Wideband interference estimation with advanced receivers
The research on load estimation presented here was performed at Ericsson AB, during 2005-2011. All material on this page can be found in the publicly avilable publications 1-8.
In systems like the cellular WCDMA system, the uplink load of a cell is expressed in terms of the measured wideband interference divided by the thermal noise power floor. This quantity is denoted the rise over thermal (RoT). The RoT is of central importance when scheduling users in the enhaced uplink (EUL) of the WCDMA system. The reason is that the cell stability and coverage of the cell is closely related to the RoT. The scheduler must therefore schedule as much traffic as possible, while at the same time keeping the RoT below a threshold that secures stability and planned coverage. Please refer to 6-8 and the references therein for a more detailed description of WCDMA stability and coverage.
In order to measure the RoT when using the conventional RAKE receiver, the thermal noise power floor and the received total wideband power need to be estimated. Although estimation of the momentary received total wideband power is straightforward, there are several problems associated with this seemingly easy problem
- The wideband power and the nosie floor are defined at the antenna connector, however they are best measured in the digital receiver.
- Scale factor errors in feeder cables and front end electronics introduce an uncertainty of the measuered quantities of 1-3 dB, i.e a substantial amount of a typical headroom of 10 dB.
- Neighbor cell interference makes it difficult to measure the thermal noise power floor .
Now, since RoT is the quotient between the received total wideband power and the thermal noise power floor, the scale factor errors cancel when the RoT is formed in the digital receiver (8). However it still remains to measure the thermal noise power floor. This is a non-trivial task since the neighbor cell interference adds to the thermal noise power floor. The paper 7 proves formally that the thermal noise power floor is mathematically unobservable in case of a neighbor cell interference power with non-zero mean, while 2 and 5 provide efficient computational algorithms. The final step in solving this problem is presented in 1, where a new algorithm for interference power splitting is presented. The scheme efficiently separates the own cell interference, the neighbor cell interference and the thermal noise power floor with an inaccuracy of only about 10-15%. The algorithm is based on an extended Kalman filter of order 4, resulting in a very low complexity. This scheme is believed to open up for SON based interference management in HetNets.
The papers 5-8 therefore address the problem of thermal noise power floor estimation, by application of non-linear techniques for minimum estimation. For reasons that become clear below, a Bayesian probabilistic (soft) approach is used.
The paper 8 describes the method applied in the field today, see figure 1. The received total wideband power is first processed to produce Gaussian power pdf:s. A Kalman filter is used for this purpose. The power pdf:s are then processed by the Bayesian minimum estimation algorithm of 7. This scheme estimates the minimum of the power as a conditional mean, conditioned on the power pdf:s in a sliding window, thereby providing an approximation of the thermal noise power floor. The RoT follows by a division of the received total wideband power with the estimated minimum power. An example of the operation of the algorithm appears in figure 2.
Figure 1: Block diagram of the Bayesian RoT estimation algorithm.
Figure 2: Noise floor estimation. The measured received total wideband power (blue) and the estimated noise power floor (red) for two antenna branches.
One of the advantages with the use of a Bayesian minimum estimation algorithm is that it allows a recursive formulation, see 6. There is hence no need to store power samples for processing, as would be the case if a non-probabilistic (hard) approach would be used. Furthermore, the Bayesian approach allows the uncertainty of the estimated RoT to be calculated as the conditional variance, see 6 and 7 for details.
When it comes to advanced receivers, like interference suppressing ones, the situation becomes more intricate. First, each user sees different interference powers, i.e. received wideband powers, after the IS gains have materialized in the receiver. Noting that the power control loops are closed at the SIR target reference, after IS processing, it is clear that it is the interference level after IS gains that matters for the stability of the uplink. Further, the coverage with IS receivers is improved since the interference level at the antenna can be increased without impairing IS receiver performance, below the performance of the RAKE receiver. The consequences of the above observed in 4 include
- The use of a single uplink RoT measure for each cell breaks down. RoT becomes user individual with IS.
- The RoT needs to be measured also after the IS receiver, to capture IS processing gains.
The paper 4 adresses these consequences by a derivation of new noise rise measures for stability and coverage, valid after IS processing for each user. These new noise rise measures are easy to compute at slot rate since they only require inner product evaluations. The computation require estimates of the symbol power, channel estimate, combining weights and the SINR. Techniques to combine the user individual noise rise measures into quantities valid for the whole cell are also described, see figure 3. Link simulation results (figure 4) show that the new algorithms perform as intended.
Figure 3: Block diagram of the load processing with a G-rake+ IS receiver.
Figure 4: Time evolution of the RoT at the antenna receive port (green), together with the estimated RoT after G-rake processing (blue), for user 1 of the four users. The IS gain also appears in the plot (red).
Recently, rise over thermal estimation has also been considered for interference cancellation receivers. In that case it is the signal cleaned from interference in a buffer that is of interest to assess the load. The paper 3 discusses the details of this problem.
1. T. Wigren, "Wireless interference estimation for inter-cell interference coordination", IET Communications, vol. 9, no. 12, pp. 1539 - 1546, 2015. DOI: 10.1049/iet-com.2014.0831.
2. O. Rosen, A. Medvedev and T. Wigren, "Parallelization of the Kalman filter for banded systems on multicore computational platforms", Contr. Eng. Practice, vol. 21, pp. 1188-1194, 2013.
3. G. Xinyu, Z. Zhang, S. Grant, T. Wigren, N. Johansson, A. Kangas, "Load control for multi-stage interference cancellation", PIMRC 2012, Sydney, Australia, pp. 355-360, Sep. 9-12, 2012.
4. T. Wigren, "WCDMA uplink load estimation with generalized rake receivers", IEEE Trans. Vehicular Tech., vol. 61, no. 5, pp. 2394-2400, 2012.
5. T. Wigren, "Low Complexity Kalman Filtering for Inter-Cell Interference and Power Based Load Estimation in the WCDMA Uplink", in Proc. 5th International Conference on Signal Processing and Communication Systems, ICSPCS 2011, Honolulu, HI, December 12-14, 2011.
6. T. Wigren, "Recursive noise floor estimation in WCDMA", IEEE Trans. Veh. Tech., vol. 59, no. 5, pp. 2615-2620, 2010.
7. T. Wigren, "Soft uplink load estimation in WCDMA", IEEE Trans. Vehicular Tech., vol. 58, no. 2, pp. 760-772, February, 2009.
8. T. Wigren and P. Hellqvist, "Estimation of uplink WCDMA load in a single RBS", IEEE VTC2007-Fall, Baltimore, MD, U.S.A., October 1-3, 2007.