jackrahor3
2013-02-21, 02:49 AM
Abstract— Overhead power control parameters in CDMA and W-CDMA networks can be used to shift load from congested cells to cells with available capacity. This improves traffic distribution and network efficiency. Trial results demonstrated a marked improvement in RF network performance and carrying capacity.
Keywords-CDMA Optimization, RF Network Capacity Improvment, CDMA Power Control, Optimum Operating Parameters
I. IntroductionThis paper reports on analysis and trial results associated with the development of algorithms for automatic RF operations systems software for CDMA networks. The results show that RF network performance can be significantly improved by adjusting operating parameter levels to more efficiently use existing capacity and improve the distribution of users among sectors. This work grew out of efforts to develop algorithms and operations support capabilities for efficient frequency allocation and site planning in GSM and TDMA networks [1], [2].
In CDMA and W-CDMA networks the overhead power control parameters can be used to shift load from congested cells to cells with available capacity. Trial results demonstrated a marked improvement in RF network performance and carrying capacity.
Operating an RF network requires constant attention to diagnose and correct a huge array of complex issues that confront the performance managers on a daily basis. RF networks are extremely dynamic with hardware upgrades, new cells, equipment outrages, shifting customer usage patterns, and new data services being introduced. The overall performance of an RF network that may have started out with an optimal configuration will degrade over time due to changes in the network and customer usage patterns. The trial results reported here indicate that operating parameter optimization needs to be done every few weeks to maintain capacity and efficiency benefits.
II. Background: CDMA Downlink PerformanceCapacity in a CDMA or W-CDMA network is usually limited by the downlink rather than the uplink because soft handoff reduces overall capacity for the downlink. This is why we focus our attention on power control to improve the overall downlink quality.
In the following equation Pi,l is the power received at mobile station (MS} i from sector l:
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image002.gif (1)
An outage will occur on sector l if file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image004.gif.
The ratio of bit energy to interference and noise on the downlink, as measured at user k’s MS receiver, denoted by (Eb/Io)k, determines link performance. In addition to power received from the user’s desired sector, L(k), the user also receives downlink power from L-1 other sectors. The term Pk,L(i) represents the power received by user k from the the sector serving user i, namely L(i). The spreading factor for user k on the downlink, which is the ratio of the user’s occupied bandwidth to the user’s data rate, is given by Fk. The users associated with sector l are contained in the set K(l).
Each sector only allocates a certain amount of its downlink power to a particular subscriber. As can be seen by the equation above, any power that the base station transmits degrades the performance of users in other cells. Therefore, sector l ideally only allocates enough power on the downlink to achieve target value of by (Eb/Io)k at user k’s MS receiver. The fraction of the total possible output power of the base station devoted to this user is gl,k .
Each sector also transmits a certain fraction of its power in overhead channels. These include paging channels and broadcast channels. The fraction of sector l’s potential output power devoted to these overhead channels is given by b l.
Depending on propagation conditions downlink signals for users in the same cell may or may not interfere with each other. This is modeled using the term n, which is equal to 0 if all downlink signals from the same sector are orthogonal.
Traffic hotspots can degrade the capacity of CDMA systems. In the example shown in figure 1, a hot spot has overloaded Cell 1B, and subscribers may be dropped, even though the network as a whole has adequate capacity to handle the aggregate traffic.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image006.jpg
Figure 1. CDMA cells with a traffic “hot spot”
We can adjust the radius and size of cells to accommodate hot spots as shown in figure 2. Here, Cell 1B shrinks to minimize its traffic load, while cells 3A and 3C help to handle the hot spot. Cells 1A, 1C, 3B, and 3D are also expanded, equalizing the traffic load on all sectors.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image008.jpg
Figure 2. Cell sizes adjusted to better utilize available capacity
.
III. The Optimization AlgorithmAn algorithm for downlink overhead power optimization was developed at Telcordia Technologies for CDMA IS-95, 1xRTT, and UMTS networks to compute optimal parameter settings based on usage patterns, network configuration, received signal strength maps, and existing operating parameter settings.
We adjust the overhead power factor (b
) so that each sector sheds or attracts subscribers. Reducing b has two effects. First, it reduces the cell radius so that subscribers making pilot strength measurements don’t see the sector as strongly as they did before, so they prefer other sectors, this sheds subscribers. Reducing b also frees up additional amplifier “headroom” to support the remaining served sectors. Additional operating parameters are looked at and are adjusted by the algorithm in order to mitigate the probability of introducing coverage holes or other undesirable effects when the cell sizes are changed.
This approach is attractive because it more effectively utilizes existing RF network resources with no changes in hardware. Increased efficiency is accomplished by selecting b for each sector so the overall RF network capacity is maximized. This is done, first, by shifting users from heavily loaded cells to cells that are less loaded. Secondly, the algorithm actually increases the effective carrying capacity of some sectors by lowering the average power per user.
IV. RF Network Field TrialA field trial was conducted in a US Carrier’s CDMA network serving a metropolitan area with a population of about 250,000. The trial consisted of collecting network usage and performance data for a reference period before making operating parameter changes, deploying the parameters computed by the algorithm, then observing the effects on usage patterns and network performance.
The trial demonstrated that the algorithm can shift load from heavily loaded cells to lightly loaded cells and provided measurable improvement in RF network carrying capacity. Network performance improved with improved blocking and dropped call rates. The trial also provided a basis for a full economic evaluation of the benefits of the algorithm.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image010.jpg
Figure 3. Location of sectors in trial area
The trial layout is seen in figure 3 as it was before the optimized parameters were implemented. The symbol size for each sector is drawn proportional to overhead power settings, and the colors indicate the number of carriers deployed at each sector. In this case there were two distinct settings, 19% and 22%, based on the manufacturer’s recommendations. Additionally, several cells in the central trial area were overloaded at peak hours, while cells on the periphery had spare capacity..
The algorithm was used to develop recommended overhead power settings for each sector to optimize RF network capacity while maintaining or improving blocked and dropped call rates. The resulting customized power settings are shown in figure 4. Note that the previously lightly loaded peripheral cells have increased their power levels, thereby attracting additional users from the higher usage density core area. The cells in the core area that previously were heavily loaded now have reduced power levels effectively shedding users, especially users that are further from the cells’ centers.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image012.jpg
Figure 4. Customized overhead power settings at each sector.
These changes in power parameters did shift traffic in a measurable way from the core cells to the peripheral cells, thereby utilizing the available capacity m ore efficiently.
The effect of the optimized parameter changes on the carrying capacity on the core cells was even more dramatic. By shifting more distant, higher power users to the peripheral cells the average power per user was lowered in the core cells increasing the effective carrying capacity of these cells. This capacity improvement is seen in figures 5 and 6.
The trial RF network employed a sequential traffic allocation algorithm to flow traffic onto the carriers at each sector. Sequential traffic allocation means that traffic first is assigned to FA-1 until its carrying capacity is reached, and then newly arriving calls are assigned to FA-2, and so on. A reserve is maintained on each FA to support soft hand-off; this reduces the number of hard hand-offs that are more prone to failures. The MCTA algorithm computes the soft hand-off reserve by dividing the average power per call into the specified reserve level.
The carrying capacity of the carriers at a sector can be estimated by observing the traffic load at which traffic overflows to FA-2 after saturating FA-1. In Figure 5 the carrying capacity of the carriers in this three-FA sector is seen to be about 17 Erlangs. This was the case before the optimized parameters were applied.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image014.jpg
Figure 5. Carrying capacity of a 3-FA sector before optimization.
The carrying capacity in the same sector after optimization is increased to about 26 Erlangs per carrier as seen in Figure 6.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image016.jpg
Figure 6. Optimization increases sector carrying capacity
These improvements in capacity and traffic balancing were realized immediately upon installing the optimized parameters.
This capacity estimation method was applied to all 2 and 3 carrier sectors in the trial area. The average improvement in carrying capacity was 30% for sectors with 3 FAs and 8% for sectors with 2 FAs. The load on single FA sectors was increased by 19%.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image018.jpg
Figure 7. The optimization effect wears off in time.
The capacity improvement from optimization wears off over time due to the time-varying traffic patterns as seen in figure 7. This result strongly suggests that network operating parameters must be adjusted every few weeks to maintain optimal capacity and performance levels. An in-line, autonomous operations support system that continually monitors network capacity and performance and regularly re-computes the optimized parameters will maintain the RF network at or near optimal performance, efficiency and capacity.
V. SummaryThe field trial showed the ability of the optimization algorithm to balance the network load by shifting call traffic from overloaded sectors to sectors with capacity. This reduces the call load in overloaded sectors and actually increases these sectors’ carrying capacity by lowering the average power per call. The effects of parameter optimization fade over time as user patterns change and the network evolves. The field trial found that fresh operating parameter optimization was needed every 3 or 4 weeks.
These findings underscore the need for an automatic RF performance management system that makes it feasible to change the operating parameters frequently enough to maintain the RF network at near optimal levels. An in-line, automatic RF operations support system can continually assesses areas where network performance can be improved, develop revised parameter sets, reconcile possible changes with trouble tickets, and notify human operators. The revised parameter sets can be autonomously implemented through interfaces to the requisite RF network elements, verification logged, reports issued, and resulting performance changes tracked.
AcknowledgmentThe author acknowledges the Telcordia™ Auto RF Team who conceptualized, designed and developed the Auto RF product that provided the motivation and capability for the field trial and subsequent analysis.
References
[1] C.C. Yu, D. Morton, C. Stumpf, R.G. White, J.E. Wilkes, M Ulema, “Low-tier wireless local loop radio systems. I. Introduction,” IEEE Communications Magazine, Mar 1997; pp 84-92, Volume: 35, Issue: 3
[2] C.C. Yu, D. Morton, C. Stumpf, R.G. White, J.E. Wilkes, M Ulema, “Low-tier wireless local loop radio systems .2. Comparison of systems”, IEEE Communications Magazine, Mar 1997; pp 94-98; Volume: 35, Issue: 3
Abstract— Overhead power control parameters in CDMA and W-CDMA networks can be used to shift load from congested cells to cells with available capacity. This improves traffic distribution and network efficiency. Trial results demonstrated a marked improvement in RF network performance and carrying capacity.
Keywords-CDMA Optimization, RF Network Capacity Improvment, CDMA Power Control, Optimum Operating Parameters
I. IntroductionThis paper reports on analysis and trial results associated with the development of algorithms for automatic RF operations systems software for CDMA networks. The results show that RF network performance can be significantly improved by adjusting operating parameter levels to more efficiently use existing capacity and improve the distribution of users among sectors. This work grew out of efforts to develop algorithms and operations support capabilities for efficient frequency allocation and site planning in GSM and TDMA networks [1], [2].
In CDMA and W-CDMA networks the overhead power control parameters can be used to shift load from congested cells to cells with available capacity. Trial results demonstrated a marked improvement in RF network performance and carrying capacity.
Operating an RF network requires constant attention to diagnose and correct a huge array of complex issues that confront the performance managers on a daily basis. RF networks are extremely dynamic with hardware upgrades, new cells, equipment outrages, shifting customer usage patterns, and new data services being introduced. The overall performance of an RF network that may have started out with an optimal configuration will degrade over time due to changes in the network and customer usage patterns. The trial results reported here indicate that operating parameter optimization needs to be done every few weeks to maintain capacity and efficiency benefits.
II. Background: CDMA Downlink PerformanceCapacity in a CDMA or W-CDMA network is usually limited by the downlink rather than the uplink because soft handoff reduces overall capacity for the downlink. This is why we focus our attention on power control to improve the overall downlink quality.
In the following equation Pi,l is the power received at mobile station (MS} i from sector l:
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image002.gif (1)
An outage will occur on sector l if file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image004.gif.
The ratio of bit energy to interference and noise on the downlink, as measured at user k’s MS receiver, denoted by (Eb/Io)k, determines link performance. In addition to power received from the user’s desired sector, L(k), the user also receives downlink power from L-1 other sectors. The term Pk,L(i) represents the power received by user k from the the sector serving user i, namely L(i). The spreading factor for user k on the downlink, which is the ratio of the user’s occupied bandwidth to the user’s data rate, is given by Fk. The users associated with sector l are contained in the set K(l).
Each sector only allocates a certain amount of its downlink power to a particular subscriber. As can be seen by the equation above, any power that the base station transmits degrades the performance of users in other cells. Therefore, sector l ideally only allocates enough power on the downlink to achieve target value of by (Eb/Io)k at user k’s MS receiver. The fraction of the total possible output power of the base station devoted to this user is gl,k .
Each sector also transmits a certain fraction of its power in overhead channels. These include paging channels and broadcast channels. The fraction of sector l’s potential output power devoted to these overhead channels is given by b l.
Depending on propagation conditions downlink signals for users in the same cell may or may not interfere with each other. This is modeled using the term n, which is equal to 0 if all downlink signals from the same sector are orthogonal.
Traffic hotspots can degrade the capacity of CDMA systems. In the example shown in figure 1, a hot spot has overloaded Cell 1B, and subscribers may be dropped, even though the network as a whole has adequate capacity to handle the aggregate traffic.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image006.jpg
Figure 1. CDMA cells with a traffic “hot spot”
We can adjust the radius and size of cells to accommodate hot spots as shown in figure 2. Here, Cell 1B shrinks to minimize its traffic load, while cells 3A and 3C help to handle the hot spot. Cells 1A, 1C, 3B, and 3D are also expanded, equalizing the traffic load on all sectors.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image008.jpg
Figure 2. Cell sizes adjusted to better utilize available capacity
.
III. The Optimization AlgorithmAn algorithm for downlink overhead power optimization was developed at Telcordia Technologies for CDMA IS-95, 1xRTT, and UMTS networks to compute optimal parameter settings based on usage patterns, network configuration, received signal strength maps, and existing operating parameter settings.
We adjust the overhead power factor (b
) so that each sector sheds or attracts subscribers. Reducing b has two effects. First, it reduces the cell radius so that subscribers making pilot strength measurements don’t see the sector as strongly as they did before, so they prefer other sectors, this sheds subscribers. Reducing b also frees up additional amplifier “headroom” to support the remaining served sectors. Additional operating parameters are looked at and are adjusted by the algorithm in order to mitigate the probability of introducing coverage holes or other undesirable effects when the cell sizes are changed.
This approach is attractive because it more effectively utilizes existing RF network resources with no changes in hardware. Increased efficiency is accomplished by selecting b for each sector so the overall RF network capacity is maximized. This is done, first, by shifting users from heavily loaded cells to cells that are less loaded. Secondly, the algorithm actually increases the effective carrying capacity of some sectors by lowering the average power per user.
IV. RF Network Field TrialA field trial was conducted in a US Carrier’s CDMA network serving a metropolitan area with a population of about 250,000. The trial consisted of collecting network usage and performance data for a reference period before making operating parameter changes, deploying the parameters computed by the algorithm, then observing the effects on usage patterns and network performance.
The trial demonstrated that the algorithm can shift load from heavily loaded cells to lightly loaded cells and provided measurable improvement in RF network carrying capacity. Network performance improved with improved blocking and dropped call rates. The trial also provided a basis for a full economic evaluation of the benefits of the algorithm.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image010.jpg
Figure 3. Location of sectors in trial area
The trial layout is seen in figure 3 as it was before the optimized parameters were implemented. The symbol size for each sector is drawn proportional to overhead power settings, and the colors indicate the number of carriers deployed at each sector. In this case there were two distinct settings, 19% and 22%, based on the manufacturer’s recommendations. Additionally, several cells in the central trial area were overloaded at peak hours, while cells on the periphery had spare capacity..
The algorithm was used to develop recommended overhead power settings for each sector to optimize RF network capacity while maintaining or improving blocked and dropped call rates. The resulting customized power settings are shown in figure 4. Note that the previously lightly loaded peripheral cells have increased their power levels, thereby attracting additional users from the higher usage density core area. The cells in the core area that previously were heavily loaded now have reduced power levels effectively shedding users, especially users that are further from the cells’ centers.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image012.jpg
Figure 4. Customized overhead power settings at each sector.
These changes in power parameters did shift traffic in a measurable way from the core cells to the peripheral cells, thereby utilizing the available capacity m ore efficiently.
The effect of the optimized parameter changes on the carrying capacity on the core cells was even more dramatic. By shifting more distant, higher power users to the peripheral cells the average power per user was lowered in the core cells increasing the effective carrying capacity of these cells. This capacity improvement is seen in figures 5 and 6.
The trial RF network employed a sequential traffic allocation algorithm to flow traffic onto the carriers at each sector. Sequential traffic allocation means that traffic first is assigned to FA-1 until its carrying capacity is reached, and then newly arriving calls are assigned to FA-2, and so on. A reserve is maintained on each FA to support soft hand-off; this reduces the number of hard hand-offs that are more prone to failures. The MCTA algorithm computes the soft hand-off reserve by dividing the average power per call into the specified reserve level.
The carrying capacity of the carriers at a sector can be estimated by observing the traffic load at which traffic overflows to FA-2 after saturating FA-1. In Figure 5 the carrying capacity of the carriers in this three-FA sector is seen to be about 17 Erlangs. This was the case before the optimized parameters were applied.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image014.jpg
Figure 5. Carrying capacity of a 3-FA sector before optimization.
The carrying capacity in the same sector after optimization is increased to about 26 Erlangs per carrier as seen in Figure 6.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image016.jpg
Figure 6. Optimization increases sector carrying capacity
These improvements in capacity and traffic balancing were realized immediately upon installing the optimized parameters.
This capacity estimation method was applied to all 2 and 3 carrier sectors in the trial area. The average improvement in carrying capacity was 30% for sectors with 3 FAs and 8% for sectors with 2 FAs. The load on single FA sectors was increased by 19%.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image018.jpg
Figure 7. The optimization effect wears off in time.
The capacity improvement from optimization wears off over time due to the time-varying traffic patterns as seen in figure 7. This result strongly suggests that network operating parameters must be adjusted every few weeks to maintain optimal capacity and performance levels. An in-line, autonomous operations support system that continually monitors network capacity and performance and regularly re-computes the optimized parameters will maintain the RF network at or near optimal performance, efficiency and capacity.
V. SummaryThe field trial showed the ability of the optimization algorithm to balance the network load by shifting call traffic from overloaded sectors to sectors with capacity. This reduces the call load in overloaded sectors and actually increases these sectors’ carrying capacity by lowering the average power per call. The effects of parameter optimization fade over time as user patterns change and the network evolves. The field trial found that fresh operating parameter optimization was needed every 3 or 4 weeks.
These findings underscore the need for an automatic RF performance management system that makes it feasible to change the operating parameters frequently enough to maintain the RF network at near optimal levels. An in-line, automatic RF operations support system can continually assesses areas where network performance can be improved, develop revised parameter sets, reconcile possible changes with trouble tickets, and notify human operators. The revised parameter sets can be autonomously implemented through interfaces to the requisite RF network elements, verification logged, reports issued, and resulting performance changes tracked.
AcknowledgmentThe author acknowledges the Telcordia™ Auto RF Team who conceptualized, designed and developed the Auto RF product that provided the motivation and capability for the field trial and subsequent analysis.
References
[1] C.C. Yu, D. Morton, C. Stumpf, R.G. White, J.E. Wilkes, M Ulema, “Low-tier wireless local loop radio systems. I. Introduction,” IEEE Communications Magazine, Mar 1997; pp 84-92, Volume: 35, Issue: 3
[2] C.C. Yu, D. Morton, C. Stumpf, R.G. White, J.E. Wilkes, M Ulema, “Low-tier wireless local loop radio systems .2. Comparison of systems”, IEEE Communications Magazine, Mar 1997; pp 94-98; Volume: 35, Issue: 3
Keywords-CDMA Optimization, RF Network Capacity Improvment, CDMA Power Control, Optimum Operating Parameters
I. IntroductionThis paper reports on analysis and trial results associated with the development of algorithms for automatic RF operations systems software for CDMA networks. The results show that RF network performance can be significantly improved by adjusting operating parameter levels to more efficiently use existing capacity and improve the distribution of users among sectors. This work grew out of efforts to develop algorithms and operations support capabilities for efficient frequency allocation and site planning in GSM and TDMA networks [1], [2].
In CDMA and W-CDMA networks the overhead power control parameters can be used to shift load from congested cells to cells with available capacity. Trial results demonstrated a marked improvement in RF network performance and carrying capacity.
Operating an RF network requires constant attention to diagnose and correct a huge array of complex issues that confront the performance managers on a daily basis. RF networks are extremely dynamic with hardware upgrades, new cells, equipment outrages, shifting customer usage patterns, and new data services being introduced. The overall performance of an RF network that may have started out with an optimal configuration will degrade over time due to changes in the network and customer usage patterns. The trial results reported here indicate that operating parameter optimization needs to be done every few weeks to maintain capacity and efficiency benefits.
II. Background: CDMA Downlink PerformanceCapacity in a CDMA or W-CDMA network is usually limited by the downlink rather than the uplink because soft handoff reduces overall capacity for the downlink. This is why we focus our attention on power control to improve the overall downlink quality.
In the following equation Pi,l is the power received at mobile station (MS} i from sector l:
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image002.gif (1)
An outage will occur on sector l if file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image004.gif.
The ratio of bit energy to interference and noise on the downlink, as measured at user k’s MS receiver, denoted by (Eb/Io)k, determines link performance. In addition to power received from the user’s desired sector, L(k), the user also receives downlink power from L-1 other sectors. The term Pk,L(i) represents the power received by user k from the the sector serving user i, namely L(i). The spreading factor for user k on the downlink, which is the ratio of the user’s occupied bandwidth to the user’s data rate, is given by Fk. The users associated with sector l are contained in the set K(l).
Each sector only allocates a certain amount of its downlink power to a particular subscriber. As can be seen by the equation above, any power that the base station transmits degrades the performance of users in other cells. Therefore, sector l ideally only allocates enough power on the downlink to achieve target value of by (Eb/Io)k at user k’s MS receiver. The fraction of the total possible output power of the base station devoted to this user is gl,k .
Each sector also transmits a certain fraction of its power in overhead channels. These include paging channels and broadcast channels. The fraction of sector l’s potential output power devoted to these overhead channels is given by b l.
Depending on propagation conditions downlink signals for users in the same cell may or may not interfere with each other. This is modeled using the term n, which is equal to 0 if all downlink signals from the same sector are orthogonal.
Traffic hotspots can degrade the capacity of CDMA systems. In the example shown in figure 1, a hot spot has overloaded Cell 1B, and subscribers may be dropped, even though the network as a whole has adequate capacity to handle the aggregate traffic.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image006.jpg
Figure 1. CDMA cells with a traffic “hot spot”
We can adjust the radius and size of cells to accommodate hot spots as shown in figure 2. Here, Cell 1B shrinks to minimize its traffic load, while cells 3A and 3C help to handle the hot spot. Cells 1A, 1C, 3B, and 3D are also expanded, equalizing the traffic load on all sectors.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image008.jpg
Figure 2. Cell sizes adjusted to better utilize available capacity
.
III. The Optimization AlgorithmAn algorithm for downlink overhead power optimization was developed at Telcordia Technologies for CDMA IS-95, 1xRTT, and UMTS networks to compute optimal parameter settings based on usage patterns, network configuration, received signal strength maps, and existing operating parameter settings.
We adjust the overhead power factor (b
) so that each sector sheds or attracts subscribers. Reducing b has two effects. First, it reduces the cell radius so that subscribers making pilot strength measurements don’t see the sector as strongly as they did before, so they prefer other sectors, this sheds subscribers. Reducing b also frees up additional amplifier “headroom” to support the remaining served sectors. Additional operating parameters are looked at and are adjusted by the algorithm in order to mitigate the probability of introducing coverage holes or other undesirable effects when the cell sizes are changed.
This approach is attractive because it more effectively utilizes existing RF network resources with no changes in hardware. Increased efficiency is accomplished by selecting b for each sector so the overall RF network capacity is maximized. This is done, first, by shifting users from heavily loaded cells to cells that are less loaded. Secondly, the algorithm actually increases the effective carrying capacity of some sectors by lowering the average power per user.
IV. RF Network Field TrialA field trial was conducted in a US Carrier’s CDMA network serving a metropolitan area with a population of about 250,000. The trial consisted of collecting network usage and performance data for a reference period before making operating parameter changes, deploying the parameters computed by the algorithm, then observing the effects on usage patterns and network performance.
The trial demonstrated that the algorithm can shift load from heavily loaded cells to lightly loaded cells and provided measurable improvement in RF network carrying capacity. Network performance improved with improved blocking and dropped call rates. The trial also provided a basis for a full economic evaluation of the benefits of the algorithm.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image010.jpg
Figure 3. Location of sectors in trial area
The trial layout is seen in figure 3 as it was before the optimized parameters were implemented. The symbol size for each sector is drawn proportional to overhead power settings, and the colors indicate the number of carriers deployed at each sector. In this case there were two distinct settings, 19% and 22%, based on the manufacturer’s recommendations. Additionally, several cells in the central trial area were overloaded at peak hours, while cells on the periphery had spare capacity..
The algorithm was used to develop recommended overhead power settings for each sector to optimize RF network capacity while maintaining or improving blocked and dropped call rates. The resulting customized power settings are shown in figure 4. Note that the previously lightly loaded peripheral cells have increased their power levels, thereby attracting additional users from the higher usage density core area. The cells in the core area that previously were heavily loaded now have reduced power levels effectively shedding users, especially users that are further from the cells’ centers.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image012.jpg
Figure 4. Customized overhead power settings at each sector.
These changes in power parameters did shift traffic in a measurable way from the core cells to the peripheral cells, thereby utilizing the available capacity m ore efficiently.
The effect of the optimized parameter changes on the carrying capacity on the core cells was even more dramatic. By shifting more distant, higher power users to the peripheral cells the average power per user was lowered in the core cells increasing the effective carrying capacity of these cells. This capacity improvement is seen in figures 5 and 6.
The trial RF network employed a sequential traffic allocation algorithm to flow traffic onto the carriers at each sector. Sequential traffic allocation means that traffic first is assigned to FA-1 until its carrying capacity is reached, and then newly arriving calls are assigned to FA-2, and so on. A reserve is maintained on each FA to support soft hand-off; this reduces the number of hard hand-offs that are more prone to failures. The MCTA algorithm computes the soft hand-off reserve by dividing the average power per call into the specified reserve level.
The carrying capacity of the carriers at a sector can be estimated by observing the traffic load at which traffic overflows to FA-2 after saturating FA-1. In Figure 5 the carrying capacity of the carriers in this three-FA sector is seen to be about 17 Erlangs. This was the case before the optimized parameters were applied.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image014.jpg
Figure 5. Carrying capacity of a 3-FA sector before optimization.
The carrying capacity in the same sector after optimization is increased to about 26 Erlangs per carrier as seen in Figure 6.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image016.jpg
Figure 6. Optimization increases sector carrying capacity
These improvements in capacity and traffic balancing were realized immediately upon installing the optimized parameters.
This capacity estimation method was applied to all 2 and 3 carrier sectors in the trial area. The average improvement in carrying capacity was 30% for sectors with 3 FAs and 8% for sectors with 2 FAs. The load on single FA sectors was increased by 19%.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image018.jpg
Figure 7. The optimization effect wears off in time.
The capacity improvement from optimization wears off over time due to the time-varying traffic patterns as seen in figure 7. This result strongly suggests that network operating parameters must be adjusted every few weeks to maintain optimal capacity and performance levels. An in-line, autonomous operations support system that continually monitors network capacity and performance and regularly re-computes the optimized parameters will maintain the RF network at or near optimal performance, efficiency and capacity.
V. SummaryThe field trial showed the ability of the optimization algorithm to balance the network load by shifting call traffic from overloaded sectors to sectors with capacity. This reduces the call load in overloaded sectors and actually increases these sectors’ carrying capacity by lowering the average power per call. The effects of parameter optimization fade over time as user patterns change and the network evolves. The field trial found that fresh operating parameter optimization was needed every 3 or 4 weeks.
These findings underscore the need for an automatic RF performance management system that makes it feasible to change the operating parameters frequently enough to maintain the RF network at near optimal levels. An in-line, automatic RF operations support system can continually assesses areas where network performance can be improved, develop revised parameter sets, reconcile possible changes with trouble tickets, and notify human operators. The revised parameter sets can be autonomously implemented through interfaces to the requisite RF network elements, verification logged, reports issued, and resulting performance changes tracked.
AcknowledgmentThe author acknowledges the Telcordia™ Auto RF Team who conceptualized, designed and developed the Auto RF product that provided the motivation and capability for the field trial and subsequent analysis.
References
[1] C.C. Yu, D. Morton, C. Stumpf, R.G. White, J.E. Wilkes, M Ulema, “Low-tier wireless local loop radio systems. I. Introduction,” IEEE Communications Magazine, Mar 1997; pp 84-92, Volume: 35, Issue: 3
[2] C.C. Yu, D. Morton, C. Stumpf, R.G. White, J.E. Wilkes, M Ulema, “Low-tier wireless local loop radio systems .2. Comparison of systems”, IEEE Communications Magazine, Mar 1997; pp 94-98; Volume: 35, Issue: 3
Abstract— Overhead power control parameters in CDMA and W-CDMA networks can be used to shift load from congested cells to cells with available capacity. This improves traffic distribution and network efficiency. Trial results demonstrated a marked improvement in RF network performance and carrying capacity.
Keywords-CDMA Optimization, RF Network Capacity Improvment, CDMA Power Control, Optimum Operating Parameters
I. IntroductionThis paper reports on analysis and trial results associated with the development of algorithms for automatic RF operations systems software for CDMA networks. The results show that RF network performance can be significantly improved by adjusting operating parameter levels to more efficiently use existing capacity and improve the distribution of users among sectors. This work grew out of efforts to develop algorithms and operations support capabilities for efficient frequency allocation and site planning in GSM and TDMA networks [1], [2].
In CDMA and W-CDMA networks the overhead power control parameters can be used to shift load from congested cells to cells with available capacity. Trial results demonstrated a marked improvement in RF network performance and carrying capacity.
Operating an RF network requires constant attention to diagnose and correct a huge array of complex issues that confront the performance managers on a daily basis. RF networks are extremely dynamic with hardware upgrades, new cells, equipment outrages, shifting customer usage patterns, and new data services being introduced. The overall performance of an RF network that may have started out with an optimal configuration will degrade over time due to changes in the network and customer usage patterns. The trial results reported here indicate that operating parameter optimization needs to be done every few weeks to maintain capacity and efficiency benefits.
II. Background: CDMA Downlink PerformanceCapacity in a CDMA or W-CDMA network is usually limited by the downlink rather than the uplink because soft handoff reduces overall capacity for the downlink. This is why we focus our attention on power control to improve the overall downlink quality.
In the following equation Pi,l is the power received at mobile station (MS} i from sector l:
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image002.gif (1)
An outage will occur on sector l if file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image004.gif.
The ratio of bit energy to interference and noise on the downlink, as measured at user k’s MS receiver, denoted by (Eb/Io)k, determines link performance. In addition to power received from the user’s desired sector, L(k), the user also receives downlink power from L-1 other sectors. The term Pk,L(i) represents the power received by user k from the the sector serving user i, namely L(i). The spreading factor for user k on the downlink, which is the ratio of the user’s occupied bandwidth to the user’s data rate, is given by Fk. The users associated with sector l are contained in the set K(l).
Each sector only allocates a certain amount of its downlink power to a particular subscriber. As can be seen by the equation above, any power that the base station transmits degrades the performance of users in other cells. Therefore, sector l ideally only allocates enough power on the downlink to achieve target value of by (Eb/Io)k at user k’s MS receiver. The fraction of the total possible output power of the base station devoted to this user is gl,k .
Each sector also transmits a certain fraction of its power in overhead channels. These include paging channels and broadcast channels. The fraction of sector l’s potential output power devoted to these overhead channels is given by b l.
Depending on propagation conditions downlink signals for users in the same cell may or may not interfere with each other. This is modeled using the term n, which is equal to 0 if all downlink signals from the same sector are orthogonal.
Traffic hotspots can degrade the capacity of CDMA systems. In the example shown in figure 1, a hot spot has overloaded Cell 1B, and subscribers may be dropped, even though the network as a whole has adequate capacity to handle the aggregate traffic.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image006.jpg
Figure 1. CDMA cells with a traffic “hot spot”
We can adjust the radius and size of cells to accommodate hot spots as shown in figure 2. Here, Cell 1B shrinks to minimize its traffic load, while cells 3A and 3C help to handle the hot spot. Cells 1A, 1C, 3B, and 3D are also expanded, equalizing the traffic load on all sectors.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image008.jpg
Figure 2. Cell sizes adjusted to better utilize available capacity
.
III. The Optimization AlgorithmAn algorithm for downlink overhead power optimization was developed at Telcordia Technologies for CDMA IS-95, 1xRTT, and UMTS networks to compute optimal parameter settings based on usage patterns, network configuration, received signal strength maps, and existing operating parameter settings.
We adjust the overhead power factor (b
) so that each sector sheds or attracts subscribers. Reducing b has two effects. First, it reduces the cell radius so that subscribers making pilot strength measurements don’t see the sector as strongly as they did before, so they prefer other sectors, this sheds subscribers. Reducing b also frees up additional amplifier “headroom” to support the remaining served sectors. Additional operating parameters are looked at and are adjusted by the algorithm in order to mitigate the probability of introducing coverage holes or other undesirable effects when the cell sizes are changed.
This approach is attractive because it more effectively utilizes existing RF network resources with no changes in hardware. Increased efficiency is accomplished by selecting b for each sector so the overall RF network capacity is maximized. This is done, first, by shifting users from heavily loaded cells to cells that are less loaded. Secondly, the algorithm actually increases the effective carrying capacity of some sectors by lowering the average power per user.
IV. RF Network Field TrialA field trial was conducted in a US Carrier’s CDMA network serving a metropolitan area with a population of about 250,000. The trial consisted of collecting network usage and performance data for a reference period before making operating parameter changes, deploying the parameters computed by the algorithm, then observing the effects on usage patterns and network performance.
The trial demonstrated that the algorithm can shift load from heavily loaded cells to lightly loaded cells and provided measurable improvement in RF network carrying capacity. Network performance improved with improved blocking and dropped call rates. The trial also provided a basis for a full economic evaluation of the benefits of the algorithm.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image010.jpg
Figure 3. Location of sectors in trial area
The trial layout is seen in figure 3 as it was before the optimized parameters were implemented. The symbol size for each sector is drawn proportional to overhead power settings, and the colors indicate the number of carriers deployed at each sector. In this case there were two distinct settings, 19% and 22%, based on the manufacturer’s recommendations. Additionally, several cells in the central trial area were overloaded at peak hours, while cells on the periphery had spare capacity..
The algorithm was used to develop recommended overhead power settings for each sector to optimize RF network capacity while maintaining or improving blocked and dropped call rates. The resulting customized power settings are shown in figure 4. Note that the previously lightly loaded peripheral cells have increased their power levels, thereby attracting additional users from the higher usage density core area. The cells in the core area that previously were heavily loaded now have reduced power levels effectively shedding users, especially users that are further from the cells’ centers.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image012.jpg
Figure 4. Customized overhead power settings at each sector.
These changes in power parameters did shift traffic in a measurable way from the core cells to the peripheral cells, thereby utilizing the available capacity m ore efficiently.
The effect of the optimized parameter changes on the carrying capacity on the core cells was even more dramatic. By shifting more distant, higher power users to the peripheral cells the average power per user was lowered in the core cells increasing the effective carrying capacity of these cells. This capacity improvement is seen in figures 5 and 6.
The trial RF network employed a sequential traffic allocation algorithm to flow traffic onto the carriers at each sector. Sequential traffic allocation means that traffic first is assigned to FA-1 until its carrying capacity is reached, and then newly arriving calls are assigned to FA-2, and so on. A reserve is maintained on each FA to support soft hand-off; this reduces the number of hard hand-offs that are more prone to failures. The MCTA algorithm computes the soft hand-off reserve by dividing the average power per call into the specified reserve level.
The carrying capacity of the carriers at a sector can be estimated by observing the traffic load at which traffic overflows to FA-2 after saturating FA-1. In Figure 5 the carrying capacity of the carriers in this three-FA sector is seen to be about 17 Erlangs. This was the case before the optimized parameters were applied.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image014.jpg
Figure 5. Carrying capacity of a 3-FA sector before optimization.
The carrying capacity in the same sector after optimization is increased to about 26 Erlangs per carrier as seen in Figure 6.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image016.jpg
Figure 6. Optimization increases sector carrying capacity
These improvements in capacity and traffic balancing were realized immediately upon installing the optimized parameters.
This capacity estimation method was applied to all 2 and 3 carrier sectors in the trial area. The average improvement in carrying capacity was 30% for sectors with 3 FAs and 8% for sectors with 2 FAs. The load on single FA sectors was increased by 19%.
file:///C:/Users/mwx70322/AppData/Local/Temp/msohtmlclip1/01/clip_image018.jpg
Figure 7. The optimization effect wears off in time.
The capacity improvement from optimization wears off over time due to the time-varying traffic patterns as seen in figure 7. This result strongly suggests that network operating parameters must be adjusted every few weeks to maintain optimal capacity and performance levels. An in-line, autonomous operations support system that continually monitors network capacity and performance and regularly re-computes the optimized parameters will maintain the RF network at or near optimal performance, efficiency and capacity.
V. SummaryThe field trial showed the ability of the optimization algorithm to balance the network load by shifting call traffic from overloaded sectors to sectors with capacity. This reduces the call load in overloaded sectors and actually increases these sectors’ carrying capacity by lowering the average power per call. The effects of parameter optimization fade over time as user patterns change and the network evolves. The field trial found that fresh operating parameter optimization was needed every 3 or 4 weeks.
These findings underscore the need for an automatic RF performance management system that makes it feasible to change the operating parameters frequently enough to maintain the RF network at near optimal levels. An in-line, automatic RF operations support system can continually assesses areas where network performance can be improved, develop revised parameter sets, reconcile possible changes with trouble tickets, and notify human operators. The revised parameter sets can be autonomously implemented through interfaces to the requisite RF network elements, verification logged, reports issued, and resulting performance changes tracked.
AcknowledgmentThe author acknowledges the Telcordia™ Auto RF Team who conceptualized, designed and developed the Auto RF product that provided the motivation and capability for the field trial and subsequent analysis.
References
[1] C.C. Yu, D. Morton, C. Stumpf, R.G. White, J.E. Wilkes, M Ulema, “Low-tier wireless local loop radio systems. I. Introduction,” IEEE Communications Magazine, Mar 1997; pp 84-92, Volume: 35, Issue: 3
[2] C.C. Yu, D. Morton, C. Stumpf, R.G. White, J.E. Wilkes, M Ulema, “Low-tier wireless local loop radio systems .2. Comparison of systems”, IEEE Communications Magazine, Mar 1997; pp 94-98; Volume: 35, Issue: 3