Go Faster WLAN 802.11n Jakob Ström, Jing Wang, Elpidoforos Arapantonis, Sareh Talebi Chalmers University of Tecnhology, Göteborg Email: {jakstr, jingwa, elpara, sareht}@student.chalmers.se

Abstract—The 802.11n protocol is an enhancement from the ten year older protocols 802.11b and 802.11g. The major change is the drastic increase in throughput - from 54 MBps to up to 600 MBps, depending on what optional parts of the protocol are used. This paper is the outcome of a project made to determine and analyze the parts of the 802.11n protocol that is responsible for the speed increase. The first part of the paper contains a litterature study and provides background information about 802.11n as well as comparisons to previous protocols. The second part contains the setup and results of various tests using a regular access point in the 2.4 GHz band intended for domestic use. The tests are made by transfering files from one computer to another using the access point. The tests were able to get results for the increase of using a Short Guard Interval (SGI) which proved to be 5% on average and 11% maximum. Testing double channels of 40 MHz failed due to nearby disturbance of a legacy WLAN (Wireless Local Area Network) which supports the conclusion that using 40 MHz channels in the 2.4 GHz is not recommended.

I. I NTRODUCTION Wireless Fidelity (Wi-Fi) is an industrial term of 802.11 IEEE Standard that released in 1997. Like cell phones and Televisions, 802.11 uses radio waves to provide a reliable, fast and secure wireless connection. This connection can be provided between electronic devices to each other, to Internet and even to a wired Ethernet network. [1] Communication through a wireless network is similar to a two-way radio communication. Using an antenna, a computer’s wireless adapter encodes data into a radio signal and transmits it to a wireless router, or access point. This router receives the signal and decodes it. To provide internet to the network, the router can be connected to the Internet by means of a Digital Subscriber Line (DSL) modem or a cable. Typically, it covers an area with 61 meters length around the router and obviously serves a better service to the computers closer to the router. [2] The 802.11 standards focus on the two bottom levels of the International Organization for Standardization (ISO) model. Maximum data rate of 2 Mbps supported by Wi-Fi was too slow for most applications. Over time the 802.11 family progressed and represented several versions. Since the first part of this paper is intended to peruse data rate improvement in 802.11n it is vital to have a general knowledge of its previous versions. The following protocols are the ones that improve speed. [3]

A. 802.11a - 1999 This protocol is a Physical Layer (PHY) standard. It supports data rate up to 54 Mbps. It uses Orthogonal Frequency Division Multiplexing (OFDM) modulation. Operating in the 5 GHz radio frequency band there is less risk for radio frequency Interference compared to following protocols. On the other hand, utilizing this radio frequency means that it provides a narrow network as well higher obstruction by obstacles such as walls. Due to its higher cost, 802.11a is usually used by corporations. The network can cover an area with a radius of 30 meters. 802.11a uses Binary Phase Shift Keying modulation (BPSK), Quadrature Phase Shift Keying (QPSK), 16-QAM (Quadrature Amplitude Modulation) and 64-QAM sub-carrier. The total band width is 20 MHz. [1] B. 802.11b - 1999 802.11b supports a maximum throughput of 11 Mbps which although lower than that of 802.11a is still comparable to the wired Ethernet. It utilizes 2.4 GHz radio signaling frequency which is same as the original 802.11 standard. Although it solves the problem of penetrating obstructions, most of the home appliances like microwave ovens and cordless phones use the same frequency band. Consequently this protocol suffers the interference problem. Low price of this frequency motivates vendors to prefer using it to lower their production costs. 802.11b can an area with a radius of 300 meters. [1] To provide higher data rates, 802.11b uses CCK (Complementary Code Keying) which is a modulation technique that makes efficient use of the radio spectrum. [3] Since 802.11a and 802.11b use different frequencies, it is said that they are compatible which means they can be implemented side by side not together. In fact, each connected device can merely use one. [1] The 802.11b specification affects only the physical C. 802.11g - 2003 802.11g is the third version of the 802.11 series. It uses the best parts of its predecessors. Consequently, it supports the data rate up to 54 Mbps like 802.11a, and uses a radio frequency of 2.4 GHz like 802.11b. Although this frequency band would not be obstructed, the problem of interference of home appliances still resists. The 802.11g standard is said to be backwards-compatible with the 802.11b standard, which means the devices that support the 802.11g standard can also work with 802.11b. [1] Since access points using 802.11g use OFDM, they cannot hear 802.11b. A mixture of 802.11b and

802.11g requires RTS / CTS (Request-To-Send / Clear-ToSend) to avoid collision. These messages provide substantial overhead and lowers outcome significantly for both 802.11b and 802.11g users. [3] Like 802.11a, 802.11g uses OFDM modulation scheme and supports data rates of 6, 9, 12, 18, 24, 36, 48, and 54 Mbps. It reverts to complementary code keying (CCK) (like the 802.11b standard) for 5.5 and 11 Mbps. Differential Binary Phase Shift Keying (DBPSK) modulation, Differential Quaternary Phase-Shift Keying (DQPSK) modulation and Direct-Sequence Spread Spectrum (DSSS) modulation technique are used in 1 and 2 Mbps data rate. [1] D. 802.11n - 2009 802.11n not only dramatically increases WLAN speed by more than 10 times up to 600 Mbps but also improves reliability and extends the range of wireless transmission. Although the Multiple Input / Multiple Output (MIMO) technology plays the most significant role - boosting the speed with a factor of 4, there are some other changes in comparison with the previous versions. The next parts of this paper will be an analysis of these improvements.

Fig. 1.

DPZ Allocation Diagram [4]

This section will cover the smaller improvements that together increase the bitrate from 54 Mbps to 72 Mbps. The parts that are improved are DPZ (Data, Pilot, Zero), BCC (Binary Convolutional Code) and SGI (Short Guard Interval). [4]

which is removing some of the redundant bits. The parameter that is responsible for the removal of the redundant bits is the coding rate. The definition of the coding rate can be expressed, “as the number of data bits transmitted as a ratio of the total number of coded bits”. As an example a convolutional code with R=3/4 has only 25% of thee bitsredundant [5], [6] The improvement in 802.11n in comparison with the older protocols, is on the puncturing because we can use 40% with R = 5/6 and as a result of this is to increase the data rate from 58.5Mbps to 65Mbps. [4]

A. Data Pilot Zero

C. Short Guard Interval

Data, Pilot, Zero is a way to increase the data rate, by discarding some zeros between the channels. In Figure 1, the DATA frequencies are represented with blue color, the PILOT frequencies in red color and finally between the channels we have the ZERO frequencies. [4] The DATA frequencies are carrying the data of the transmission, PILOT frequencies-signals are used, for synchronization or control reasons and usually have a single frequency. In 802.11n 4 sub-carriers, used as PILOT for “phase and frequency tracking and training ”. [1] Finally the ZERO are some frequencies which are not include any data inside them, and that’s why we represent them with zeros. [1] In the older protocols (802.11a/g), between the channels the gap was 11 ZERO frequencies, but in the 802.11n, this number is decreased in 7 ZERO frequencies. With this change the data rate is improved from 54 Mbps to 58.5 Mbps. [4]

The last improvement is the implementation of the SGI (Short Guard Interval). The GI (Guard Interval) is the time space between 2 OFDM symbols (Figure 2). We add this time space between them because during the transmission echoes, or reflections from other signals or environmental parameters could occur and this could lead, to ISI (Intersymbol Interference), which is the result when one symbol interfere with another. [1] The procedure to create a GI is not complicated because it is uses the output of the IDFT (Inverse Discrete Fourier Transform) of the signal. In the older protocols a percentage of this output (25%) it is attached from the end to the beginning of the symbol. [4] In 802.11n the data symbol has 4µs duration, which is the sum of the 800 ns of GI plus 3.2µs of data. As an optional imrpovement SGI can be activated to increase the data rate. [4] SGI reduces the 800 ns GI to a 400 ns GI, which boosts the data rate by 11%. Instead of attaching the 25% of the output of the IDFT, in 802.11n only attaches 12.5%. [1], [4]

II. S MALL I MPROVEMENTS

B. Binary Convolutional Code Another improvement is the BCC. In telecommunication these codes are used as a type of controlling the data transmission. If we want to analyze more, we can say that “permit reliable communication of an information sequence over a channel, which has noise, and provokes the distortion of the transmitted signal”. [5] So the function of this code is to protect the information, by adding redundant bits. After the BCC exist a puncturer

III. OFDM AND BANDWIDTH INCREASE A. OFDM DPZ and GI are part of OFDM and to understand these parts as well as OFDM, it is important to first understand FDM (Frequency Division Multiplexing).

IV. M ULTIPLE I NPUT M ULTIPLE O UTPUT

Fig. 2.

Guard Interval [7]

When transmitting over a channel, one can using FDM divide the channel into smaller sub-carriers. Since the single carrier is split up, this means that the data stream must be modulated across these sub-carriers. For example, if a subcarrier experiences fading at some point that sub-carrier can be amplified without affecting other sub-carriers, or less data could be sent over it. [1] For FDM to be possible the sub-carriers have to be nonoverlapping with each other. This means introducing a guard band between each pair of sub-carriers which will spread them out over the channel. The downside of this is that a significant amount of bandwidth will be unused to serve as separators. In OFDM, each sub-carriers are transmitted in mutually orthogonal frequencies, meaning that the top of one sub-carrier will coincide with the null of an adjacent sub-carrier. Because of this, there will be no ISI between them which diminishes the need for a guard band. This way a larger part of the band can be used for transmitting data thereby increasing the Spectral Efficiency of the channel. B. Bandwidth Increase 802.11n is the first of the 802.11-series that allows merging two adjacent 20 MHz channels as one 40 MHz channel. This effectivly doubles the throughput, and is the change that gives the second largest increase in throughput in 802.11n, after MIMO. The 2.4 GHz band has room for one 40 MHz channel, while the 5 GHz band can fit 11 non-overlapping channels. This makes the 5 GHz band vastly superior, although the 2.4 GHz band is still useful to provide legacy support. [8] Of course, having channels that are twice as wide as before will limit the amount of channels that are available. However, the cost of the equipment is barely increased with this change, and as long as there is enough free spectrum the bandwidth increase is the easiest and most efficient change in 802.11n [1]. If the limitation on the number of channels proves to be a problem, 20 MHz can be used since the increase is optional. There may also be a problem with interference using 40 MHz channels in the 2.4 GHz band, as is shown in the testing later in this paper.

The final part of the study is going to be about Multiple Input Multiple Output (MIMO), which is the most important improvement of the data transfer rate in the 802.11n and boosts the speed up to four times (4x4 MIMO). [9] MIMO technology is used in 802.11n to evolve the existing OFDM physical interface presently implemented with legacy 802.11a/g. MIMO exploits the use of multiple signals transmitted into the wireless medium and multiple signals received from the wireless medium to improve wireless performance. [4] MIMO can provide many benefits, all derived from the ability to process spatially different signals simultaneously. Two important benefits explored here are antenna diversity to improve system performance and spatial multiplexing used to significantly increase data rate. [9] Using multiple antennas, MIMO enables the opportunity to spatially resolve multipath signals, providing diversity gain that contributes to a receiver’s ability to recover the intelligent information. Another valuable opportunity MIMO technology provides is Spatial Division Multiplexing (SDM). [9] SDM spatially multiplexes multiple independent data streams, transferred simultaneously within one spectral channel of bandwidth. MIMO SDM can significantly increase data throughput as the number of resolved spatial data streams are increased. Please refer to Figure 3.

Fig. 3.

Basic MIMO system diagram [4]

Consider the Basic MIMO system diagram in Figure 3. A compressed digital source is fed into a simplified transmitting block with the functions of error control coding and mapping to complex modulation symbols. The latter produces several separate symbol streams which are independent. Each is then mapped onto one of the multiple TX antennas. Mapping may include linear spatial weighting of the antenna elements or linear antenna space-time precoding. After frequency conversion and amplification, the signals are launched into the wireless channel. At the receiver, the signals are captured by possibly multiple antennas and demodulation and demapping operations are performed to recover the message. The level of intelligence, complexity, and a priori channel knowledge used in selecting the coding and antenna can be varies depend one real system requirement. [9] In order to implement MIMO system into 802.11n, some new function blocks are introduced such as Stream Parser (SP) and Spatial Mapper (SM). It is showed in Figure 4. All these new blocks are aimed to improve the system performance in an intelligent way. At the end, the data rate in 802.11n can be boosted up to four times to 600 Mbps which is a huge improvment compared to the previous protocols.

Fig. 4.

Impletemented blocks for MIMO (4 x 4) [4]

V. T ESTING

C. Test III: 40 MHz The last test involves increasing the channel bandwidth from 20 MHz to 40 MHz. • 40 MHz • SGI (Short Guard Interval) 400 ns • 2x2 MIMO Doubling the bandwidth provides an increase of 100% in throughput which brings the expected theoretical throughput up to 300 Mbps. VI. T EST RESULTS AND CONCLUSION

To find out how much difference the optimizations in the new protocol would do in practice, the second part of the project involved testing using two computers and an access point. The access point was a Netgear WN802Tv2 supporting 802.11n-draft in the 2.4 GHz band. It was placed in the middle of two computers placed 25 meters from each other. The test was done by transfering a large file for a duration of approximately two minutes from one computer to the other, and using a network traffic monitoring tool to measure the throughput. The test was set up in a basement of Chalmers with thick concrete walls to attempt to limit the amount of radio signals from other WLANs. The signal strength of the large NOMADnetwork was barely noticable, below -80 dB. However the signal of a smaller 802.11g network in an adjacent room was measured to -69 dB which proved to be fatal to one of the tests. The tests were made by turning the optional settings of 802.11n on and off using configuration settings in the software of the access point. In this access point, there exists configurations for GI and channel bandwidth but unfortunately there was no way to alter MIMO-settings, so all of the tests were conducted with 2x2 MIMO. Fig. 5. Results of Test I and II. First curve is Test I, second curve is Test II.

A. Test I: Default test The first test used the following settings: • • •

20 MHz LGI (Long Guard Interval) 800 ns 2x2 MIMO

The theoretical speed of 802.11n using none of the optional settings is 65 Mbps, and using 2x2 MIMO increases this speed by 100% to 130 Mbps. B. Test II: Short Guard Interval The second test involved using the same settings as before, but using a short guard interval of 400 ns: • • •

20 MHz SGI (Short Guard Interval) 400 ns 2x2 MIMO

Using SGI over LGI gives a theoretical increase of 11% [1] bringing throughput up to a maximum of 144.4 Mbps.

The results of Test I and II can be viewed in Figure 5. The throughput of Test I was measured to an average of approximately 25.5 Mbps with a maximum of 26.3 Mbps. Test II measured a slightly higher throughput of 27 Mbps on average and a maximum of 29.3 Mbps. This means that the average bit rate was increased by 5%, and the maximum by 11%, the latter being the theoretical increase of using SGI. The reasons why the measured bit rate is so much lower than the expected theoretical bit rate are that there other WLANs present - albeit with low signal strength. Also, the walls of the corridor reflected the test signals which provided even more obstruction. Studying the curves further, one can see that the curve of Test II was more unstable than that of Test I. This might point to the downside of using a short guard interval which is the fact that it is more likely that ISI is introduced leading to a higher error rate. Test III using 40 MHz channels did not render any result at all due to that it proved to be impossible to get a stable enough

connection to be able to transfer files. Pinging one computer from the other showed a 50-80% packet loss. This can be explained by the presence of the other 802.11g networks. It

Fig. 6. Figure of the three non-overlapping 20 MHz channels (red blocks) of the 2.4 GHz band, as well as the right-most and left-most 40 MHz channels (green blocks).

turns out that the presence of a 802.11g network can degrade the signal of a 802.11n network by up to 85% [10]. As can be seen in Figure 6, any 40 MHz channel covers a large part of the 2.4 GHz band. Since channel 1, 6 and 11 are the three non-overlapping 20 MHz channels they are also the most commonly used. Channel 6 being in the middle overlaps with all of the 40 MHz channels and therefore heavily obstructs the 802.11n network despite having a low signal strength. This fact along with the fact that there is only one non-overlapping 40 MHz available in the 2.4 GHz band, makes this band lacking for full use of 802.11n. To sum up, the following improvements make up for the throughput increase of 802.11n. 1) Using MIMO. Data rate is increased by 100% for every additional antenna used. 2) Decreasing the number of zeros in the DPZ part. 3) Increasing the puncturing from 25% in 802.11g to 40% in 802.11n in the BCC part. 4) Using spatial multiplexing which means that multiple data streams are transmitted at the same time and on the same channel. 5) Using Channel bonding which means combining two adjacent channels, effectively doubles the amount of available bandwidth. However, using double channels (40 MHz) in the 2.4 GHz band is not recommended as the network will become easily obstructed by any presence of legacy WLAN. 6) Using an SGI of 400 ns instead of an LGI of 800 ns. R EFERENCES [1] E. Perahia and R. Stacey, Next Generations Wireless LANs Throughput Robustness and Reliability in 802.11n. Cambridge University Press, September 2008. [2] J. Geier, “How Wireless Works,” October 2004. [Online]. Available: http://www.ciscopress.com/articles/article.asp?p=344242 [3] ——, “802.11 Alphabet Soup,” August 2002. [Online]. Available: http://www.wi-fiplanet.com/tutorials/article.php/10724_1439551_1 [4] F. Brännström, “802,11a/b/g/n,” Lecture in Introduction to Communication Engineering, October 2003. [5] R. D. Wesel, “Convolutional Codes,” Wiley Encyclopedia of Telecommunications, 2003. [Online]. Available: http://www.ee.ucla. edu/~wesel/documents/Misc/eot309.pdf [6] M. S. Gast, 802.11 Wireless Networks: The Definitive Guide, Second Edition. O’Reilly Media, Inc, April 2005. [7] N. I. Corporation, “Tutorial for OFDM and Multi-Channel Communication Systems,” January 2007. [Online]. Available: http://zone.ni.com/devzone/cda/tut/p/id/3740

[8] F. Networks, “802.11n Primer,” August 2005, airmagnet. [Online]. Available: http://www.airmagnet.com/assets/whitepaper/WP-802. 11nPrimer.pdf [9] A. Goldsmith, Wireless Communications. Cambridge University Press, May 2005. [10] V. Shrivastava, S. Rayanchu, J. Yoon, and S. Banerjee, “802.11n under the microscope,” Proceedings of the ACM/USENIX Internet Measurement Conference (IMC ’08), 2008.

A PPENDIX R EVIEW QUESTION Q: What are the two most significant changes with the new WLAN protocol 802.11n that make up for the speed increase up to 600 Mbps? A: A bandwidth increase from 20 MHz to 40 MHz and the introduction of MIMO.