Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)

1
Sept 2003
doc. IEEE 15-03-0337-01-003a
Project IEEE P802.15 Working Group for Wireless
Personal Area Networks (WPANs) Submission Title
Techniques for MB-OFDM improvement Date
Submitted 5 September 2003 Source Mitsuhiro
Suzuki, Chihiro Fujita, Michio Yotsuya,
Kazuhisa Takamura, Takashi Usui Bob
Huang Company Sony Corporation

Sony Electronics of America Address
6-7-35 Kitashinagawa Shinagawa-ku,Tokyo. Japan
141-0001 One Sony Drive TA-1 Voice
81-3-6409-3201, FAX [phone] Park
Ridge, NJ 07656 E-Mail suzuki_at_wcs.sony.co.jp,
chihiro_at_ wcs.sony.co.jp,
V 201-358-4409 yotuya_at_
wcs.sony.co.jp, takamura_at_wcs.sony.co.jp,
F 201-930-6397
usui_at_ wcs.sony.co.jp
EMail
robert.huang_at_am.sony.com Re none Abstract This
presentation introduces the unique techniques for
MB-OFDM , ranging, null prefix,
preamble waveform, coding, tracking Purpose Techn
ical contribution to MB-OFDM proposal. Notice Thi
s document has been prepared to assist the IEEE
P802.15. 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 acknowledges and accepts that this
contribution becomes the property of IEEE and may
be made publicly available by P802.15.
Suzuki, et al, Sony Corp., Sony Electronics
Slide 1
Submission
2
Techniques for MB-OFDM improvement
Sept 2003
Ubiquitous Technology Labs Sony Corporation
3
Contents
1. Ranging Techniques -Clocking burst
transmission -Coherent sub-banding
-Frequency domain smoothing 2. Modified Tx
preamble waveform 3. Null prefix for OFDM 4.
Coding scheme 5. Tracking method without pilot
4
Clocking Burst Transmission
(How to measure the turn around time)
5
Clocking Transmission Mechanism
Define clocking period T which is longer than
2x(Max propagation delay) Device A start to
clock. Device A transmit signal to Device B at
the clocking timing. Device B detect the signal
arrival time from A. Device B start to clock
from the arrival time. Device B transmit to
Device A at the clocking timing. Device A detect
the signal arrival time from B. Time lag between
the arrival time and the last clocking timing is
2x(Propagation delay)
6
In MB-OFDM case, T which is same as time
slot length ( 312.5ns ) is convenient.
Physical layer system closed (with no MAC help
) No negotiation is necessary.
7
Coherent sub-banding
(How to improve measuring precision)
8
Current problem for ranging
Frequency characteristics of each band can be
obtained. However, no phase coherency between
each band. Ranging resolution is determined by
each bandwidth (528MHz).
9
Solution for increase measured bandwidth
1. Define reference time of t0 during the burst.
( It is the best that t0 is set to head of the
channel estimation part.) 2. Define phase of each
carrier set to 0deg. ?Phase coherency between
each band is achieved. (even if the difference
between t0 of Tx and t0 of Rx exists.)
10
Setting phase of each carrier
Fortunately, every frequency can be synthesized
from one oscillator (4224MHz). All carriers
phase is set to 0deg at the time of 264MHz
phase 0deg
11
Frequency domain smoothing
(How to reduce complexity to calculate time
response from frequency response )
12
Current problem of calculate time response
Large size FFT is needed ( In 3 band case,
128x3384 )
13
Frequency domain smoothing and decimation
For the purpose of ranging, time response around
time 0 is important. Therefore, smoothing and
decimation in frequency domain is allowed. FFT
size can be reduced.
14
Conclusion for ranging techniques.
Turn around time can be measured only by physical
layer implementation by clocking burst
transmission. 1.5GHz bandwidth channel
response measuring is possible by coherent
sub-banding (3 band case) Calculation complexity
can be reduced by frequency domain smoothing.
15
Tx preamble waveform improvement
16
Current preamble (time domain)
Hierarchical code is convenient to reduce
complexity
Preamble No.1
A little bit bad auto-correlation
Auto-correlation
17
Current preamble (frequency domain)
Power spectrum shape is not flat. (Tx power have
to be reduced 5dB!!! in FCC rule point of view)
Preamble part power spectrum
Power spectrum Shape is different from OFDM part.
How to shape the spectrum into 500MHz with 528
Mchip/s chip rate ?
OFDM part power spectrum
18
Modification of Tx Preamble
( original preamble )
force to set amplitude 0
force to set amplitude 1 keeping phase
information
Preamble at Tx side is regarded as OFDM signal
which symbols are in frequency domain.
Power Spectrum shape is (of course) same as
OFDM.
Power spectrum is (of course) flat.
Modified power spectrum
19
In Rx side, original preamble is used
Tx preamble is no longer binary ( still real
number)
Original binary preamble is used at the
receiver not to increase correlation complexity
Preamble for Tx(red) and Rx(blue)
Correlation characteristics improves. ( approx.
3dB )
Cross-correlation
20
Conclusion for Tx preamble improvement
Preamble at Tx amplitude modification in
frequency domain. same power spectrum as
OFDM no-necessity to care about extra
spectrum shaping flat power spectrum and
allowed max Tx power.
Preamble at Rx same as original preamble
not increase correlation complexity better
correlation performance.
21
OFDM with null prefix
22
Conventional cyclic prefix GI
Cyclic prefix is to avoid ICI (
Inter-subCarrier-Interference)
23
Proposed null prefix
Null prefix (or postfix) OFDM can also avoid
ICI by cyclic adding at the receiver.
24
Necessary processing for null prefix
Need detecting multi-path time dispersion ( TMP
) ( This may be done during CCA?
) Need ADC during TMP longer period. Need cyclic
adding during TMP period.
25
Picked up noise
Null prefix case, the receiver will pick up same
noise or (TEMTMP)/TEM larger noise.
(Depend on multi-path time dispersion 0 lt TMP lt
TGI )
However,,,
26
Tx power can be increased (prefix energy
consumption aspect )
UWB Tx power is specified by power density.
Null prefix OFDM can have (TEMTGI)/TEM higher Tx
power because it does not spend the energy
for prefix.
TGIgtTMP (typical) S/N at receiver improves.
TGITMP (worst) S/N is same as cyclic prefix.
27
Tx spectrum shape will be flat
Cyclic prefix generates ripple in spectrum.
Null prefix generates no ripple in spectrum.
28
Tx power can be increased (power density in MHz
aspect)
UWB Tx power is specified by power density in MHz
Null prefix OFDM is allowed (TEMTGI)/TEM higher
Tx power keeping FCC spectrum regulatory.
Null prefix has higher allowable Tx power when
sub-carrier spacing is not less than 1MHz
29
Conclusion for null prefix
Null prefix improves from 2dB to 1dB
link performance.
Null prefix can be applied to modified Tx preamble
30
Coding scheme
31
Coding scheme
Decoder power consumption can not be neglected.
Decoding latency is not so critical in payload
part.
K 7 convolutional coding (current )
low latency, but high power consumption not so
good BER/PER
K 4 CC RS(255,239) concatenation (proposed)
K 4 gate count is 1/8 smaller than K7 K 4
power consumption is 1/8 smaller than
K7 RS(255,239) gate count is comparable to K7
RS(255,239) power consumption is 1/8 smaller
than K7
32
Performance comparison
To be prepared
33
Conclusion for coding scheme.
K4 CC RS concatenation
a few dB better BER/PER performance than K7 gate
count is comparable to K7 power consumption is ¼
lower than K7
All of considerable coding scheme should be
studied. Low power consumption coding scheme is
desired as mandatory.
34
Tracking method without pilot
35
Synchronization situation
Fortunately, sub-carrier of OFDM and center
frequency are synchronized.
Modulation timing error and carrier phase error
are synchronized
Timing error ( carrier phase error ) is caused by
reference frequency difference between Tx and Rx.
36
Rotation in frequency domain
OFDM signal is described.
OFDM with timing error is described.
Timing error is observed rotation in symbol on
each sub-carrier.
The phase value of rotation is represented by
center frequency, because bandwidth is small
compared to center frequency.
37
Tracking method block diagram
Timing error detection part (red) is very
conventional
Pilot is not necessary
38
Conclusion for tracking using pilot
More precise timing error detection by using many
(122) sub-carrier information. (current number of
pilot symbol is 12)
0.5dB (112/100) link performance will be
improved because of no energy lost for pilot.
Coding rate for high bit rate (e.g.
480Mbps_at_r3/4 ) can be decreased and
improve Eb/No performance. -More link
performance will be improved. -SOP
performance will be improved
Pilot is not necessary
( The idea is under qualitative consideration.
Quantitative value by simulation is
needed. )
39
Total conclusion
This document introduced
1. Ranging Techniques -Clocking burst
transmission -Coherent sub-banding
-Frequency domain smoothing 2. Modified Tx
preamble waveform 3. Null prefix for OFDM 4.
Coding scheme 5.Tracking method without pilot
These achieve better performance in MB-OFDM system
40
THE END OF SLIDES Thank
you!