Click here for the enhanced PDF version of this article
For a pdf version of this article, including diagrams.

In a typical indoor 802.11 wireless network, a single access point can cover 2500 to 10,000 square feet (based on typical access point placement, which is normally 50-100 feet apart). Picocells are substantially smaller RF coverage areas — on the order of 150 to 1000 square feet (cells on the order of 10-35 feet in diameter). They are used to provide high data throughput for dense wireless terminals and appliances by minimizing the amount of bandwidth that is shared among wireless users. A sample application for picocells would be a sporting arena, lecture hall, dense cubicle farm, or the trading floor of a stock exchange, where hundreds of wireless users require simultaneous high-bandwidth wireless services while confined within a localized space.

Creating a single picocell is relatively easy. However, it is difficult to deploy large numbers of non-overlapping picocells in a tightly packed space.

The basic design elements available in creation of picocells are radio transmit power control and receiver sensitivity control at the wireless access point (AP) and a mobile station (i.e. “client”).

The primary issues associated with picocell design are:

• RF multipath and scattering.
• 802.11 protocol non-ideal results.
• number of available channels.
• high-gain antennas.
• balancing of RF link symmetry.
• coverage redundancy.
• user performance and fairness.

The coverage area of an AP can be reduced by decreasing the RF link budget (transmit power and receiver sensitivity) of the AP and the mobile station. In addition to controlling the link budget, the effect of co-channel interference and adjacent-channel RF interference must be dealt with before bringing APs and clients closer together. The strength of interference between cells will affect the quality of the RF link and the throughput of the cell(s) due to 802.11's built-in interference mitigation: carrier-sense multiple access (CSMA).

Generally the size of an RF coverage cell, with the center being the AP, is based on the maximum capabilities of the radio design of the access point and the client in terms of transmit power, receiver sensitivity, antenna gain and RF environment. Typical 2.4 GHz and 5 GHz transmit power, which is regulated by the Federal Communications Commission (FCC), is around +17 to +30 dBm (50 mW to 1 W). Typical radio receiver demodulation maximum sensitivity is around -90 dBm for base data rate modulations and all data packet headers for 802.11a/b/g. Antenna gain maximums are limited by the FCC.

Transmit power control and receiver sensitivity control — Adjusting transmitter power and receiver sensitivity is one method to create a picocell. These parameters may be adjusted via hardware or software. Software adjustments of hardware can be made in terms of transmitter power control, implemented in adjustable hardware amplifier feedback loops. Software adjustments of hardware receivers can be made by setting the receiver (at baseband demodulation) to immediately ignore the packet if it does not exceed a minimum signal-to-noise ratio (SNR) threshold.

Figure 1 is a typical ceiling-mounted relationship between APs, comparing attenuators with distance. The strength of the signal is quantified by the received signal strength intensity (RSSI).Table 1 is a typical ceiling-mounted relationship between APs (with no attenuators) and the co-channel and adjacent interference, as measured by signal-to-interference ratio (SIR), from the clients' perspective.

Special high-gain antennas — Special high-gain antennas can help control the area of RF coverage by focusing the main portion of energy on a particular area, thereby creating spatial diversity and more available bandwidth for other areas. Antenna gain is generally measured in terms of an antenna's 3 dB beamwidth in both the horizontal plane (azimuth) and the vertical plane (elevation) relative to an isotropic radiator. Typical 802.11 low-gain antennas are 2.2 to 6 dBi omni-directional dipole\patch antennas with sizes in the range of 2 inches × 2 inches, but a high-gain antenna can be 2 to 10 times larger depending on the amount of directivity. Other types of modern antennas can use phased arrays or Multi-Input Multi-Output (MIMO) smart antenna technology to enhance directivity.

Protocol reality (data rate and minimum SNR\SIR) — One key constraint in creation of picocells is the choice of data rate. The higher the data rate, the higher the SNR and SIR requirements for both the AP and client (see Figure 1).

Protocol reality (CSMA/CA) — All 802.11 packets have packet headers that are at the minimum data rate and have minimum SNR requirements (i.e. 802.11a is 6 dB SNR). When decoding a distant packet header on the same channel, the 802.11 protocol requires a radio (via CSMA/CA) to defer transmission of any packet, reducing its throughput. CSMA/CA constricts throughput of any 802.11 device because it will defer transmission if it “hears” a packet header. The RF reality mentioned above creates a strict limitation to the problem of spacing so that picocells sharing the same channel do not hear each other yet have enough SNR power to decode high date rates. Furthermore, the distance that the header can be heard compared to the data is far greater due to the data payload's high SNR requirement when >6 Mbps (see Figure 1). Receiver sensitivities, usually implemented in software, can make decisions to ignore packets on computation of header SNR below a user-defined threshold.

RF interference and cell overlap — Interference in a picocell is a critical attribute to be considered as the maximum density of APs and clients will be constrained by this. Interference is determined by the strength of same-channel or adjacent-channel RF power in a channel due to frequency and physical overlap. The strength of RF overlap is controlled by transmit power/ receiver sensitivity, antennas and number of channels.

Location of adjacent channels cannot necessarily be placed right next to each other. This results in a shuffling of channels to ensure that immediately adjacent channels do not overlap from one cell to another. Adjacent channels have about 30 dB worth of isolation, given all measurements are equal (see Figure 1 and 2).

Number of available channels — Another constraint that comes to bear is the number of channels available. The spacing density of reused channels is a function of the number of available channels. At this point in time, 802.11b has three channels (1, 6 and 11) available. 802.11a has a total of 23 channels with 12 standard channels (36,40,44,48,52,56,60,64,149,153,157,161) and 11 recently added channels (100, 104,108,112,116,120,124,128,132,136,140).

High-gain antennas — With regard to high-gain antennas, the directivity may help contain the focus of the main beam, but the moment that the energy hits any object, scattering occurs. Scattering (reflection, refraction or diffraction) off of walls, floors, and miscellaneous objects will result in some power loss. More importantly, the picocell designer will lose control over the direction of RF energy and hence, cell coverage, size and the possibility for one picocell on channel A to affect another picocell several cells away on channel A.

Balancing the symmetry of client and AP cell size — Balancing the symmetry of the AP-client RF link is also an important issue. For example, simple attenuation added between an AP and its antenna or reduction of AP transmit power will shrink the cell size. However, balancing of the RF link reduction ensures that the client receiver does not have a much larger receiver range than the AP and vice versa. In this example, an AP with a reduction in transmit power and\or receiver sensitivity will not be able to see its co-channel AP, but the clients may be able to see the distant AP or other clients if it also does not have a reduction in power and receiver sensitivity. In addition, throughput will suffer due to collision avoidance CSMA/CA mechanism in the 802.11 protocol.

Given user density and throughput per user, basic equations that must be solved are cell size, SNR, and SIR requirements.

General equations:

Cell Size (estimate) = DataRateThroughput_per_Cell/Throughput_per_client_Spec*Protocol_efficiency * space_per_client

Downlink dB Link Budget AP (estimate) = TxPower(AP) — Attenuator(AP) — Attenuator(Client) — Client(Noise Floor) > Data Rate SNR

Uplink dB Link Budget (estimate) = TxPower(Client) — Attenuator(Client) — Attenuator(AP) — AP(Noise Floor) > Data Rate SNR

SIR (estimate) = TxPower (AP/Client) — greater of [Co-Channel_Power(AP/Client), Power of adjacent-channel (AP/Client)]

Redundancy (to ensure network uptime) — Generally, because RF spectrum is a fixed resource, cell overlap is kept to a minimum. In some applications, redundancy may be considered. In this case, RF redundancy is the assurance that a coverage hole does not occur if an AP goes down. The spacing of access points and their antennas should not create a coverage hole if an AP or APs go down. The straightforward answer is to duplicate APs, one for one, on the same channel and duplicate network redundancy within the infrastructure. A less expensive approach would be to interlace APs with semi-overlapping coverage given wired side redundancy.

Fairness — In some applications, the effect of bandwidth sharing for more users under one AP relative to another AP with less users results in better user throughput under the less loaded AP. To ensure 100 percent balance among all users, one method — which can be a tricky task — is to equally balance all users based on the lowest performing user. Another option is to hard-fix all users at a minimum fixed link data rate and throughput.

Picocells are not as easy to create as back-of-the-napkin-engineering would lead you to believe. Some factors might require a decrease in picocell sizes, such as client density, throughput per user or total system capacity. Other factors typically result in an increase in picocell size and density, such as co-channel reuse, data rate SIR/SNR requirements, and the 802.11 CSMA/CA protocol mechanisms. Picocell creation is a delicate balancing act, requiring careful analysis of budgets, data rates, density, available channels, redundancy and throughput. However, when implemented appropriately, picocells can provide significant performance gains in many different types of indoor WLAN deployments.

Gregory Davi is a systems engineer at Airespace,with numerous years of experience dealing with RF technology, including three patents pending on RF location tracking and one on RF security. He's authored several papers on wireless networking while at Airespace and in his previous role as -senior systems engineer at Metricom.

RSS    Save to Del.icio.us  Digg This