Sie sind auf Seite 1von 44

Informal Training on VzW/VF LTE

Phase3 DL Metrics

Alex Anokye

The downlink physical layer parameters

2 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

DL Frame Structure
Picture below shows a pair of Physical Resource Blocks (PRB) (1 PRB = 0.5 ms, 2 PRB per TTI=1ms)
In frequency domain, there are 50 PRBs in a 10 MHz LTE channel
PRB in frequency domain: 12 carriers = 180 kHz
PRB in time domain: 7 OFDM symbols = 0.5 ms

One antenna poert

R0

R0

R0

R0

R0

R0
l 6

l 6l 0

Time
R0

R0

R0

Resource element k , l

R0

R1

R0

R0

l0

3 | Presentation Title | Month 2006

R0

R0

l 0

Two antenna poert

Frequency

In current implementation, Scheduler grants a minimum of 3 (or 2) PRBs per user (more on that later)

R1

R0

R1

R0
l 6l 0

R1

R1

R1
l6

l0

Not used for transmission on this antenna port

R1

Reference symbol on this antenna port

R1
l 6l 0

l 6

All Rights Reserved Alcatel-Lucent 2006, #####

OFDM overview
Lets try to have a better understanding of the physical layer parameters and
frame structure with an overview of OFDM

4 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Whats OFDM?
OFDM stands for Orthogonal Frequency Division Multiplex
Any two carriers with an integer number of cycles during the symbol time are
orthogonal to each other
This plot illustrates OFDM baseband carriers in time domain:

Carrier #1

Carrier #2

Carrier #3

OFDM Symbol Time


5 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Micros oft Excel


Works heet

Whats OFDM? (cont)


Try it by yourself:
Open the attached Excel file

Change the frequency (carrier #), and play with the phase and the
amplitude of the two carriers (columns C and D)
Note that, independent of the phase and amplitude you select, the two
carriers remain orthogonal (integration of product equal to zero: in cell M3)
Since the carriers are orthogonal we can transmit information simultaneously in
the different carriers by changing their phase and/or the amplitude, and we will
be able to recover the information in the reception
The phase and amplitude are defined by the selected modulation scheme and
transmitted information bits. UMB supports up to 64QAM, which can transmit 6
bits per OFDM symbol time, so it can pump data 6 times faster than BPSK

!
BPSK
6 | Presentation Title | Month 2006

QPSK

8PSK
All Rights Reserved Alcatel-Lucent 2006, #####

16QAM

64QAM

FFT & DSPs make OFDM Practical


Fast Fourier Transform mathematical algorithms (see
http://en.wikipedia.org/wiki/FFT), and current state of the art in Digital
Signal Processing make OFDM practical
FFT algorithms can reduce significantly the amount of operations required to
modulate/demodulate OFDM signals
Still, for a large number of carriers (e.g. 512) a very large number of real time
computations need to be performed every OFDM symbol time (e.g. 110
microseconds). This can be accomplished with DSP integrated circuits

7 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Why not just modulate a single carrier?


A single carrier requires a very small symbol time in order to transmit at high speed data
rates
A small symbol time requires complex equalization and/or spreading and a rake
receivers to handle Inter Symbol Interference (ISI) caused by multi-path propagation
By using multiple carriers we can make the symbol time of each carrier large enough.
This way the individual carriers transmit at low speed data rate and so are robust to ISI,
and the aggregate of all carriers still give us a high speed data pipe
With low speed data rates in the sub-carriers, we dont need complex equalization or
spreading/rake receiver
Still we can not eliminate the ISI 100% (see next page)

8 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

ISI
Even though the individual sub-carriers transmit at relatively low symbol
rates, there is still some amount of ISI, as illustrated below

OFDM Symbol Time. Example: 113 microseconds

Multi-path delay.
Example: 6
microseconds

Inter symbol
interference. Here,
the delayed multipath causes
interference in the
adjacent symbol time

9 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

The solution for ISI: Cyclic prefix

Add a cyclic prefix

Now there is no ISI within the OFDM symbol time

Note that, even though we can eliminate ISI with the cyclic prefix, we still have multi-path copy of the OFDM symbol within the OFDM symbol
time
Depending on its phase, a multi-path copy can interfere constructively or destructively
The net effect is that for a large number of sub-carriers (e.g. 512), a number of them will be impacted destructively by multi-path
propagation effects, however we are left with enough good carriers, and by making use of coding, interleaving and frequency hopping
schemes we can live with the fact that some sub-carriers are severely attenuated by multi-path propagation

10 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Exercise
The 10 MHz LTE channel is implemented with a 1024 sample FFT. What would be the sampling rate if no cyclic prefix
was used?
The cyclic prefix in our current implementation is 80 samples in the first OFDM symbol and 72 samples in the
remaining 6 symbols of a slot. How much bandwidth penalty this represents?
A reflected signal reaches the UE antenna as shown in the picture below. Is the cyclic prefix in the first OFDM symbol
capable of canceling the ISI? How about the cyclic prefix in the remaining 6 symbols of a slot?
What could be the reason why the first OFDM symbol is protected with a larger cyclic prefix than the remaining 6
symbols of a slot?
How many OFDM sub-carriers are there in our 10 MHz implementation? For a FFT of 1024, how many zeros are padded
in the FFT?

5 km
3.5 km
UE
11 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

DL Scheduler Overview

LTE Trial Team


March 6th, 2009

DL Scheduler Functions
DL Scheduler Functions
UE1 packets

UE2 packets

UE1 buffer status

DL Link Adaptation
Rank Adaptation
DL MAC Processing

UE2 buffer status

CQI processing

Fairness
UEn buffer status

DL TBs

UE1

CQIs
DL TBs

Default Frequency
Selective Scheduler

DL RB assignment
UEn packets

CQIs

UE2
CQIs
DL TBs

DL grant formatting

DL HARQ processing
OCNS

UEn

13 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

LA0.1 Maximum PDSCH Resource Block Utilization


47.2 Resource Blocks are available for PDSCH in LA0.1 solution
Therefore maximum utilization of 94.4%
SFs 1,2,3,4,7,8 and
9
70%

SFs 0 and 5
PDSCH/SCH sharing
SF 5 every 80 ms
reserved for
D-BCH

18.75%

PDSCH/RACH
Msg 2 sharing
10%

1.25%

14 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

DL Link Adaptation Maximum RB Utilization and Maximum Throughput


The number of RBs per frame (10 TTIs) available for data:
41 RBs x 2 TTIs (9 RBs of Synchronization info in TTIs 0 and 5)
45 RBs x 1 TTI (5 RBs reserved for RACH message 2)
50 RBs x 7 TTIs (7 TTIs without any overhead)
Max RB Utilization = (2*41 + 1*45 + 7*50)/10 = 47.7 RBs (95%)

Max Throughput = (2*27376 + 1*30000 + 7*33368)/10 = 31.8 Mbps


Max Throughput = 63.6 Mbps in MIMO mode
Note: actual max values are a bit lower due to overheads for TA messages and D-BCH
15 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

CFI1 and CFI2 PDCCH Bandwidth


8 REs per RB
400 REs per TTI
CFI1

288 REs used for PDCCH


1 Mini-CCE = 4 REs
1 CCE = 9 mini-CCEs
288/(4*9) = 8 CCEs per TTI
20 REs per RB
1000 REs per TTI

CFI2

900 REs used for PDCCH


1 Mini-CCE = 4 REs
1 CCE = 9 mini-CCEs
900/(4*9) = 25 CCEs per TTI

16 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

PDCCH Format 2 / CFI 2 (Selected for Scheduler Test Cases)


The L1/L2 DL Control Channels RE are grouped 4 by 4
into mini-CCE, themselves distributed among the
PCFICH, PHICH and PDCCH
For PDCCH, the mini-CCE are then grouped into CCE,
one CCE being composed of 9 mini-CCE, thus 1 CCE is
composed of 36 RE, themselves conveying 72 coded bits
(QPSK is used over PDCCH).
CFI2 Supports 3 DL grants and 3 UL grants per TTI
DL-1 every 80 ms due to D-BCH, UL-1 every 50 ms
due to UL PC
CFI2 supports 25 CCE
Each DCI 0/1/1A/2/3 transmission requires 4 CCEs
288 coded bits
64 DCI2 bits (Code Rate = 0.2222)
50 DCI1 bits (Code Rate = 0.1736)
40 DCI0/1A bits (Code Rate = 0.1528)
17 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Rank Adaptation
Original Rank Adaptation (RA) tunable parameters
thresholdSinrMimoNotMimo: Signal to noise ratio threshold for switching mode [7 dB]
rankAveragingCoefficient: Forgetting factor for rank switching. No MIMO if output of IIR
filter = rank 1 [13 -> 0.95]
mimoSinrSwitchAveragingCoefficient: Forgetting factor for SINR. No MIMO if output of IIR
filter is lower than thresholdSinrMimoNotMimo [13 -> 0.95]
New RA tunable parameters (added as part of MIMO optimization effort)
txdiv_sinrOffset: SINR offset applied to rank2 CQI when eNB grants rank1 while UE reports
rank2 [6 dB]
mimo_sinrOffset: SINR offset applied to rank1 CQI when eNB grants rank2 while UE
reports rank1 [-10 dB]
mimoblerth: rank 2 BLER threshold above which RA switches to rank 1 [100%]

18 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

CQI Processing Rank 1 CQI Format

MCS selection is based on both, WB and NB CQI values


Default Scheduler configuration is frequency selective scheduler
The Subband CQI values are sent as a differential on the WB CQI. Differential values are
from the following set: {-2,0,1,2}

19 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

CQI Processing Rank 2 CQI Format

CW0 and CW1 MCS selection is based on both, WB and NB CQI values

20 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Sub-Band to RB Mapping
In 10 MHz bandwidth, 8 sub-bands are reported, each sub-band representing
the CQI for 6 RBs (except the last one which measures the CQI over 8 RBs).

21 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

DL Resource Block Groups


17 addressable RBG are defined as follows:

Note: RBG16 mapped to


only 2 PRBs. All other
RBGs mapped to 3 PRBs.

Sample bit map RBG assignments


Single User
11111111111111111 = 50 RBs
11111110001111111 = 41 RBs (in the case of TTIs 0 or 5)

Two users
10101010101010101 = user1 (26 RBs)
01010101010101010 = user2 (24 RBs)

Three users
10010010010010010 = user1 (18 RBs)
01001001001001001 = user2 (17 RBs)
00100100100100100 = user3 (15 RBs)

22 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Alpha-Fairness Scheduler
The Alpha-Fairness scheduler is an extension of the Proportional Fair scheduler:

Find the rate allocation r1 t ,...,rn t per TTI which


maximize

Users

i1

w i t
ri t ,

Ri

ri t is the instantaneous rate


R iis the average rate
w i t is a QoS weight

is the alpha fairness factor


Test cases are based on values for of 0, 1, and 2, where
Aggressive (maximum C/I scheduler) more resources are assigned to users with
favorable RF conditions
Proportional Fair (P-F) all users are assigned equal resources regardless of RF conditions
Conservative more resources are assigned to users with unfavorable RF conditions

23 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Preliminary Four User Scheduler Test Proportional-Fair (CE2, MC1, MC4, NC2)

24 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Preliminary Four User Scheduler Test Proportional-Fair (CE1, MC3, MC3, NC1)
Comparative Single User Test Case:

25 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

DL Link Adaptation Transport Block Sizes (in bits) 1 to 25 RBs

26 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

DL Link Adaptation Transport Block Sizes (in bits) 26 to 50 RBs

27 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

LGE SINR Measurement Scheme


Impact on DL Test Cases
01/09/09

LTE Trial Team

Objective
Purpose of the presentation is to
Explain UE approach to measuring SINR
Understand potential limitations for loaded and unloaded test cases
Propose methodology to ensure SINR measure and CQI reports properly
reflect PDSCH loading for all test cases

29 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

LGE SINR Measurement Scheme

The LGE UE measures SINR


in the 200 RS resource
elements of the first OFDM
Symbol

30 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Six Reference Signal Frequency Shifts

31 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Cases of no RS Interference from a Time Aligned Neighbor Cell

Serving Cell

Neighbor Cell

OFDM
Symbol 0

OFDM
Symbol 0

RS Freq.
Shift 0

RS Freq.
Shift 1

32 | Presentation Title | Month 2006

Serving Cell
RS of neighbor
cells with one
and two
frequency
shifts away do
not interfere
with RS of
serving cell

Neighbor Cell

OFDM
Symbol 0

OFDM
Symbol 0

RS Freq.
Shift 0

RS Freq.
Shift 2

All Rights Reserved Alcatel-Lucent 2006, #####

Cases of Full RS Interference from a Time Aligned Neighbor Cell

Serving Cell

Neighbor Cell

OFDM
Symbol 0

OFDM
Symbol 0

RS Freq.
Shift 0

RS Freq.
Shift 3

33 | Presentation Title | Month 2006

Serving Cell
RS of neighbor
cells with three
and six
frequency
shifts away
causes full
interference on
RS of serving
cell

Neighbor Cell

OFDM
Symbol 0

OFDM
Symbol 0

RS Freq.
Shift 0

RS Freq.
Shift 6

All Rights Reserved Alcatel-Lucent 2006, #####

Non Time Aligned Neighbor Cell with 100% OCNS Load


A non time aligned neighbor
cell with 100% OCNS load
causes full interference on
RS in first OFDM symbol of
Serving
serving cell
Cell

Note that in this case not


all REs in the first OFDM
symbol of the neighbor cell
interferes with the PDSCH
of the serving cell.
Neighbor
Cell
Even if all neighbor cells
were in this condition the
SNR error would be less
than 0.2 dB.

34 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Time Aligned Neighbor Cell with Same RS Frequency Shift


A time aligned neighbor cell
with same RS frequency
shift causes full
interference on RS in first
OFDM symbol of serving cell

35 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Proposed Approach for 100% Load Test Cases


Use same RS frequency shift in all cells
Load all neighbor cells with 100% OCNS load
By using same RS frequency shift and 100% OCNS load in neighbor cells we can
force LGE UE to report SINR/CQI equivalent to actual 100% loading
independent of the cells time alignment

36 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

PCFICH Resource Element Mapping


PCFICH is transmitted in 16 resource elements in the first OFDM symbol
4 groups of 4 REs
This picture illustrates mapping of a single PCFICH RE group for different RS
frequency shifts
The 4 PCFICH RE group are well separated in frequency to obtain good
diversity

37 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Cases of PCFICH Interference on First OFDM Symbol RS in Time Aligned Cells

Serving Cell

Neighbor Cell

OFDM
Symbol 0

OFDM
Symbol 0

RS Freq.
Shift 0

RS Freq.
Shift 1

38 | Presentation Title | Month 2006

Serving Cell
RBs carrying a
PCFICH RE
group in a
neighbor cell
interferes with
two RS of the
serving cell

Neighbor Cell

OFDM
Symbol 0

OFDM
Symbol 0

RS Freq.
Shift 0

RS Freq.
Shift 2

All Rights Reserved Alcatel-Lucent 2006, #####

LGE SINR Reporting for Interference Generated by Time Aligned Cell with One or
Two RS Frequency Shifts Away
Assume no PDCCH loading on the neighbor cell
8 out of 200 RS in the first OFDM of the serving cell receive interference
from the PCFICH of the neighbor cell
With per RE interference level equal to signal level, and assuming PCFICH
power boosted by 3 dB in relation to RS power
LGE SINR = 10*Log(200/8) - 3 dB = 11 dB
This SNR measurement is equivalent to PDSCH SINR in unloaded system
when per RE interference level is equal to signal level see next slide

39 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Actual Interference from Unloaded Time Aligned Neighbor Cells


12 RS of the neighbor cell interfere with PDSCH REs of the serving cell
There are 144 PDSCH REs per resource block with CFI=1
With per RE interference level equal to signal level, PDSCH SINR =
10*Log(144/12) = 10.8 dB

40 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Proposed Workaround for 0% Load Test Cases


Use different RS frequency shifts in neighbor cells
Set all cells with 0% OCNS load
Since all neighbor cells are time aligned in the 0% OCNS test cases, by using
the two configurations above we can force LGE UE to report SINR/CQI
equivalent to actual 0% loading

41 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Actual Interference from 50% loaded Neighbor Cells


Approximately half of the PDSCH REs in the
serving cell will receive interference from
OCNS of neighbor cell
Therefore actual PDSCH SINR ~ 3 dB in the
case of per RE signal level equal to
interference level

42 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Proposed Workaround for 50% Load Test Cases


Use different RS frequency shifts in neighbor cells
Set all neighbor cells with 50% OCNS load
Set the PCFICH to RS power ratio to 11 dB
The objective is to create a level of interference in the first OFDM symbol that
corresponds to 50% loading, i.e., which corresponds to SNR equal to 3 dB when receiving
per RE interference level equal to signal level
Result for time aligned cells:
As discussed in a previous slide, with PCFICH to RS power ratio set to 3 dB we obtained
11 dB SNR in the first OFDM symbol. Therefore, adding 8 dB to PCFICH power level
lowers the SNR to 3 dB
Result for non-time aligned cells
50% of the RS REs in the first OFDM symbol will receive interference from OCNS REs of
the neighbor cells

43 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####

Conclusions
LGE SINR measurement only takes into account 1st OFDM symbol
need to ensure SNR measure (and associated CQI reports) properly reflect
PDSCH load
By using same RS frequency offset and 100% OCNS loading in neighbor cells LGE
UE reports for SINR/CQI will reflect actual100% loaded case
By using different RS frequency offsets and 0% OCNS loading in neighbor cells of
same NodeB, LGE UE reports for SINR/CQI will reflect unloaded case
By using different RS frequency offsets and 50% OCNS loading in neighbor cells,
LGE UE reports for SINR/CQI will reflect actual 50% loaded case

44 | Presentation Title | Month 2006

All Rights Reserved Alcatel-Lucent 2006, #####