Beruflich Dokumente
Kultur Dokumente
Channels
Amir D. Dabbagh, Rapeepat Ratasuk, and Amitava Ghosh
Motorola Networks Advanced Technology
1501 West Shure Drive, Arlington Heights, IL 60004, USA
adabbagh@motorola.com, ratasuk@motorola.com, amitava.ghosh@motorola.com
INTRODUCTION
th
order polynomial
contiguous subcarriers follow an N
function. The number L and the polynomial order generally
depend on the coherence bandwidth and can be estimated using
techniques found in the literature (e.g. [8]).
In the proposed channel estimation technique, the ML based
channel estimate
~
hk = y k p k* = hk + nk p k* , k = 1,2,, N sc
y k is the received signal at the k th
subcarrier after cyclic prefix removal and FFT, hk is the true
channel response, n k is the additive noise, p k is the
transmitted pilot symbol, and N sc is the total number of
~
hk = a N k N + a N 1 k N 1 +
where the set of parameters
+ a1k + a0 + nk p k* ,
a 0 , a1 , , a N is estimated using
the least squares criterion. For all data OFDM symbols the
channel estimate is obtained by linear interpolation and
extrapolation of the two pilot based channel estimates when
frequency hopping is not enabled between slots otherwise the
channel estimate is fixed across the OFDM symbols within
each slot. Compared to the IFFT based channel estimation
technique, this channel estimation method offers a factor of
three to four reduction in computational complexity.
In Figure 3, the performance of the proposed channel
estimation method is compared with the ML and the IFFT
based methods for the 5MHz PedB-3 channel model with one
RB allocation and a first order model based polynomial
with L equal to 12. In this simulation, one transmit and two
receive antennas were assumed and the MCS is QAM16 rate
3/4 LTE turbo code with IR HARQ. As can be seen from the
figure, at 10% frame error rate a 3dB gain in SNR is achieved
with the proposed channel estimation method and the
performance gap increases with the SNR. The performance
gap, however, is significantly reduced when a larger number of
RBs are allocated for a user and eventually with full RB
channel allocation the results in Figure 4 show that the IFFT
based estimation technique is marginally better than the first
order polynomial channel estimation method. At 10% FER a
gain of 0.3dB in SNR can be achieved with the IFFT technique
but at the expense of a much higher computational complexity.
Finally, we note that the case of one RB allocation is very
important since it should be the typical type of allocation for
low data rate applications such as VoIP.
In Figure 5, the achievable throughput of PUSCH as a
function of the SNR is shown for a system operating at 10MHz
channel bandwidth with full RB allocation and 2GHz carrier
frequency. In this simulation, MMSE based equalization with
true noise variance estimation is performed for the detection of
10
2.2
x 10
EPA-5
EVA-5
EVA-70
ETU-70
2
1.8
1.6
-1
FER
Throughput (Kbps)
10
-2
10
1.4
1.2
1
0.8
-3
10
0.6
IFFT
LS
ML
0
0.4
10
12
SNR (dB)
14
16
18
0.2
20
10
12
SNR (dB)
14
16
18
20
is used also here, and the results are shown for different
10
35
-1
Throughput [Mbps]
FER
10
-2
10
30
25
20
-3
10
IFFT
LS
-4
10
EPA-5
EVA-5
EVA-70
ETU-70
15
10
4
8
10
SNR (dB)
12
14
16
18
10
12
14
16
18
SNR (dB)
20
22
24
26
1
12
12
l
k
+ nkl p k* wl*
-1
-2
10
l k =1
10
Pe
h=
10
-3
10
10
-30
-25
-20
-15
-10
-5
SNR
Probability of Ack Miss Detection vs. SNR at 1% DTX to Ack Error Rate
10
-1
10
Pe
-4
Ack Misdetect
Nack->Ack
DTX->Ack
Ack Misdetect w/LS
Nack->Ack w/LS
DTX->Ack w/LS
-2
10
-3
10
-4
10
-16
-12
-10
-8
-6
-4
-2
SNR
signals are received with 0dB and 3dB increased power offsets.
Furthermore, simulation results are shown for the case when
there are timing misalignments between the received users
signals. From the simulation results, it can be seen that a
performance degradation of up to 3dB in SNR is present with
three interfering users when 3dB power offset is present. The
system shows robust performance for up to 1us of timing
offsets.
Table 1. Uplink control signaling target quality.
Event
Target quality
(1e-2)
(1e-2)
(1e-4)
10
-1
FER
10
REFERENCES
-2
10
-3
10
-4
10
-18
1.4MHz
3MHz
5MHz
10MHz
15MHz
20MHz
-16
-14
-12
-10
SNR (dB)
-8
-6
-4
Fig. 9. Four bits CQI FER as a function of the SNR with an LTE
Reed-Muller block code
CQI are transmitted over the PUCCH using an LTE ReedMuller block code. The four bits of data are encoded into 20
coded bits. The coded bits are then mapped into 10 QPSK
modulation symbols, and subsequently each modulation
symbol is spread using a length 12 ZC sequence which are then
mapped to a corresponding set of subcarriers for transmission.
At the receiver end, the averaging based channel estimation
method is used over the two pilot OFDM symbols of each slot
and the channel response is then fixed across the subcarriers
within a slot. After the derotation of the spreading sequences,
the 10 soft symbol estimates are finally generated by
correlating the received signal of each data OFDM symbol
with the corresponding channel response. Given the above soft
symbol estimates and the channel responses, the information
bits are detected using an ML detection technique since only
16 codewords are considered in this case and an ML based
detection is still possible with low computational complexity.
Figure 9, shows the performance of the CQI transmission
scheme for all different scalable channel bandwidths supported
in LTE. An ETU-70 channel model is simulated again with one
transmit and two receive antennas. From Figure 9 it can be
easily seen that on the average a 1% FER is achieved at an
SNR of -6.7dB. This shows the robustness of the CQI
transmission scheme under the proposed detection method.
V.
CONCLUSIONS