Performance analysis for IEEE 802.11 distributed coordination function in radio-over-fiber-based distributed antenna systems

Yuting Fan; Jianqiang Li; Kun Xu; Hao Chen; Xun Lu; Yitang Dai; Feifei Yin; Yuefeng Ji; Jintong Lin

doi:10.1364/OE.21.020529

1. Introduction

Distributed antenna systems (DASs) using radio-over-fiber (RoF) links have been demonstrated and deployed to achieve broadband wireless access and improve wireless coverage in buildings with the features of low attenuation, large capacity, small-size remote antenna units (RAUs) and centralized management [1–4]. In a RoF-DAS, multiple RAUs are fiber-connected to a center unit where base station facilities are placed. In the downlink, the RAUs receive the optical signals carrying the radio-frequency (RF) signals from the center unit and convert them into electrical signals, then radiate them into air without any signal processing. A reverse process happens in the uplink. In some DAS applications, a single set of base-station facility in the center unit is connected to multiple RAUs to extend the indoor coverage of one base-station and to share the bandwidth resource. This kind of DAS is often called a simulcast DAS where a single base station simultaneously broadcast wireless signals to multiple RAUs in the downlink. In the uplink, the user stations covered in different RAUs contend for the shared transmission medium and base-station facilities. However, not every established wireless communication standard is suited to the simulcast DAS architecture, such as IEEE 802.11 distributed coordination function (DCF). IEEE 802.11 DCF is a random channel-access mechanism based on carrier sense multiple access with collision avoidance (CSMA/CA) where each user station has a right to initiate its transmission and decides the proper transmitting occasion itself. A number of previous studies have exclusively noticed the performance constraint imposed by the media access control (MAC) layer based on IEEE 802.11 DCF in WLAN-RoF systems and identified the fiber length upper bounds for IEEE 802.11b/g/n protocols in theory [5–9]. However, the inherent restraint by the hidden-node (HN) problem is not taken into account. For traditional wireless local-area network (WLAN), many theoretical efforts have been made on investigating HN problems and other MAC-related issues [10–16] including the classic Bianchi’s model [10]. In [17, 18], the Bianchi’s model has been extended to WLAN-RoF-DASs by considering the fiber delay in the calculation of T_s and T_c, which are the period of time when the channel is sensed busy due to a successful transmission or a collision. However, in the previous works [5–8, 17, 18], a dedicated RAU is assumed to be connected to a WLAN access point (AP) and thereby only the intra-RAU HN problem was considered. When fitting an IEEE 802.11-compliant WLAN into a simulcast RoF-DAS for in-building coverage, one WLAN AP is expected to connect multiple RAUs of which the coverage area is separate from each other. Therefore, the user stations under one RAU cannot identify whether a station under other RAUs has occupied the transmission medium or channel, which leads to more frequent inter-RAU HN problem than the intra-RAU HN problem. In addition, another practical issue needs to be addressed that the RAUs are distributed in locations at different distances away from the center unit. Hence, it is inevitable to lead to an unbalanced fiber length, which is referred to as the fiber-length difference effect in this paper.

To the best of our knowledge, no one has presented an analytical model to predict the throughput of IEEE 802.11 DCF in a simulcast RoF-DAS in the presence of both the fiber-length difference effect and the inter-RAU HN problem. In this paper, we derive an analytical model to evaluate the throughput of IEEE 802.11 DCF in a simulcast RoF-DAS where multiple distributed RAUs are connected to a single WLAN AP with different-length fiber links. The developed model allows for a quick and accurate prediction on the throughput performance of an IEEE 802.11 DCF simulcast RoF-DAS, which facilitates the design of practical WLAN-RoF-DASs. In addition, we analyze the performance of DCF in both basic access and RTS/CTS exchange modes with the help of the derived model.

2. Simulcast WLAN-RoF-DAS architecture and assumptions in the theoretical analysis

Two optional architectures for simulcast WLAN-RoF-DASs are shown in Fig. 1. The simulcast is achieved in the RF domain by microwave power dividers and combiners in the RF simulcast architecture, while the simulcast is implemented in the optical domain by optical couplers in the optical simulcast architecture. There is no essential difference in functionality between these two architectures. However, in practical deployments, the RF simulcast architecture is preferred especially for the cases in which the number of RAUs is large. This is because a 1dB loss of optical power will translate to a 2dB loss of RF power, which makes the optical simulcast architecture suffer from relatively high RF power loss. In both architectures, a number of RAUs distributed in separate locations are connected with different-length fiber links to a single WLAN AP located in the central unit.

Fig. 1 The typical simulcast WLAN-RoF-DAS architectures.

Download Full Size | PDF

In the theoretical analysis, we assume the following conditions:

(a). Ideal channel condition, i.e., no capture effect.
(b). No timing boundary. In a WLAN-RoF system, a fast decline of the throughput happens when the inserted fiber delay exceeds the MAC timing boundary commonly determined by the Acknowledgment (ACK) timeout and CTS timeout values. Usually, the ACK timeout and CTS timeout values are set to 316 μs [8], but most WLAN product vendors always set them configurable according to the users’ specific requests. In our analysis, we assume that the ACK and CTS timeout are both 100 μs which are sufficient to cover the involved fiber delay in our investigations (i.e. no timing boundary for all RAUs). This ensures that the transmit failure is solely induced by the packet collisions.
(c). Saturated load: there is always a packet in the buffer waiting for transmitting after a successful transmission.
(d). Dual-RAU configuration: two RAUs are involved in our analysis for convenience. We further assume that a short length (e.g. 100m fiber only leads to a 0.5 μs delay which can be ignored in the analysis comparing to the typical wireless propagation delay 1 μs) to one RAU (say RAU-A) and a variable large length to the other RAU (say RAU-B). The one-way differential fiber delay between RAU-A and RAU-B is denoted as F slots (1 slot = 9 μs for IEEE 802.11g).

3. Theoretical analysis

3.1 Inter-RAU HN problem in a simulcast WLAN-RoF-DAS

In the mechanism of IEEE 802.11 DCF, the stations under different RAUs which are distributed in different locations cannot hear from each other and thereby access the channel blindly. Enormous collisions might occur during the transmission and a backoff stage starts in each station’s MAC layer after the collisions. As Fig. 2 shows, transmissions of the stations under one RAU have high probability to impact transmissions of stations in other RAUs, which results in so-called inter-RAU HN problem. The inter-RAU HN problem can be defined as coupling effect, which has been investigated in [19].

Fig. 2 Backoff stage of two stations in different RAU with differential fiber delay.

Download Full Size | PDF

However, in a simulcast WLAN-RoF-DAS with different-length fiber links, the stations in the RAU with longer fiber link have to wait for relatively longer propagation time, which makes the stations in the RAU with shorter link have more chance to occupy the channel. As Fig. 2 Case 1 illustrates, Station 1 under RAU-A with negligible fiber delay transmits frames after its backoff counter finishes. Nevertheless, a collision which should take place is missed due to the differential optical delay for Station 2 under RAU-B with large fiber delay. Similarly in Case 2, when Station 2 under RAU-B grabs the channel firstly, it might miss this chance owing to the additional fiber delay. In this respect, we can see that the stations under RAU-A suffer less interference than the stations in RAU-B, which ultimately translates to a higher throughput in RAU-A. Therefore, the inter-RAU HN problem may result in unbalanced throughput among RAUs with different-length fiber links, which has not been discussed before.

3.2 The performance of the basic access and RTS/CTS exchange mechanisms in a simulcast WLAN-RoF-DAS

The RTS/CTS exchange mechanism adopts the reservation method to against HN problems at the cost of a decrease of the average throughput. Many reports have proved that the RTS/CTS exchange mechanism is effective to mitigate the HN problem in both traditional WLAN and WLAN-RoF-DASs with uniform-length fiber links. However, as explained above, the situation of unbalance throughput performance may occur in a simulcast WLAN-RoF-DAS. Therefore, in the presence of inter-RAU HN problem, we here discuss how much the fiber-length difference effect impact the throughput unfairness among RAUs in both the basic access and RTS/CTS exchange modes.

We define a vulnerable period T_v which is the interval between the time a station in one RAU starting its transmission and the time other stations in the other RAU receiving an ACK or a CTS frame from the AP. A collision among the stations in different RAUs tends to happen during the vulnerable period. We can define T_v in two modes as:

T_{v b a s i c} = D A T A + A C K + S I F S + F

T_{v R T S} = R T S + C T S + S I F S + F

where DIFS and SIFS is the DCF interframe space and short interframe space respectively. In Fig. 3, a frame is transmitted after the backoff counter finishes. In the RTS/CTS exchange mode, the stations can identify if other stations in the other RAU have occupied the channel after receiving a CTS frame, while the stations only know the channel state after an ACK frame in basic access mode. As a RTS frame length is much shorter than a maximum data frame, the vulnerable period in RTS/CTS mode is much shorter than the basic access mode. Therefore, the collision probability caused by the HN problem in the basic access mode is much higher than that in the RTS/CTS exchange mode. Especially when the number of stations increases, the impact of the inter-RAU HN problem is stronger in the basic access mode. Hence, the RTS/CTS exchange mode is supposed to perform better against the HN problem. Additionally, the period of RTS frame transmission including waiting F slots is still averagely shorter than the data frame transmission period in the basic access mode in the same network. Therefore, the fiber delay may impact less the RTS/CTS exchange mode than the basic access mode.

Fig. 3 The vulerable period in basic access and RTS/CTS exchange modes.

Download Full Size | PDF

3.3 Theoretical throughput model considering fiber length difference

As show in Fig. 2, as fiber delay F increases, a settled delay exists for the stations under RAU-B. Even though the backoff procedures of the stations in RAU-B have finished the countdown, it should take another F slots to contend for the channel. So we reconsider the two-dimensional Markov chain model as Fig. 4 illustrates. Let W₀ and m be the minimum contention window and the maximum backoff stage, respectively. The backoff range changes to [F, F + W_i −1] instead of [0, W_i −1], where W_i = 2ⁱW₀ represents the contention window of i-th backoff stage.

Fig. 4 Markov Chain Model considering fiber delays.

Download Full Size | PDF

Let n_A and n_B be the number of stations within a carrier sensing range of RAU-A and RAU-B respectively. We set b(t) be the stochastic process representing the timer for a given slot. The premise of Markov chain model [10] to describe the backoff process is the assumption that the probability p_R (R∈{A,B}, which stands for the RAU-R of a transmitting station’s backoff stage s(t), is independent of the other station’s backoff stage. In our analysis, we consider the station’s backoff process {b(t), s(t)} which is independent on the others in the same group (or RAU) and dependent on the HNs in the other group. Therefore, we firstly consider the situation that no inter-RAU HN effect happens for a RAU, then adding the coupling effect to calculate the inter-RAU HN effect. As different-length fiber links involved in our model, each RAU’s throughput has to be calculated separately.

Absence of inter-RAU HN problem

Firstly, we discuss the situation when a transmitting station is not interfered by the inter-RAU HNs. We define thatσ is the system slot time and h is the number of slots to transmit the packet. Let L = hσbe the packet size and l be the minimum backoff stage which satisfies 2^lW₀ + F−1≥h. If the backoff counter of a station in HN-group is bigger than h, then there is no interference between the groups. We denote δ_dR as the probability that a HN backoff counter is bigger than h. As there is no interference in this situation, we can assume that the HN’s backoff stage is independent on the transmitting station’s one approximately. Hence, the stationary probability τ_R that a station transmits a packet in a randomly chosen time slot and the RAU-R’s conditional collision probability p_R can be represented as:

τ_{R} = \sum_{i = 0}^{m} b_{i, 0_{R}} = \frac{2 (1 - 2 p_{R})}{p_{R} W_{0} (1 - {(2 p_{R})}^{m}) + (1 + W_{0}) (1 - 2 p_{R})}

p_{R} = 1 - {(1 - τ_{R})}^{(n_{R} - 1)} g_{R o}

Furthermore, we calculate

g_{R} = {(δ_{R})}^{n_{R}} q_{R}

, and we calculate δ_R and q_R as follows:

\begin{matrix} δ_{d R} = \sum_{k = h}^{W_{i} + F - 1} b_{m, k_{R}} \\ = \frac{b_{0, 0}_{_{R}}}{2} {W_{0} [\frac{{(2 p_{R})}^{l} - {(2 p_{R})}^{m}}{1 - 2 p_{R}} + \frac{{(2 p_{R})}^{m}}{1 - p_{R}}] + \frac{(2 F - 2 h + 1) p^{l}}{1 - p_{R}} \\ + \frac{(F^{2} + h^{2} + 2 F h + F + h)}{W_{0}} [\frac{{(\frac{p_{R}}{2})}^{l} - {(\frac{p_{R}}{2})}^{m}}{1 - \frac{p_{R}}{2}} + \frac{{(\frac{p_{R}}{2})}^{m}}{1 - p_{R}}]} \end{matrix}

b_{0, 0_{R}} = \frac{2 (1 - 2 p_{R}) (1 - p_{R})}{p_{R} W_{0} (1 - {(2 p_{R})}^{m}) + (1 + W_{0}) (1 - 2 p_{R})}

q_{R} = \frac{σ}{(1 - P_{t r R}) σ + P_{t r R} P_{S R} T_{S} + P_{t r R} (1 - P_{S R}) T_{C}}

where b_i,kR is the probability that the station is in the i-th backoff stage and k-th backoff counter. We denote g_R as the probability that a transmitting station in RAU-R (trans-group) does not suffer from an interference with the stations in the other RAU-R_o which could be taken as HN-group (if R = A, R_o = B, for example). And we define q_R is the probability that a station in HN-group starts to transmit on the generic slot timing of the HN-group.

Moreover, we need to obtain P_trR (the probability of at least one transmission in each slot time) and P_sR (the probability of a successful transmission in each slot time):

P_{t r R} = 1 - {(1 - τ_{R})}^{(n_{R} - 1)}

P_{s R} = \frac{n_{R} τ_{R} {(1 - τ_{R o})}^{(n_{R_{o}} - 1)} g_{R_{o}}}{P_{t r R}}

As a result, q_A, q_B and δ_dA, δ_dB can be given by Eq. (3)-(9) using numerical methods with the corresponding parameter

F

.

Presence of inter-RAU HN problem

In this part, we investigate the inter-RAU problem between the two groups (i.e. trans-group and HN groups) based on our model. It is difficult to make the model completely accurate. Thus, we only consider the primary situations involved in our discussion. We partition the Markov chain into two parts, the front part (back stage 0 to m-1) and the rear part (back stage m). As we aforementioned, we assume that the backoff stages of the station in one RAU are dependent upon that of other stations in different RAUs when there is an inter-RAU HN problem. However, more specifically in our discussion, they are dependent in the front part and almost independent in the rear part.

Hence, under this approximation, the transmission probability and conditional collision probability τ_fR, p_fR in the front part and τ_rR, p_rR in the rear part can be represented as:

τ_{f R} = \sum_{i = 0}^{m - 1} p_{f R}^{i} b_{_{i, 0 c_{R}}}^{} = b_{_{0, 0_{c R}}}^{} \frac{1 - p_{f R}^{m}}{1 - p_{f R}}

τ_{r R} = p_{r R}^{m} b_{_{m, 0 c_{R}}}^{} = b_{_{0, 0_{c R}}}^{} \frac{p_{r R}^{m}}{1 - p_{r R}^{m}}

With b_{0, 0_{c R}}^{} = \frac{2}{W_{0} (\frac{1 - {(2 p_{f R}^{})}^{m}}{1 - 2 p_{f R}} + \frac{{(2 p_{f R}^{})}^{m}}{1 - p_{r R}}) + \frac{1 - p_{f R}^{m}}{1 - p_{f R}} + \frac{p_{f R}^{m}}{1 - p_{r R}}}

p_{f R} = 1 - {(1 - τ_{c R})}^{n_{R} - 1} {(δ_{c R_{O}})}^{n_{R_{o}}} q_{R_{O}}

p_{r R} = 1 - {(1 - τ_{c R})}^{n_{R} - 1} {(δ_{d R_{O}})}^{n_{R_{o}}} q_{R_{O}}

where δ_cR represents the probability that the transmitting station does not suffer HN’s interference which is in a high backoff stage (we assume

2

to

m

) and it can be calculated as:

δ_{c R} = \frac{δ_{d R}}{1 - b_{0, 0_{R}} \frac{W + 1 + 2 F}{2} - p_{R} b_{0, 0_{R}} \frac{2 W + 1 + 2 F}{2}}

The equations from Eq. (8)-(13) represent a nonlinear relation among unknowns p_fA, p_rA, τ_cA, p_fA, p_rA, τ_cA, which can be solved using numerical methods. In addition, we obtain P_trR and P_sR as:

P_{t r R} = 1 - {(1 - τ_{c R})}^{n_{R}}

P_{s R} = \frac{n_{R} {(1 - τ_{c R})}^{(n_{R} - 1)} (τ_{f R_{O}} {(δ_{c R_{O}})}^{(n_{R_{o}} - 1)} q_{R_{o}} + τ_{r R_{O}} {(δ_{d R_{O}})}^{(n_{R_{o}} - 1)} q_{R_{o}})}{P_{t r R}}

A successful transmission in a slot time occurs with the probability of P_trRP_sR. The normalized system throughput $S_{R}$ can be calculated based on Eq. (16) and (17):

S_{R} = \frac{2 P_{s R} P_{t r R} E_{R} [p a c k e t]}{(1 - P_{t r R}) σ + P_{t r R} P_{s R} T_{s} + P_{t r R} (1 - P_{s R}) T_{c}}

And the throughput of system will be obtained by:

S = S_{A} + S_{B}

The packet payload size in bit is E_R[packet]. For the basic access mode and RTS/CTS exchange mode, T_smode and T_cmode can be calculated respectively by:

\begin{array}{l} T_{s b a s i c} = D I F S + P L C P + H + C e i l i n g (E_{R} [p a c k e t] + T r a i l) \\ + S I F S + P L C P + H + C e i l i n g (A C K + T r a i l) \\ + 2 \times 2 (ν + F) \end{array}

\begin{array}{l} T_{c b a s i c} = D I F S + P L C P + H + C e i l i n g (E_{R} [p a c k e t] + T r a i l) \\ + 2 (ν + F) + A C K t i m e o u t \end{array}

\begin{array}{l} T_{s R T S} = D I F S + 4 \times P L C P + 4 \times H + 4 \times 2 (ν + F) + 3 \times S I F S \\ + C e i l i n g (R T S + T r a i l) + C e i l i n g (C T S + T r a i l) \\ + C e i l i n g (E_{R} [p a c k e t] + T r a i l) + C e i l i n g (A C K + T r a i l) \end{array}

\begin{array}{l} T_{c R T S} = D I F S + P L C P + 2 (ν + F) + C e i l i n g (R T S + T r a i l) \\ + C T S t i m e o u t \end{array}

We define PLCP and H as the time of physical layer (PHY) preamble and PHY header. The ceiling function is to make sure the number of the symbols transmitted is an integer and Trail is the trail bits after the OFDM symbol. The wireless delay here is represented by v.

4. Simulations and discussion

4.1 The performance in basic access mode

We set n_A = n_B = n to reduce the complexity in our simulations. The simulation parameters are listed in Table 1. Then, we take F = 0 which represents no consideration of optical delay effect in backoff stage as the reference. Firstly, we validated the normalized throughput of our analytical model with the Matlab results in the basic mode to show the severe inter-RAU HN problem in the presence of the fiber delay difference effect.

Table 1. System Parameters

View Table | View all tables in this article

In Fig. 5, as F in RAU-B grows, we observe that when the fiber effect are considered in both T_sbasic and T_cbasic and backoff procedure, the total normalized throughput gradually increases and a more distinct discrepancy appears between the two RAUs’ throughput, which shows a good agreement with the simulation results in the case of n = 1. It is clear to understand that a fiber delay factor F considered in backoff process influences the two groups’ conditional collision probabilities. Therefore, the conditional probabilities are unequal in this condition, but they needn’t to be distinguished in previous studies. More specifically, the group with longer fiber length suffers from more interference by other groups, and the stations in this group are more difficult to occupy the channel. In contrast, it makes the opposite effect on the shorter-fiber-employed groups. For the whole system, the proportion of the success transmissions gradually rises. As a result, the system throughput increases as well. However, we have to stress that the increase of system throughput is not the main point in our validation, which is only led by the increase of the unfairness of stations between different RAUs. Obviously, the consideration of optical delay both in backoff stages and T_sbasic and T_cbasic can more accurately describe the throughput performance of WLAN-RoF-DAS based on IEEE802.11 DCF MAC mechanism.

Fig. 5 Normalized throughput via fiber delay when n = 1. Dashed lines indicate traditional methods considering fiber effect only in T_s and T_c. Solid lines represent our model.

Download Full Size | PDF

Figure 6 shows the detrimental effect of HNs on the throughput of system as the number of stations grows. The solid lines represent the model calculations, the asterisks which have the same color with the line express the simulation results of it. As F goes up, we can see the gradual increase of throughput given the same n. We can imagine that if F goes to infinite, inter-RAU HN problem will disappear and thereby the throughput in RAU-B is zero. At this time, the system throughput is contributed only by RAU-A. It is equal to the throughput in typical WLANs where all the stations are in the carrier sensing range and no HNs exist. Furthermore, the system throughput is composed of two parts which we can see in Fig. 7 when n = 5 for example. It has a general meaning to reflect the unfairness of different group’s throughput owing to the fiber delay difference. As the simulation results proved, the RAU with larger fiber delay has little chance to occupy the channel.

Fig. 6 Fiber effect via different number of stations: the solid lines represent the model calculations, the asterisks which have the same color with the line express the simulation results of it.

Download Full Size | PDF

Fig. 7 Normalized throughput in different RAU versus Fiber delay when n = 5.

Download Full Size | PDF

4.2 The performance in RTS/CTS exchange mode

As we have discussed in the former parts, the basic access mode is shown to have a degressive tendency of system throughput with the increase of the station number. Also, the results indicated two detrimental effects from both the inter-RAU HN problem and the different-fiber-length distribution. We know that the RTS/CTS mode has become a feasible method to address the HN problem in both traditional WLAN and WLAN-RoF-DAS with uniform-length fiber links. Here, we discuss the performance of the RTS/CTS mode considering fiber-length difference effect. We assumed a saturated payload and each packet with equal maximum frame length. A RTS/CTS exchange happens before every packet transmits. The simulation results are shown in Fig. 8.

Fig. 8 Comparison of fiber effect via different number of stations. Solid lines represent basic access mode and dashed lines represent RTS/CTS exchange mode.

Download Full Size | PDF

By comparing the throughput performances of the RTS/CTS exchange mode and basic access mode in Fig. 8, it can be seen that the RTS/CTS exchange mode is more effective against both inter-RAU HN problem and fiber-length difference effect. Similar to basic mode, as the inserted delay increases, the interval between a station successfully transmitting a RTS packet and other stations in the other RAU receiving a CTS packet from AP is gradually rising. Then, the stations in a RAU with longer fiber length induced are more difficult to occupy the channel than the shorter-employed ones. From Fig. 8, we can see that the maximum throughput difference between RAU-A and RAU-B is 87% in basic access mode and 56.9% in RTS/CTS exchange mode. As the fiber length increases, the difference in throughput between the two RAUs is significantly reduced as compared to the basic access mode. The results have the same trend with the analysis in part 3.2. In summary, the RTS/CTS exchange mode performs better than the basic access mode against inter-RAU HN problem and fiber-length difference effect, when saturated load is assumed in the systems.

5. Experimental results

5.1 Experimental setup

The experimental setup of a dual-RAU WLAN-RoF-DAS using the optical simulcast architecture is depicted in Fig. 9. The distance between the two antennas is more than 40m in typical office environment with several rooms and walls separated. Further distance and some obstacles like walls ensure a small inference between the two RAUs. The AP connects to a PC using Ethernet cable. Ethernet limits the maximum MAC payload size to 1500 bytes rather than 2312 bytes. AP and each RAU are connected by RoF links. In the downlink, the RF signals transmitted from AP are converted into the optical signal by the distributed feedback (DFB) laser diode (LD) in the RF-optic transceiver. The downlink optical signal is equally split by an optical coupler with ~3.1 dB loss. Then they go through the RoF link, and they are converted back to the electrical signals by the photodetectors in the RF-optic transceivers located in RAUs. Finally, the electrical signals are amplified by a bi-directional amplifier and radiated from each antenna. Reversed process happens in the uplink. At the terminal end, up to 8 laptops (Dell Vostro series with Intel Dual-Core CPU @ 1.8GHz; 2GB DDR3 RAM; MS Windows XP Professional) was used as the user stations in the experiment. The WLAN AP was Broadcom 802.11n wireless card, and the AP was using Open Source firmware OpenWrt (version 3.18_4M). The other parameters in the experiment are listed in Table 2.

Fig. 9 Experimental setup for a RoF-based dual-RAU DAS

Download Full Size | PDF

Table 2. System Parameters

View Table | View all tables in this article

5.2 Experimental results

As the firmware in AP we used is limited by the fixed parameter ACK timeout, the throughput decreases quickly when the inserted fiber length is longer than 2km. So in our experiment, we will discuss the two modes’ performance and validate the model with 100m short fiber links and 2km long fiber links.

There are up to 8 wireless terminals (i.e. laptops) involved in the test, which are averagely allocated to the two RAUs. A throughput testing software IxChariot is used and all working wireless terminals are 2m away from the AP to ensure a fair opportunity to access the wireless channel. Figure 10 shows the uplink results of a RoF-DAS dual-RAU system with identical short fiber links and different fiber links, where stations transmit UDP traffics to the AP with saturated payload (1500 bytes). All testing data were obtained from 10-minute average results. From the results, we can find inter-RAU HN problems are very obvious, system throughput decreased quickly in basic mode as contending stations increasing. Also, it can be observed that RTS/CTS mode is efficient against inter-RAU HN problems under DAS architecture. However, the fiber-length difference effects are not reflected in system throughput performance. Figure 11 demonstrates solitary-RAU throughput performance. With the analysis in previous part, fiber-length difference may lead to an unfairness of contention between stations in different RAUs. In the experimental results, a similar conclusion can be obtained. When using the same short-length fiber links, the throughput value in each RAU are almost equal (dotted lines in Fig. 11). Nevertheless, an obvious difference (an average of 0.52Mbps) of throughput between RAU-A and RAU-B can be observed in basic access mode, while a relative small difference (an average of 0.21Mbps) in RTS/CTS exchange mode. Due to the 2km fiber-length limit, the experiment only validates the situation with 10 μs (1.1 slot time) inserted delay. Compared with the previous validation, as the increase of inserted fiber delay, the unfairness of contention opportunity between stations in different RAUs becomes stronger, and the RTS/CTS exchange mode performs better than the basic access mode against the fiber-length difference effect.

Fig. 10 Basic access mode and RTS/CTS exchange mode system throughput in RoF-Based Dual-RAU DAS. The solid lines indicate experimental results with the same-length (100m) fiber links connected to each RAU. And the dotted lines indicate experimental results with 100m-fiber connected to RAU-A and 2-km fiber links connected to RAU-B.

Download Full Size | PDF

Fig. 11 Basic access access mode and RTS/CTS exchange mode solitary-RAU throughput in RoF-Based Dual-RAU DAS. The solid lines indicate theory results. The dotted lines indicate experimental results with the same fiber length. The dashed lines indicate experimental results with different fiber length.

Download Full Size | PDF

6. Conclusion

In this paper, we take into account the inter-RAU HN problem and different fiber length effect in the performance analysis of a simulcast WLAN-RoF-DAS and derive an analytical model to evaluate the IEEE 802.11 DCF performance in terms of network throughput. It is shown that the inter-RAU HN problem could result in unbalanced throughput performance among RAUs with different-length fiber links. We also prove that the RTS/CTS exchange mode is still effective in a simulcast WLAN-RoF-DAS in the presence of inter-RAU HN problem and fiber length difference effect. With the developed model, detailed analysis has been conducted for the performance of DCF in both basic access and RTS/CTS exchange modes. Extensive numerical and experimental investigations show a good agreement with the derived model.

Acknowledgment

This work was supported in part by National 973 Program (2012CB315705), National 863 Program (2011AA010306), NSFC Program (61271042, 61107058 and 61120106001), the Cooperation Project between Province and Ministries (2011A090200025), the Fundamental Research Funds for the Central Universities (2013RC1203).

References and links

1. C. Liu, N. Cvijetic, K. Sundaresan, M. Jiang, S. Rangarajan, T. Wang, and G.-K. Chang, “A novel in-building small-cell backhaul architecture for cost-efficient multi-operator multi-service coexistence,” in Proceedings of OFC/NFOEC2013, Anaheim, California, United States, Mar.2013, paper OTh4A.4.

2. M. J. Crisp, S. Li, A. Wonfor, R. V. Penty, and I. H. White, “Demonstration of radio over fibre distributed antenna network for combined in-building WLAN and 3G coverage,” in Proceedings of OFC2007, Anaheim, California, United States, Mar.2007, paper JThA81.

3. N. Wei and I. B. Collings, “Indoor wireless networks of the future: adaptive network architecture,” IEEE Commun. Mag. 50(3), 130–137 (2012). [CrossRef]

4. J.-Z. Wang, H.-L. Zhu, and N. J. Gomes, “Distributed antenna systems for mobile communications in high speed trains,” IEEE J. Sel. Areas Comm. 30(4), 675–683 (2012). [CrossRef]

5. B. Kalantarisabet and J. E. Mitchell, “MAC constraints on the distribution of 802.11 using optical fibre,” in Proceedings of the 9th European Conference on Wireless Technology, Manchester, United Kingdom, Sep.2006, 238–240. [CrossRef]

6. A. Das, M. Mjeku, A. Nkansah, and N. J. Gomes, “Effects on IEEE 802.11 MAC throughput in wireless LAN over fiber systems,” J. Lightwave Technol. 25(11), 3321–3328 (2007). [CrossRef]

7. B. Kalantari-Sabet, M. Mjeku, N. J. Gomes, and J. E. Mitchell, “Performance impairments in single-mode radio-over-fiber systems due to MAC constraints,” J. Lightwave Technol. 26(15), 2540–2548 (2008). [CrossRef]

8. S. Deronne, V. Moeyaert, and S. Bette, “Analysis of the MAC performances in 802.11g radio-over-fiber systems,” in Proceedings of the 18th IEEE Symposium on Communications and Vehicular Technology in the Benelux (SCVT), Ghent, Belgium, Nov. 2011. [CrossRef]

9. S. Deronne, V. Moeyaert, and S. Bette, “Impact of the slottime parameter value on the MAC performances in IEEE 802.11 wireless systems using radio-over-fiber technology,” in Proceedings of the 17th IEEE Symposium on Communications and Vehicular Technology in the Benelux (SCVT), Enschede, Netherlands, Nov. 2010. [CrossRef]

10. G. Bianchi, “Performance analysis of the IEEE 802.11 distributed coordination function,” IEEE J. Sel. Areas Comm. 18(3), 535–547 (2000). [CrossRef]

11. I. Tinnirello, S. Choi, and Y. Kim, “Revisit of RTS/CTS exchange in high speed IEEE 802.11 networks,” in Proceedings of the IEEE 2005 Int. Conf. World of Wireless, Mobile, Multimedia Networks (WoWMoM), Taormina, Italy, Jun. 2005, 240–248. [CrossRef]

12. C.-G. Wang, B. Li, and L. Li, “A new collision resolution mechanism to enhance the performance of IEEE 802.11 DCF,” IEEE Trans. Vehicular Technol. 53(4), 1235–1246 (2004). [CrossRef]

13. X. Wang, P. H. J. Chong, and L. W. Yie, “Evaluation of performance on random back-off interval and multi-channel CSMA/CA protocols,” in Proceedings of TENCON 2009 -IEEE Region 10 Conference, 2009, pp. 1–5.

14. N. Wattanamongkhol, W. Srichavengsup, S. Nakpeerayuth, and L. Wuttisiittikulkij, “Performance analysis of modified backoff algorithm in IEEE 802.11 networks,” in Proceedings of the 3rd IEEE/IFIP International Conference in Central Asia on Internet, Tashkent, Sep.2007. [CrossRef]

15. Y.-S. Kim, J.-Y. Yu, S.-Y. Choi, and K.-H. Jang, “A novel hidden station detection mechanism in IEEE 802.11 WLAN,” IEEE Commun. Lett. 10(8), 608–610 (2006). [CrossRef]

16. F.-Y. Hung and I. Marsic, “Access delay analysis of IEEE 802.11 DCF in the presence of hidden stations,” in Proceedings of GLOBECOM 2007, Washington, DC, United State, Nov. 2007.

17. A. Pal and A. Nasipuri, “Performance analysis of IEEE 802.11 distributed coordination function in presence of hidden stations under non-saturated conditions with infinite buffer in radio-over-fiber wireless LANs,” in Proceedings of the 18th IEEE Workshop on Local & Metropolitan Area Networks (LANMAN), Chapel Hill, NC, Oct. 2011. [CrossRef]

18. M. Mjeku and N. J. Gomes, “Analysis of the request to send/clear to send exchange in WLAN over fiber networks,” J. Lightwave Technol. 26(15), 2531–2539 (2008). [CrossRef]

19. T. Kim and J. T. Lim, “Throughput analysis considering coupling effect in IEEE 802.11 networks with hidden stations,” IEEE Commun. Lett. 13(3), 175–177 (2009). [CrossRef]

Parameters	Value
Slot Time (σ)	9 μs
DIFS	28 μs
SIFS	10 μs
PHY header	20 μs
ACK timeout	100 μs
CTS timeout	100 μs
RTS Frame	16 × 8 bits
CTS Frame	14 × 8 bit
ACK Franme	14 × 8 bit
Paketlength	12312bits
Transmission Rate	54 × 10^6 bps
M	7
CWmin	15
CWmax	1023
Wireless delay	1 μs
Fiber delay	0~10 slot time

Parameters		Value
Frequency Band		2.4GHz
AP Output Power		10 dBm
RF-Optic Transceiver Output Optical Power		4 dBm
Antenna distance		40m
RoF link B2B Gain		0 dB
Bi-directional amplifier	Downlink Gain	15 dB
Bi-directional amplifier	Uplink Gain	25 dB
RF Power fed to each antenna		20dBm
Antenna		Omni Directional (3dBi)
RTS threshold		1000 Bytes

Parameters	Value
Slot Time (σ)	9 μs
DIFS	28 μs
SIFS	10 μs
PHY header	20 μs
ACK timeout	100 μs
CTS timeout	100 μs
RTS Frame	16 × 8 bits
CTS Frame	14 × 8 bit
ACK Franme	14 × 8 bit
Paketlength	12312bits
Transmission Rate	54 × 10^6 bps
M	7
CWmin	15
CWmax	1023
Wireless delay	1 μs
Fiber delay	0~10 slot time

Parameters		Value
Frequency Band		2.4GHz
AP Output Power		10 dBm
RF-Optic Transceiver Output Optical Power		4 dBm
Antenna distance		40m
RoF link B2B Gain		0 dB
Bi-directional amplifier	Downlink Gain	15 dB
Bi-directional amplifier	Uplink Gain	25 dB
RF Power fed to each antenna		20dBm
Antenna		Omni Directional (3dBi)
RTS threshold		1000 Bytes

Performance analysis for IEEE 802.11 distributed coordination function in radio-over-fiber-based distributed antenna systems

Abstract

1. Introduction

2. Simulcast WLAN-RoF-DAS architecture and assumptions in the theoretical analysis

3. Theoretical analysis

3.1 Inter-RAU HN problem in a simulcast WLAN-RoF-DAS

3.2 The performance of the basic access and RTS/CTS exchange mechanisms in a simulcast WLAN-RoF-DAS

3.3 Theoretical throughput model considering fiber length difference

Absence of inter-RAU HN problem

Presence of inter-RAU HN problem

4. Simulations and discussion

4.1 The performance in basic access mode

4.2 The performance in RTS/CTS exchange mode

5. Experimental results

5.1 Experimental setup

5.2 Experimental results

6. Conclusion

Acknowledgment

References and links

Cited By

Figures (11)

Tables (2)

Equations (23)

Optics Express