This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
Abstract-The hype with respect to MIMO systems is due to the theoretical analysis which have been reported, stating that such systems may increase the channel capacity by a factor equivalent to the smallest value between Nt and Nr, where Nt and Nr are the number of transmit and receive antennas respectively. Moreover, an improvement in the range may be observed if the number of antennas is increased at both the transmitter and the receiver, without demanding more transmit power. However, point-to-point long distance MIMO links are limited by the "keyhole" effect which results in a degenerate channel.
The use of polarization diversity or artificial scatterers was proposed to eliminate the "keyhole" effect. In this analysis the effect of polarization diversity on the throughput of an IEEE 802.11n Draft 2.0 MIMO system for point-to-point long distance links is reported. IEEE 802.11n is an amendment to the IEEE 802.11 family of standards employing MIMO technology at the PHY layer besides improving the MAC layer to reduce the overhead incurred from acknowledgements and taking advantage of frame aggregation .
The outcomes prove the hypothesis that a 2x2 MIMO system consisting of cross polarization at both ends of the communication link will mitigate the "keyhole" effect. Moreover, it also highlights the efficiency of IEEE 802.11n draft 2.0 via comparison of SISO systems utilizing IEEE 802.11g and IEEE 802.11n. Another point which resulted from the measurement campaign adopted was the improvement with respect to throughput that one can achieve by using 40 MHz channel spacing.
A communication system with more than one antenna at both the transmitter and the receiver end, known as a Multiple-Input Multiple-Output (MIMO) system, has the potential of augmenting the channel capacity, reducing the bit error rate, increasing the coverage area and decreasing the transmit power requirement. One must keep in mind that all the advantages mentioned, some of which cannot be present at the same instance, are at a cost of increasing the number of antennas and the radio frequency (RF) chains. This increase in cost is especially highlighted if despite the correct installation of a MIMO system, which must consist of decorrelated antennas, only one spatial path is present between the two ends of the communications system leading to the "keyhole" effect. This will occur when the communications range considered is large and hence one has to opt for other techniques that will aid the system to achieve the increase in range and throughput promised by MIMO systems.
MIMO technology has already been employed in the 3rd Generation Partnership Project (3GPP) to achieve 20Mbps of data transmission in high speed downlink packet access (HSDPA). Moreover it is one of the techniques adopted by the IEEE 802.16 and the IEEE 802.11n standards. Before the ratification of the 802.11n standard was through, 802.11n draft devices were already on-the-shelf solutions for home Wi-Fi systems, with the number of antennas ranging from 2 to 3. Before the use of multiple antenna elements to receive different spatial streams at the same instance, diversity techniques were employed via transmitting or receiving different copies of the same signal using more than one antenna at the transmitter or receiver respectively, resulting in diversity gain. The importance of such techniques is due to the small probability that different spatial paths are affected by the same amount of fade, hence the communications system can be less sensitive to fading, co-channel interference and error bursts. The theoretical throughput of 802.11n using a 2x2 MIMO system is predicted to be 100Mbps as a result of the implementation of spatial multiplexing and other physical and medium access control layer amendments. The use of 40MHz channels in both the 2.4GHz and 5GHz band will also augment the throughput.
The "Keyhole" Effect
Winters , Foschini  and Telatar  were the pioneers who highlighted the potential of MIMO systems, predicting an increase in the spectral efficiency given that the wireless system has a rich scattering environment. Initially the work was focused on an analysis of the effect of MIMO on Shannon's capacity proving that this shall increase implying that the maximum data rates that can be used in wireless communications yielding a small error of probability have increased . Applications such as broadband wireless access  and HSDPA for the 3GPP  demand high data rates.
The rank of the channel matrix, H, which determines the number of spatial stream which can be received depends on the correlation of the antenna elements at both the transmitter and the receiver. However, decorrelated antennas do not guarantee that the rank of H will be greater than one and hence a low ergodic capacity is still possible. This was first analysed in 2000 by a group from Bell-Labs . Eventually other research groups published work in this area with the aim of analysing what is the cause of such an event, modelling of the channel and how to eliminate such an outcome i.e. the "keyhole" effect known also at the "pinhole" effect.
It was proved that the "keyhole" effect can be observed in both indoor and outdoor environment and hence this degenerate channel is not limited to long distances only. This effect can be present in hallways and tunnels. As explained in  this environment acts like an "overmoded" waveguide limiting the degrees of freedom of the channel matrix, H. For the outdoor environment it has been explained that this effect is controlled by the scatterers around the mobile stations (MSs) affecting the angular spread of the base station (BS).
Chizhik et al. have stated in  that the "keyhole" effect occurs if there is diffraction at a horizontal edge for a vertically separated antenna array and at a vertical edge for horizontally separated antennas. Examples of such edges are roof top and corners of tall buildings. In such cases increasing the separation of the antenna elements will not increase the channel capacity. One may opt for antenna elements which are separated in the same alignment as the edge which will be causing the diffraction, however the probability is that both horizontal edges and vertical edges will cause the diffraction as it was mentioned in . This increase in diffraction will aid to improve the rank.
A mathematical approach was adopted by Gesbert et al.  to analyse the "keyhole" effect. The parameters which are considered by the model proposed are the effect of scatterers, scatterer radii at the transmitter and the receiver, antenna beamwidths, antenna separation, angle spreads and range. It was stated that an environment with substantial amount of scatterers does not necessary result in a high rank channel. Gesbert et al. also stated that MIMO systems have a limitation with regards to capacity for a green-field channel i.e. a line of sight (LOS) environment such that if the distance between the transmitter and the receiver is large, H will have only one degree of freedom. In this paper it was highlighted that in order for spatial streams to maintain their spatial signature when the environment consists of a considerable number of scatterers, the multiplexing gain is maintained by the scatterers not by the antenna elements. However, the elements still need to be decorrelated. Gesbert et al. also proposed in  that cross polarization is one of the ways how to eliminate the "keyhole" effect, even though this will limit the channel capacity increase to a doubling.
The aim of the work in this paper is to prove that using two pairs of antennas with different linear polarization would achieve approximately double the throughput that can be observed in a MIMO system with the same number of antennas but without polarization diversity for a long distance link; hence overcoming the "keyhole" effect. Both indoor and outdoor measurements were taken during this measurement campaign .
A wireless network was setup for this measurement campaign, composed of a Conceptronic C300BRS4 802.11n wireless broadband router, a Conceptronic C300Ri 300Mbps Wireless Peripheral Component Interconnect (PCI) card, a laptop, and a computer. Independent of the environment considered, the apparatus was always the same except for the antennas used i.e. 2 dBi antennas for the indoor environment and 4 high gain antennas for the outdoor environment.
The Conceptronic C300BRS4 wireless router and C300Ri Wireless PCI card have similar specifications both operating at a wireless frequency range of 2.412GHz to 2.4672GHz. In the C300BRS4 user manual , it is stated that a range of 100m in the indoor environment and 400m for the outdoor one can be reached using the full transmit power and the three 2dBi antennas that are enclosed with the router. The Ralink chipset used by these devices implements spatial diversity using Maximum Ratio Combining (MRC) and Orthogonal Frequency Division Multiplexing (OFDM) (MRC-OFDM). During the analysis, it was discovered via observing the transmitted signals using a high frequency oscilloscope, that the Ralink chipset has only 2 transceiver RF chains. Hence the router and the PCI card can form a 2x3 system in which both spatial multiplexing and receive diversity can be implemented.
Irrespective of whether the setup was used indoors or outdoors, the C300BRS4 router was connected to the laptop via a 100Mbps Ethernet cable. The PCI card was installed in the computer to connect the computer to the router via a wireless link. As already mentioned the 2dBi dipole antennas were used for the indoor measurements whereas for the outdoor environment four 2.4GHz polarized Yagi directional antennas were needed. The specifications of the outdoor antennas are specified in the Table I:
Properties of the Outdoor Antennas
12dBi - 15dBi
Voltage Standing Wave Ratio
Half Power Beamwidth in the horizontal plane and in the vertical plane
From 25Â°- 50Â°
The throughput, delay jitter and the percentage of datagrams lost were measured using Iperf  opting for User Datagram Protocol (UDP) data streams. Netstumbler 0.4.0  was used to observe the signal strength and the SNR of the communication link between the PCI card and the router and any other surrounding wireless networks which act as noise and cause interference to the analysis. Throughout all the measurement campaign the computer transmitted the datagrams, i.e. the computer was set as the client and the laptop was set as the server.
Iperf is capable of transmitting both UDP data streams and Transmission Control Protocol (TCP) data streams however; the former type was opted for because of the reduced overhead incurred. In this analysis the focus is on the maximum throughput that can be achieved using Single-Input Single-Output (SISO) or MIMO systems with or without cross polarization. Hence, as was stated in , UDP data traffic results in a throughput closer to the analytical one than TCP does.
The hypothesis on which this measurement campaign was built is that if one uses two pairs of antennas exhibiting both vertical and horizontal polarization, a doubling in throughput can be observed for a long distance MIMO link; hence overcoming the "keyhole" effect . The theoretical throughput that can be achieved for a 2x2 system with 20MHz 802.11n channels and a PHY data rate of 144Mbps is 100Mbps . The first step of the practical side of the analysis was to measure the maximum throughput of the WLAN built using the mentioned devices in Section III-A, in a controllable environment i.e. indoors. This was required so that comparisons of the measured parameters in the outdoors to those obtained in the indoors could be performed. Hence, the environments in which the measurements were obtained can be divided in two:
The indoor environment;
The outdoor environment.
Independently of the environment the measurements obtained were the following:
1x1 802.11n with 20MHz channels
1x1 802.11n with 40MHz channels
2x2 802.11n with 20MHz channels
Indoor Measurements: The indoor measurements were performed in a domestic setting. Three sets of experiments were performed with the objective of
Analyzing the ideal buffer size and UDP datagram size to be used for the described WLAN in order to observe the maximum throughput possible. The buffer sizes, for which measurements were obtained, are 0.06Mbyte, 1Mbytes and 2Mbytes. The datagrams used had size of 0.25Kbytes, 0.5Kbytes, 0.75Kbytes, 1Kbytes and 1.25Kbytes.
Observing the variations of the throughput to changes in UDP datagram size for SISO system considering both the legacy 802.11g and 802.11n standard. For the latter standard both the 20MHz channel and the 40MHz channel were considered. For this and the following measurement the buffer size used was that of 1Mbytes.
Studying the effect of distance between the transmitter and the receiver for both SISO and MIMO system. The distance was increased by 4m in a corridor with the total range varying from 4m to 12m. The buffer size and the UDP datagram size used are 1Mbytes and 1Kbytes respectively.
Throughout these measurements the inter-element spacing at the broadband router measured at the tips of the 2dBi dipole antennas was that of 25cm whereas for the PCI card 17cm separations were used. For the first two analysis, i.e. the two which do not consider the effect of distance on the throughput, delay jitter and percentage of datagrams lost, the distance between the two ends of the communications system built was that of 2m.
Outdoor Measurements: The outdoor antennas used were 4 directional Yagi high gain antennas. The effect of different polarizations on throughput as the distance is increased was observed by obtaining measurements for 
A SISO system with vertical polarization - Fig. 1.(a)
A SISO system with horizontal polarization - Fig. 1.(b)
A 2x2 system with only vertical polarization - Fig. 1.(c)
A 2x2 MIMO system with horizontal polarization and vertical polarization at each end of the communications system - Fig. 1.(d).
Fig. 1. The different polarization configurations tested
4 different environments were analysed
The rooftop of the Engineering Faculty at the University of Malta at tal-Qroqq;
A link between the University of Malta and Hastings Gardens in Valletta;
The inter-element spacing used was that of 1.40m.
The first two outdoor locations proved to be highly noisy as a result of the densely populated areas . At Ä¦al Far three distances were considered (i) 230m, (ii) 700m and (iii) 1.17Km. For the first two locations set-up at Ä¦al Far, a LOS link was possible unlike the last location, where a construction site was hindering this link. For the remaining location, i.e. Ä¦ad Dingli, a link of 1.50km was setup, with the server end located near the radar dome atop the Dingli cliffs at Panoramika Street .
The WLAN setup had only one wireless link between the C300BRS4 and the C300Ri PCI card, hence the throughput, delay jitter and percentage of datagrams lost measured were related solely to this wireless link. Only the outcome of those commands which resulted in the maximum observed throughput for the described location is listed in this section. A number of tests were conducted for each scenario so that the highest throughput possible with the smallest percentage of lost datagrams could be reported. When this was determined the command was executed twice for 10 minutes. In this section the IP address of the server is referred to as 192.168.0.100 in the listed commands. The unit of throughput is Mbps and that of the delay jitter is ms.
Fig. 2. Map showing the geographical layout at Ä¦ad Dingli
Fig. 3. The MIMO system at Ä¦ad Dingli (a) the server and (b) the client side
The indoor results have proved that with a 2x2 MIMO system using the 802.11n Draft 2.0 standard and 20MHz channel spacing, a maximum throughput of 93Mbps can be achieved. This was observed at a distance of 2m. The delay jitter was that of 0.914ms and the percentage of datagrams lost was that of 0.13%. This outcome was obtained using a buffer size of 1Mbyte and a UDP datagram size of 1Kbyte. The maximum throughput observed is close to the theoretical throughput predicted by  of 100Mbps. The results of the throughput obtained using different buffer sizes and UDP datagram size in the mentioned environment are depicted in Fig. 4.
Fig. 4. The effect of the buffer size and UDP datagram size on the throughput for 802.11n 2x2 MIMO systems with 20MHz channels
The physical data rate of the system during the above measurements was that of 144Mbps. The graphical representation of the results show that increasing the packet size improves the throughput obtained. One must highlight the fact that the 100Mbps Ethernet cable was imposing a limitation on the maximum data rate that could be measured, hence a rate higher than that of 100Mbps could not be achieved. This is the reason why a decrease in throughput was observed when the packet size was increased to 1.25Kbyte datagrams using 1Mbyte and 0.06Mbyte buffer sizes.
31.3Mbps was the maximum throughput for a SISO system with 20MHz channel spacing. Another outcome of the indoor environment measurements was the almost doubling effect that 40MHz channel spacing has over the same SISO system but with a smaller spacing. The measured throughput for the 802.11n 40MHz SISO system was that of 57.6Mbps obtained using 1.25Kbytes datagram sizes and 1Mbyte buffer size. The last set of measurements focused on the effect of range on the throughput, highlighting the fact that the throughput decreases as the range increases independently of the 802.11x standard used.
The two links implemented at Ä¦al Far i.e. a distance of 230m and another of 700m were successful, confirming the
Fig. 5. The signal strength at Ä¦al Far at a distance of 1.17km for a MIMO system
hypotheses that were stated in the beginning. Also at this location one cannot argue that the degenerate channel observed for the MIMO system without cross polarization was a result of noise interference from other WLANs since this location had only three other networks, all operating in a different channel than the one used by our MIMO 802.11n system. Hence one can report that the "keyhole" effect was observed at both distances mentioned. From the SISO system one can note that the throughput to the PHY data rate for 802.11n is higher than that for 802.11g as shown in Table II. The Modulation and Coding Scheme (MCS) that the system used for a SISO system using 802.11n was MCS-4 i.e. the PHY data rate for the 20MHz channel spacing and the 40MHz channel spacing were 39Mbps and 81Mbps, respectively.
Comparison of the Ratio of Throughput to the PHY Data Rate for 802.11g and 802.11n SISO Systems
PHY data rate (Mbps)
From Table II, one can note that the throughput to PHY data rate ratio did not change as the distance increased for all three variants of the SISO system. However it is clear that the 802.11n reduces the amount of overhead via the green field preamble and MAC amendments and is more efficient than the legacy 802.11g since the throughput to PHY data rate ratio for the new standard is higher than for 802.11g. However even though the 802.11n using 20MHz channels system was more efficient than the same setup using 40MHz channel spacing, one should still opt for the 802.11n using 40MHz channel spacing since the throughput for a such a system was the highest for both distances.
As mentioned already this system did not suffer from noise but still the MIMO system without polarization diversity exhibited a degenerate channel. At 700m using a 2x2 802.11n MIMO system without polarization diversity having 20MHz channel spacing, the maximum SNR reported by NetStumbler 0.4.0 was 49dB and the results which Iperf measured were as tabulated in Table III. Since NetStumbler 0.4.0 did not report any interfering WLANs using the same channel, this result can be accredited to the "keyhole" effect.
MIMO System Without Dual Polarization at a Distance of 700m
Datagrams Lost/Transmitted Datagrams
0/2211051 = 0%
However having the same scenario but with polarization diversity the results were as tabulated in Table IV.
MIMO System with Dual-Polarization at a Distance of 700m
Datagrams Lost / Transmitted Datagrams
1435/4100010 = 0.035%
1214/4110444 = 0.03%
One can observe that the throughputs reported for a MIMO system with no polarization diversity at a distance of 700m, as tabulated by Table III, are equivalent to those reported by a SISO system despite of the added pair of high gain outdoor antennas, which increases the expenses to implement this WLAN. However from Table IV, one can note that using a 2x2 MIMO system but having different polarization diversity results in an almost doubling of the throughput that can be obtained from a SISO system with 20MHz channel spacing hence making it possible to transmit and receive two different spatial streams at the same instance for the mentioned distance. One must note that the PHY data rate at this distance had dropped from 144Mbps to 104Mbps.
Another experiment was performed using a separation of 700m between the transmitter and the receiver, to observe, the throughput of a 2x2 MIMO system with an inter-element spacing of 0.40m instead of 1.40m. The outcome of this showed that this reduction in the inter-element spacing did not degenerate the channel, as expected since
There is still a 3.2Î» separation, and
Independent channels are created when cross polarization is used.
For the last distance at Ä¦al Far, the signal was not a strong one due to
The distance between the client and the server, and
The construction site which was exactly between the two ends.
Moreover another disadvantage of the system was the fact that the antennas were not located at a height but held in bases relatively close to the ground. NetStumbler 0.4.0 monitored and displayed for a MIMO system with polarization diversity at 1.17km, the signal strength profile shown in Fig. 5.
As can be seen from Fig. 5, the system wasn't continually connected to the AP eliminating the possibility of obtaining measurements averaged over 10 minutes. Fig. 6 shows the throughput variations reported by Iperf at the server for a MIMO system with polarization diversity at 1.17km at Ä¦al Far.
At the last location considered, i.e. Ä¦ad Dingli, the system broke down. This cannot be a result of interference since at the rural location chosen no WLAN were detected besides the one which was setup. Hence the reason for this break down is the non-configurable acknowledgement (ACK) timeouts of the Conceptronic devices. Each datagram transmitted requires that the receiver acknowledges it. This feedback mechanism will either cause a retransmission of the lost datagram, which is not the case for UDP data streams or will be used to chose a data rate lower than the one used. For a long distance link, since the datagrams and the ACKs have to travel a longer range, the ACK timeout should be increased so that they arrive in time, avoiding unnecessary retransmissions and reduction in the PHY data rate, resulting in a higher throughput even though the same range, power and antenna gain are used. Unfortunately, the Conceptronic devices used do not allow the ACK timeouts to be altered.
Fig. 6. Measured throughput variations for a 2x2 MIMO system at Ä¦al Far at a distance of 1.17km
Although the default ACK timeout of the Conceptronic devices was not specified in the user manual, at a range of 700m it functioned properly therefore it can cater for ranges higher than the stated 400m. The ACK timeout, especially for devices which are intended for ranges up to 400m as specified in the user manual , cannot be too large since this would delay the realisation of a lost packet and the retransmission process. Therefore the Conceptronic devices have a default ACK timeout optimised for a range of around 400m to 700m.
This experimental study proved that the "keyhole" effect is present in long distance MIMO link, degrading the quality and hindering the benefits that one can achieve from increasing the number of antennas at both the transmitter and the receiver. It also has shown that in order to avoid this degenerate channel one must opt for cross polarization diversity, which resulted in better throughputs independently of the distance considered. Hence given that attention is paid to ACK timeouts and the use of the highly congested 2.4GHz ISM band , one can conclude that cross-polarization allows the implementation of long distance MIMO point-to-point links with the attendant spatial multiplexing gain, as has previously been confirmed in the short-haul scenario in .
Besides the importance of cross polarization diversity combined with MIMO technology, this paper also highlights the advantages that IEEE 802.11n has over the legacy standard 802.11g. This standard is more efficient than the 802.11g by up to 23.6% independently of whether 20MHz or 40MHz channel spacings are used. It was also proved that the expectation of a throughput reaching up to 100Mbps using a 2x2 MIMO system can be realized in practice reaching 93Mbps at a distance of 2m in the indoor environment and 88.3Mbps at a distance of 230m at Ä¦al Far.