The active set size distribution can be computed and displayed for each cell of the network.
The purpose of this measurement is to find out how many radio links UEs have in their active set during the observation period of a full measurement session. Aggregation of measure- ment results for an unlimited number of individual calls shall be done at cell level. It can be computed and displayed how often UEs using a radio link of cell X have one, two, three or more radio links simultaneously in their active set.
The maximum active set size for UEs defined by 3GPP is six radio links. Due to capacity reasons most RNCs in today’s networks are configured to allow a maximum of three radio links in the active set. In addition, 3GPP 25.942RF System Scenariosrefers to a number of laboratory simulations and tests and summarises test results as follows: ‘. . .assuming ideal handover measurements by UE and delay free handover procedure the gain of having more than 3 best cells in the active set is minimal. Thus, including extreme cases it can be concluded that UE does not have to support more than 4–6 as the maximum size of the active set.’ Following this for the active set size distribution per cell it can also be assumed that the maximum number of radio links in the active set is six. The measurement can be dis- played as a percentage value in a table in which each column represents a different active set size. An applicable graphic presentation format of this measurement is the histogram. Two case studies will help to explain how active set size distribution is calculated.
Case study 1: it is assumed that during the whole observation period of a monitoring session two UEs are served by a network cell. One of these UEs is in soft handover with a neighbour cell (active set sizeẳ2) while the other UE is not in soft handover and has a single dedicated radio link in use provided by the cell for which the active set size distri- bution shall be computed. In this case the sample table would look as shown in Table 2.20.
However, at any time new radio links can be added to or existing radio links can be deleted from active sets of UEs. Knowing that this occurs Table 2.20 only depicts a momentary situation in a cell. To analyse active set sizes used by all UEs in a cell at any time it is necessary to take samples, but in contrast to BLER samples (discussed in section 2.1) there is no sampling period defined for this measurement and hence there is no reset of counter values if the sampling period timer expires. For the active set size distribution the first sample is taken at the beginning and the last sample is taken at the end of a monitoring session. The total number of samples depends on session duration.
Now another problem arises: which protocol events will be counted? The first idea is to look for NBAP radio link setup and radio link addition procedures, especially since Request messages of these procedures contain a radio link ID information element. The first radio link setup within an active set is indicated by Radio link IDẳ0, the second one by Radio link IDẳ1 and so on. . .until the first radio link is deleted from the active set while the link identified by Radio link IDẳ1 remains in service. The next radio link to be added to the active set will be identified by Radio link IDẳ0. And following this it becomes evident that the NBAP radio link ID cannot be used to distinguish how many radio links belong to the
Table 2.20 Active set size distribution example 1
Active set size 1 2 3 4 5 6
Average percentage 50% 50% — — — —
current active set of a single UE. Hence, a so-called invisible background application must run in performance measurement software that stores and tracks the changing active set size of each UE monitored in network. From this invisible background application the active set size of each UE can be polled periodically. In addition to the size of the active set the application must also know the identity of cells involved in the active set at any time.
The more samples polled during the observation period the higher the precision of measurement values. Thinking about the fact that the average usage time of a radio link in an active set in an optimised network is between 15 and 20 seconds it can be assumed that a polling period of one or two seconds provides a good minimum granularity for an active set size distribution measurement.
Case study 2: in the following example the session duration is 10 minutes and the sampling period is 2 seconds. As a result, it is expected to get a total number of 300 samples within the observation period (duration of monitoring session). If there is no UE active in the monitored cell the ‘0’ counter values will be ignored.
Note: UEs that are in CELL_FACH state using common transport channels RACH and FACH of the monitored cell do not have a dedicated radio link in the active set. Hence, the active set size of UEs in CELL_FACH is zero.
It is assumed in this case study that within the observation period (10 minutes) three different UEs (represented by different line styles in Figure 2.25) have active dedicated radio links in this cell. UEs represented by dotted and dashed lines set up an initial dedicated radio link in the monitored cell, later go into soft handover with neighbour cells before the last radio link of the active set is deleted in the same cell in which the dedicated radio link has initially been established. The UE represented by a solid line enters and leaves the monitored cell following soft handover radio link addition and subsequent radio link deletion.
The active set size distribution percentage for each possible number of radio links in an active set is computed as the ratio of total counts of each possible active set size divided by the sum of all total counts as polled from the invisible background application multiplied by 100. If only the six samples shown as arrows in Figure 2.25 are taken into account, this will deliver the results shown in Table 2.21.
Figure 2.25 Three UEs with changing active set size are monitored in a single UTRA cell
The sum of total counts in Table 2.21 is seven. The sum of total counts in Table 2.21 is seven and calculation results are as follows:
Average percentage active set size 1ẳ(2/7)100%ẳ28.6%
Average percentage active set size 2ẳ(1/7)100%ẳ14.3%
Average percentage active set size 3ẳ(4/7)100%ẳ57.1%
The histogram for this case study is shown in Figure 2.26.
In a real network environment the number of polls and the number of total counts is of course much higher due to longer monitoring sessions and to higher number of calls per cell.
It can also be noticed that the application for measuring the active set size distribution of cells is an excellent opportunity to calculate average active set sizes of single calls as required for calculation of primary/secondary traffic.
The average active set size can be computed for each active set, which means for each particular radio connection. But the aggregation of average active set size at cell level also makes sense to analyse the soft handover situation in the cell. Table 2.22 gives an example for three selected cells of the network.
Table 2.21 Active set size distribution example 2
Active set size 1 2 3 4 5 6
Sample 1 Ignored, because zero UEs in cell
Sample 2 1 0 0 0 0 0
Sample 3 0 1 1 0 0 0
Sample 4 0 0 2 0 0 0
Sample 5 1 0 0 0 0 0
Sample 6 0 0 1 0 0 0
Total counts 2 1 4 0 0 0
Average percentage 28.6% 14.3% 57.1% — — —
Figure 2.26 Histogram for active set size distribution case study 2
The changing active set size of a single call can be correlated to many different radio quality measurements to find out how soft handover situations interfere with the quality of service. Figure 2.27 shows the correlation of the active set size and uplink BLER in a single call. The active set size varies between one and three radio links in the active set. There are some steps in the active set size graph at approximately active set sizeẳ1.6. These steps occur if radio link addition or radio link deletion is monitored in the middle of a sampling period and value 1.6 represents the average active set size within a sampling period of 2 seconds. This effect can be minimised by choosing a shorter sampling period as long as it is not compensated by the graphic plot function of the GUI.
The combined graph analysis shows that UL BLER is especially high (up to 0.375%) if the UE has two or three radio links in the active set for a longer time. Which cells have been involved in such a soft handover situation can be found out by choosing the cell aggregation level (dimension) in the analysis table shown in Figure 2.28.
Indeed the cell with the highest average active set size also has the highest uplink block error rate. The network operator now wants to know if this measurement result can be seen as an exception or do such measurement results occur generally when this cell is a member
Table 2.22 Average active set size on cell level (three selected cells)
Cell Average active set size
Cell 10001 1.8
Cell 10002 1.6
Cell 10003 2.1
Figure 2.27 UL BLER in % correlated with active set size of call
of a UE’s active set. The active set size distribution histogram for this particular cell can help to answer this question.
The histogram in Figure 2.29 shows that UEs using this cell always have either two or three radio links in the active set. Hence, all users of this cell are in soft handover.
However, the total number of polls have only reached 40: 34 times active set sizeẳ2 and 6 times active set sizeẳ3 has been polled.
Figure 2.28 Active set size and UL BLER per cell involved in a single call
Figure 2.29 Active set size distribution for cell involved in soft handover
The measurement results presented in the histogram indicate that either the observation period has been relatively short and/or that not many UEs have used this cell during the monitoring session. Without doubt a more detailed analysis of the radio environment situa- tion in this cell is necessary, which means the cell must be observed for a longer time and if possible more radio connections using this cell should be monitored. The latter objective can be realised by sending drive test equipment to the desired location, which shows by the way how the benefits of performance measurement software can be combined with drive tests and how this software can help to optimise drive test plans in a very efficient way.
Last but not least it should be mentioned that a further analysis of the overlapping situa- tion in this specific cell can also be done using cell matrices (see section 2.14).