doc.: ieee 802.22-07/0002r5 submission march 2007 edward au, huawei technologies slide 1 ieee...

doc.: IEEE 802.22-07/0002r5

Slide 1Submission

March 2007

Edward Au, Huawei Technologies

IEEE P802.22 Wireless RANs Last Revised: 2007-03-11

Notice: This document has been prepared to assist IEEE 802.22. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein.

Release: The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE 802.22.

Patent Policy and Procedures: The contributor is familiar with the IEEE 802 Patent Policy and Procedures http://standards.ieee.org/guides/bylaws/sb-bylaws.pdf including the statement "IEEE standards may include the known use of patent(s), including patent applications, provided the IEEE receives assurance from the patent holder or applicant with respect to patents essential for compliance with both mandatory and optional portions of the standard." Early disclosure to the Working Group of patent information that might be relevant to the standard is essential to reduce the possibility for delays in the development process and increase the likelihood that the draft publication will be approved for publication. Please notify the Chair Carl R. Stevenson as early as possible, in written or electronic form, if patented technology (or technology under patent application) might be incorporated into a draft standard being developed within the IEEE 802.22 Working Group. If you have questions, contact the IEEE Patent Committee Administrator at [email protected].

Name Company Address Phone email

Wai Ho Mow HKUST Hong Kong, China 852-2358-7070 [email protected]

Vincent K. N. Lau HKUST Hong Kong, China 852-2358-7066 [email protected]

Roger S. Cheng HKUST Hong Kong, China 852-2358-7072 [email protected]

Ross D. Murch HKUST Hong Kong, China 852-2358-7044 [email protected]

Khaled Ben Letaief HKUST Hong Kong, China 852-2358-7064 [email protected]

Linjun Lu Huawei Technologies Shenzhen, China 0086-755-28973119 [email protected]

Soo-Young Chang Huawei Technologies Davis, CA, U.S. 1-916 278 6568 [email protected]

Jianwei Zhang Huawei Technologies Shanghai, China 86-21-68644808 [email protected]

Lai Qian Huawei Technologies Shenzhen, China 86-755-28973118 [email protected]

Jianhuan Wen Huawei Technologies Shenzhen, China 86-755-28973121 [email protected]

Authors:

Modified CAZAC Sequences Based Low PAPR PreamblesFinal Summary: Pages 24 - 27

http://standards.ieee.org/guides/bylaws/sb-bylaws.pdf

mailto:[email protected]

mailto:[email protected]

doc.: IEEE 802.22-07/0002r5

Slide 2Submission

March 2007


Name Company Address Phone email

Jianhua Sun HKUST Hong Kong, China 852-2358-7086 [email protected]

Edward K. S. Au HKUST Hong Kong, China 852-2358-7086 [email protected]

Zhou Wu Huawei Technologies Shenzhen, China 86-755-28979499 [email protected]

Jun Rong Huawei Technologies Shenzhen, China 86-755-28979499 [email protected]

Jian Jiao Huawei Technologies Beijing, China 86-10-82882751 [email protected]

Meiwei Jie Huawei Technologies Shenzhen, China 86-755-28972660 [email protected]

Co-Authors:

doc.: IEEE 802.22-07/0002r5

Slide 3Submission

March 2007


Background: Impact of Payload PAPR Reduction (1)

• Preamble transmission power is typically higher than the payload transmission power to provide sufficient preamble SNR.

– E.g. In 802.16, downlink preambles are transmitted at an average power of 3~4dB higher than the data payload (considering the mode that preamble subcarriers boosting = 9dB and decimation factor = 3)

• Insufficient preamble SNR will cause– lower accuracy in timing synchronization and channel estimation

– degraded system FER performance

• In Draft v0.2, two binary PN Sequences are used to generate the I and Q components of QPSK symbols which form preambles in the frequency domain (c.f. Section 8.3)

• Superframe and frame preambles currently specified have very high PAPR (> 7.8 dB)

• New proposals of binary preambles can reduce this PAPR to about 5 dB, which provides sufficient preamble SNR only for power amplifiers with a clipping level of 9dB or higher.

• However, many effective methods [1] (e.g. clipping and coding) for reducing the payload PAPR to 5~7dB at various tradeoffs have been reported to reduce power amplifier backoff and thus improve system power efficiency.

[1] S.H. Han, J.H. Lee, “An Overview of Peak-to-Average Power Ratio Reduction Techniques for Multicarrier Transmission,” IEEE Wireless Comm., pp.56-65, Apr 2005.

doc.: IEEE 802.22-07/0002r5

Slide 4Submission

March 2007


Background: Impact of Payload PAPR Reduction (2)

• In Samsung’s doc IEEE802.16e-03/60r1, the tone reservation method was proposed to reduce payload PAPR to ~7dB by reserving about 1.5% of the 2048 sub-carriers.

• Constrained clipping method [2] proposed for 802.16 can reduce payload PAPR to 6.5dB with no sacrifice in bandwidth efficiency and no modification at the receiver.

[2] R.J. Baxley and et al., “Constrained Clipping for Crest Factor Reduction in OFDM,” IEEE Trans Broadcasting, vol.52, Dec 2006, pp.570-575.

• It is practical to achieve a payload PAPR of < 6dB payload PAPR of < 6dB with a combination of known PAPR reduction techniques.

• With preamble boostingWith preamble boosting, the use of binary preambles with PAPR ~5dB may limit the improved power efficiency (power amplifier backoff) achievable with payload PAPR reduction methods.

• To remove this limitation, we propose here a class ofa class of polyphase preambles with max PAPR polyphase preambles with max PAPR ~2dB~2dB and show that, relative to the use of binary preambles, the extra percentage extra percentage complexity incurred at both the transmitter and receiver are practically negligiblecomplexity incurred at both the transmitter and receiver are practically negligible.

doc.: IEEE 802.22-07/0002r5

Slide 5Submission

March 2007


Background: CAZAC sequences• Recently, polyphasepolyphase preambles (Frank & Chu sequencesFrank & Chu sequences) belonging to the class

of Constant Amplitude Zero Auto-Correlation (CAZAC)Constant Amplitude Zero Auto-Correlation (CAZAC) sequences are adopted in IEEE802.16a and IEEE802.15.3 due to their perfect autocorrelation property.

• When the effect of adjacent cell interference (ACI) on preambles has to be considered (c.f. Runcom’s doc IEEE802.22-06/0223r0), a set of preambles with low time-domain cross-correlation energy is desirable.

• These requirements for preambles are very similar to those for channel sounding sequences. Actually, sets of CAZAC sequences, called GCL (Chu) sequencesGCL (Chu) sequences, are also specified in Draft v0.2 as sounding sequences (c.f. Section 8.10.5.4.4).

• Note that only one sequence for one preamble of a type is specified in Draft v0.2. i.e. the effect of ACI was not considered for preambles.

• In this proposal, we modifymodify the CAZAC sequences to obtain (sets of) preambles with very low PAPR and low cross-correlation energy.

doc.: IEEE 802.22-07/0002r5

Slide 6Submission

March 2007


Unified Construction of M-Phase CAZAC Sequences (1)

• The preambles proposed in this contribution are based on a very general construction of M-phase CAZAC (in the M-PSK format), i.e. the unified perfect unified perfect roots-of-unity sequences (PRUS)roots-of-unity sequences (PRUS) [4].

[4] W.H. Mow, “A new unified construction of perfect root-or-unity sequences,” Proc. IEEE 4th International Symposium on Spread Spectrum Techniques and Applications (ISSSTA'96), Germany, September 1996, pp. 955-959.

• It includes the Frank, Chu, Milewski, and GCL sequences and more.

• It was proved by an exhaustive search that the unified PRUS construction includes all M-phase CAZAC sequences with M 15, sequence length L 20 and LM 1111.

• It was conjectured that no more unknown M-phase CAZAC sequences exist [5]. [5] H.D. Lüke, et al. “Binary and quadriphase sequences with optimal autocorrelation properties: a survey,” IEEE Transactions on Information Theory, Vol. 49, Dec. 2003, pp.3271-3282.

doc.: IEEE 802.22-07/0002r5

Slide 7Submission

March 2007


Unified Construction of M-Phase CAZAC Sequences (2)

• The unified CAZAC sequence sCAZAC of length L = sm2 is

• By By modifyingmodifying a properly selected sCAZAC and optimizingoptimizing the parameters s, m, α(l), β(l), (l), low PAPR sequences can be obtained.

number. rationalany is ,

of npermutatio a is that such functionany is ,

withfunctionany is ,

otherwise mod( if /2

where

and

m

mmsm

ms

smm

CAZAC

ZllZmlZlZl

slZlZl

ssc

ZkZl

smlklklsmcilkms

),(,)),(mod(,)(

,1)),(gcd(,)(

,0

0)2,1)(

,

)1.......(..............................)()()()(2exp)(2

doc.: IEEE 802.22-07/0002r5

Slide 8Submission

March 2007


Generation of Modified CAZAC Sequences with M = 2n phases

• Storage requirement for a 2n-phase preamble is nNnNusedused bits bits, where Nused = no. of usable subcarriers. E.g. a 32-phase preamble requires 5 bits per used subcarrier.

• Initial generation of the proposed preamble involves 2 steps:

1. Generate the integer phase indices based on Equation (1) → requires 1 multiplication and 2 additions per index

2. Perform table lookup to obtain the corresponding I and Q representations→ only need to store M/4 = 2n-2 pairs of I/Q values (i.e. phase angles in [0, π/4)) as multiplication by ±1 or ±j can be computed with little complexity. This lookup table can be shared for generating IFFT/FFT coefficients and/or M-phase sounding sequences.

• Proposed sequences can be extended to form a sequence set with low cross-correlation energy for use as a set of ACI-resistant preambles.

– Require only 2 additional cyclic shift operations to generate another sequence in the set.

– Need to store 2 cyclic shift values per sequence

• There is no additional complexity at the transmitter, once the allocated preamble is generated and stored.

doc.: IEEE 802.22-07/0002r5

Slide 9Submission

March 2007


Receiver Complexity: Polyphase vs. Binary Preambles (1)

• The choice of a polyphase or binary preamble has little impact on the timing sync processing, because after taking FFT, the time samples of the preamble used in the correlator are complex-valued anyway. It only makes a difference in the channel estimation block of the receiver.

• Channel estimation requires 1 complex multiplication per used subcarrier with polyphase preambles, while a negation is needed with binary preambles.

doc.: IEEE 802.22-07/0002r5

Slide 10Submission

March 2007


Receiver Complexity : Polyphase vs. Binary Preambles (2)

• To estimate the extra # complex multiplications incurred, let us consider the processing of a framea frame in 2K-FFT mode

– There are 26 OFDM symbols per frame since Frame duration = 10 ms (c.f. Section 6.7.1.1.1) and symbol duration = 373.33 µs (c.f. Section 8.1.2.3.1)

– Channel estimation is performed only using the long frame preamble, so it is only performed for 1/26= 3.85%3.85% of all OFDM symbols in a frame

– FFT is performed for all 26 OFDM symbols (i.e. for both preambles and data).

– An N-point FFT requires (N/2)log2(N) complex multiplications and Nlog2(N) complex additions. For N=2048, 11,264 complex multiplications per OFDM symbol are performed.

– Channel equalization for the 24 data symbols requires 1 complex multiplications per usedused sub-carrier

– With polyphase preamble, channel estimation for the preamble requires 1 complex multiplication per usedused sub-carrier

doc.: IEEE 802.22-07/0002r5

Slide 11Submission

March 2007



• Assuming that 1680 subcarriers are all used, the percentage increase in # complex multiplications due to the use of polyphase preambles is

• The extra complexity incurred is actually much less than 0.5% because the following have not been considered in the above analysis.

– The no. of used subcarriers at CPE is generally <<1680 and can be as small as 18.

– Frequency offset compensation: 1 complex multiplication per time sample for every OFDM symbol

– Packet detection: 2 real multiplications and 3 real additions per time sample

– Viterbi decoding of the 64-state convolutional code for each data OFDM symbol

%5.02416802611264

1680

doc.: IEEE 802.22-07/0002r5

Slide 12Submission

March 2007



• Last but not the least, the complex multiplications required for channel estimation with a polyphase preamble are only phase rotation, which can be very efficiently implemented without any real multiplication by applying the famous shift-and-add-only CORDIC algorithm.

• Reference: http://www.dspguru.com/info/faqs/cordic.htm

• In conclusion, the use of the proposed polyphase preambles only incurs negligible extra percentage complexitynegligible extra percentage complexity at the receiver, compared with the binary preambles.

http://www.dspguru.com/info/faqs/cordic.htm

doc.: IEEE 802.22-07/0002r5

Slide 13Submission

March 2007


Results on Low PAPR Preambles (1)

• Results are presented for the setting: – 2K, 4K and 6K FFT modes, null subcarriers [L=184n, DC, R=184n-1] (n

= no. of bonded TV channels)

– Decimation factor = 1, 2, 3 or 4

– Number of bonded TV channels = 1, 2 or 3

• Here, all PAPR values are estimated for continuous-time waveforms using an oversampling factor of 4. Without oversampling, the computed PAPR values may be over-optimistic.

doc.: IEEE 802.22-07/0002r5

Slide 14Submission

March 2007


Results on Low PAPR Preambles (2)• The following table lists the PAPR values of the proposed modified CAZAC

sequences for different modes of preambles in draft v0.2 with 32 and 128 phases, respectively.

• The proposed preambles can be used to replace preambles in the current draft with a PAPR gain of >5.6dB.

FFT Size = 2048 Null subcarriers [L=184, DC, R=183]

Modified CAZAC(128 phases)


Superframe Long Preamble (Decimation factor = 1)

1.63 dB 1.88 dB

Superframe Long Preamble (Decimation factor = 2)

1.81 dB 1.97 dB

Superframe Preamble (Decimation factor = 3) 1.76 dB 2.03 dBSuperframe Short Preamble (Decimation factor =

4)1.93 dB 2.07 dB

Frame Long Preamble (Decimation factor = 1) 1.74 dB 1.99 dBFrame Long Preamble (Decimation factor = 2) 1.81 dB 2.02 dB

Frame Preamble (Decimation factor = 3) 1.91 dB 1.98 dBFrame Short Preamble (Decimation factor = 4) 1.88 dB 2.03 dB

doc.: IEEE 802.22-07/0002r5

Slide 15Submission

March 2007


Results on Low PAPR Preambles (3)• The following table lists the PAPR values of the modified CAZAC sequences

for frame preambles when the number of bonded TV channels n = 2 or 3.

Null subcarriers [L=184n, DC, R=184n-1]



Frame Long Preamble (2 bonded channels, FFT size = 4096, Decimation factor = 1)

1.74 dB 2.05 dB


1.69 dB 2.09 dB

Frame Preamble (2 bonded channels, FFT size = 4096, Decimation factor = 3)

1.76 dB 1.95 dB

Frame Short Preamble (2 bonded channels, FFT size = 4096, Decimation factor = 4)

1.79 dB 2.08 dB


1.41 dB 1.78 dB


1.75 dB 2.14 dB

Frame Preamble (3 bonded channels, FFT size = 6144, Decimation factor = 3)

1.69 dB 2.09 dB

Frame Short Preamble (3 bonded channels, FFT size = 6144, Decimation factor = 4)

1.75 dB 2.04 dB

doc.: IEEE 802.22-07/0002r5

Slide 16Submission

March 2007


• PAPR reduction as compared to the preambles based on PN sequences in the current Draft v0.2 is at least 5.6 dB.

• By reducing M from 128 to 32 and hence the lookup table size from 32 to 8 pairs of I/Q values, the resultant PAPR values are still very low and the worst-case PAPR is only increased mildly from 1.93dB to 2.14dB.

Results on Low PAPR Preambles (4)

doc.: IEEE 802.22-07/0002r5

Slide 17Submission

March 2007


• When adjacent cell interferenceadjacent cell interference is a concern, we propose a set of modified CAZAC sequences with low PAPR and low cross-correlation levels as preambles (and sounding sequences).

• The average energy of the time-domain cross-correlation functions is as low as that of the GCL (Chu) set specified in Draft v.0.2, leading to same adjacent cell interference power.

• Next, the PAPR values of a set of 114 modified CAZAC sequences are evaluated.

• The worst case PAPR of the proposed set is 2.55dB, which is about 2.2dB better than the Chu set.

Modified CAZAC Sequence Set (1)

doc.: IEEE 802.22-07/0002r5

Slide 18Submission

March 2007


0 20 40 60 80 100 1201.5

2

2.5

3

3.5

4

4.5

5

Sorted Sequence Index

PA

PR

(dB

)

FFT Size = 2048, Decimation Factor = 4

Chu setModified CAZAC Set (128 phases)

Modified CAZAC Sequence Set (2)

doc.: IEEE 802.22-07/0002r5

Slide 19Submission

March 2007


CDF of PAPR of Sequence #114 in the Modified CAZAC Set

-4 -3 -2 -1 0 1 2 30

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

PAPR in dB

Pro

babi

lity


CD

F

doc.: IEEE 802.22-07/0002r5

Slide 20Submission

March 2007


CDF of PAPRs of the 114 Sequences in the Modified CAZAC Set

1.8 1.9 2 2.1 2.2 2.3 2.4 2.5 2.60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

PAPR (dB)

CD

F


doc.: IEEE 802.22-07/0002r5

Slide 21Submission

March 2007


Summary1. We proposed the use of modified CAZAC sequences to replace the existing preambles specified in draft v0.2.

– The proposed polyphase preambles can attain very low PAPR (≤2.14dB for 2K, 4K and 6K FFT) so that

1. More robust synchronization and better detection performance can be achieved by having higher tolerance on AGC error and faster AGC convergence, as well as less sensitive to amplifier nonlinearity.

2. With preamble boosting, detection performance will not be limited by the preambles , even if advanced PAPR reduction methods are applied to reduce the payload PAPR to <6dB.

– It was demonstrated that sets of modified CAZAC sequences can also attain very low PAPR (≤2.55dB for 2K, 4K and 6K FFT & set size = 114), while having the same time-domain cross-correlation energy as that of the GCL (Chu) set specified in Draft v0.2.

2. Compared with binary preambles, the extra percentage complexity of using the proposed polyphase preambles is practically negligible (<<0.5%) at both the transmitter and receiver, because

– At the transmitter, preamble generation needs only be done once and then stored.

– At the receiver, channel estimation operations are performed very infrequently relative to the complicated payload processing like IFFT and the involved complexity multiplications are only phase rotations which can be easily implemented with shifts and adds only via the CORDIC algorithm, rather than hardware multipliers .

3. Low PAPR results for the proposed type of preambles have been obtained for various decimation factors 1,2,3,4 (corresponding to 1, 2, 3 and 4 repetitions of PN sequences) and for FFT size = 2K, 4K and 6K. These confirm that the desirable properties of the proposed preambles can be maintained independent of the choice of preamble structures.

doc.: IEEE 802.22-07/0002r5

Slide 22Submission

March 2007


Recommendation

Since the extra complexity of applying polyphase preambles is not an issue, we recommend the use of the modified CAZAC preambles which enable more robust synchronization and better detection performance, compared with the binary preambles.

doc.: IEEE 802.22-07/0002r5

Slide 23Submission

March 2007


A Table Suggested for Comparing Different Proposals

Null subcarriers [L=184n, DC, R=184n-1], Oversampling factor = 4Frame Long Preamble: 1 - 3 bonded channels; FFT size = 2048, 4096, 6144; Decimation factors = 1, 2

Frame Preamble: 1 - 3 bonded channels; FFT size = 2048, 4096, 6144; Decimation factor = 3

Frame Short Preamble: 1 - 3 bonded channels; FFT size = 2048, 4096, 6144; Decimation factor = 4

1) Per Sequence:

2) Per a set of sequences:

Null subcarriers [L=184n, DC, R=184n-1] ], Oversampling factor = 4Frame Short Preamble: 1 - 3 bonded channels; FFT size = 2048; Decimation factor = 4; 114 sequences

3) Receiver Complexity

doc.: IEEE 802.22-07/0002r5

Slide 24Submission

March 2007


Final Summary (1): Preamble Boosting

• Preamble transmission power is typically higher than the payload transmission power to provide sufficient preamble SNR.

– E.g. In 802.16, downlink preambles are transmitted at an average power of 3~4dB higher than the data payload (considering the mode that preamble subcarriers boosting = 9dB and decimation factor = 3)

• Insufficient preamble SNR will cause– lower accuracy in timing synchronization and channel estimation

– degraded system FER performance

• New proposals of binary preambles can reduce this PAPR to about 5 dB, which provides sufficient preamble SNR only for power amplifiers with a clipping level of 9dB or higher.

• With preamble boostingWith preamble boosting, the use of binary preambles with PAPR ~5dB may limit the improved power efficiency (power amplifier backoff) achievable with payload PAPR reduction methods.

• To remove this limitation, we propose here a class ofa class of polyphase preamblespolyphase preambles with with max PAPR max PAPR ~2dB ~2dB for various decimation factors 1,2,3,4 (corresponding to 1,2,3,4 repetition(s) of PN sequences) and for FFT size = 2K, 4K, 6K. These confirm that the desirable properties of the proposed preambles can be maintained independent of the choice of preamble structures.

doc.: IEEE 802.22-07/0002r5

Slide 25Submission

March 2007


Final Summary (2): Negligible Extra Complexity

• Compared with binary preambles, the extra percentage complexity of using the proposed polyphase preambles is practically negligible (<<0.5%) at both the transmitter and receiver, because

– At the transmitter, preamble generation needs only be done once and then stored.

– At the receiver, channel estimation operations are performed very infrequently relative to the complicated payload processing like IFFT

– The complex multiplications required for channel estimation with a polyphase preamble are only phase rotation, which can be very efficiently implemented without any real multiplication by applying the famous shift-and-add-only CORDIC algorithm, rather than hardware multipliers.

doc.: IEEE 802.22-07/0002r5

Slide 26Submission

March 2007


Final Summary (3): More Robust Synchronization

• More robust synchronization and better detection performance can be achieved by use of superframe/frame short preambles with lower PAPR because of larger tolerance on AGC error and faster AGC convergence.

– When preamble is received, the signal level normally increases rapidly causing the ADC to saturate. The AGC gain is then increased in discrete steps until the ADC operates in its linear range and a proper power estimate can be made.

– Preambles with lower PAPR require lower ADC resolution (relative to that of the payload) and hence provides larger tolerance on AGC error. (Positive AGC error will increase false alarm in packet detection, while negative AGC error will increase miss rates.)

– Lower PAPR (smaller amplitude variation) leads to faster AGC convergence, which reduces the distortion due to the transient effect on the leading edges of the received preambles. (Ref:http://www.freepatentsonline.com/20030108129.html)

• E.g. These advantages of the short training sequence with PAPR ~2dB in 802.11a facilitates the realization of high-accuracy all-digital AGC and more robust synchronization (c.f. [3]).

[3] H.-Y. Liu and et al., "A COFDM Baseband Processor with Robust Synchronization for High-Speed WLAN Applications," 2004 Symposium on VLSI Circuits, 17-19 June 2004, pp.156-159.

• Preambles with lower PAPR also have less distortion due to the non-linearity of amplifiers. In general, amplifiers with high linearity are costly and less power efficient.

• Even without preamble boosting, it is very advantageous to use polyphase preambles with PAPR ~2dB to achieve more robust synchronization and better detection performance. Note that these advantages, which are basically why constant envelop modulations are traditionally preferred for wireless, are obtained at practically negligible extra complexity.

http://www.freepatentsonline.com/20030108129.html

doc.: IEEE 802.22-07/0002r5

Slide 27Submission

March 2007


Since the extra complexity of applying polyphase preambles is not an issue, we recommend the use of the modified CAZAC preambles which enable more robust synchronization and better detection performance, compared with the binary preambles.

Recommendation

doc.: ieee 802.22-07/0002r5 submission march 2007 edward au, huawei technologies slide 1 ieee...

Documents