Академический Документы
Профессиональный Документы
Культура Документы
1xEVDV, An Overview
Vivek P. Mhatre
04/13/04
Talk for EE 647
School of Electrical and Computer Engineering
1
Presentation Outline
Objective: To provide an overview of networking-related aspects of
CDMA Data networks
• CDMA Basics
• CDMA-HDR, i.e., IS-856 or 1xEVDO
– Forward link: Rate Control
– Reverse link: Power Control
• 1xEVDV
• Recent Work
• Research Challenges
−−++
−+−+
++++
3
CDMA Basics (contd.)
• Perfect orthogonality not possible due to channel variations
• Interference limited capacity
PG
γi = Pi i
η + j6=i θPj Gj
Ri = W log(1 + φγi )
• All about Power control and Rate control!!!
• Reverse link power control to counter “near-far” effect
BS
N F
4
IS-95
• First CDMA system
• Mainly for voice
• One CDMA code per user on forward and reverse link
• Forward and reverse links separated in frequency
• Limited data rate (64 Kbps)
5
1xEVDO or IS-856 or CDMA-HDR
• 1xEVDO stands for 1xRTT (Radio Transmission Technology)
EVolution to Data Only
• Qualcomm Inc. was the driving force
• Need to get ahead of other players ⇒ Quick deployment
• Philosophy:
– Do not tamper with the spectrum allocated to voice traffic of
IS-95
– Use separate spectrum for data traffic
– Use CDMA for data communication as well
– Forward and reverse links separated in frequency
• Resources available to the base station on the forward link
– A set of orthogonal codes (64 per sector)
– Base station transmit power (PT )
6
1xEVDO (contd.)
• FLa as perceived by each user is time-varying
• Question: For a given user, to achieve the same average rate, R̄,
should the BS use power control or rate control on FL?
• Assume linear model for rate vs SINR
P0 g R̄ 1
Rate Control: R=K ⇒ P0 = g
I KE I
R̄ I R̄ I
Power Control: P= ⇒ P̄ = E
Kg K g
P̄
⇒ = E[X]E[1/X] ≥ 1
P0
P̄ ≥ P0 ⇒ Rate Control better than power control
a FL = Forward Link, RL = Reverse Link
7
1xEVDO (contd.)
• In reality, R varies as log(SINR), not linearly
• Answer: For practical CDMA network settings (noise levels, path
losses etc.) P̄ /P0 ≥ 1
• Hence CDMA systems use a bunch of modulation schemes to adapt
rate to the current channel state
• Conclusion: Rate control is better than power control on FL
• Use channel state feedback about FL from each user for rate control
8
Channel State Feedback: FL
Feedback about FL: FL is rate controlled
• Inner Loop Rate Control
– FL pilot signal sent by BS, monitored by all active data users
– DRC (Data Rate Control) channel on RL (one code per active
user) used by each user to indicate supportable FL rate and best
sector
– Coarse adjustments of rate control
• Outer Loop Rate Control
– During actual packet receptions, measure PER, and request
increase in data rate if P ERmeas < P ERtarget and vice-versa
– Fine adjustments of rate control
9
Channel State Feedback: RL
Unlike FL, RL is power controlled to counter the “near-far” effect
• Coarse power control
– RA (Reverse Activity) channel of FL used by BS to request user
to ↑ / ↓ its rate depending on the total interference at the BS
– Objective: Keep the total interference level (load) at BS below a
threshold T
∗ Maximum user transmit power, link budget analysis ⇒ cell
coverage area
∗ Fixed margin is alloted to in-cell interference
∗ Hence must ensure that intra-cell interference is below a
threshold
– Slow, time averaged decisions over 128 slots
• On RL, power control dictates rate control
10
Channel State Feedback: RL (contd.)
• Fine power control
– RL pilot signal sent by user on RL, monitored by BS
– RPC (Reverse Power Control) channel of FL used by BS to
request user to ↑ / ↓ its power level (target PER 1%)
– Neighboring BSs also have a say
– Fast, every slot
11
Reverse Link Structure
RL
13
1xEVDO FL in details
• Question: How to serve multiple data users on FL?
• Answer: Opportunistic scheduling
• Philosophy: Serve the user with the best channel
• Since channel is time-varying for everyone, some user will be in
good state with high probability
• Since we only serve “good users”, system throughput improves
• Everyone will be in good state some time or the other, so long term
fairness issue solved
• Short term fairness can also be dealt with through appropriate
modifications
14
FL Opportunistic Scheduling, A Toy
Problem
Total bandwidth B User A User B
Channel state Good, p 0.8 0.8
Channel state Bad, q 0.2 0.2
• Conclusion: Choose the best user, and serve him at full rate
• Gains of opportunistic scheduling are better with more users, since
higher probability of finding a “good user”
• In reality, channel (and hence the sustainable rate) varies over a
range, not just “good” or “bad”
Ri = W log(1 + φγi )
16
Opportunistic Scheduling (contd.)
• Hence in CDMA-HDR, on FL, BS serves the “best” user with
maximum power and maximum bandwidth (use all 64 codes)
• Essentially, FL is TDM!!
• No need to account for fading power margin as in IS-95, since only
one data user served at a time, no voice users
Control Channel
MAC Channel
Pilot Channel
Total Traffic
Total Traffic
Sync Channel
Paging Channel
Pilot Channel
17
Performance of 1xEVDO
• FL maximum bandwidth = 2.5 Mbps
• RL maximum bandwidth = 180 Kbps per user
• Maximum number of data users = 16
• Soft handoff of data users possible
EOF
18
1xEVDV
1xEVDO 1xEVDV
19
1xEVDV (contd.)
Operating point
for 1xEVDV
MAX MAX
Overhead channels Actualdata
FL BS transmit power
(data)
FL BS transmit power
Channels
Dynamic
Operating
Point
Voice calls
Overhead channels
(static)
MAX MAX
Walsh codes Walsh codes
20
1xEVDV (contd.)
• No soft handoff for data users
• Opportunistic scheduling on FL
• One or Two users served on FL simultaneously
• Improved ARQ
• Improved RL rate 1.5 Mbps
• Details to come out soon in Revision D
EOF
21
Recent Work
Opportunistic scheduling over FL
• Qualcomm’s HDR algorithm
– Single user chosen for FL serving
– Maximize system utility over each time slot
• Utility function based opportunistic scheduling, [Xin Liu, Chong and
Shroff]
– Single user chosen for FL serving
– Resource sharing constraints
– A Utility function associated with each user
– Maximize long term system utility
22
Recent Work (contd.)
Opportunistic scheduling over FL (contd.)
• Opportunistic power scheduling for multiple-server systems,
[Jang-Won Lee, Mazumdar and Shroff]
– Maximize long term system utility
– Determine power allocation to users
– Multiple users could be chosen for FL service
• Opportunistic scheduling over multiple interfaces, [Kulkarni and
Rosenberg]
– Multiple interfaces correspond to e.g. 802.11 interface, 3G link,
satellite link etc.
– Joint scheduling to choose the best user for each interface
23
Recent Work (contd.)
RL power control and scheduling
• Power constrained RL scheduling [Kumaran and Qian]
– Some user terminals constrained by maximum transmit power
– Two classes of users: strong and weak
– Original RL power control aims to solve near-far (strong-weak)
problem
– Conclusion: Weak users scheduled simultaneously, and strong
users one at a time
EOF
24
Research Challenges
My viewpoint:
• Inter-cell interference issues
– Law of large numbers on RL
– Law of small numbers on FL
– Co-ordination between multiple BSs difficult
• 1xEVDV, several research challenges
– Resource allocation is difficult since voice and data integrated
– “Movable boundary” problem in two dimensions: bandwidth
(codes) and power
– Hard SINR requirements for voice ⇒ power margin to overcome
deep fades (IS-95 like)
• New technology brings new problems with it ⇒ No dearth of
problems!
25
Acknowledgments
• Sunil Kulkarni and Aravind Iyer for useful discussions in MSEE 318
• Sunil Kulkarni for providing several pointers, in particular 1xEVDV
tutorial by Ericsson in Globecomm 2004 and [Kumaran and Qian]
paper
26
References
• Qualcomm Inc. website for white papers (1xEVDO)
• An Algorithm for Reverse Traffic Channel Rate Control for cdma2000 High
Rate Packet Data Systems, Chakravarty et al. (Qualcomm Inc.)
27