A Radio Channel Estimation Scheme Using the
CQI Feedback Information in High Speed
Downlink Packet Access
Junsu Kim, YoungJun Hong, and Dan Keun Sung Department of Electrical Engineering and Computer Science Korea Advanced Institute of Science and Technology (KAIST)
Channel Quality Indicator (CQI)
• Mobile terminals do not report the measured SINR directly to base station
• Mobile terminals measure their own radio channel status and report the quality in uplink direction to base station using a Channel Quality
Indicator (CQI).
• Channel Quality Indicator (CQI) which is a sampled and quantized version of the measured SINR.
– CQI is another version of SignaltoInterference and Noise Ratio (SINR), more information about radio channel states of mobile terminals can be extracted
– CQI is an indicator carrying the information on how good/bad the communication channel quality is.
• Base station utilizes CQI feedback information for scheduling, Adaptive Modulation and Coding (AMC)
• Factors play important roles to CQI measurement.

SignaltoNoise Ratio (SNR)

SignaltoInterference plus Noise Ratio (SINR)

SignaltoNoise plus Distortion Ratio (SNDR)
Channel Quality Indicator (CQI)
• Based on careful observation of reported CQI information by base station, the mobile speed and geographical environment of each mobile terminal can be estimated.
• Depending which value User Equipment (UE) reports, network transmit data with different transport block size.
– If network gets high CQI value from UE, it transmits the data with larger transport block size and vice versa.
• In this case, network would send a large transport block size according to the CQI value and it would become highly probable
that UE failed to decode it (cause CRC error on UE side) and UE
send NACK to network and the network have to retransmit it which in turn cause waste of radio resources.
Proposed Scheme for Channel estimation
• An effective channel estimation scheme in which large scale and small scale fading are estimated simultaneously at a base station using the feedback information from mobile terminals.
• The proposed scheme decomposes a radio channel model into a largescale fading component and a smallscale fading component using the feedback information.
• The LargeScale fading component is determined by the distance from base station and its surrounding geographical environment.
• The Small Scale fading component depends on the mobile speed. These components affect the Signal to Interference
and Noise Ratio (SINR) measured in a mobile terminal.
• The analysis with the measured SINR values is still valid for CQI with some quantization errors.
Proposed Scheme for Channel estimation
• The mean of SINR represents the largescale fading and the level crossing rate (LCR) of the normalized SINR, which is the SINR value divided by its average value, only depends
on the smallscale fading.
• A twodimensional 'MeanLCR Plane' to efficiently estimate the radio channel status of mobile terminals.
• Proposed estimation scheme is able to estimate the radio channel status of mobile terminals including the small and largescale fading components
Overview of HSDPA
• The HSDPA (High Speed Downlink Packet Access) system consists of a single NodeB (base station) and multiple user equipments (UE, mobile station) in a cell.
• A UE measures its channel quality using a known pilot channel and determines a CQI value to obtain the target QoS, such as Frame Error Rate (FER) or BLock Error
Rate (BLER), and the CQI value is reported to NodeB.
• Then, the scheduler in the NodeB selects a UE based on an appropriate policy using the reported CQI values.
• In HSDPA, the CQI value ranges from 0 ~ 30. • Value 30 indicates the best channel quality and 0.1 indicates the poorest channel quality.
Measured CQI and SINR values in HSDPA
• Figure shows the relation between the SNR and CQI values, when the UE's mobile speed is 5km/h. • In the figure, the solid line indicates the measured SINR and the discrete steps indicate the corresponding CQI values. • Since only 30 CQI values are defined in HSDPA, the CQI values cannot indicate the whole range of SINR values
Mathematical Analysis of SINR and CQI
A. Environment and Assumptions
• A cell model which consists of a
homecell and outer cells.
• We assume that all the NodeBs
transmit constant power and the
transmitted power from the outer
cell NodeBs becomes outercell
interference to a UE in the home
cell.
• The cell radius is R km and the
distance from NodeB to the UE
is r km.
• Single path Rayleigh fading as a
smallscale fading model and the
path loss as a largescale fading
model.
Mathematical Analysis of SINR and CQI
B. SINR Formulation
The path loss from the NodeB in the cell (i, j) can be expressed as
where K is a constant factor, r _{o} is a reference distance in km, and n is a path loss exponent.
Mathematical Analysis of SINR and CQI
B. SINR Formulation
cell can be written as:
Where No is the gaussian thermal noise
Mathematical Analysis of SINR and CQI
B. SINR Formulation
• This equation includes random variables in both denominator and numerator, it is very complex to derive the probability density function (PDF) of SINR directly. • In order to simplify this the variances of the signal component and the interference component are investigated. • If the variance of the interference component is much smaller than that of the signal component, then the interference component can be approximated as a constant in the viewpoint of the signal component. • In other words, the signal component dominates the statistical characteristics of the SINR.
Mathematical Analysis of SINR and CQI
B. SINR Formulation
Where Ω _{i}_{,}_{j} is the mean power of fading coefficients i.e. α ^{2} _{i}_{,}_{j} thus we can assume Ω _{i}_{,}_{j} = 1. The ratio of the above variances is given by
Mathematical Analysis of SINR and CQI
B. SINR Formulation
• Simulation result in a 2tier cell model, where the cell radius is 1 km, and xaxis indicates the distance from the homecell
NodeB to the UE.
• Variance of the signal component is 10 times larger than that of the interference component, at a location of 0.9 km apart from the home cell NodeB, and even 100 times larger at a
location of 0.6 km. • This means that the characteristics of the signal component dominate the statistical characteristics of entire SINR in the inner area of the home cell. • I ^{O}^{C} (r) ~ E[I ^{O}^{C} (r)]. • Since No is relatively smaller than the interference component, we can neglect the thermal noise component.
Mathematical Analysis of SINR and CQI
B. SINR Formulation
Rewriting the SINR formula with only one random variable
Mathematical Analysis of SINR and CQI
B. SINR Formulation
Mean and Standard Deviation of SINR
Mathematical Analysis of SINR and CQI
determined by the Rice formula
• Normalized SINR is independent of distance between the home cell NodeB and UE.
Mathematical Analysis of SINR and CQI
• Normalized SINR is independent of distance between the home cell NodeB and UE. • It only depends on the small scale fading coefficients • LCR of the normalized SINR would be determined by smallscale fading regardless of the large scale fading in other words, regardless of the geographical characteristics
Mathematical Analysis of SINR and CQI
Mathematical Analysis of SINR and CQI
• This equation is a function of threshold, Y _{t}_{h} , and the mobile speed inherent in σ. • It always has the maximum value at Y _{t}_{h} = 0.5 regardless of mobile speeds. • Figure shows the analytical LCR value of Y(r) for various thresholds and various mobile speeds. • LCR always has the maximum value at a threshold of 0.5. Therefore, it is reasonable to use the threshold value of 0.5 to classify the mobile speeds.
Mathematical Analysis of SINR and CQI
• When the distance between the NodeB and
the UE is less than 0.3km, the LCR of CQI
becomes zero regardless of mobile speeds. • This is due to the accuracy of CQI. Since there are only 30 CQI values, all the measured SINR values should be classified
into one of 30 CQI values. Therefore, the
UEs near the NodeB tend to have very high
probabilities to report a maximum CQI value of 30. • Although the SINR fluctuates as the mobile speed increases, the CQI value has a
maximum value of 30. Therefore, the level crossing of CQI values hardly happens. However, if the distance between the NodeB and the UE is larger than 0.3km,
Two Dimensional Channel Estimation
A. Channel Estimation in the MeanLCR Plane
• Mean of SINR depends on the largescale fading including the distance from the NodeB to the UE.
• The LCR of the normalized SINR depends on the smallscale fading including the mobile speed.
• Mean SINR and the LCR of the normalized SINR together, can be used to estimate the wireless channel status of a certain UE.
Two Dimensional Channel Estimation
A. Channel Estimation in the MeanLCR Plane
• Channel Classification based on the Mean and LCR values of SINR
• Figure shows the LCR values of the normalized SINR according to the mean values of SINR for 20 UEs with different mobile speeds and different locations. • In this figure, we can classify each UE using the position on the plane of both the mean and
the LCR values of SINR. • UEs in the rightbottom corner of the plane have better channel status than those in other parts of the plane.
Two Dimensional Channel Estimation
A. Channel Estimation in the MeanLCR Plane
• Channel Classification based on the Mean and LCR values of CQI
• The mean values of SINR are replaced by the
mean values of CQI and the LCR values of the
normalized SINR are substituted by the LCR values of CQI. • Here the dots concentrated in the rightbottom corner of the plane. These dots are for the UEs located near the NodeB. The UEs near the NodeB have very low LCR values of CQI. • Although these UEs have relatively higher mobile speeds, their channel status is good enough to be served by the NodeB.
Two Dimensional Channel Estimation
A. Channel Estimation in the MeanLCR Plane
• Mean – LCR plane
• Here the plane is divided into 9 regions. The x axis indicates the distance and the yaxis indicates the mobile speed. • Each region is numbered according to the mean and LCR values of CQI/SINR about current wireless channel status. • UEs in the smaller numbered regions have better wireless channels. Therefore, if the Mean
LCR plane is managed in the NodeB, the Node B can estimate not only the UE's current channel status but also the channel variation history in the past. • This information is very valuable in HSDPA systems.
Conclusions
•

A UE specific radio channel is mainly characterized by smallscale fading due to its geographical conditions and large scale fading due to the mobile speed of the UE.

•

In this paper, it is possible to estimate the smallscale and largescale fading using the mean values of SINR and the LCR values of the normalized SINR,

respectively.
• The approximated SINR distribution is found out and analyze the mean and LCR values.
•

A twodimensional channel estimation scheme based on a 'MeanLCR Plane' is proposed which is used to represent the UE's current radio channel status and its history.

•

In HSDPA, CQI values are reported to NodeB instead of the measured SINR.

•

CQI is a sampled and quantized version of the measure SINR.

•

The analysis based on SINR is still valid for CQI with some quantization error by comparing this result with the simulation result. Therefore, we can efficiently manage the UE's radio channel status based on the 'MeanLCR Plane' of the CQI

version at the NodeB.
• Finally, we verify the performance of the proposed scheme using intensive systemlevel simulations. It is possible to develop more efficient scheduling algorithms or handoff schemes using the proposed channel estimation scheme for further work.