MIMO OFDM

Transcription

MIMO OFDM

MIMO OFDM
Space Time Coding – Spatial Multiplexing
Increasing Performance and Spectral
Efficiency
in Wireless Systems
Part I
Technical Basis
Part II
Draft-N (IEEE 802.11n) Performance
Measurements of Practical
Implementations
at the Institut für Rundfunktechnik
(Institute for Broadcast Technologies)
Hermann Lipfert, August 2007
Broadcast Networks and Servers
1
Summary........................................................................................................ 3
2
Comparative Measurements at IRT ............................................................... 4
2.1 Measurement Layout ............................................................................. 5
2.2 Pre-n Products....................................................................................... 8
2.2.1 Belkin MIMO Pre-n .................................................................... 10
2.2.2 Linksys MIMO Pre-n .................................................................. 11
2.2.3 Buffalo MIMO Pre-n................................................................... 12
2.2.4 D-Link MIMO Pre-n.................................................................... 13
2.2.5 Netgear Pre-n "RangeMax" ....................................................... 14
2.2.6 Edimax MIMO Pre-n .................................................................. 16
2.2.7 Level One MIMO Pre-n.............................................................. 17
SMC Barricade MIMO Pre-n ................................................................ 18
2.3 Differences in throughput by choosing different security methods ...... 19
2.4 Draft-n 1.0 Products............................................................................. 20
2.4.1 Linksys MIMO Draft-n................................................................ 22
2.4.2 Trendnet MIMO Draft-n ............................................................. 22
2.4.3 Netgear Draft-n.......................................................................... 24
2.4.4 Buffalo Draft-n ........................................................................... 27
2.5 Some detailed measurement results of a Draft-n-Product................... 28
2.5.1 Trendnet: TCP with Line-of-Sight .............................................. 28
2.5.2 Trendnet: Triple-Play measurements - Test Execution ............. 30
3
A concrete example : pre-n versus Draft-n................................................... 33
4
Conclusion ................................................................................................... 34
Page 2 of 34
© IRT – Broadcast Networks and Servers Hermann Lipfert
1
Summary
First implementations of Multiple Input Multiple Output (MIMO) technology in conjunction with
Orthogonal Frequency Division Multiplexing (OFDM) in next-generation Wireless Local Area
Networks (WLAN) show a significant increase in performance and spectral efficiency in
comparison to previous WLAN systems.
MIMO-OFDM also presents an exciting solution for improving the air interface performance for
Wireless Metropolitan Area Networks (WMAN) that have been set up based on the international
standard Wireless Interoperability for Microwave Access (WiMAX) and for fourth-generation (4G)
mobile cellular wireless systems.
In January 2006 the US standard-setting body, IEEE, approved a proposal of the Enhanced
Wireless Consortium (EWC) IEEE working group as IEEE-802.11n Proposal Draft.
This proposal required for the first time that MIMO-OFDM technology shall be used in wireless
communications (in addition to numerous other innovations). In the first and second quarters of
2006 the first so-called pre-n WLAN products were already appearing on the market. The core of
these products are the highly integrated MIMO chipsets from Airgo Networks Inc. and Ralink
technology Corp. The IRT (www.irt.de) compared eight of these products with each other and with
a standard 802.11g product.
The first "draft-n" products have been available since the third quarter of 2006 after the
acceptance of a new IEEE 802.11n draft. The basic difference between pre-n and draft-n is in the
fact that only the latter can be updated to the IEEE 802.11n standard (presumably ratified not until
mid-2009) by means of firmware and driver updates alone, without having to tamper with the
hardware.
Moreover, draft-n works towards raising transmission capacity by further increasing the number of
sending and receiving antennas and towards establishing cross-manufacturer methods for
securing minimum quality of service (QoS). Draft-n-compatible chipsets are offered by the
companies Broadcom Corp., Atheros Communications Inc., Marvell Technology Group Ltd. and
MIPS Technologies Inc. The IRT has conducted and is conducting performance measurements on
the latest draft-n 2.0 products available on the market (see Report Part II
http://www.irt.de/en/themengebiete/digitale-netze/mimo-ofdm.html and [13]).
Page 3 of 34
2
Comparative Measurements at IRT
The Institut für Rundfunktechnik (IRT) has conducted and is conducting performance
measurements on the latest pre-n and draft-n products available on the market (see also
www.tomsnetworking.de ). A real comparison of the new technologies with the existing 802.11g
standard can be made only if all measurements are performed at channel bandwidths of 20 MHz –
that is without channel bonding. Unfortunately, several implementations do not allow the restriction
of the sending spectrum to one WLAN channel. In selecting the sending location and the different
receiving locations, it was ensured that all site situations that occur in practice were mapped
against each other as far as possible. The measurements were performed in a two-dimensional
area, in other words on one floor of the building. For the duration of the measurement, all other
access points in the measurement range were switched off or converted to sending channels
whose spectral masks do not overlap in any way with the spectral masks of the measurement
channel.
Figure 1 shows the floor plan. The location of the MIMO router is shown in green. There are in
total seven locations and a measurement laptop with the corresponding MIMO PC card was
placed at each one. The separating walls of the individual rooms, as can be seen from the floor
plan, consist of brickwork about 15 centimetres thick and four fire protecting walls constructed out
of poured concrete 30 centimetres thick. The damping of these fire barriers thus corresponds to
the damping provided by conventional concrete ceilings or floors or two layers of thermal double
glazing.
Measurement locations (LOS, LOC1...6) MIMO Wireless LAN
Beton
Loc4
Standort
MIMO AP
concrete
Beton
LOS
Loc5
Loc6
Loc3
Loc2
concrete
Loc1
concrete
concrete
concrete
Wall thickness (brick + plaster) about 15 cm
Wall thickness concrete about 30 cm
Steel cabinets in many rooms
Figure 1: The WLAN measurement environment at the IRT on the upper floor of building 17B, with
its concrete and brick walls and book shelves with glass doors, represents a good model of the real
conditions of European, and in particular German, construction.
Page 4 of 34
2.1
Measurement Layout
The MIMO-AP was placed at a height of about 1.8 meters at the spatial position that can be seen
in Figure 1. The measurement locations of the MIMO laptops were each at about one meter above
the floor. A measurement location was located in the sending room at a three-meter distance with
direct line of sight (LOS); all other six measurement locations were non-line-of-sight locations
(NLOS).
A view of the building plan shows that wave propagation in the 2.4 GHz range within the floor is
marked with manifold dispersions and reflections. Since the send signal reaches the recipient at
NLOS in the most varied and indirect ways, this automatically leads to phase displacements and
variable weakening of the individual signals. The overlapping of the individual signals at the
receiving location (Figure 2) leads to signal fall-offs that fluctuate over time and frequency
(multipath fading). However, these negative effects of multipath fading in SISO systems can be
turned to a positive by using space time coding or spatial multiplexing in MIMO OFDM systems.
Multipath
Scattering
WLAN
Reflection
Receive
Figure 2: Principle of multipath propagation
Page 5 of 34
Figure 3 shows the measurement equipment used. The measurement software "IxChariot from
IXIA was used. This software is designed for professional performance and QoS measurement
based on protocols such as TCP or UDP.
IxChariot
Console
IxChariot
Endpoint 1
MIMO WLAN
802.11Draft-n
802.11 pre-n
Laptop PC
With MIMO card
IxChariot
Endpoint 2
Figure 3: The measurement equipment for the MIMO measurements at IRT
consists of a Gigabit Ethernet switch, to which the WLAN access point to be
tested is connected, a laptop with the counterpart to be tested – the WLAN client card,
as well as another laptop on which the measurement software is running.
[Independently of this system, a third laptop with WLANadapter monitors the WLAN spectrum.]
Channel 13 was selected for the measurement. However, for purposes of comparison each
product was randomly tested again on channel 6. There were no noteworthy changes in the
results. The MIMO PC card to be measured was inserted into a Siemens/Fujitsu Lifebook (model
S7010D S-Series Centrino). Random measurements were repeated and compared here as well,
using a Samsung X10 laptop. The results of the measurement were not dependent on the laptop
used here either. Windows XP Professional with current patches was used as the operating
system in all PCs.
Page 6 of 34
The driver and utility software for each individual MIMO PC card was installed on the respective
measurement laptop from the accompanying software CD and deinstalled after completion of the
measurements. Prior to the draft-n measurements conducted in November/December 2006, the
then most current firmware for the MIMO-WLAN router and the most current driver and utilities for
the MIMO client cards were installed from the respective manufacturer websites. The individual
measurements with the first five pre-n products took place in the first quarter of 2006. The
products which were available at the end of the 2nd quarter, and the draft-n products which were
available in the 4th quarter of 2006, and the draft-n 2.0 products which were available during the
2nd quarter of 2007 are shown and discussed in detail in this Part II of the report. Finally a Part III
will be available on the IRT Web Server. This Part shows the newest most udated measurements
with Draft-n 2.0 products available on the market. All measurements are done according to the
strict defined and crucial testing conditions specified by IRT together with the tomsnetworking guys
from germany (www.tomsnetworking.de). Part II is available and Part III will be available from
October 2008 in an always updated version at the IRT web-server:
http://www.irt.de/en/themengebiete/digitale-netze/mimo-ofdm.html)
TCP performance measurements were conducted at all seven locations. The transmission
direction was "downstream", in accordance with the expected applications, that is, directed from
the access point to the client. Each individual measurement lasted three minutes and each one
was performed three times. Bar graphs show the average values from the three measurements in
each location (and this is an average of 9 minutes length of time which means IRT examined very
carefully). By using the newest version of the measurement software, IxChariot (V6.4), test
transmissions could be done on the pre-n, draft-n 1.0/2.0 products themselves especially using the
“High Performance Throughput” script which is provided by IxChariot. This script is perfectly suited
for doing comparative measurements of maximum TCP performance over certain kinds of
networks. With predefined scripts from IxChariot also IPTV streaming, MPEG2 video streaming
and VoIP measurements can be done and highlighted according to quality criteria such as delay
and jitter. Several SD- and HDTV streaming measurements were carried out as well at the IRT
using data rates reaching from 4Mbps up to 20Mbps per stream (these measurements and graphs
are not added to the report Part II but were carried out in a diploma thesis in 2007 at IRT).
Page 7 of 34
2.2
Pre-n Products
Table 1 and Table 2 show the Pre-n products tested by IRT.
Manufacturer
Model APRouter
Model ClientCard
ChipManufacturer
Concept of
Improvement
Netgear
WPN 824
RangeMax
WPN 511
RangeMax
Atheros
Video54
ChannelBonding
Beamforming
Buffalo
Airstation
WZR-G108
Airstation
WLI-CB-G108
Airgo Networks
Airgo AGN-103
Spatial
Multiplexing
Belkin
F5D8230-4
F5D8010
Airgo Networks
Airgo AGN-103
Spatial
Multiplexing
D-Link
DI-634M
DWL-G650M
Atheros
AR5005VL
(Super-G)
ChannelBonding
Spatial
Multiplexing
Linksys
WRT54GX
SRX400
WPC54GX
SRX400
Airgo Networks
Airgo AGN-103
Spatial
Multiplexing
ChipManufacturer
Concept of
Improvement
Ralink
RT2529
RT2661
Spatial
Multiplexing
Table 1:
first test session of Pre-n-MIMO-products
Manufacturer
Model APRouter
Edimax
BR-6216Mg(G)
Level One
WBR-5400
WPC-0500
Ralink
RT2529
RT2661
Spatial
Multiplexing
SMC Barricade
SMCWBR14GM
SMCWCB-GM
Ralink
RT2529
RT2661
Spatial
Multiplexing
Table 2:
Page 8 of 34
Model ClientCard
EW-7608Pg
second test session of Pre-n-MIMO-products
Standard-IEEE-802.11g-Referece: Funkwerk Artem W1000
Figure 4: the model Artem W1000 from Funkwerk AG - CPD-XT-g.
In order to compare performance of the new MIMO-products with a standard 802.11g product, the
TCP performance measurements were conducted at all seven depicted measurement locations,
applying the same criteria as in the later MIMO measurements. These reference results are shown
in Figure 5. For this purpose the access point, Artem W1000 from Funkwerk, which is equipped
with two antennas, came into operation.
Standard 11g AP mit WPA-TKIP
25
20
15
Standard 11g AP mit WPA-TKIP
10
5
0
LOS
3m
Ort 1
ca.12m
Ort 2
Ort 3
ca.16m ca.16m
Ort 4
ca.18m
Ort 5
Ort 6
ca.28m ca.40m
Figure 5: standard 802.11g – WLAN Access-Point Artem W1000 from Funkwerk with its
TCP/IP throughput at 7 different locations
Page 9 of 34
2.2.1
Belkin MIMO Pre-n
Figure 6: Belkin Wireless Pre-N RouterF5D8230-4 with Airgo-Chipsatz AG-103 and 3 antennas
The WLAN-Pre-n-Duo from Belkin, consisting of access point F5D8230-4 and pc card F5D8010,
shows no significant performance improvement for LOS and NLOS-close up range, in comparison
to the standard WLAN G reference product. However for reaching locations 2 through 6, which are
subject to high attenuation and fading, the employed MIMO technique from Airgo Networks shows
highly improved throughput rates due to multipath receiving (Figure 7). For location 5 the
measured TCP throughput stays constant above 10 Mbps. The structure of the Belkin access
point equals exactly the Pre-n-routers of Linksys, as shown in the Test of Tom's Networking Guide
Germany. [11]
25
Belkin
20
15
Funkwerk Artem 11g WPA-TKIP
Belkin Wireless Pre-N MIMO WPA-TKIP
10
5
0
LOS
3m
Ort 1
ca.12m
Ort 2
ca.16m
Ort 3
ca.16m
Ort 4
ca.18m
Ort 5
ca.28m
Ort 6
ca.40m
Figure 7: Belkin-Acces-Point F5D8230-4 with True MIMO and Pre-n-Client-Card F5D8010 reaches higher
throughput at location 1 with NLOS than with LOS – an evidence for the operation mode spatial multiplexing.
Page 10 of 34
2.2.2
Linksys MIMO Pre-n
Figure 8: Linksys-Access-Point Linksys WRT54GX SRX – 3 antennas and built in Airgo-Chipsatz AN-103
For the Linksys Duo, consisting of access point Linksys WRT54GX SRX and client PC card
WPC54GX SRX, good performance results are measured even in the far field. For location 5
constant 15 Mbps can be obtained in comparison to zero connect ability for 802.11g, which is a
demonstrative prove of the functional capability of MIMO technique (Figure 9). These results are
also documented by the independently executed measurements from Tom’s Networking Guide
Germany. [13]
25
Linksys
20
15
Funkwerk Artem 11g
WPA-TKIP
Linksys SRX WRT54
GX WPA-TKIP
10
5
0
LOS
3m
Ort 1
Ort 2
Ort 3
Ort 4
Ort 5
Ort 6
ca.12m ca.16m ca.16m ca.18m ca.28m ca.40m
Figure 9: Pre-n-Duo from Linksys, composed of the Access-Point-Router WRT54GX SRX and
Client-PC-Card WPC54GX SRX reaches better throughput than Belkin’s Pre-n technique with
same chipset, Airgo-AG-103, especially at far distances.
Page 11 of 34
2.2.3
Buffalo MIMO Pre-n
Figure 10: MIMO Wireless Cable/DSL Router WZR-G108 from Buffalo AirStation
series is another pre-n-equipment comprising theTrue-MIMO-Chipset from Airgo Networks.
While the capability of the Buffalo pre-n-products in the close-up range turns out to be rather
moderate, the products demonstrate high throughput in reflection and multipath situations
(Figure 11).
25
Buffalo
20
15
Buffalo Airstation MIMO WPA-TKIP
10
5
0
LOS
3m
Ort 1
ca.12m
Ort 2
Ort 3
ca.16m ca.16m
Ort 4
ca.18m
Ort 5
ca.28m
Ort 6
ca.40m
Figure 11: Buffalo Airstation Router WZR-G108 combined with Airstation WLI-CB-G108 PC-Card pre-n
Page 12 of 34
2.2.4
D-Link MIMO Pre-n
The MIMO pair from D-Link (Figure 12) demonstrates very good results in the close up range.
Despite suffering from some performance dents in the mid range the overall picture of the D-Link
pair exhibits good throughput, also in the far field (Figure 13).
Figure 12: WLAN-Access-Point-Router DI-634M includes two external and
two internal sending and receiving antennas for MIMO-mode
35
D-Link
30
25
20
15
D-Link 108G MIMO DI-634M ohne Turbo
WPA-TKIP
10
5
0
LOS
3m
Ort 1
ca.12m
Ort 2
ca.16m
Ort 3
ca.16m
Ort 4
ca.18m
Ort 5
ca.28m
Ort 6
ca.40m
Figure 13: D-Link employs the "Super-G"-Chipset AR5005VL“ from Atheros. With its pre-n access
devices like access point router DI-634M and (pc card?)DWL-G650M it outperforms the Airgo technique
even at LOS
Page 13 of 34
2.2.5
Netgear Pre-n "RangeMax"
Figure 14a: Inside Access-Point WPN 824 from Netgear all antennas are fixed.
The rotating "clavilux" on top is cosmetic.
Figure 14b: Beneath its animated cover a PCB from Video54 is implemented. With its seven various arranged
stripline antennas it also allows beamsteering.
The RangeMax technique from Netgear works with channel bonding and beamforming (Figure 14
a/b). In the close up range channel bonding obviously generates high throughput but only
accounting of spectral efficiency. Due to the fact that RangeMax’s channel bonding, which
works with 40 MHz bandwidth and therefore allocates two times the original spectral range,
is not disengageable, the results of these measurements are not directly comparable with
other product’s results. For completeness reasons the Netgear results are presented in the
following.
Comparing these results exposes an interesting fact. Neither beamforming nor channel bonding
has the capability to keep up with True-MIMO in the far field. This is also proven in tests from
Tom’s Networking Guide Germany. [12]
Page 14 of 34
45
40
Netgear
35
30
25
20
Netgear Range Max WPN 824 WPATKIP Channel-Bonding
15
10
5
0
LOS
3m
Ort 1
ca.12m
Ort 2
ca.16m
Ort 3
ca.16m
Ort 4
Ort 5
ca.18m ca.28m
Ort 6
ca.40m
Figure 15: pre-n-WLAN Netgear Access Point WPN 824 Router from RangeMax series, in combination
with PC-Card WPN 511. RangeMax exposes inferior throughput at critical receiving locations compared
with True-MIMO products (like Airgo).)
The high throughput at LOS and location1 is outcome of Channel bonding and not of MIMO spatial
multiplexing. This is accounting of frequency spectrum and therefore the spectral efficiency is not
improved.
Page 15 of 34
2.2.6
Edimax MIMO Pre-n
Figure 16: With its pre-n-Router BR-6216Mg Edimax’s focus lies on the MIMO-Chipset from Ralink.
The first tested pair with Ralink chipset was the access point BR-6216Mg(G) in combination with
pc card EW-7608PG – both from Edimax (Figure16). In the close up range with LOS and NLOS
the Edimax pair achieves a throughput which is almost as high as Neatgear’s RangeMax, only
using half the bandwidth. In the far field the Ralink True-MIMO chipset is not able to keep up with
the performance of products equipped with Airgo chipset (Figure 17).
35
30
Edimax
25
20
15
Edimax MIMO Wireless Router BR6216MG(G) WPA TKIP
10
5
0
LOS
Radius
3m
Ort 1
Ort 2
Ort 3
Ort 4
Ort 5
Ort 6
Radius Radius Radius Radius Radius Radius
Figure 17: Edimax pre-n Access Point BR-6216Mg(G) with PC-Card EW-7608PG is exploiting
the spectrum very well in close up range.
Page 16 of 34
2.2.7
Level One MIMO Pre-n
Figure 18: Level One pre-n Router comprises the MIMO-Chipset from Ralink.
Although the pre-n products WBR-5400 and WPC-0500 from Level One (Figuer 18) are equipped
with the same Ralink chipset as the Edimax products, its throughput in the close up range turns
out to be smaller. On the other hand Level One’s Ralink performance is almost able to keep up
with products employing Airgo True-MIMO chipset in the far field (Figure 19).
30
Level One
25
20
15
10
Level One MIMO AP Router WBR-5400
WPA-TKIP
5
0
LOS
Ort 1
Ort 2
Ort 3
Ort 4
Ort 5
Ort 6
Radius Radius Radius Radius Radius Radius Radius
3m
Figure 19: Pre-n products from Level One with Ralink chipset – Access Point WBR-5400 and PC-Card
WPC-0500 demonstrate that in the far fieldt its throughput is by far better than Edimax’s Ralink-Chipset
offers.
Page 17 of 34
SMC Barricade MIMO Pre-n
Figure 20: SMC Barricade Router with Ralink MIMO-Chipset
The SMC-Duo Barricade WBR14-GM and SMCWCB-GM (Figure 20), also equipped with the pren chipset from Ralink demonstrates very good throughput results in close up range. In the mid
range its performance values are superior to all other tested competitors but in the far field the
throughput clearly falls below 5 Mbps (Figure 21).
35
30
SMC Barricade
25
20
15
SMC Barricade MIMO SMCWBR
14-GM WPA-TKIP
10
5
0
LOS
3m
Ort 1 Ort 2 Ort 3 Ort 4 Ort 5
Ort 6
Figure 21: access point router SMC Barricade WBR14-GM combined with pc card WCB-GM achieves
best results at medium distances.
Page 18 of 34
2.3
Differences in throughput by choosing different security methods
The performance analyses show that different security settings result in considerable differences
in net throughput performance (Figure 22). All measurement tests were performed with the
ciphering mode WPA-TKIP (WPA1). In Particular for many pre-n products a proper Hardware
support is missing, while it is already existent for mode WPA2-AES. This leads to the issue that
WPA-TKIP (preshared key) counts as most critical operate mode, because it still relies on the RC4
algorithm and (at least for all pre-n products in these tests) operates without any Hardware
support.
30
25
20
15
Buffalo Airstation MIMO WPA-TKIP
Buffalo Airstation MIMO WEP
Buffalo Airstation MIMO WPA-AES
10
5
0
LOS
Radius
3m
Ort 1
Ort 2
Ort 3
Ort 4
Ort 5
Ort 6
Radius Radius Radius Radius Radius Radius
Figure 22: Different security settings result in different performance – used modes WEP, WPA1-RC4,
WPA2-AES
Page 19 of 34
2.4
Draft-n 1.0 Products
With high expectations the tests of the first available draft n-products were carried out. Supplied
with the chipset “Atheros AR 5416”, which represents the so called XSPAN-technique, the
products are supposed to advance into new performance dimensions. For the first time this TrueMIMO chipset transmits three parallel sub data streams into one channel. All draft-n products
tested by the IRT are detailed in table 3.
Manufacturer
Linksys
Model AP-Router
WRT 300N
Model ClientCard
ChipManufacturer
Concept of
improvement
WPC 300N
Atheros
AR5416 XSPAN
Channel
bonding
Spatial
Multiplexing
TEW-631BRP
TEW-621PC
Atheros
AR 5416
XSPAN
Channel
bonding
Spatial
Multiplexing
UBICOM
StreamEngine
Technique
Netgear
RangeMax
Wireless Router
Gigabit Edition
WNR854T100ISS
RangeMax
Wireless
Notebook
Adapter
WN511T100ISS und
RangeMax
Wireless SB 2.0
Adapter
WN121T100GRS
Marvell
Topdog
88W8360
Channel
bonding
Spatial
Multiplexing
Buffalo
Nfiniti WirelessN Router &
Access Point
WZR-G300N
Nfiniti WirelessN Notebook
Adapter WLICB-G300N
Broadcom
Intensi-fi
Channel
bonding
Spatial
Multiplexing
Trendnet
Table 3: tested draft-n 1.0 MIMO-products
Colourful marketing inscriptions on top of packages, like “12 times faster” or “4 times the
distance”, or “up to 300 Mbps”, raise highest expectations, concerning performance jumps,
compared to standard 802.11 b/g. Even though the IEEE 802.11n standard does not explicitly
support Quality of Service (these features were implemented in the IEEE 802.11e standard), the
promised performance improvements can add to improved transmission properties. Preceding the
passing of the current draft-n standard was a tenacious struggle of two opposing syndicates which
lasted for more than a year.
Page 20 of 34
Only the intervention of the EWC in the end of 2005 brought movement into the effort of achieving
a standard. Also the fact that manufacturers are now pushing into the market, shows that
(especially the consumer industry) they are interested in and were waiting for improved wireless
solutions in the end consumer market.
The results the Linksys duo was showing were very disappointing, though. In the beginning the
broadband router WRT 300N and its counterpart the pc card WPC 300N together with the
provided software on the measurement laptop, were put into operation. In the close up range the
measured values stayed even below the 802.11g reference. Indeed the achieved throughput
stayed approximately constant for all measured distances. Also with an additional ap/pc card pair
no improvement was detectable. The functioning of the MIMO chip is supposedly not working
correctly. For the draft-n products the IRT was applying additional streaming tests on UDP (User
Datagram Protokol) basis, for the first time. These tests show a slightly better throughput attitude
as the TCP tests. A possible explanation for this low performance could be the TCP congestion
control with its “sliding window” mechanism in combination with MIMO. All transmitted sub data
streams are recombined into a single data stream at the receiver. In the Transport Layer all TCP
packets, which are sent block wise, are receipted with an acknowledgement (ACK) message. In
case of a transmission error the ACK message is hold back and the whole packet has to be
retransmitted. In the MIMO case a transmission error on only one of the sub data streams leads to
a transmission error of the recombined single data stream.
Page 21 of 34
2.4.1
Linksys MIMO Draft-n
Figure 23 : Linksys Draft-n Broadband Router WRT 300N
An update to the current software version (firmware version 2.0.0.17 for the router and software
version v2.14.05 respectively driver version 6.0.2.9 for the client card) brought for the Linksys
WRT 300N (Figure 23) considerable improvements in the close up range. The hoped for “11n
effect” still did not kick in, though.
The net TCP rate stays at 22 Mbps for LOS and NLOS-close up range. In mid range (location 3)
the results vary around 15Mbps and in the far field (location 4 through 6) the transmission rate
stabilizes above 6Mbps.
Since the disappointing performance in close up range seems much too low, compared to the
claims about considerably higher throughputs, by the company side, all tests with Linksys products
were discontinued. All comparing tests are going to be repeated with updated Firmware/Software
versions from Linksys and will be published in Part III.
2.4.2
Trendnet MIMO Draft-n
Figure 24: With WLAN Access Point Router TEW-631BRP Trendnet implements the StreamEngine from
Ubicom which is supposed to reduce transmission delays and jitter during streaming applications like VoIP,
video or gaming via a technique of prioritization.
Page 22 of 34
With access point Trendnet TEW-631BRP and pc card TEW-621 (Figure 24) the performance
jump, as expected from the new technique, was finally obvious. In close up range with LOS the
throughput was climbing onto more than 60 Mbps and for NLOS at location 1 it still offered 50
Mbps, which was definitely achieved with only one channel bandwidth of 20 MHz – not accounting
of spectrum (Figure 25). With this capacity the parallel transmission of a few HDTV streams is
thinkable. In the physical layer these products work with three parallel transmissions. Unfortunately
in the mid range and the far field the results came out rather meager. It seems like a lot of work
still needs to be done when it comes to the synchronization of software, firmware and hardware.
Figure 25 : Trendnet Draft-n devices show a top performance at close-up range - LOS and location 1.
At mid range and far field the deployed technique does even achieve the pre-n performance.
Page 23 of 34
2.4.3
Netgear Draft-n
The Netgear RangeMax Wireless Router Gigabit Edition WNR854T (figure 26) was tested in
combination with a pc card, as well as a USB2.0 wireless adaper. Testresults can be seen in figure
27 and 28.
While testing the Netgear WNR854T Wireless Router and WN121T USB2.0 Client-Adapter the
spectrum analyser Laptop Analyzer from Airmagnet detected that the connection is built up with a
bandwidth of 40 MHz, for LOS and location 1. For location 2 through 4 the connection is switched
to a 20 MHz bandwidth (lower S/N ratio  adaptive channel expansion). For location 5 a frequent
switching between OFDM and CCK modulation is identifiable by the spectrum displayed on the
analyzer. This implies that while still using a 20 MHz bandwidth the SINR is even lower here. For
location 6 finally, a higher modulation was used again. Here the SINR seems to be higher than at
location 5, which it also was in all other measured cases, for these two locations. The used
firmware on the router was version 1.3.44GR. Driver and configuration software was version
1.0.16.319 (31.10.2006). As pc card software, version 2.1.4.3 (4.10.2006) was used and as utility
software, version 1.1.8.11 (31.10.2006) installed. The driver version of the USB adapter was
1.0.3.7 (29.09.2006) at time of the measurement.
Figure 26: Netgear Draft-n RangeMax Wireless Router Gigabit Edition WNR854T-100ISS
is supposed to build up stable wireless connections via its Steady Stream Technique.
Page 24 of 34
100
90
80
70
60
50
40
DraftN Netgear WNR854T - WN511T PCCard im Kanal6, 40MHz-Kanalreite bei LOS
und am Ort1, 20 MHZ an den Orten 2-6,
WPA TKIP
30
20
10
0
LOS
Radius
3m
Ort 1
Radius
ca.12m
Ort 2
Radius
ca.16m
Ort 3
Radius
ca.16m
Ort 4
Radius
ca.18m
Ort 5
Radius
ca.28m
Ort 6
Radius
ca.40m
Figure 27: WNR845T AP in combination with a WN511T pc card. Netgear’s Draft-n technique comes
very close to Ethernet link speed at close-up range...
100
90
80
70
60
50
40
DraftN Netgear Router WNR854T mit
WN121T USB Adapter 40MHzKanalbreite bei LOS und Ort1, 20MHz
Kanalbreite Ort2-Ort5, WPA TKIP
30
20
10
0
LOS
Radius
3m
Ort 1
Radius
ca.12m
Ort 2
Ort 3
Ort 4
Radius Radius Radius
ca.16m ca.16m ca.18m
Ort 5
Ort 6
Radius Radius
ca.28m ca.40m
Figure 28: … and provides acceptable throughput at mid range with the USB-2.0 adapter WN121T.
The draft-n products from Netgear automatically allocate double the bandwidth (40 MHz) when
detecting good SNR values. Unfortunately the user itself may not decide how the available
spectrum should be used. The spectral diagrams underline the bandwidth usage at doubled
channel width (Figure 29 and 30).
Page 25 of 34
Figure 29: Used spectrum in case of channnel bonding (almost 40 MHz) like the example for
Netgear WNR854T shows
Figure 30: The Trendnet example shows how the spectrum looks like for a 20 MHz channel bandwidth.
Page 26 of 34
2.4.4
Buffalo Draft-n
With the Buffalo Nfiniti Router WZR-G300N (Figure 31) throughput data rates of over 30Mbps for
LOS and 25 Mbps for NLOS at location 1 (20MHz channel) are achieved in combination with client
pc card Nfiniti Wireless-N WLI-CB-G300N.
These results are not superb but still better than 802.11g. Disappointing results do the Buffalo
products show in more critical receiving situations, as for location 3 through 6 (Figure 32). A more
interesting comparison to an older Buffalo pre-n product can be seen in the later.
The access point WZR-G300N was tested with firmware version 1.44 (1.0.37-1.07-1.03) and the
client pc card Nfiniti Wireless-N WLI-CB-G300N with driver version 4.80.17.0 (15.05.2006), both
being the newest available software versions at testing time.
Figure 31: Buffalo Nfiniti Wireless-N Router & Access Point WZR-G300N exclusively operates in mixed
b/g/n-modes ( n-mode alone is not possible).
35
30
25
20
15
Buffalo WZR-G300N mit PC-Card WLI-CBG300N
10
5
0
LOS
Radius
3m
Ort 1
Radius
ca.12m
Ort 2
Radius
ca.16m
Ort 3
Radius
ca.16m
Ort 4
Radius
ca.18m
Ort 5
Radius
ca.28m
Ort 6
Radius
ca.40m
Figure 32: Buffalo Nfiniti Wireless-n Router WZR-G300N with Client PC-Card Nfiniti Wireless-N WLI-CBG300N. Results are better than for 802.11g in case of LOS and at locations 1 and 2. Performance
is meagre in more critical situations like locations 3 through 8.
Page 27 of 34
2.5
Some detailed measurement results of a Draft-n-Product
2.5.1
Trendnet: TCP with Line-of-Sight
Trendnet TEW-632 BRP shows a very high TCP throughput data rate in close up range in
cooperation with pc card TEW-621PC. While the average stays over 63.5 Mbps, peak values of
around 65 Mbps can be reached (Abbildung 27).
Figure 33: TCP-Throughput Trendnet MIMO Draft-n w LOS.
Page 28 of 34
Also with NLOS-close up range (location 1) very high performance is achieved (Figure 34).
Figure 34: Comparing TCP-Throughput measurements - Product: Trendnet MIMO at location 1.
Group/ Pair
Average
(Mbps)
Minimum
(Mbps)
Maximum
(Mbps)
Ort1_Messung1.tst
All Pairs
Pair 1
Ort1_Messung2.tst
All Pairs
Pair 1
Ort1_Messung3.tst
All Pairs
Pair 1
56,194
56,194
56,214
55,642
55,642
55,662
53,011
53,011
53,028
36,815
36,815
36,815
46,485
46,485
46,485
13,002
13,002
13,002
59,613
59,613
59,613
58,055
58,055
58,055
58,097
58,097
58,097
Throughput
95%
Confidence
Interval
Measure
d Time
(secs)
Relative
Precision
0,616
179,314
1,096
0,279
179,657
0,502
2,819
179,529
5,317
Table 4: Three performance measurements at Location1 – Product: Trendnet MIMO.
Page 29 of 34
2.5.2
Trendnet: Triple-Play measurements - Test Execution
The Trendnet TCP throughput measurements gained the highest performance values (for 20 MHz
bandwidth) with LOS and at location 1. Exemplary tables and diagrams are listed in the following.
These are showing the streaming behavior in the close up range without line of sight (location 1)
and in the mid range (location 4). In order to test QoS, Triple Play data streams were used. The
baseline group consists of DNS, HTTPtext and NNTP. The second group consists of five MPEG2
video streams which are send downlink. The third group consists of four IPTV streams, of which
two are send up- and two are send downlink. A fourth group consists of three bidirectional Voice
over IP connections with different codecs (Table 6).
Figure 35 : Throughput Triple-Play at Location1 – Product: Trendnet
The MPEG2 streams, each with 3.75 Mbps, are cleanly transmitted without data rate recession
(dashed line on the very top of figure 35). The four IPTV streams, each with 1 Mbps, are also
operated without errors (straight line at 1 Mbps in figure 35).
Page 30 of 34
The jitter is staying very low, below 2 ms – problems are only faced for values over 50 ms.
Figure 36: Jitter for Triple-Play at location 1 – Trendnet .
The MOS values of the VoIP connections are presenting excellent quality at location 1. All values
above 4 refer to excellent values on this scale.
Figure 37: MOS-values for VoIP connections at location 1 – Trendnet
Page 31 of 34
The one way delay stays below 4 ms and is therefore exceptionally small.
Figure 38: One-Way Delay – location 1 – Trendnet
Page 32 of 34
3
A concrete example : pre-n versus Draft-n
Pre-n versus Draft-n
35
30
25
20
15
Buffalo Airstation MIMO pre-n mit AigoChip WPA-TKIP
10
Buffalo Airstation Draft-N WZR-G300N PC-Card WLI-CB-G300N
5
0
LOS
Ort 1
Ort 2
Ort 3
Ort 4
Ort 5
Ort 6
Radius Radius Radius Radius Radius Radius Radius
3m
Figure 39: The knockdown results of a direct comparison: Draft-n 1.0 is only better at close-up range
(LOS, LOC1). Nevertheless in the case of difficult sending and receiving conditions the pre-n technique
from Airgo is demonstrating by far the best performance.
The blue bars in the diagram show the successfully transmitted data rates of the pre-n chipset
which’s market entry was over one year ago. All draft-n products currently available on the market,
seem to ignore that this True-MIMO technique from Airgo Networks is achieving a considerable
increase in throughput especially in difficult transmit and receive situations.
One can only hope that the promising approach from Airgo Networks will also be included into the
IEEE 802.11n standard and does not fail due to patent litigations, even after Airgo’s acquisition
through Qualcom. All comparable solutions from the draft-n standard partially stay far behind the
expectations.
Page 33 of 34
4
Conclusion
Since radio resources are scarce and data rate requirements keep increasing, spectral efficiency
is a stringent requirement in present and future wireless communications systems MIMO-OFDM
has become a new star in the constellation of wireless and mobile communications. Its potential to
increase spectral efficiency has not been reached by any other technology before. In addition to
increasing spectral efficiency, MIMO can also be used to reduce transmitting power while keeping
coverage areas constant. The use of MIMO technology in future transmission systems for
broadcasting, multicasting and unicasting represents real business logic also for broadcasting
corporations because of the possible reduction in transmission stations. The measurements
conducted at IRT also show, as described in [9], very good MIMO properties with line of sight in
the indoor area. What is important is that adequate multipath scattering and thus "gain-bringing
multiple reception" emerge. Something that has been regarded as an annoyance in transmission
in radio technology for a hundred years has become an advantage to users through the smart
application of physics and mathematics. Because signal transmission is always analogue – digital
bits come into being only through calculation.
This Part II of the report (Draft-N Performance Measurements) is also accessible under:
http://www.irt.de/en/themengebiete/digitale-netze/mimo-ofdm.html
Part III of the report (Copious Draft-N 2.0 Performance Measurements of topical 11n products) is
accessible from October 2007 under:
http://www.irt.de/en/themengebiete/digitale-netze/mimo-ofdm.html
Page 34 of 34

MIMO OFDM

Transcription

Similar documents

from archives-ouvertes.fr

Foodmaster Plaza

Centre East - Super Walmart

view the PDF - Kerfkore Company

Spring 2014 Parkside - Hermann Park Conservancy

japanese Temple Geometry Problems

Think Big_Routman

Study and Measurement of Scatter Effect on Wireless

2010 Annual Report - Farmingdale State College

COMPUTER BASED TESTING Evaluation of Question Classification