Upload
grahame
View
212
Download
0
Embed Size (px)
Citation preview
HSPA+ (2100 MHz) vs LTE (2600 MHz) Spectral Efficiency and Latency Comparison
Brendan McWilliams, Yannick Le Pézennec, Grahame Collins Vodafone Technology Networks,
Access Competence Centre, Madrid, Spain and Newbury, United Kingdom
[email protected] [email protected]
Abstract—High Speed Packet Access+ (HSPA+) has become
widely deployed whilst the more recently specified Long Term Evolution (LTE) deployments are in their infancy. The objective of this paper is to provide a side by side field performance comparison between HSPA+ and LTE deployed in the 2100 MHz and 2600 MHz bands respectively reusing the same sites and antennas and using the most advanced mobile broadband feature readily available in the same radio environment. Separately, both of these technologies have been investigated however a comparison of both as part of the same field trial has received less attention.
The performance comparison was carried out with comparable spectrum bandwidth (5 and 10 MHz) in a real network – Vodafone Spain – dedicated to the trial without any bias in terms of traffic or interference from other users.
The exercise revealed that the latency performance difference is not as high as expected with less than 10 ms difference however the connection time is significantly improved. Most distinguishing factor in LTE downlink is the more efficient and robust usage of Multiple Input Multiple Output (MIMO) and 64 Quadrature Amplitude Modulation (QAM) altogether which outperforms HSPA+ even when using a less advantageous frequency band. In uplink LTE only outperforms HSPA+ in good radio conditions due to the possibility to use 16QAM modulation.
Keywords- 3GPP, HSPA+, LTE, MIMO, Latency
I. INTRODUCTION 3GPP has in recent years developed the LTE standard as an
evolution beyond the UMTS 3G Standard, widely deployed within which a wide range of HSPA+ features form part of. “HSPA+” is generally used to refer to the HSPA capabilities that were introduced from 3GPP Release 7 and thereafter in subsequent releases amongst which 64QAM modulation, MIMO, and Dual Cell.
Mobile Broadband services over 3G networks are now already widely available using HSPA+ features and speeds up to a theoretical value of 43.2 Mbps are becoming ever more common [1][2].
TABLE I. and TABLE II. below show respectively the downlink (DL) and uplink (UL) peak rates supported with the mobile broadband capabilities introduced with HSPA+
(HSDPA MIMO refers to MIMO 2x2) in both Single Carrier (SC) using 5 MHz and Dual Carrier (DC) with 10 MHz respectively.
TABLE I. HSDPA MOBILE BROADBAND CAPABILITIES
HSDPA Capability
3GPP Release
#HSDPA Codes Modulation Peak rate
7.2 Rel’5 10 16QAM 7.2 Mbps 14.4 Rel’5 15 16QAM 14.4 Mbps
64 QAM Rel’7 15 64QAM 21.6 Mbps MIMO Rel’7 15 16QAM 28.8 Mbps
DC SIMO Rel’8 2x15 64QAM 43.2 Mbps DC MIMO Rel’9 2x15 64QAM 86.4 Mbps MC MIMO Rel’10 3x15 64QAM 129.6 Mbps MC MIMO Rel’10 4x15 64QAM 172.8 Mbps
TABLE II. HSUPA MOBILE BROADBAND CAPABILITIES
HSUPA Capability
3GPP Release
TTI Modulation Peak rate
2 Mbps Rel’6 10 ms QPSK 2 Mbps 5.7 Mbps Rel’6 2 ms QPSK 5.7 Mbps
SC16QAM Rel’7 2 ms 16QAM 11.5 Mbps DC16QAM Rel´9 2 ms 16QAM 23.0 Mbps
Where spectrum availability permits, LTE deployments are
now beginning across the world following the now stable standard, successful validation via trials and availability of corresponding devices. LTE Mobile Broadband capabilities are limited / defined in terms of peak rate by the maximum rate supported by the UE. TABLE III. shows the DL and UL UE LTE peak rate capabilities supported up to 3GPP Rel’10 [1].
TABLE III. LTE UE CAPABILITIES
Category 3GPP MIMO layers DL Peak Rate UL Peak Rate
1 Rel'8 DL: 1, UL:1 10 Mbps 5 Mbps 2 Rel'8 DL: 2, UL:1 50 Mbps 25 Mbps 3 Rel'8 DL: 2, UL:1 100 Mbps 50 Mbps 4 Rel'8 DL: 2, UL:1 150 Mbps 50 Mbps 5 Rel'8 DL: 4, UL:1 300 Mbps 75 Mbps 6 Rel'10 DL: 2/4, UL:2 300 Mbps 50 Mbps 7 Rel'10 DL: 2/4, UL:2 300 Mbps 100 Mbps 8 Rel'10 DL: 8, UL:4 3000 Mbps 1500 Mbps
978-1-4673-1391-9/12/$31.00 ©2012 IEEE
MIMO which was added in 3GPP Release 7 for 3G is a technology that forms a fundamental component of LTE thereby making it a key enabler in fulfilling the original LTE requirements. A key difference between 3G and LTE is the immediate fallback capability in LTE which permits fallback to transmit diversity – not possible in 3G (slow fallback would be possible to STTD Tx Diversity scheme, but this utilization of this brings performance issues across all radio conditions).
II. SCOPE
A. Objectives
LTE field performance has been investigated in many papers - [3] [4] [5] [9] [10] - however comparison between HSPA+ and LTE performance in the same environment i.e. a side by side direct comparison using the location and, site re-use has received less attention.
The key objectives of this trial were to compare the performance of HSPA+ and LTE using the latest available features over the same bandwidth (5 and 10 MHz in DL, 5 MHz in UL). Single user and multi-user throughput performance for FTP traffic model was measured as well as latency. This performance comparison aimed at gaining knowledge of field performance versus expected performance from simulations, and to identify the key strengths and weaknesses of both HSPA+ and LTE in the field and identify areas of improvements.
B. Field Trial Environment The field trial consisted of a cluster of 3 NodeBs and 3
eNodeBs in a semi urban environment in an industrial estate on the outskirts of the city of Madrid, Spain, using the Vodafone España network. The test cluster was closed to commercial traffic and the collaboration included leading network and device providers of HSPA+ and LTE technologies.
C. Field Trial Setup The main characteristics of the trial setup used to
benchmark HSPA+ and LTE performance are listed below:
• 5 MHz and 10 MHz blocks of spectrum were utilized in both the 2100 MHz and 2600 MHz bands for HSPA+ and LTE respectively. This permits the comparison of spectral efficiency given that both amounts of spectrum were available.
• A real network implementation using the same three sites for both HSPA+ and LTE i.e. in each site there was a NodeB and eNodeB connected to the same antenna i.e. the same sector.
For both technologies:
• Full buffer traffic model was based on downloads of large files via FTP.
• IP transport was used in the backhaul – dimensioned to support LTE peak rates and latency contribution.
• User throughput measured at MAC level using both NW and UE level traces.
A list of notable network parameter settings are listed in TABLE IV.
TABLE IV. KEY NETWORK PARAMETERS
Notable Parameters
HSPA+ LTE
Pilot Power 35 dBm 21.2 dBm (5 MHz) 18.2 dBm (10 MHz)
Downlink capabilities
SC 64QAM, SC MIMO 16QAM,
DC 64QAM.
Tx diversity (TM2) MIMO Open Loop (TM3)
Uplink capabilities SC HSUPA 2 ms TTI QPSK 16QAM
Scheduler type Proportional Fair Proportional Fair
Uplink Receiver Rake 2 with Interference Cancellation
Interference Rejection Combining
Security ON (UEA1) OFF (no IPSec)
Note: cell pilot power allocation in 3G and LTE is not comparable i.e. 35 dBm in 3G is used across all the carrier’s 5 MHz all the time, whilst in LTE the pilot signal is transmitted during predetermined time periods and frequency subcarriers - in both instances pilot power was the same proportionately.
Similarly for the UEs, a list of notable parameters is given in TABLE V.
TABLE V. KEY UE PARAMETERS
Notable Parameters HSPA+ Capabilities LTE
Rx Diversity ON ON
Equaliser ON (LMMSE) N/A
Interference Cancellation
OFF N/A
UE category HSDPA:cat.14/18/24 HSUPA: cat. 6
Category 3
UE Maximum Output Power
23 dBm 22 dBm
In addition to the usage of different frequency bands, only one other variable existed in the exercise in that the LTE only device used (multi mode devices were not available at that time) was limited to 22 dBm output power as opposed to 23 dBm in the HSPA+ device).
Same energy was used for SIMO and MIMO testing i.e. same overall power settings for Common Pilot Channel
(CPICH) and total power. Power settings are highlighted in TABLE IV. . Cross polar multi band antennas (covering both the 2100 MHz and 2600 MHz frequency bands) were used for SIMO and MIMO testing, i.e. such antennas are typical in commercial networks (no antenna upgrade for MIMO).
III. PROCEDURE As in the testing or comparison of all new technologies a
reference scenario was agreed upon with which to make subsequent comparisons. As one of the objectives of this activity was to measure LTE performance against existing 3G performance and the additional gains it might bring – despite utilisation of different frequency bands for both – 3G was selected as this reference technology.
As Channel Quality Indicator (CQI) reports differ in both calculation and scale for 3G and LTE respectively, selection of static test points were identified as reported by the 3G HSPA+ devices– see TABLE VI. .
TABLE VI. REFERENCE RADIO CONDITIONS
Radio Conditions
Average CQI
Received Signal Code Power
(RSCP) Good 28 RSCP ≥ -70 dBm
Medium 24-25 -70 dBm ≤ RSCP ≤ -90 dBm Bad 16-17 RSCP < -90 dBm
Thus static points were identified using the 3G HSPA+ network as the reference and tests thereafter were conducted at the same time in the same place using the same spectrum bandwidth in both the 2100 MHz and 2600 MHz band. The time between testing was minimised i.e. one immediately after another, to ensure radio conditions were similar as possible. The tests were performed over 5 different static points to ensure sufficient confidence in results.
Due to the propagation differences of using the 2600 MHz frequency band – less coverage versus the 2100 MHz frequency band - care was taken in the selection of both static locations and a drive testing ensuring that all were in LTE coverage and thereby avoiding dropped calls.
IV. TEST CASES
A. Throughput Testing Spectrum availability and functionality available from both
the network and device supplier meant that testing was limited to 5 MHz and 10 MHz bandwidth testing. TABLE VII. lists the throughput tests conducted during the trial.
TABLE VII. THROUGHPUT TEST SUMMARY
UL/DL Technology Feature Spectrum Used
Dow
nlin
k HSPA+ SIMO 64QAM 5 & 10 MHz
MIMO 2x2 16QAM
5 & 10 MHz
LTE MIMO 2x2
64QAM 5 & 10 MHz
SIMO 64QAM 5 & 10 MHz
Upl
ink HSPA+ QPSK (2 ms TTI) 5 MHz
LTE 16QAM (1 ms TTI) 5 MHz
All of the above tests were carried out in static points in
the aforementioned radio conditions and similarly via drive testing (approx 20km/hr).
HSPA+ MIMO was not included in any the drive tests as it is known already that this feature does not perform well in dynamic conditions [6].
B. Latency Testing Latency Testing consisted of performing ping testing of a
32 byte packet 100 times at intervals of 1 second in 5 different locations. For both technologies, latency testing focused upon:
• The transition time in going from idle to connected mode.
• Measured latency whilst already in connected mode.
V. RESULTS
A. Static Tests 1) Single User Throughput Performance in 5 MHz
Figure 1. shows the results of throughput obtained using the same spectrum comparing HSPA+ 64QAM, HSPA+ MIMO 16QAM and LTE MIMO 64QAM modulation. LTE MIMO 64QAM in the downlink outperforms HSPA+ across all radio conditions, with gains of 72%, 55% and 22% vs HSPA+ in good, medium and bad radio conditions respectively.
Figure 1. Single User downlink performance over 5 MHz
In the uplink, LTE outperforms HSPA+ quite significantly in good radio conditions, modest improvements in medium radio conditions although degradation was observed in bad radio conditions – see Figure 2.
Figure 2. Single User uplink performance over 5 MHz
2) Multi User Throughput Performance in 5 MHz
In the downlink, LTE outperforms HSPA+ across all radio
conditions – see Figure 3. - respectively in a multi user scenario (2 devices only) in a similar fashion as was shown in Figure 1 and single user testing.
Figure 3. Multi User downlink performance over 5 MHz
In the uplink, no multi user testing was performed.
3) Single User Throughput Performance in 10 MHz
LTE vs HSPA+ gains are higher in 10 MHz due to the better efficiency of LTE MIMO schemes in 10 MHz. The usage of dual stream MIMO is higher in 10 MHz allowing for more efficiency see Figure 4. Note that HSPA+ in 10 MHz (i.e. DC-HSUPA) is not yet available in uplink hence the trial focussed only on downlink for 10 MHz comparison tests.
Figure 4. Single User Downlink performance over 10 MHz
B. Drive Tests 1) Single User Throughput Performance in 5 MHz
For the drive tests, parameter settings were unchanged from those settings during the static testing and the speed of the test vehicle did not exceed 20Km/hr in the test route (using the same cell). Average throughput using HSPA+ SIMO 64QAM over the drive route was 9 Mbps, whilst using LTE MIMO it was 14 Mbps – a 56% gain in favour of LTE MIMO.
0
5
10
15
20
25
30
Good Medium Bad
Thr
ough
put (
Mbp
s)
Radio Conditions
HSPA+ SIMO 64QAM
HSPA+ MIMO 16QAM
LTE MIMO 64QAM
0
1
2
3
4
5
6
7
8
9
10
Good Medium Bad
Thr
ough
put (
Mbp
s)
Radio Conditions
HSUPA 5.7 Mbps
LTE
0
5
10
15
20
25
30
35
HSPA+ 64QAM
HSPA+ MIMO
LTE MIMO
HSPA+ 64QAM
HSPA+ MIMO
LTE MIMO
HSPA+ 64QAM
HSPA+ MIMO
LTE MIMO
Good Medium Bad
Thr
ough
put (
Mbp
s)
Radio Conditions
user 2
user 1
0
10
20
30
40
50
60
Good Medium Bad
Thr
ough
put (
Mbp
s)
Radio Conditions
HSPA+ DC 64QAM
LTE MIMO 64QAM
+136%
+12%
-24%
+72%
+55%
+22%
+72%
+76%
+78%
2) Single User Throughput Performance in 10 MHz
Average throughput using HSPA+ SIMO 64QAM feature over the drive route was 17.1 Mbps, whilst using LTE MIMO it was 24.3Mbps – a 42% gain in favour of LTE MIMO.
C. Latency The results show that the latency LTE connection time
gains is approximately 3.4 times lower compared to HSPA+ on average, see Figure 5.
Figure 5. Connection Time from Idle to Connected Mode
Latency whilst in connected mode showed gains in LTE of
below 10 ms compared to HSPA+ as shown in Figure 6. .
Figure 6. Latency in connected Mode
VI. KEY FINDINGS
A. DL Throughput Figure 7. shows the utilisation rate (y-axis) of the three
used modulation schemes across good, medium and bad radio conditions in both HSPA+ and LTE respectively. It is clear that under the same radio conditions that usage of the highest modulation scheme usage (64QAM) is significantly higher in LTE than in HSPA+ e.g. 64QAM usage in bad radio conditions still occurs in bad radio conditions in LTE but not at all in 3G. Similar modulation utilisation was observed in 10 MHz testing.
Figure 7. Modulation Usage, 5 MHz, Single User.
By design, LTE has an intrinsically better link performance
than 3G and allows not only better usage of higher order modulation schemes but increased usage of dual layer MIMO – see Figure 8.
The trial revealed that for HSPA+ “dual stream” occurs throughout good radio conditions and to a lesser extent in medium radio conditions – as expected. For LTE, “dual stream” refers to utilisation of Rank 2, and this occurs throughout good and medium radio conditions. In bad radio conditions there is no dual stream transmission neither in HSPA+ nor LTE.
It should again be noted that the radio conditions are worse for LTE due to the 2600 MHz band and were LTE transmitted in the 2100 MHz band – like HSPA+ – then some Rank 2 transmission could be expected in LTE.
Figure 8. MIMO Dual Stream Usage in 5 MHz, Single User
0
100
200
300
400
500
600
SP1 SP2 SP3 SP4 SP5
Con
nect
ion
time
(ms)
Static Point
HSPA+ SIMO 64QAMLTE MIMO
0
10
20
30
SP1 SP2 SP3 SP4 SP5
Lat
ency
(ms)
Static Point
HSPA+ SIMO 64QAMLTE MIMO
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
HSPA+ 64QAM
HSPA+ MIMO
LTE MIMO
HSPA+ 64QAM
HSPA+ MIMO
LTE MIMO
HSPA+ 64QAM
HSPA+ MIMO
LTE MIMO
GOOD MEDIUM BAD
Modulation Usage
64QAM 16QAM QPSK
0%
20%
40%
60%
80%
100%
Good Medium Bad
MIM
O D
ual S
trea
m U
se (%
)
Radio Conditions
HSPA+ MIMO 16QAM
LTE MIMO 64QAM
Similar Dual Stream Usage was observed in 10 MHz testing.
Note that for HSPA+ MIMO capability was limited here to 16QAM as MIMO 64QAM for HSPA+ was not available at the time of the trial.
B. UL Throughput For LTE, the Uplink improvements in good radio
conditions are due to the predominately use of 16QAM modulation and in addition – although to a lesser extent - with what appears to be a more efficient receiver.
It should also be noted that in addition to the coverage loss (around 2 dB) brought about due to operation in the 2600 MHz band, the LTE device used during the trial had a lesser maximum power (1 dB less) than the HSPA+ device(s) used and this is also a factor in explaining the lower UL LTE performance observed in bad radio.
C. Latency The results from the latency testing revealed that in
connected mode the difference between the two technologies was closer than expected. This shows that in a fully optimised network, latency figures of 18 ms can be achieved in a 3G network. For LTE, some improvements can be expected – around or below 10 ms – when future optimisations become available.
However, in Call establishment testing it was clearly observed that LTE was significantly quicker than WCDMA. It is expected that a 100 ms connection time for LTE is achievable – so there is room for improvement. For 3G, the introduction of additional optimisations in channel switching i.e. to and from idle mode and the channel states when in connected mode will improve call establishment times.
VII. CONCLUSIONS Side by side testing of HSPA+ vs LTE with all things being
equal was not possible in this exercise due to spectrum usage restrictions but in this paper we have discussed the results from side by side field testing of HSPA+ in the 2100 MHz band against LTE in the 2600 MHz band. This is expected to be a typical deployment across many territories and the
performance results obtained indicate that in the downlink - in spite of the different and disadvantageous frequency band utilised for LTE - LTE spectral efficiency is greater i.e. it outperforms HSPA+ over a wide variety of radio conditions. A true side by side comparison utilising the same frequency band would reveal greater gains for LTE compared to HSPA+. In the uplink, the benefits of LTE are more modest at this stage. Finally, with regards to Latency it was observed that the idle to connected mode performance of LTE significantly outperforms its HSPA+ counterpart whilst there are more modest reductions in latency whilst in connected mode.
REFERENCES [1] 3GPP Technical Specification “TS 25.214: Physical Layer Procedures”,
3GPP Standards Release 7. [2] 3GPP Technical Specification TS 36.306, “User Equipment (UE) Radio
Access Capabilities”. [3] S. Parkvall, E. Englund, A. Furuskär, E. Dahlman, T. Jönsson, A.
Paravati, “LTE Evolution towards IMT-Advanced and Commercial Network Performance”, IEEE International Conference on Communications Systems (ICCS), pp. 151–155, 2010.
[4] M.P. Wylie-Green, T. Svensson, “Throughput, Capacity, Handover and Latency Performance in a 3GPP LTE Field Trial”, IEEE Global Telecommunications Conference GLOBECOM, pp. 1-6, 2010.
[5] C. Ball, T. Hindelang, I. Kambourov, S. Eder, “Spectral Efficiency Assessment and Performance Comparison between LTE and WiMAX“, IEEE 19th International Symposium on Personal, Indoor and Mobile Radio Communications, pp. 1-6, 2008.
[6] Santiago Tenorio, Kyriakos Exadaktylos, Brendan McWilliams, Yannick Le Pézennec, “Mobile Broadband Field Network Performance with HSPA+”, 16th European Wireless Conference 2010, Lucca, Italy.
[7] Santiago Tenorio, Yannick Le Pézennec, and Manuel Sierra, "3G HSDPA evolution: MIMO and 64QAM performance in macrocellular deployments'' in Proc. European Wireless Conference 2008, Prague, Czech Republic, June 22-25. 2008, pp. 1-5.
[8] Christian Mehlführer, Sebastian Caban, and Markus Rupp, “MIMO HSDPA Throughput Measurement Results in an Urban Scenario” in Proc. IEEE Vehicular Technology Conference VTC2009-Fall, Sept. 2009, Anchorage, AK, USA.
[9] Mark Beach, and Mythri Hunukumbure, “Outdoor MIMO Measurements for UTRA Applications” in European Cooperation In The Field of Scientific and Technical Research, COST 273, May 2001, Espoo, Finland.
[10] M. Jurvansuu, J. Prokkola, M. Hanski, and P. Perala, “3G/HSPA Performance in Live Networks from the End User Perspective” in Proc. IEEE International Conference on Communications ICC’09, June 2009, pp. 1-6.