Sei sulla pagina 1di 5

MIMO Performance at 700MHz:

Field Trials of LTE with Handheld UE


Bo Hagerman, Karl Werner Jin Yang
[firstname.lastname@ericsson.com] [Jin.Yang@VerizonWireless.com]
Ericsson Research, Sweden Verizon Communications Inc., USA

Abstract— MIMO is one of the techniques used in LTE Release 8 configurations for LTE with up to 4x4 MIMO schemes and [5]
to achieve very high data rates. A field trial was performed in a for corresponding HSPA results for 2x2 MIMO.
pre-commercial LTE network. The objective is to investigate how
well MIMO works with realistically designed handhelds in band MIMO (here used to denote multi-layer transmission) is one of
13 (746-756 MHz in downlink). In total, three different handheld the techniques used in LTE Release 8 to achieve very high data
designs were tested using antenna mockups. In addition to the rates. It is well known that MIMO requires certain conditions
mockups, a reference antenna design with less stringent on the radio channel to be fulfilled in order to work well. One
restrictions on physical size and excellent properties for MIMO of these conditions is that the radio channels experienced by
was used. The trial comprised test drives in areas with different each of the multiple receive antenna branches are statistically
characteristics and with different network load levels. The effects different from the others. The longer the wavelength at the
of hands holding the devices and the effect of using the device carrier frequency is compared to the size of the terminal, the
inside a test vehicle were also investigated. In general, it is very more challenging it is to fulfill this condition from a UE
clear from the trial that MIMO works very well and gives a antenna design perspective.
substantial performance improvement at the tested carrier
frequency if the antenna design of the hand-held is well made A trial was designed to investigate how MIMO works in
with respect to MIMO. In fact, the best of the handhelds cellular systems on lower carrier frequencies, e.g. 700MHz.
performed similar to the reference antenna. Focus is on practical UE dimensions and on practical network
deployments. A secondary purpose of the campaign was to test
Keywords-LTE, MIMO, Field-measurements, UE antenna different possible dual-antenna UE designs in order to obtain
design, band 13 qualitative results on their relative performance. In the trial,
2x2 MIMO was tested, i.e. 2 receive branches at the UE and
I. INTRODUCTION two transmit branches at the eNB. This is the typical setup for
today’s deployed LTE networks. Focus of this work is on
Multiple antennas for reception and transmission at the
downlink performance. The deployed eNB antenna systems are
radio base station (eNB) and the user equipment (UE) is a key
dual-polarized antennas. The network used is a pre-commercial
enabler for the high performance offered by the 3rd Generation
LTE network.
Partnership Project (3GPP) Long Term Evolution (LTE)
standard [1]. The standard [3] supports multiple antenna The trial was to be performed in a radio environment as similar
technologies that improve both link- and system- level as possible to what UEs will experience in a live LTE network.
performance in a wide range of scenarios. In order to achieve this, different load conditions were created,
hand dummies were used, and areas with different town
In LTE, the data streams (codewords) are mapped to layers.
architectures (urban and suburban) were used for the
The number of layers transmitted is denoted the (transmission-)
measurements. Hence, results regarding the interaction of
rank. In release 8 of the standard [3], up to four layers are
hands holding the device and vehicle components surrounding
supported for downlink transmission. The layers are mapped to
it were also obtained. The campaign is intended to investigate
antenna ports in a precoding step. Hence, it should be noted
how the radio environment interacts with the antenna design at
that the use of multiple antennas at any end of the link may
the used carrier frequency. It is not intended to benchmark any
improve performance also when one layer (rank=1) is
particular vendor of UEs nor the network performance.
transmitted.
The various transmission modes defined in the standard may be II. THE FIELD TRIAL CAMPAIGN
used with any antenna setup, and the codebooks used in the
closed-loop precoding modes are designed for a wide range of A. UE antennas
antenna setups and scenarios. However, the design of the
At the tested frequency band (746-756 MHz) the
antenna system is critical to performance. The best choice of
wavelength is 0.4m. A typical smartphone is smaller than 0.2m
antenna setup depends on the target performance profile and
along its dominating dimension.
the propagation environment, including the carrier frequency as
well as on other factors. See [6], [7] for basic result from a field
trial on 2.6GHz carrier frequency evaluating antenna
behind the mockup. In the free space on roof use mode, the
mockups were placed on a block of foam that has similar radio
characteristics to air. The blocks were then placed on the roof
of the car in the same way as in the hands on roof use mode.
Pictures exemplifying the use modes and UE mockups are
shown in Figure 1. The cables connecting the antenna mockups
to the receiver circuitry were the same independent of use
mode and type of antenna used.

B. Network configuration and field trial setup


Measurement drives were conducted with antennas
mounted according to use mode, see previous section. In each
measurement drive, multiple modem boards were used. One of
them was connected to the reference antenna (always mounted
on the van roof with no hand interaction) and the others were
connected to UE mockups. Either the two smartphones (SP1
and SP2) or the featurephone (FP1) were tested in each test
drive. The network implemented round robin scheduling and
data transmission was made using UDP with a bitrate selected
to ensure full buffers for all UEs at all times. Furthermore, the
drive routes were selected and the network was configured to
ensure that there were no handovers during the test drives.
Hence all UEs were in the same cell at all times.
The network implemented the possibility to transmit user plane
data in selected cells by scheduling DL transmission to non-
Figure 1. Top left picture: SP1 UE antenna mockup. Top right picture: FP1
type of UE antenna mockups in the hands in van use mode. Bottom most
existing UEs. This feature was used in the trial to create
picture: roof of measurement van with reference antennas mounted on the interference from neighboring cells in a reproducible way in
rear side, and mockups mounted on the front side as in free space on roof and order to investigate the impact of network load on
hands on roof use modes. performance. Two load settings were used: fully loaded and no
user plane load in neighboring cells. It should be noted that
some interference was present also in the latter case as
Three different dual antenna UE mockups were tested in the reference signals and control information was continuously
trial: Smartphone 1 (SP1) and Smartphone 2 (SP2) were transmitted in all cells at all times.
designed with the size of a smartphone, 73mm x 150mm (~3x6
inches). Featurephone 1 (FB1) was designed with a smaller Drives were performed on two measurement routes: urban and
form factor, 65mm x 115mm (~2,5x4,5 inches). The antenna suburban. The area used for the suburban measurement route is
realization, such as radiating element type, orientation, and characterized by one or two storied houses (residential and
positioning, are designed on purpose for high and low antenna business). It has a typical suburban character. The terrain is
branch correlation in order to evaluate impact on performance generally quite flat but the measurement route has a leg back
in the network. In particular SP2 has a very high antenna and fourth up a rather steep hill-side towards strong interferers.
branch correlation in order to serve as a “no MIMO” (i.e. no The duration of the suburban measurement route was about 16
multi layer transmission) reference case. In addition, a minutes and the drive speed mostly kept below 20 mph
reference antenna combination comprising two pole antennas (~35kmph).
was also used. The reference was designed to have very good The urban measurement route is of a typical North American
correlation properties. In contrast to the mockups, the reference urban character: The route is in an area dominated by high-rise
was designed with no limitations on the form factor. buildings in a regular grid of streets. The general interference
Figure 1 shows pictures of the mockups, the reference and their level is significantly higher than in the suburban drive route.
mounting in and on top of the test vehicle. The urban measurement route took around 9 minutes. Traffic
had a much larger impact of the drives on this route due to the
Three use modes are included in the trial: hands on roof, hands much higher occurrence of traffic lights and congestion. Drive
in van, and free space on roof. The hands on roof and hands in speed was slightly higher but mostly under 30 mph (~45
van use modes include hand dummies designed to have the kmph).
same radio characteristics as a hand. The mockups were placed
in a two-hand grip: “browse mode”. In the hands in van use GPS information was logged, and post-processing was
mode, the mockups and hands were placed to the right of the performed with the purpose of removing some of the traffic
driver’s seat; in the hands on roof use mode they were placed effects that disturb comparability. With such normalization,
on the roof of the car. Note that the body style of the vehicle is repeatability was in general good.
probably an important aspect when characterizing the in- On each measurement route and with each use mode, two test
vehicle radio channel, hence results obtained may only be valid drives were performed, with and without neighboring cell load.
for the type used in the trials: a panel truck with no windows
Figure 2. Left plot: Average RSSI for mockup FP1 in tests drives. Right plot:
average SNR for FP1 in the same test drives.
Figure 3. CDFs of SNR for the suburban drive route and all mockups in hands
in van use mode. Left plot: CDFs for unloaded network. Right plot: CDFs for
loaded network.
In total [3 use modes] x [2 measurement routes] x [2 network
load settings] x [2 mockup arrangements (smartphone or
featurephone)] = 24 test drives are included in this paper.
Measurement results were logged out from the UE modems
and they are inevitably depending on vendor specific
algorithms. However, the same type of modem boards was
used in all tests. Hence absolute numbers are of limited
interest, and focus is on comparison between the use-modes
and mockups.

III. MEASUREMENT RESULTS


Figure 2 (left) shows measured average received signal Figure 4. CDFs of throughput for the suburban drive route and all mockups with
strength indicator (RSSI) values for mockup FP1 in all the hands in van use-mode. Throughput is normalized in the same way as in
measurement drives. Note that these values also include Figure 5 below. Left plot: CDFs for unloaded network. Right plot: CDFs for
loaded network.
attenuation due to cables and not only due to propagation
losses and antenna efficiency. The attenuation caused by the
hands holding the device in the hands on roof use mode as well
attenuation due to the hand. SP1 and SP2 on the other hand
as the attenuation cased by surrounding vehicle components in
show very similar SNR; this indicates similar antenna
the hands in van use mode is evident in these bars. The signal
efficiency. In the loaded case, where interference is limiting, all
strength loss due to use mode is consistent over the
mockups experience similar SNR.
measurement drives.
Figure 4 shows CDFs of normalized throughput for the same
Figure 2 (right) shows estimated SNR levels over the
measurement drives as in Figure 3. In the unloaded scenario,
measurement drives. It is clear that the distinct drops in signal
the reference outperforms the mockups. In the loaded scenario,
strength due to hand and van attenuation is not visible in the
mockups FP1 and SP1 perform similar to the reference: SNR is
SNR, the only exception being drives on the suburban
more interference limited and vehicle attenuation is less
measurement route in the unloaded case. Clearly, signal
important. The higher received power in the unloaded scenario
strength only determines SNR in scenarios with low
translates into an SNR difference and hence throughput is
interference. In loaded, interference limited, scenarios, antenna
better. While FP1 shows much higher SNR in the unloaded
efficiency is less important.
case than SP1 (Figure 3) it shows similar throughput (Figure
Figure 3 shows CDFs for SNR over two test drives on the 4). On the other hand, SP1 has similar SNR to SP2 but much
suburban measurement route with the hands in van use-mode. better throughput. Differences are smaller in the loaded case
The SNR varies in an interval between 2dB and 20dB in the but SP2 stands out in this case as well with much lower
loaded scenario, and between 5dB and 30dB in the unloaded throughput than the others. With the possibility of rank two
scenario. Comparing the right and left plots show how SNR is transmission (MIMO), UEs experiencing highest SNR do not
pushed down by the interference. SNR is interference limited necessarily enjoy the highest throughput; it is not possible to
in the loaded scenario; there are only small differences between map SNR directly to throughput, antenna correlation and
the in-car mounted mockup-antennas and the reference antenna channel properties are equally important.
mounted on the roof. Conversely, in the unloaded scenario, the
Figure 5 shows throughput plotted against SNR for an
reference antenna clearly has the highest SNR. This is expected
aggregation of all test-drives with the hands in van use mode.
given its placement on the roof of the van. Similarly, in the
The left plot shows throughput plotted against own SNR. The
unloaded case, the FP1 mockup gets higher SNR than the two
right plot shows throughput plotted against reference antenna
smartphone mockups (SP1 and SP2). This is because the
SNR. The former way of plotting hides the effect of SNR
antennas of FP1 are placed so that they are subject to less
differences between UEs while the later includes it. It is clear
Figure 5. Throughput as a function of SNR for the hands in van use mode for all Figure 7. Rank two selection ratio as a function of SNR. Left plot: Free-space on
tested mockups. Throughput is normalized to maximum throughput for the roof use-mode. Right plot: Hands in van use mode.
reference antenna. Left plot: throughput as a function of SNR measured by same
antennas. Right plot: throughput as a function of SNR measured by the roof
mounted reference antenna (at same location on drive route).
Figure 7 shows ratio of rank two selection as a function of SNR
(measured by each mockup). SP2 is practically never selecting
rank two while SP1 and FP1 selects multi stream transmission
from the right plot that the SNR drop associated with
almost as often as the reference. It is also interesting to note
placement inside the van causes performance to deteriorate for
that placement inside van and with hands does not change this
both FP1 and SP1, and more so for SP1. However, the left plot
picture very much. Note that the way of plotting hides the
does not show such drop. This indicates that correlation
effect of SNR differences between mockups as well as between
properties are not affected as much as SNR by placement
mockups and reference.
inside the van. This is particularly true for SP1, which on the
other hand has lower efficiency than FP1 when held by hands.
SP2 is clearly outperformed: the difference between SP2 on the IV. SUMMARIZING COMPARISON
one hand and FP1 and SP1 on the other hand is larger than the Figure 9 shows mean normalized throughput
difference between the mockups and the reference. (normalization with reference in each drive) for all
Figure 6 shows mockup FP1 in all use modes. When plotted measurement drives. As a complement, Figure 8 shows rank
against reference antenna SNR (right plot) there is as expected two selection ratios for the same measurement drives. Mockup
a performance drop associated with both addition of hands SP1 and FP1 consistently outperforms mockup SP2. The UEs
(compare free space on roof and hands on roof) and with connected to SP1 and FP1 are able to select multi stream
placement inside the van (compare hands on roof and hands in transmission to a large extent while the UE connected to SP2 in
van use modes). However, when plotted against own SNR (left practice never selects multi-stream transmission. In
plot), difference is slim. This indicates that correlation measurement drives with large load, resulting in high
properties do not deteriorate as much as SNR when adding interference, SP1, FP1, and the reference perform similarly.
hands and placing the mockup inside the van. Figure 6 also The reference is significantly better only when compared to
shows SP2 in the free space on roof use-mode. Obviously, SP2 hands in van use mode in the least interference limited scenario
is outperformed (even by FP1 when placed inside van), in (suburban, no load).
particular at high SNR, and that this is mostly due to poor
correlation properties and to a smaller extent due to lower [FSR] [urban] [L]
antenna efficiency yielding lower SNR. [HR] [suburban] [NL]
1
0.9 [FSR] [suburban] [L]
0.8
0.7
0.6
[HR][urban] [NL] 0.5 [FSR] [urban] [NL]
0.4
0.3
0.2
0.1
[HR] [suburban] [L] 0 [FSR] [suburban] [NL]

[HR] [urban] [L] [HV] [urban] [L]


Ref
SP1
[HV] [suburban] [NL] [HV] [suburban] [L] SP2
FP1
[HV] [urban] [NL]

Figure 6. Throughput as a function of SNR for all use-modes with the FP1
mockup and (for extra reference) SP2 mockup in the free-space on roof use-
mode. Throughput is normalized to maximum throughput for the reference Figure 8. Average rank two (MIMO) selection ratio for free space on roof
antenna. Left and right plots show SNR reference similar to Figure 5. (FSR), hands in van (HV), hands on roof (HR) use modes in combination with
loaded (L) and unloaded (NL) network settings. Measurements using
reference and all mockups are included.
less attenuation due to hands which yields better SNR at low
[FSR] [urban] [L] network load levels.
1.4
[HR] [suburban] [NL] 1.2 [FSR] [suburban] [L] The most prominent effect of placement of the UE inside the
1
test vehicle was loss of signal strength. The effect on
0.8
[HR] [suburban] [L] 0.6 [FSR] [urban] [NL]
correlation was small. Signal strength loss may be very
0.4 significant in a power limited scenario; the same applies to
0.2 antenna efficiency. Hence, most on the tested cases were not
0 mainly power limited, and hence the effect on performance of
[HR] [urban] [L] [FSR] [suburban] [NL]
the in-van placement was relatively small.

[HV] [suburban] [NL] [HV] [urban] [L] SP1


VI. ACKNOWLEDGEMENTS
SP2
The authors wish to thank the Ericsson FOA testing team
FP1
[HV] [urban] [NL] [HV] [suburban] [L] headed by Brendan O’Connor, Can Hatipoglu, Anders Otto,
Lars Näslund, Henrik Asplund for all help during the
Figure 9. Throughput normalized to reference antenna throughput (on same campaign. Further, the SonyEricsson team Thomas Bolin,
drive). Measurements for free space on roof (FSR), hands in van (HV), hands Vanja Plicanic Samuelsson, Zhinong Ying and Peter Karlsson,
on roof (HR) use modes in combination with loaded (L) and unloaded (NL) for providing mockups, and Anders Derneryd, Ericsson. The
network settings are included. Measurements using reference and all mockups support from Lee Tjio, Verizon Corporate Technology, Gunnar
are included. The measurement data for [HR][urban][NL] was corrupt due to
backhaul issues.
Borg, Ericsson, and Hiep Pham, Ericsson, is also
acknowledged.

SP1 is generally able to select two stream transmission more REFERENCES


often than FP1. This is consistent with the larger physical size [1] E. Dahlman, S. Parkvall, J. Sköld, P Beming; 3G Evolution: HSPA and
of SP1. However, this difference is not as evident in LTE for mobile broadband, second edition, Academic Press.
throughput: there is a range of channel conditions where rank [2] M. Riback, J. Karlsson; Initial Performance Measurements of LTE,
two and rank one transmission can be expected to perform Ericsson Review, March 2008.
similarly. [3] 3GPP TS 36.211 3rd Generation Partnership Project; Technical
Specification Group Radio Access Network; Evolved Universal
Terrestrial Radio Access (E-UTRA); Physical Channels and Modulation
V. CONCLUSIONS (Release 8)
It is clear from the trial that MIMO does give a substantial [4] 3GPP TS 36.213 3rd Generation Partnership Project; Technical
Specification Group Radio Access Network; Evolved Universal
performance improvement with handheld UEs also at the Terrestrial Radio Access (E-UTRA); Physical layer procedures (Release
carrier frequencies in question. In fact, the best of the 8)
handhelds performed similar to the reference antenna, in [5] M. Riback et al.; “MIMO-HSPA Testbed Performance Measurements”
particular in free space configurations. This was the case even in proceedings of PIMRC 2007, Athens, 2007.
though the reference was designed with much looser size [6] K. Werner, J. Furuskog, M. Riback, B. Hagerman; Antenna
restrictions. In scenarios with low load, MIMO was selected by configurations for 4x4 MIMO in LTE – Field Measurements,
the terminal more than 80% of the time. In a highly loaded Proceedings of VTC Fall 2010, Taipei 2010
scenario, MIMO was selected more than half of the time. [7] J. Furuskog, K. Werner, M Riback, B. Hagerman; Field trials of LTE
Similar results have been obtained at 2.6GHz, see [6]. While with 4x4 MIMO, Ericsson Review, issue 2, 2010
such results are highly implementation and network
deployment dependent - there are channel conditions where
MIMO and SIMO have similar performance – they may still
give an indication that 2x2 MIMO works equally well at
2.6GHz and 700MHz.
If the antenna system is not designed targeting low correlation
properties, then MIMO performance will be very poor.
Mockup SP2 had much lower performance than the other
mockups and the reference due to the design of SP2 giving
intentionally high correlation. This is manifested in both its low
ability to select rank two transmission (MIMO) and in the
actual throughput figures.
The trial also showed that the effect of hands holding the
device on MIMO performance was relatively large with certain
antenna designs, even if the MIMO gains were still significant.
This further emphasizes the importance of a careful antenna
design, especially at lower carrier frequencies. Mockup FP1
had a smaller performance loss than mockup SP1: It showed

Potrebbero piacerti anche