Influence of Inter Carrier Interference on Link Adaptation Algorithms in OFDM Systems Suvra Sekhar Das1,2 , Muhammad Imadur Rahman1 , Yuanye Wang1 , Fleming B Frederiksen1 , Ramjee Prasad1 1
Center for TeleInFrastruktur (CTiF), Aalborg University,Denmark. 2 Tata Consultancy Services, India. e-mail:
[email protected]
Abstract— The performance of Link Adaptation (LA) under the influence of inter carrier interference (ICI), which is cause by carrier frequency offset (CFO) and Doppler frequency spread due to mobility, in orthogonal frequency division multiplexing (OFDM) based wireless systems is analyzed in this work. LA maximizes throughput while maintaining a target bit error rate (BER)or Block Error Rate (BLER) at the receiver using the signal to noise ratio (SNR) fed back from the receiver to the transmitter. Since ICI power is proportional to the signal strength, the implications of such an impairment on LA OFDM systems is not obvious and has not received much attention till now. It is seen that when ICI is not captured in the SNR measured then the LA system fails to meet the target error rate. However, it is shown that with suitable modification to the SNR switching threshold for changing the adaptive modulation can over the problem. It is also noted that ICI severely reduces the spectral efficiency of OFDM systems even when LA is used.
I. I NTRODUCTION OFDM is widely used as a physical layer technology for high data rate wireless access because of its high spectral efficiency and ability to tolerate the harsh wireless frequency selective fading channel [1]. LA techniques, using adaptive modulation, have also been proposed to overcome the maligning effects of time and frequency selective fading conditions of the wireless channel [2]. Many algorithms have been developed which bring the benefits of adaptive modulation in OFDM systems [3]–[5]. Though OFDM has many benefits, one of its disadvantage is its sensitivity to synchronization errors. It is mainly the frequency synchronization errors that affect the performance severely [6], [7]. Doppler conditions giving rise to time varying channel coefficients and frequency spread also introduce ICI effects [8], [9]. There are several carrier frequency synchronization algorithms but they are not perfect since the estimates of the frequency offset are noisy. Moreover, the frequency synchronization algorithms are designed to track a single carrier offset and hence fail under Doppler frequency spread conditions. Therefore there is always some residual offset and spread which causes ICI in the signal. The amount of ICI generated is dependent on the relative frequency offset to the sub carrier spacing and it is also proportional to the received signal strength [10]. This leads to a difficult situation in LA since different modulations are used in different signal to noise ratio conditions. Therefore it is important to analyze the impact of ICI in OFDM systems using adaptive modulations. To the best of the authors’ knowledge,
there is little focus in the available literature on the performance of LA algorithms when effects of ICI set in. The LA algorithms for OFDM need a lot of consideration under such environments, neglecting which would mean faltering of the LA algorithms. Accordingly, investigations have been made to find the performance of LA schemes under such conditions, which have been detailed in the section below. The organization of the rest of the paper is as follows. The Link adaptation algorithm is described in Section II. The impact of ICI is discussed in Section III. The conclusion is in Section IV which is followed by the Acknowledgement. II. L INK A DAPTATION A LGORITHM There are several LA algorithms for OFDM systems. These algorithms are either very complex and achieve high efficiency [5] or are simple in implementation but have lower throughput [4]. An algorithm for LA is used in this article which has the highest possible throughput of the two algorithms referred, while the complexity is between the two. The proposed algorithm is referred to as simple adaptive modulation and power distribution algorithm (SAMPDA). It is designed based on the two algorithms, simple rate adaptation algorithm (SRA) [5] and adaptive power distribution (APDA) [4]. In the proposed algorithm, greedy approach is used, but unlike the APDA, instead of starting from 0 power and 0 bits as the beginning, the algorithm is initiated with equal power for all sub-carriers. Then by comparing received signal to noise ratio (SNR) with the SNR-lookup table, loaded bits for each subcarrier can be found, and power required for each subcarrier is recalculated. The flow chart for the algorithm is in Figure 1. The details of the algorithm can be found in the work by the authors in [11]. III. I MPACT OF I NTER C ARRIER I NTERFERENCE Figures 2-7 shows the impact of effective offset for different modulations for different signal to noise ratio (SNR) conditions (i.e. different receive signal strength conditions) for uncoded and rate half FEC coded system, in fading channel. The effective frequency offset value is translated to the corresponding velocity considering 2 GHz carrier frequency. In the measurements, 3GPP-LTE parameters such as 512 sub carriers in 5MHz bandwidth have been considered. In all of the above cases the adaptation rate is considered to be
−0.5
Transmit same power on all sub carriers
−1
Feedback measured SINR on each sub carrierto the transmitter
−1.5
−2.5
10
Choose allowable modulation for the received SINR
10*Log (BER)
−2
−3
50 Hz 150 Hz 225 Hz 300 Hz 400 Hz
−3.5
−4
Check if total loaded power exceeds total available transmit power
−4.5
−5
0
5
10
15
20
Fig. 2.
Find the best sub carrier and re calculated the power and modulation
0
Updateparameters, bit and power are distributed to the best sub carrier
−1.5
10
No
50 Hz 150 Hz 225 Hz 300 Hz 400 Hz
−1
10*Log (BLER)
Check if total allocated power is greater than total allowed transmit power
30
Impact of frequency offset on 4-QAM in fading channel.
−0.5
Exclude impossible modulation modes
25
POST−SNR
No
−2
−2.5
Caclculate number of bits loaded for each subcarrier and stop
−3
−3.5
0
5
10
15
20
25
30
POST−SNR
Fig. 1.
Flow diagram of the proposed algorithm Fig. 3. Impact of frequency offset on 4-QAM, FEC rate 1/2, in fading channel.
0.5ms. The channel rms delay spread of 2µs is used. It can be realized that for un coded system a target BER of 10−3 can hardly be maintained at high mobility conditions for high order modulations. For coded case the metric is BLER. In this case the error floor is not easily seen except for 64-QAM. But the impact of high Doppler frequency spread is evident. For QPSK, the performance degradation from 50Hz to 400Hz is about 5 dB at 10−1 BLER level. At the same BLER threshold, the performance degradation for 16-QAM is about 6dB and that for 64-QAM is about 17 dB. At lower Block Error Rate (BLER) thresholds the performance degradation is even worse. A. LA under undetected ICI If the Look Up Table (LUT) for Link Adaptation (LA) is generated based only one Doppler condition, say the 50Hz case, then the BLER performance for other Doppler conditions can be seen from Fig. 8. It can be seen that higher the Doppler frequency spread, the larger is the deviation of the
BLER performance from the target BLER of 10−1 . This can be understood in the light of the increasing Inter Carrier Interference (ICI) with increase in Doppler frequency spread. It has been described before that the threshold to switch from one modulation to another or to select a modulation for a given coding rate depends on the Signal to Interference plus Noise Ratio (SINR). In the above situation, it is assumed that the true SINR is not measured in the receiver. i.e. the ICI component has not been captured at the receiver in the SINR meaurement, i.e. SIN R =
E[|X|2 ]|H|2 σ2
(1)
where σ 2 is the noise power. Here the deterioration in SINR due to ICI is not considered and hence the wrong decision is made in selecting the modulation level. If receivers do not have the capability to estimate the residual carrier offset or
0
0
−0.5
−1 −1
−1.5
10*Log (BER)
−2
10
10*Log10(BER)
−2
−3
−3
50 Hz 150 Hz 225 Hz 300 Hz 400 Hz
−4
−5
−2.5
50 Hz 150 Hz 225 Hz 300 Hz 400 Hz
−3.5
−4
−4.5
−6
−5
0
5
10
15
20
25
30
35
40
0
5
10
15
Fig. 4.
Fig. 6.
Impact of frequency offset on 16-QAM in fading channel.
20
25
30
35
40
45
50
POST−SNR
POST−SNR
Impact of frequency offset on 64-QAM in fading channel.
0
0
−0.5
50 Hz 150 Hz 225 Hz 300 Hz 400 Hz
−1
−1.5
10
10*Log10(BLER)
−1
10*Log (BLER)
−0.5
−1.5
−2
50 Hz 150 Hz 225 Hz 300 Hz 400 Hz
−2 −2.5
−2.5
−3
−3
−3.5 0
5
10
15
20
25
30
35
40
POST−SNR
Fig. 5. Impact of frequency offset on 16-QAM, FEC rate 1/2, in fading channel.
the maximum Doppler spread then the receivers are expected to be in this situation. Following the above formula, the estimate of SINR is much more optimistic than the actual SINR. However as mentioned earlier it is assumed that the maximum Doppler offset is measurable at the Base Station (BS). Then to make the system meet the target BLER, the SNR thresholds must be modified. The updated LUT is given in Table I. Though the mobile stations may not have the capability to measure the residual offset, the base station may be able to do so and indicate the mobile station to use a different SNR threshold corresponding to the Doppler condition. The BLER of the system using the new threshold as defined in Table I is given in Fig. 9. It can be seen that the system now satisfies the target BLER threshold at all Doppler conditions. The spectral efficiency curves for this system is in Fig. 10.
0
5
10
15
20
25
30
35
40
45
50
POST−SNR
Fig. 7. Impact of frequency offset on 64-QAM, FEC rate 1/2, in fading channel.
At 20dB, there is about 17% decrease in spectral efficiency when the effective Doppler increases from 50 to 150 Hz, and the decrease is about 30% when the Doppler is about 225 Hz and when the Doppler by 400 Hz, the decrease in spectral efficiency is about 70%. The decrease in performance is due to the increase thresholds which can be easily seen from the Table I. IV. C ONCLUSION In this section, it is shown that if LA systems are used, when the user equipment which feedback the SNR to the BS cannot estimate the Doppler frequency spread, then the systems fail to meet the target error rate. However, it is also shown that if the BS at least has the required capability to measure the Doppler frequency spread of a user, then a suitable adjustment of the SNR switching threshold can be used for the adaptive modulation and coding, so that LA system can maintain the
0
3
10
2.5
SE(b/s/Hz)
BLER
2
−1
10
1
Fd 50Hz Fd 150Hz Fd 225Hz Fd 300Hz Fd 400Hz 6
8
10
12
14
16
18
20
22
24
PRE−SNR
Fig. 8.
Bler performance without additional margin for coding rate 1/2
5
Fig. 10. rate 1/2
0
Fd 150Hz Fd 225Hz Fd 300Hz Fd 400Hz
BLER
0
10
15
20
TABLE I
10
−2
15
20
25
30
35
PRE−SNR
Fig. 9.
35
Spectral efficiency performance with additional margin for coding
Modulation SNR threshold in dB at 50 Hz SNR threshold in dB at 150 Hz SNR threshold in dB at 255 Hz SNR threshold in dB at 300 Hz SNR threshold in dB at 400 Hz
10
30
LUT FOR DIFFERENT D OPPLER SPREAD CONDITIONS FOR CODE RATE OF 1/2 AT BLER THRESHOLD OF 10−1
−1
5
25
PRE−SNR
10
10
Fd 50Hz Fd 150Hz Fd 225Hz Fd 300Hz Fd 400Hz
0.5
−2
10
1.5
Bler performance with additional margin for coding rate 1/2
target error rate while maximizing the spectral efficiency. Even though the LA system with the new thresholds can meet the required error rate there is a severe impact on the spectral efficiency performance of the systems. It is seen that when the Doppler frequency spread increase from 50Hz to 150Hz, the spectral efficiency drops by about 17%, while if the Doppler is about 400Hz, the performance drops by about 70%. V. ACKNOWLEDGEMENT The authors are grateful to CTIF, Aalborg University and TCS India for funding the project. R EFERENCES [1] Prasad R., OFDM for Wireless Communications. Artech House Publishers, 2004. [2] Chung S. T., Goldsmith A. J., “Degrees of freedom in adaptive modulation: a unified view,” IEEE Trans. Commun., vol. 49, no. 1, pp. 1561– 1571, Sep. 2001.
QPSK 8.52 10.52 12.27 13.52 17.22
16-QAM 14.67 16.67 18.42 19.67 13.37
64-QAM 20.46 22.46 24.12 25.46 29.16
[3] Keller T., Hanzo L.,, “Adaptive modulation techniques for duplex OFDM transmission,” IEEE trans. Veh. Tech., vol. 49, no. 5, pp. 1893 – 1906, Sep. 2000. [4] Toyserkani A. T. et al., “Sub-carrier based Adaptive Modulation in HIPERLAN/2 System,” in IEEE International Conference on Communications, 2004 , June. [5] Lei M., et al., “An Adaptive Power Distribution Algorithm for Improving Spectral Efficiency in OFDM,” IEEE Trans. Broadcast., vol. 50, no. 3, pp. 347 – 351, Sep. 2004. [6] Pollet T., Bladel V. M., Moeneclaey M., “BER sensitivity of OFDM systems to carrier frequency offset and Wiener phase noise,” IEEE Trans. Commun., vol. 43, no. 234, pp. 191 – 193, Feb.-March-April 1995. [7] Sathananthan K. et al., “Analysis of OFDM in the presence of frequency offset and a method to reduce performance degradation,” vol. 1. IEEE Globecom 2000. [8] Zhao Y., Haggman S. -G., “Sensitivity to Doppler shift and carrier frequency errors in OFDM systems-the consequences and solutions,” in IEEE VTC 1996, vol. 3, pp. 1564 – 1568. [9] Tiejun Wang, et al., “Performance Degradation of OFDM Systems Due to Doppler Spreading,” IEEE Trans. Wireless Commun., vol. 5, no. 6, pp. 1422–1432, June 2006. [10] e. a. Das S. S., “Variable subcarrier bandwidths in ofdm systems,” in IEEE VTCS 2007. [11] Das S.S., M.I. Rahman et al., “Influence of PAPR on Link Adaptation Algorithms in OFDM Systems,” in IEEE VTC Spring’07, Dublin, Ireland, 22-25 April 2007.
BLER BS
Block Error Rate Base Station
ICI LA LUT
Inter Carrier Interference Link Adaptation Look Up Table
SINR
Signal to Interference plus Noise Ratio