Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
1
Introduction to Shortened TTI And Processing Time for LTE
Sam Meng
HTC
2
• Background
• Design Considerations
• Specification
• Concluding Remarks
Table of Contents
3
Background
4
• Short for Transmission Time Interval
• But it can mean different things, e.g., TTI for
PBCH is 40 ms
• Here it refers to the smallest scheduling unit in
the time domain
• One subframe <=> 1 ms
TTI in LTE
#0 #1 #2 #3 #19#18
One radio frame, Tf = 307200Ts = 10 ms
One slot, Tslot = 15360Ts = 0.5 ms
One subframe
[1]
5
• Like TTI, it does not have a precise definition
• Here it refers to the standardized time delay for
UE/eNB to report HARQ-A/N for DL/UL data
transmissions
• FDD: 4m
• TDD: defined in a table
Processing Time in LTE
[2]
6
• Reduce packet data latency
• Directly affect perceived responsiveness;
video conferencing/driving/real-time
gaming/VR
• Indirectly determine throughput
• Improve resource efficiency
• Improve error rate given delay requirement
• Lower buffer requirement
• Advancements in hardware and processing
capability
Motivation to Reduce Latency
[9]
7
• Directly shorten the time unit for data
scheduling
• Has various impacts on the physical control
channel (PDCCH) and physical data channel
(PDSCH)
Reduce Latency by Shortening TTI
D
ow
nlin
k s
yste
m b
an
dw
idth
1 subframe
1st slot 2nd slot
Unicast PDSCH
PD
CC
H
1 subframe
1st slot 2nd slot
sP
DS
CH
PD
CC
H
sP
DS
CH
sP
DS
CH
sP
DS
CH
sP
DC
CH
sTTI sTTI sTTI sTTI
sP
DC
CH
sP
DC
CH
sP
DC
CH
Unicast PDSCH
D
ow
nlin
k s
yste
m b
an
dw
idth
1 subframe
1st slot 2nd slot
PD
CC
H
sP
DS
CH
sP
DC
CH
sTTI
1 subframe
1st slot 2nd slot
PD
CC
H
sP
DS
CH
sP
DC
CH
sTTI
Unicast PDSCH
(a) Cases corresponding to Alt 1. (b) Cases corresponding to Alt 2 and Alt 3.[4]
8
• Reduce latency for normal TTI too
• Easier to multiplex with legacy UEs’ data
Reduce Latency by Shortening Processing Time
0 1 2 3 4 5 6 7 8 9
0 1 2 3 4 5 6 7 8 9
Legacy
UE
UE with reduced
processing time
PUCCH resource
collision
DL
UL
HARQ
A/N
9
• For the actual time spent on processing data, yes
• But the catch is that UE can’t just report
HARQ-A/N to or received from eNB earlier
even if the HARQ-A/N is ready
Don’t We Already Have A Shorter Processing Time?
10
• This allows both sides to agree on a schedule (where to expect data, where to
expect HARQ-A/N)
• Lower complexity, facilitate resource scheduling
• Also affects the soft buffer management at the UE side
HARQ Timeline in LTE
[2]
11
Design Considerations
12
• Might not be that straightforward because
• Boundary alignment
• Control channel (PDCCH)
• HARQ timing does not proportionally
scale
• Limited UE tx power (UL coverage)
• Increased overhead of control signals
So, Just Shorten The TTI?
D
ow
nlin
k s
yste
m b
an
dw
idth
1 subframe
1st slot 2nd slot
Unicast PDSCH
PD
CC
H
1 subframe
1st slot 2nd slot
sP
DS
CH
PD
CC
H
sP
DS
CH
sP
DS
CH
sP
DS
CH
sP
DC
CH
sTTI sTTI sTTI sTTI
sP
DC
CH
sP
DC
CH
sP
DC
CH
Unicast PDSCH
D
ow
nlin
k s
yste
m b
an
dw
idth
1 subframe
1st slot 2nd slot
PD
CC
H
sP
DS
CH
sP
DC
CH
sTTI
1 subframe
1st slot 2nd slot
PD
CC
H
sP
DS
CH
sP
DC
CH
sTTI
Unicast PDSCH
(a) Cases corresponding to Alt 1. (b) Cases corresponding to Alt 2 and Alt 3.[4]
13
• For ease of scheduling and multiplexing with
legacy UE data, aligning sTTI at subframe
boundary is desirable
• Can sTTI length change within a subframe?
• More flexibility
• Higher complexity/signaling overhead
Boundary Alignment
14
• Inform UE there is a DL data scheduled for it
and where/how to receive the data
• For efficiency, PDCCH is designed in a
somewhat convoluted way
• Variable length
• Blind decoding
• E-PDCCH
PDCCH in LTE
[2]
15
• Reuse PDCCH for sTTI?
• Contradicts the intention to have a lower
latency (more granular scheduling
decision)
• A new (shortened) PDCCH has to be
designed and placed within each sTTI
sPDCCH for sTTI
[6]
16
• Span how many symbols?
• Unified or separate design for different
sTTI lengths?
• Same number of blind decoding in every
sPDCCH?
• Can the first sTTI be scheduled by the legacy
PDCCH?
sPDCCH for sTTI
sPD
CC
H
PD
CC
H for
2-sy
mbo
l sT
TI
sPD
CC
H sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H sPD
CC
H
PD
CC
H for
2-sy
mbo
l sT
TI
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
[7]
17
• HARQ timing is 4 ms in legacy LTE systems
• Processing capability
• Its advancement is what motivated this
feature
• Timing advance
• Determined by the supported cell size
HARQ Timing
18
• TA is determined by distance (cell size), and
does not scale with the length of TTI.
HARQ Timing
[2]
19
• UL coverage is limited by UE tx power
• TTI bundling is supported in LTE
• Allow combinations of a longer UL sTTI than
DL sTTI?
• Complicates system design and HARQ
Asymmetric DL/UL sTTI
2-symbol sTTI
DL
UL
7-symbol sTTI
Minimum processing time of n+k, where k=8
2-symbol sTTI
DL
UL
7-symbol sTTI
Minimum processing time of n+k, where k=4
[8]
20
• 8-layer MIMO transmission is supported in
LTE-A
• Extra control information
• More reference signal (RS)
• The added control information and RS
overhead eats into the available time frequency
resource for data transmission
Multi-layer Transmission
1 P
RB
subframe
2-symbol TTI
1 P
RB
subframe
1 P
RB
subframe
Subframe TTI
1 P
RB
subframe
Slot TTI
1 P
RB
subframe
3/4-symbol TTI 1-symbol TTI
1 P
RB
subframe
2-symbol TTI
1 P
RB
subframe
1 P
RB
subframe
Subframe TTI
1 P
RB
subframe
Slot TTI
1 P
RB
subframe
3/4-symbol TTI 1-symbol TTI
[10]
21
Specification
22
• No sTTI spans over subframe boundary
• Support DL/UL sTTI combinations
{2,2} {7,7} {2,7}
sTTI Arrangement
[5]
23
• Configurable bandwidth
• 1~2 OFDM symbols (configurable) in each
sTTI
• sDCI2 is still under discussion
sPDCCH
sPD
CC
H
PD
CC
H for
2-sy
mbo
l sT
TI
sPD
CC
H sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H sPD
CC
H
PD
CC
H for
2-sy
mbo
l sT
TI
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
sPD
CC
H
[7]
24
• For DL/UL sTTI 7 OFDM symbols, n+4 assuming a max TA of 0.33 ms is
supported
• For DL/UL sTTI 2 OFDM symbols, support
• n+6 assuming a maximum TA of 0.33 ms
• n+4 assuming a maximum TA of 0.067 ms
• For 1 ms TTI, n+3 assuming a maximum TA of 0.33 ms is supported
HARQ Timing
25
• 4-layer MIMO transmission with single codeword is supported for sTTI 2 OFDM
symbols
• 4-layer MIMO transmission with single codeword is supported for sTTI 7 OFDM
symbols
Multi-layer Transmission
26
Concluding Remarks
• Packet data latency, along with data rate, are the most important performance
metrics
• Before 3GPP Rel-14, enhancements on LTE/LTE-A systems are mostly on
increasing data rates (300 Mbps -> 4 Gbps -> 25 Gbps)
• Support of sTTI can potentially reduce latency to around 1/7th
The specification of sTTI is still in progress.
• The support of different TTI values is also more aligned with the design principle of
5G NR systems.
28
References
[1] “Physical channels and modulation,” 3GPP TS 36.211, V14.0.0, 2016-09.
[2] 4G: LTE/LTE-Advanced for Mobile Broadband, 2nd ed., Erik Dahlman, Stefan Parkvall, Johan Skold
[3] Draft Report of 3GPP TSG RAN WG1 meeting #86b.
[4] R1-164795, “Discussion on sPDSCH,” Samsung
[5] R1-1613691, “Way Forward on 2-symbols DL sTTI layout structure”
[6] R1-164068, “The co-existence consideration of legacy TTI and sTTI in one carrier ,” Huawei
[7] R1-166661, “Discussion on sPDCCH for shortened TTI ,” Sony
[8] R1-1609871, “Association timing for sTTI with different UL and DL lengths,” Sharp
[9] ITU-R, IMT Vision e Framework and Overall Objectives of the Future Development of IMT for 2020 and beyond, Recommendation ITU-R
M.2083, September 2015.
[10] R1-162706, “Downlink physical channel design for TTI shortening,” Samsung