Coordinated Multipoint Tx and Rx

Document Sample
Coordinated Multipoint Tx and Rx Powered By Docstoc
					                                               White Paper

Coordinated Multipoint Tx and Rx
By: Nagi Mahalingam, Chief Architect

Overview                                                                          CONTENTS
LightRadio, Liquid Radio and Cloud RAN—the fundamentals behind these three        Central and Distributed CoMP Processing pg. 3
seemingly incongruent solutions are very sound but the commonalities between
                                                                                  CoMP in Downlink pg. 4
these solutions are strikingly similar. Over the past few years, end users of
services have become impatient for increases in per-user throughput with an       CoMP in Uplink pg. 5
underscored demand for consistency in service quality. The Release 11 work
                                                                                  Carrier and Data Aggregation pg. 6
item in 3GPP, Coordinated Multipoint TX and RX (CoMP), aims at addressing
three of the most important problems plaguing the telecoms industry:              Adaptive Traffic Shapping pg. 6
(a) spectral efficiency, (b) cell-edge throughput and (c) coordinated coverage.
                                                                                  CoMP: Why and How Does it Matter to Us? pg. 7
                                Coordinated Multipoint Tx and Rx | Radisys White Paper       2

At the cell-edge, typically, downlink signals from
many evolved NodeBs (eNBs) could be received at
the mobile while the mobile’s uplink transmissions
could also be received by these various eNBs.
• If the downlink transmissions to a given mobile
  from the various cell towers can be carefully
  ‘coordinated’, performance metrics for the downlink
  can be significantly boosted
• If the uplink transmissions from the mobile can
  be carefully ‘scheduled’ by the various cell towers,
  the performance metrics for the uplink can be
  significantly boosted

CoMP addresses mundane issues such as ‘interference
mitigation’ as well as tightly coordinated bit transmis-
sions over the air interface. A farm of baseband
processors centrally located and controlling digital                                      
In-phase and Quadrature (I/Q) data to and from various      Figure 1. Intra-cell CoMP
remote radio heads will enable this joint coordination
and scheduling intra-cell as well as inter-cell. Figures
1 and 2 show the principles between intra-cell and
inter-cell CoMP.

In both Intra-cell and Inter-cell schemes shown in
Figures 1 and 2, the operator would deploy a number
of remote radio heads (RRH) and have the entire
baseband processing moved into a farm of baseband
processors somewhere in the ‘cloud.’ To ensure
better coupling of the signal with the transceiver, the
amplifier and gain control functions are built into the
RRH. Since the processing is moved into the ‘cloud,’
the antenna elements require less power and hence
a smaller, less expensive, amplifier could be utilized.
It becomes an implementation issue whether the
antenna elements need to offer wideband operation
or need to support multiple frequencies/bands.
The common theme in the figures above is that the
processing is ‘centrally’ controlled. This, however,
is not the only possible way to realize CoMP as will
be discussed in sections to follow.                         Figure 2. Inter-cell CoMP
                                Coordinated Multipoint Tx and Rx | Radisys White Paper                                3

The RRH to baseband processing farm link will
be Common Public Radio Interface (CPRI)-based,
typically carried over an optical network. The latest
version of CPRI, v4.2, has all the provisions detailing
synchronization and timing, delay calibration, link
accuracy and I/Q bit widths. These apply equally to
UMTS Terrestrial Radio Access (UTRA) and Evolved
UTRA (E-UTRA). A properly dimensioned network
can see a farm of baseband processors serving RRHs
placed as far away as 30 to 40 kilometers while still
meeting dimensioned latency budgets.

Central and Distributed
CoMP Processing
For CoMP to work, the downlink channel should be
                                                            Figure 3. Central and Distributed CoMP processing
known prior to the transmission. Mobiles perform
channel estimations and provide Channel quality,
Rank Index, Precoding matrix index and Channel state        In the distributed approach, however, processing
information to the eNB serving the connected mode           is done at more than one eNB but with joint
mobile. There are two possible methods for CoMP             coordination. This enables them to control multiple
to be realized:                                             users as well, if required. In this approach CQI, RI,
                                                            PMI and Channel state information is being received
• Centrally controlled CoMP processing
                                                            from the mobile at each of these coordinating eNBs.
• Distributed CoMP processing                               Distributed CoMP is illustrated by the ‘Blue’ colored
                                                            mobile controlled by eNB#1 and eNB#2 in Figure 3. In
In the central approach, a common farm of processors—
                                                            the distributed approach, the transmission waveforms
centrally located—makes use of Channel State
                                                            are computed at each of the eNBs and hence there
information, Channel Quality Indicator (CQI), Pre-coding
                                                            is no need for an RRH. Further, there is no latency
Matrix Index (PMI) and Rank Indicator (RI) reported by
                                                            associated with this process at the front end. Since
the mobile to decide which sets of RRH are best suited
                                                            the signal processing is done locally at each eNB,
for serving the mobile. It is important to note at this
                                                            the radio front ends at each of the eNBs need to be
point that since the processing is centrally controlled,
                                                            very tightly bound for synchronization. The precise
the central processors will handle all the digital signal
                                                            clock reference can either be from a global positioning
processing. The I/Q waveforms for transmission are
                                                            system (GPS) tick or network-based protocols such as
pre-computed at the central farm of processors and
                                                            SyncE and Precision Time Protocol (PTP). It becomes
distilled into the set of RRHs that are best suited to
                                                            necessary that two eNBs in distributed CoMP must
serve the mobile. The data hauling to and from the
                                                            transmit the same data symbol at the same resource
RRH is via a CPRI interface as mentioned earlier and
                                                            element in the time-frequency grid.
the bandwidth requirement to handle I/Q waveforms in
either direction can be in gigabits per second depending
on the capacity the RRH is dimensioned to provide.
Centrally controlled CoMP is illustrated by the ‘Green’
colored mobile centrally controlled by eNB number 1
(eNB#1) in Figure 3.
                                  Coordinated Multipoint Tx and Rx | Radisys White Paper                                      4

Figure 4. Distributed CoMP processing at eNB Tx and precoding

For distributed CoMP, the precoding is done at each             Since the MAC layer processing is already completed
of the eNBs. Figure 4 shows the use case of two                 in the respective eNB, latencies due to data exchange
eNBs in distributed CoMP handling two separate                  are compensated for by implementing buffers before
users enabling joint scheduling and beamforming.                the precoder. On the user equipment (UE) side, the
Data bits coming into the Media Access Control (MAC)            precoded channel is estimated at each mobile. The
layer at both eNBs are formed into transport blocks             channel estimation is multi-cell, multi-user and
after HARQ (Hybrid Automatic Repeat ReQuest)                    symbol equalization is performed by Interference
processing. These transport blocks are encoded                  Rejection Combining. There is a downside to this in
(forward error correction, interleaving), mapped into           the sense that channel estimation at the mobile is
symbols and then into appropriate resource elements             a lot more complex, but the net benefit is a system-
within a resource block. Note the mobile-specific               wide improvement in capacity.
and cell-specific pilot insertion in Figure 4; the link
between the two eNBs is, for instance, X2. Data                 CoMP in the Downlink
exchanged between the two eNBs will be over the                 Generally speaking, downlink CoMP is performed
‘wire’ (Ethernet, for example) and some header-like             as follows:
information is signaled to indicate which resource
blocks have CoMP mode enforced.                                 • A transmission to a given mobile is done by the
                                                                  cell serving the mobile (connected mode mobile).
A point to note from above is that the precoding vector           However, the transmissions are ‘beamformed’ from
is nothing but a multiplicative complex number. The I/Q           the serving cell to the mobile such that interference
bits at the eNB are precoded before the IFFT (Inverse             due to that downlink transmission to other network
Fast Fourier Transform) operation. The feedback from              elements is either eliminated or reduced. This
the mobiles (CQI, PMI, RI and Channel state information)          ‘beamforming’ is performed by means of joint
is required at the MIMO (multiple-input and multiple-             coordination between the various cell towers.
output) input stage of the eNB. ACK/NACK                        • A transmission to a given mobile will be performed
(acknowledgement/negative acknowledgement)                        by many cell towers (at least two). However, the
processing is at the HARQ with the output of the                  transmission is scheduled such that it is seen by
CODEC unit in the mobile feeding back that information.           the mobile as a single transmitter with spatially
                                                                  diverse antenna elements.
                                 Coordinated Multipoint Tx and Rx | Radisys White Paper                       5

Looking into this deeper, there are three methods
discussed in the 3GPP standards to address
downlink CoMP:

In CoMP schemes, control channels such as PDCCH
are only ever transmitted from the serving cell. In
Figure 5, ‘Coordinated Scheduling’ is shown in which
data channel PDSCH is transmitted only from one
cell. The scheduling is coordinated among multiple
cells. This has the benefit of increased SINR (Signal to
Interference plus Noise Ratio) which directly translates
to an increase in cell-edge performance metrics.
                                                             Figure 5. Coordinated Scheduling (beamforming)
Fast Selection
In Fast Selection, the eNB transmitting the PDSCH
having the lowest path loss is instantaneously selected
through fast scheduling at the central eNB. Since only
one eNB is selected for transmission, all the other
transmitters among the coordinated cells are not
scheduled. The advantage of this method is that the
muting of cells (non-scheduling) will have the net result
of decreased interference at the mobile, thus providing
maximum received signal power at the mobile.

Joint Scheduling
In Joint Scheduling, the PDSCH is transmitted from
multiple cells with codebook base precoding, at each
coordinated cell, using demodulation reference pilots.
Specific lobe weights (beamforming weights) are              Figure 6. Fast Selection
added before transmission. The throughput at the
cell edge sees a significant improvement as more than
one cell participates in the coherent transmission of
resource elements. The mobile combines the received
signal coherently using a coherent combining receiver.

CoMP in the Uplink
On the uplink, the PUSCH is received at multiple cells.
In Figure 8, the two possible CoMP methods on the
uplink are illustrated. In the left side of Figure 8, only
one mobile transmits the PUSCH and is scheduled by
the coordinated eNBs. On the right side of Figure 8,
multiple mobiles transmit the PUSCH simultaneously
using same resource block. In this method, upon
reception of PUSCH the cells use various estimation          Figure 7. Joint Scheduling
schemes to recover the PUSCH. Looking at it from
another view point, limiting interference from the
                                Coordinated Multipoint Tx and Rx | Radisys White Paper                   6

mobile at the receiver, whether intended of unintended,
is purely a scheme that will involve implementing clever
schedulers and receivers. Since it can be addressed
entirely by a given ‘implementation,’ uplink CoMP
does not have much standards impact.

Carrier and Data Aggregation
Carrier Aggregation is the method in which two or
more component carriers are combined to increase
the total system capacity. It is to be noted that
non-contiguous carrier aggregation is likely to be
the common method for frequency aggregation.
This poses a problem if, for different component
carriers, differing methods should be implemented
for resource allocation, modulation and coding scheme
support selection, transmit power setting, etc. Data
Aggregation can either be handled at the physical
(PHY) layer or at the MAC layer. If the PHY layer
aggregation option is chosen, HARQ must be used
for the ‘aggregated data’ rather than the ‘component
data’ from the component carriers. In practice, this
is very inefficient to support and, as a result, it is
more than likely for data aggregation to take place
at the MAC layer as shown in Figure 9.

Carrier aggregation on the uplink and downlink can be
asymmetric; the number of component carriers on the
uplink and downlink can be different in size and number.
In TDD (time division duplexing) mode, asymmetric                                                     
carrier aggregation can be achieved by allocating
different number of time slots for uplink and downlink.     Figure 8. Uplink CoMP

Adaptive Traffic Shapping
Improvements to backhaul capacity will lag all
advancements made on the Radio Access Network
(RAN) side by a number of years. Even taking current
High Speed Packet Access (HSPA) deployments into
account, the packet networks form the bottleneck
and the promise of gigantic increases in per-user
throughput is far from reality—and LTE and LTE-
Advanced deployments will accentuate this problem.
The logical answer, of course, is to augment wired
backhaul capacity by installing or adding passive
optical networks. But this is not a singular solution
that will cap the problems at hand. Quality of Service
and Quality of Experience need to be guaranteed for         Figure 9. Carrier and Data Aggregation
                                 Coordinated Multipoint Tx and Rx | Radisys White Paper                                                                    7

subscribers that pay a premium. Prioritizing traffic         • Cloud RAN calls for real, carrier-grade joint
on a subscription basis or on traffic class/traffic type       schedulers, interfaces to content servers,
basis requires inspection of traffic from various users        manipulation of content at line speed, true
in either direction, at line speed, taking into account        distributed MIMO processing and several other
capacity on the uplink and downlink directions in              value added services.
addition to application type, time of day and other
service level principles.                                    At Radisys, we are aware of the constant need to keep
                                                             abreast of rapid, radical changes in the communications
Bringing this ‘intelligence’ into the eNB itself helps       industry. The company’s Trillium® Advanced Portability
unblock congestion to and from eNBs and facilitates          Architecture (TAPA®) is platform-agnostic and is
user and eNB load balancing from the Evolved Packet          readily portable across all platforms and processor
Core (EPC) view point. A traffic-shaped RAN grants the       architectures. TAPA is woven into the fabric of every
operator better Key Performance Indices (KPI).               Trillium software product, individual or combined, and
                                                             hence forms the cornerstone of ‘product consistency.’
CoMP: Why and How Does it                                    The fundamental design of TAPA enables itself to
Matter to Us?                                                auto-align and scale to function on a single core or
                                                             across multiple cores. Radisys welcomes CoMP and
Moving the baseband processing to a ‘farm’ in the            sees it as a logical progression to providing better and
cloud would mean larger, more powerful processors,           greater service to end users. From a practical viewpoint,
possibly with multiple cores. However, as hinted             subscribers should see corner scenarios disappear
in sections above, this calls for changes that the           as CoMP-enabled networks exhibit tighter control of
industry should adopt but has yet to do so:                  mobiles, thereby giving rise to increased reliability of
• CoMP hints at the possibility of multiple ‘radio access    services (for example, significantly reduced call drops
  technologies’ to co-exist at the front end: this is        due to mobility).
  essentially the software defined radio (SDR) concept.
  As a result, the baseband as well as the radio stacks
  running on the processor farms will have to become
  multi-mode capable to address this need.

                                                                                                                      Corporate Headquarters
                                                                                                                      5435 NE Dawson Creek Drive
                                                                                                                          Hillsboro, OR 97124 USA
                                                                                                                 503-615-1100 | Fax 503-615-1121
                                                                                                                          Toll-Free: 800-950-0044

                                                                                                                                       ©2011 Radisys Corporation.
                                                                                                            Radisys, Trillium, Continuous Computing and Convedia
                                                                                                                are registered trademarks of Radisys Corporation.
                                                                                              *All other trademarks are the properties of their respective owners.
                                                                                                                                                  September 2011

Shared By:
Description: LightRadio, Liquid Radio and Cloud RAN—the fundamentals behind these three seemingly incongruent solutions are very sound but the commonalities between these solutions are strikingly similar. Over the past few years, end users of services have become impatient for increases in per-user throughput with an underscored demand for consistency in service quality. The Release 11 work item in 3GPP, Coordinated Multipoint TX and RX (CoMP), aims at addressing three of the most important problems plaguing the telecoms industry: