1. IEEE 802.16 Presentation Submission Template (Rev. 9) Document Number:
Detaining and dropping signaling for HARQ feedback in relaxing necessary HARQ soft buffer requirement
IEEE Presentation Submission Template (Rev. 9)
Document Number:
IEEE S802.16m-08/1415r3
Date Submitted:
Source:
Zheng Yan-Xiu, Ren-Jr Chen, Chang-Lan Tsai, Chung-Lien Ho, Richard Li, ITRI
 Yih-Shen Chen, Kelvin Chou, I-Kang Fu and Paul Cheng, MTK
Tsuguhide Aoki, Yong Sun, Toshiba
Xin Qi, Jianfeng Kang, Shaohua Li, Chao Wei, Nokia Siemens Networks
Zexian Li, Nokia
Venue:
Re: TGm SDD: Other (reply to IEEE m-08/040 Call for Comments and Contributions on Project m SDD
TGm SDD, HARQ)
Base Contribution:
C80216m-08/1415
Purpose:
To be discussed and approval by IEEE m TG
Notice:
This document does not represent the agreed views of the IEEE Working Group or any of its subgroups. It represents only the views of the participants listed in the “Source(s)” field above. It is offered as a basis for discussion. It is not binding on the contributor(s), who 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
Patent Policy:
The contributor is familiar with the IEEE-SA Patent Policy and Procedures:
< and <
Further information is located at < and < >.

2. Background
In IEEE m-08/003r4, “The Draft IEEE m System Description Document”
10.2 HARQ Functions
The HARQ operation relies on N-process (multi-channel) stop-and-wait protocol
N-process requires N soft buffers or enough shared buffer to stored received samples and this has been specified in IEEE Std e-2005
Basic scheduling rule: In order to avoid the risk of buffer overflow, maximum transmitted coded bits can not exceed the buffer in an MS.
HARQ soft buffer might dominate MS baseband cost due to large soft buffer necessary for an MS to support high throughput.

3. Necessary Soft Buffer for N-process HARQ
N-process HARQ throughput:
Necessary buffer:
The necessary buffer increases with system throughput SMbps.
The buffer also increases with the round trip delay Tms.
Higher code rate Rc leads less storage due to coded bits stored in the soft buffers
Higher modulation order Morder also reduces memory storage if symbol-level storage is considered
Bit-level buffer:
Symbol-level buffer:
When HARQ throughput becomes large, the necessary buffer may be incredible large
IEEE e turbo decoder requires 2,880 soft bits for received samples and extrinsic information
HARQ soft buffer dominates baseband complexity
Bit-Level Buffer
SMbps=100Mbps
Tms=5ms
SMbps=1000Mbps
Rc=1/2, Morder=1
1000K [Soft Bits]
10M [Soft Bits]
Rc=4/5, Morder=1
625K [Soft Bits]
6.25M [Soft Bits]
Symbol-Level Buffer
SMbps=100Mbps
Tms=5ms
SMbps=1000Mbps
Rc=1/2, Morder=2 (QPSK, One retransmission)
500K [Symbols]
5M [Symbols]
Rc=1/2, Morder=4 (16QAM, One retransmission)
250K [Symbols]
2.5M [Symbols]
Rc=1/2, Morder=6 (64QAM, One retransmission)
167K [Symbols]
1.67M [Symbols]

4. FACT: Low Utilization of Soft Buffer
H-ARQ operation scenario:
Packet error rate=0.1
Buffer usage is geometrically distributed given N-processes HARQ
In most cases, half of buffers are vacant
For example: 6-channel HARQ buffer usage is shown in below able and three soft buffers are enough for 99.3% occasions
FACT:
At least half soft buffers can be reduced is expected
Existing buffer design strategy is for the worst case
Significant complexity reduction can be achieved by implementing less soft buffer
An example of the occurrence of number of HARQ buffers in use for 6-process HARQ
Number of H-ARQ buffer in use 1 2 3 4 5 6
Occurrence
0.53
0.35
0.098
0.015
0.0012
5.4×10-5
10-6

5. Solution: HARQ Buffer Management for 802.16m
Aggressive HARQ transmission
Enhance HARQ throughput with less soft buffer
MS-assisted HARQ flow control
Avoid HARQ failure due to buffer overflow
HARQ buffer synchronization
Acknowledge BS MS buffer state by an extra HARQ feedback
Clean up the obsolete soft buffer while necessary (C802.16m-08/1328)

6. MS-assisted HARQ Flow Control (1/2)
Concept: MS feedbacks current buffer status to BS to facilitate HARQ scheduling
Dropping signal (RESTART): to inform BS the occurrence of soft buffer overflow
Dropping signal can further reinitiate HARQ process if DL allocation signal might be missed by MS, or MS ACK/NACK signal is missed in UL
Detaining signal (DET): to inform BS the congestion status of soft buffer

7. MS-assisted HARQ Flow Control (2/2)
Scenario 1: If BS can real-time schedule the HARQ processes
BS traces HARQ processes by the number of received NACK
BS halts partial HARQ processes when MS faces buffer busy due to receiving large number of NACKs
If all packets are not correctly received, buffer overflow occurs. MS sends dropping (RESTART) signaling to notify BS which HARQ process is dropped in MS side. BS can reinitiate the dropped HARQ process
Scenario 2: If BS can not real-time schedule the HARQ processes
MS sends detaining (DET) signaling to halt partial HARQ processes when MS faces buffer busy
MS restarts ACK if buffer busy scenario disappears
If all packets are not correctly received, buffer overflow stills occur. MS sends dropping (RESTART) signaling to notify BS which HARQ process is dropped in MS side. BS can reinitiate the dropped HARQ process.
Do you think “on the fly” is a better term than “real-time”?
- if BS can schedule HARQ processes on the fly
- if BS can’t schedule HARQ process on the fly

8. An Example for N-Process HARQ Flow Control
N-processes HARQ with soft buffers capable of storing coded bits of R HARQ processes
Postponing finished HARQ processes when most soft buffers are in use
Extra S soft buffers are introduced to avoid buffer outage and the number of S is proprietary and can be 0.
Examples:
6-processes H-ARQ provides extra 2 transmission chances
R=4 memory buffers
S=1 of R=4 memory buffers is used to avoid buffer outage
As R-S=3 buffers are in use, CRC passed HARQ processes are detained
Can we put this page to Appendix?

9. Examples 4 H-ARQ soft buffers
6-channel H-ARQ is used for the proposed method
IEEE e
MPDU=2048 bytes=16,384 bits, Tms=20ms, Rc=0.5, Bbits=131,072 soft bits=4*MPDU/Rc,
Three mechanisms:
4-channel H-ARQ mechanism
6 channel H-ARQ mechanism with detaining signaling, R=4 and S=1
Packet Error Rate=0.1
IEEE e
4 H-ARQ processes
Throughput= 2.95Mbps
6 H-ARQ processes with detaining signaling and R=4, S=1
Throughput= 4.36Mbps
Throughput gain=47.8%
Buffer saving under the same throughput=
32.39%
Can we put this page to Appendix?

10. Feedback Channel Design and Overhead
Main stream design: 12 orthogonal sequences for 6 for ACK and 6 for NACK
One exemplary design: 12 orthogonal sequences for 4 ACK and 4 NACK and 4 RESTART
Compatible with existing popular design with decreased throughput
Overhead
Applying this new feedback when aggressive HARQ transmission is used
DL resource occupied by only less number of long HARQ bursts
HARQ feedback channel is almost vacant
Since the HARQ feedback channel is almost vacant, overhead is not an issue
Transmitting larger packet (larger than 50Kbits) with more overhead is also reasonable comparing with VoIP packet

11. Conclusions
Aggressive HARQ transmission can enhance the MS throughput with less soft buffer and improves user experience
Aggressive HARQ transmission can significantly reduce 16m system cost
The proposed HARQ feedback alleviates buffer overflow issue due to less soft buffers
Dropping and detaining feedbacks are introduced for HARQ flow control in case of buffer overflow
Dropping signaling can further reinitiate HARQ process if DL allocation signal might be missed by MS, or MS ACK/NACK signal is missed
If a BS can on-fly trace buffer usage and real-time schedule, the detaining signaling is unnecessary

12. Proposed text Option 1 11.x.1.2.6 HARQ feedback Option 2
A multi-state HARQ feedback channel is used in 16m. Basic ACK/NACK HARQ feedbacks are supported. The number of state is FFS. Dropping HARQ feedback (RESTART) is used to avoid buffer outage when aggressive HARQ transmission is used. It also restarts HARQ process when control signaling is missed.
Detaining HARQ feedback (DET) regulates the buffer usage and is FFS when aggressive HARQ transmission is used and BS can not real time schedule HARQ processes.
Option 2
11.x HARQ feedback
A multi-state HARQ feedback channel is used in 16m. Basic ACK/NACK HARQ feedbacks are supported. The number of state is FFS. Dropping HARQ feedback (RESTART) is used to avoid buffer outage when aggressive HARQ transmission is used. It also restarts HARQ process when control signaling is missed.

13. Example of HARQ Flow Chart in Scenario 1
Objective:
Avoiding buffer outage introduced HARQ failures
Safely reducing necessary soft buffer
Solution:
Dropping and reinitiating the HARQ process as no soft buffer for the unfinished HARQ process
Postponing the finished H-ARQ processes when unfinished HARQ processes occupy all soft buffers

14. Example of HARQ Flow Chart in Scenario 2
When most HARQ processes are incorrect received concurrently, buffer outage still occurs.
When the case occurs, dropping HARQ feedback indicates the process is dropped to assist BS to reschedule
Extra Advantages:
The successful transmission would not be dropped
The failed transmission could be safely reinitiated
ARQ-introduced latency is further reduced
Dropping signaling can further reinitiate HARQ process if DL allocation signal might be missed by MS, or MS ACK/NACK signal is missed

15. Controlling Signaling Errors and Response
ACK => RESTART : BS/MS restarts the process in the next transmission (no significant influence)
ACK => NACK : send ACK at the next transmission (same as conventional HARQ feedback)
NACK => RESTART : BS/MS restarts the process in the next transmission (no significant influence)
NACK => ACK : apply ARQ mechanism to restore the lost packet (same as conventional HARQ feedback)
RESTART => ACK : apply ARQ mechanism to restore the lost packet (same as conventional HARQ feedback)
RESTART => NACK : Send RESTART in the next transmission to reinitiate the process (no significant influence)