations to and from which base station i can send and receive synchronisation transmissions respectively.

Let A= (diag(M) - L) This matrix is singular, i.e. has no inverse. This reflects the fact that any common value can be added to all compensation values, Ci, without affecting the sum square error. A reasonable constraint to apply to the compensation values is that their sum should be zero so as to minimise the overall drift. Tbus,, we have an additional equation:- N YC =0 This can be reflected in the matrix equation by adding a row of ones to any of the rows in A to form A'.

We can now solve the equation to obtain the compensation values.

However, we can note that A (and therefore A') does not change very rapidly, if at all, since it is a function only of the base station connectivity. Thus, it may be more efficient to compute the inverse of A' which need only be updated infrequently. We thus obtain C = (A5')-'. D - I I - Having these compensation values, Ci, each base station in the telecommunications system can be synchronised with every other base station. In addition to fulfilling the requirements of UTRA TDD mode, synchronisation is important in locating mobile stations.

However, as discussed above, each clock has a particular rate inaccuracy, the timings of the base stations will tend to drift apart between updates. In accordance with the present invention, synchronisation accuracy is improved by adjusting not only the clock timings at the various base stations, but also the clock rates.

When a base station receives a clock update, it can note the period of time since the previous update. The effective adjustment rate is then given by the size of this adjustment divided by the time since the previous adjustment.

Essentially, if the clock rate had been altered by this amount, the adjustment would not have been necessary. In accordance with the present invention, the clock rate of a base station is to be adjusted in dependence upon the computed effective adjustment ratio.

During the initial phase of synchronisation, the clock adjustments are related to removing potentially large time offsets rather than making adjustments for drift. It is not appropriate to adjust the clock rate during this phase, and any adjustments to the clock rate needs to be made in or approaching a steady state phase.

Therefore,, it is necessary to determine whether the synchronisation process is in the initial phase or when it is in or approaching the steady state phase.

In a preferred embodiment of the present invention, it is necessary to compute the effective adjustment rate and compare its magnitude with the known range of errors for the clock in use. For example, if the clock has an accuracy specified as within 0. 1 ppm, then an effective adjustment rate of 1 Oppm cannot arise primarily from clock rate errors. It would therefore be inappropriate to attempt to adjust the rate by this amount. Thus, rate adjustments should only be applied when the effective rate adjustment is substantially within the range of possible clock rate effors.

In a synchronised or synchronising network, the timing adjustments are applied to all base stations and there may be some interdependency in the updates computed. It is therefore appropriate to stabilise the network by making the clock rate adjustment using only a fraction of the measured effective adjustment ratio. This allow the network synchronisation to settle into a stable equilibrium.

In one approach to synchronisation described above, every base station acts autonomously on the basis of the infort-nation it has received to adjust its clock timing in such a way that, given that all other base stations operate similarly, they will come into synchronisation.

In a preferred approach described above, all base stations report their results regarding timing to the RNC which then computes a set of timing adjustments and signals these adjustments individually to the relevant base stations. Additionally, the RNC could also send out frequency adjustments as well as the timing adjustments.

Once synchronisation has been achieved for all the base stations, adjustments can be made to correct for frequency errors in the clocks themselves. It will be appreciated that the frequency errors are relatively small when compared to the timings of the base stations and that if a steady state has not been reached, very large adjustments will need to be made to correct the frequency errors which may put the system further out of synchronisation.

Generally, the frequency error is quite small compared to the timing errors and can be corrected by an iterative process when it is less than the clock error. This means that the frequency is adjusted by a fraction of the computed error until full correction is obtained. For example, if the error is determined to 0.7ppm, an adjustment of 0.2ppm is made and then iterated.

In a system having a reference frequency of 5MHz (corresponding to a frame as described above with reference to Figure 3, that is, each frame having 5 million pulses), the frequency error may be 0.2ppm, requiring that the reference frequency is changed by I Hz. This can be implemented by the insertion or removal of one pulse at the end of every frame can provide the required correction.

Additionally, if smaller adjustments are required, a pulse can be added or removed every other frame or other suitable multiple number of frames.

Claims (10)

CLAIMS: 1. A method of adjusting frequency errors in a telecommunications system comprising a radio network controller and a plurality of base stations each having its own clock, the method comprising the steps of.- a) providing a timing signal from the radio network controller for the plurality of base stations; b) determining a timing difference at each base station relative to the timing signal; c) determining a timing update for each of the base stations based on the timing difference; d) using the timing update to determine the effective frequency error for each of the base stations; and e) using the effective frequency error to effect frequency error correction for each of the clocks at the base stations. 2. A method according to claim 1, further comprising the step of substantially synchronising the base stations prior to carrying out step d). 3. A method according to claim I or 2, wherein the effective frequency error calculated for each clock is within tolerance for that clock. 4. A method according to any one of the preceding claims, wherein step e) is an iterative process. 5. A method according to any one of the preceding claims, wherein step c) is carried out by each base station. 6. A method according to any one of claims I to 5, wherein step c) is carried out by the radio network controller. I& Amendments to the claims have been filed as follows CLABIS:

1. A method of adjusting frequency errors in a telecommunications system comprising a radio network controller and a plurality of base stations each having its own clock, the method comprising the steps of- (a) providing a timing update signal from the radio network controller for the plurality of base stations; (b) using the timing update signal to determine the effective frequency error for each of the base stations; and (c) using the effective frequency error to perform frequency error correction for each of the clocks at the base stations.

2. A method according to claim 1, further comprising the step of substantially synchronising the base stations prior to carrying out step (b).

3. A method according to claim I or 2, wherein the effective frequency error calculated for each clock is within tolerance for that clock.

4. A method according to any one of the preceding claims, wherein steps (a) to (c) are repeatedly performed in an iterative process.

5. A method according to claim 4 wherein the frequency error correction of step (b) comprises adjusting the frequency by an amount less than the determined error, then performing a further iteration of steps (c), (a) and (b).

6. A method according to any one of the preceding claims, wherein steps (b and (c) are carried out by each base station.

7. A method according to any one of claims I to 5, wherein step (b) is carried out by the radio network controller.

8. A method according to any preceding claim wherein step (c) comprises inserting or removing a number of pulses at an end of certain frames.

9. A method according to any preceding claim wherein step (a) itself comprises the steps of- - providing a timing signal to the radio network controller from each of the plurality of base stations; - determining a timing difference for each base station according to the timing signal; determining a timing update for each of the base stations based on the timing difference; and - providing the timing update signal in accordance with the determined timing update.

10. A method substantially as described.