After another long and productive circuit-simulation session, Joe Engineer needed a break. "To reward myself, I'll download the complete works of Britney Spears MPEG collection," thought Joe. He figured that transferring the several-hundred-megabyte video file would tie up the office's 802.11b network for several minutes. To occupy himself, Joe began sending e-mails. He sent some large attached files over the Bluetooth link from his PDA to his 3G cell phone.
Did the interference between the Bluetooth and 802.11b networks frustrate the data transfers? Or did Joe have a positive wireless-networking experience? Don't underestimate the importance of these questions. The co-existence of IEEE 802.11b and 802.11g with Bluetooth wireless personal-area networks will be crucial to user satisfaction. These networks use the same 2.4-GHz spectrum. Often, they operate in the same locations. To determine how the networks can be used simultaneously, it is necessary to evaluate how they interact with one another. It then becomes possible to maintain acceptable performance on each network.
The IEEE 802.11b and 802.11g standards (referred to here as 802.11b/g) for WLANs will replace many wired-LAN computer networks. Specifically, the 802.11b standard provides for payload data rates of 1, 2, 5.5, and 11 Mbps. For 1- and 2-Mbps data transmission, respectively, it uses Differential Binary Phase Shift Keying (DBPSK) and Differential Quadrature Phase Shift Keying (DQPSK) direct-sequence spread-spectrum modulation. For data transmission at 5.5 and 11 Mbps, 802.11b specifies Complementary Code Keying (CCK) modulation.
In contrast, the 802.11g standard uses Orthogonal Frequency Division Multiplexing (OFDM) modulation to extend 802.11b with payload data rates up to 54 Mbps. In these WLANs, an access-point radio wirelessly connects terminal devices, such as personal computers, to each other and to the wired network. The maximum distance of terminal devices from the access point is 30 to 100 m, depending on the data rate.
According to the transmit spectrum mask of the IEEE 802.11b/g standards, a channel's occupied bandwidth must be less than 22 MHz. Three non-overlapping 25-MHz spaced channels can co-exist in the 80-MHz-wide ISM band. Although channel agility is an option for 802.11b and 802.11g access points, many implementations are expected to be fixed on a single channel.
Bluetooth wireless personal-area networks (WPANs) are intended to provide wireless datalinks between cell phones, wireless headsets, personal digital assistants, personal computers, and other devices in PANs. These devices communicate with one another within a range of approximately 10 m. Bluetooth signals are FSK modulated with a bit rate of 1 Mbps. The 2.4-GHz ISM band is divided into 79 Bluetooth channels, which are spaced 1 MHz apart. Bluetooth networks frequency hop through a pseudo-random selection of these channels at 1600 hops per second.
Typically, 802.11b and 802.11g WLAN access points will be stationary. Operating frequencies can then be planned to minimize interference between WLANs. This isn't the case for interaction between Bluetooth and 802.11b/g networks, however. Even if they're in areas where 802.11b/g networks are operating, devices like cell phones need to maintain their Bluetooth links to other devices. When Bluetooth networks hop onto a channel used by an 802.11b/g network, disruption of either or both networks is possible. Adaptive-hopping procedures are being considered for a new Bluetooth standard. They would help it avoid the frequencies being used by 802.11b/g networks.
Standards and designs must be created that will most effectively allow the two networks to co-exist. To begin this development, it's necessary to know the conditions that cause the interference between networks. The connected solutions of Agilent EEsof EDA's Advanced Design System (ADS) and Agilent's Vector Signal Analyzer software (virtual VSA) offer capabilities for the analysis of both Bluetooth and 802.11b/g networks.
Using ADS and the software-based VSA, several combinations of networks and interfering signals were simulated and analyzed. Here are the results for the following combinations of desired and interfering signals:
- Bluetooth performance with 802.11b interference
- Bluetooth performance with 802.11g interference
- 802.11g performance with Bluetooth interference
In all of these simulations, the results show bit-error rate (BER) when collisions occur between the desired and the interfering networks. The interfering signal is applied at 100% duty cycle at a constant frequency. In this worst-case scenario, every packet that is transferred on the network collides with interfering-signal packets. In actual applications, the probability that a collision will occur is less than 100%. To predict performance, one needs to know the interfering-signal power that will result in the specified BER when network collisions occur.
Over an 8-hr. day, an average 802.11b/g network may be expected to transmit only a small percentage of the time. The average BER that a network experiences during a day is therefore low, even though it is high during collisions. However, the average BER over a day may be a poor indicator of user satisfaction. The user may find network performance unacceptable if it's severely degraded during periodic intervals of simultaneous heavy activity.
The bandwidth of a Bluetooth channel is less than 1 MHz. Meanwhile, 802.11b/g signals may be as wide as 22 MHz. To simplify the calculation of a Bluetooth network's performance with 802.11b/g interference, a broadband noise source is sometimes used to represent the interference. Simulations also may include the details of the 802.11b/g modulation along with full models of the transmitter, receiver filter amplifiers, and mixers. These simulations show the extent of validity for broadband noise assumptions. They also allow circuit designers to identify modulation- and filtering-dependent effects.
The performance of a Bluetooth network was simulated with interference from an 802.11b source, an 802.11g source, and broadband noise sources. The simulations determine the network's BER, as the interfering signal's power and frequency offset are both varied. The simulations include the transmit filter and modulation characteristics, which determine the interfering-signal power density as a function of frequency offset.
To analyze the performance of the Bluetooth network, the threshold between acceptable and unacceptable performance of a Bluetooth raw BER is set to 0.001. With Bluetooth packet lengths of 366 bits, this BER produces a raw packet error rate (PER) of 31%. Network performance with 31% raw PER is worse than insignificantly degraded. Yet it is better than complete network failure.
Bluetooth hops frequency through 79 channels over an 80-MHz frequency span. With the transmit filters used in this analysis, the 802.11b signal occupies 15 of those 79 channels. At the most, collisions between an actual Bluetooth network and an 802.11b network will occur 19% of the time. With a BER of 0.001 during collisions, the total PER will be about 5.9% (assuming the error rate is much smaller with no collisions). Although this PER may not be acceptable for Bluetooth voice applications, it should be sufficient for data transmission.
The simulation results of a Bluetooth network with 802.11b, 802.11g, and filtered broadband noise interference can be seen in Figure 1. The plot shows the minimum interference power that degrades a Bluetooth receiver's BER to 0.001. Note what happens when the center frequencies of the Bluetooth and 802.11b signals are the same. An interfering 802.11b signal power, which is 8 dB lower than the Bluetooth power, then degrades the Bluetooth BER to 0.001.
The 802.11b interference power that's required to degrade the Bluetooth BER to 0.001 remains fairly constant as the signal offset increases to 4 MHz. As the frequency offset increases to 7 MHz, the 802.11b interference power—which is required to cause 0.001 BER—rises at about 2.6 dB per MHz. As frequency offset increases, the 802.11b transmitter filter is the primary factor that determines the change in interference power for constant BER.
The plot also shows the results of simulations in which a broadband noise source interferes with the Bluetooth network. That noise source has the same transmit filters as the 802.11b source. Look at the plot for the broadband-noise-source power that is required to degrade Bluetooth BER to 0.001. The broadband noise power level is about 1 dB less than the 802.11b power that produces the same BER.
The simulation results for 802.11g interference with a Bluetooth network also are plotted in Figure 1. At up to a 4-MHz offset, the 802.11g signal power—which degrades Bluetooth BER to 0.001—is greater than it was for 802.11b interference. This difference occurs because the 802.11g signal is spread over a wider bandwidth than the 802.11b signal. As a result, the spectral power density of the 802.11g signal is lower when the total signal power is the same. At 0-MHz offset, the 802.11g power is particularly high. The power rises because the center OFDM subcarrier isn't used.
A simulation also is done for the interference produced in a Bluetooth network by the broadband noise source that uses the same transmit filters as the 802.11g source. The results are shown in Figure 1. The transmit filter passband of the 802.11g source is 20 MHz wide—significantly wider than the 802.11g occupied bandwidth. The filters have little effect on the spectral shape of the 802.11g signal within the occupied bandwidth. The total noise-source power is measured over a 17-MHz bandwidth. As a result, the average noise power density in the 17-MHz span equals the density of a 802.11g source with equivalent total power. The broadband-noise-source power that produces 0.001 BER in the Bluetooth network is close to the power of an interfering 802.11g source that produces the same degradation.
To calculate the carrier-to-interference ratio in the Bluetooth receiver, divide the interference power in the Bluetooth-receiver noise bandwidth by the received Bluetooth signal power. To produce 0.001 BER for all of the interference sources in these tests, the Bluetooth receive band requires an average carrier-to-interference ratio of 18.6 ±0.3 dB.
By using broadband noise sources as interference instead of 802.11b/g modulated sources, it's possible to get a reasonable estimate of Bluetooth performance. When making this approximation, the power of the noise source in the Bluetooth-receiver bandwidth must be the same as the power that would be produced by an 802.11b/g source. To guarantee that interference spectral densities are accurately modeled, designers often include the full details of 802.11b/g modulations and filters in simulations.
IEEE 802.11a and 802.11g both use OFDM modulation, which divides 16.25 MHz of bandwidth into 52 subcarriers that are 312.5 kHz wide. An OFDM data packet consists of a preamble, a header, and a data block. In the data block, 48 subcarriers are used to transmit data. These carriers may be modulated with BPSK, QPSK, 16-QAM, or 64-QAM, depending on the data rate. Four subcarriers serve as pilot signals in the data block. By using the pilot signal as a reference for phase and amplitude, the 802.11g receiver demodulates data in the other subcarriers.
The pilot signals allow the receiver to compensate for the OFDM signal's phase and amplitude distortion. When numbering a channel's OFDM subcarriers from −26 to +26, the pilot signals are on channels −21, −7, +7, and +21. If the 1-MHz wide Bluetooth interference is applied to the 802.11g signal, only a few subcarriers are directly affected. Any Bluetooth interference that falls on the pilot subcarrier can produce errors in the phase and amplitude correction, which the receiver uses when demodulating the data subcarriers.
Figure 2 shows the test schematic used to simulate 802.11g BER with interference from a Bluetooth source. The 802.11g RF waveform is generated from a pseudo-random bit stream. An interfering Bluetooth RF signal is then added to the 802.11g RF waveform. The combined RF signal is input to an 802.11g receiver. That 802.11g receiver demodulates the signal and outputs a bit stream. The bit stream is compared to the bit stream that was used to create the 802.11g RF signal. In the simulations presented here, the data rate of the 802.11g signal is 48 Mbps. At that rate, the OFDM data subcarriers use 64-QAM modulation.
In Figure 3, the simulation results are depicted for an 802.11g network with interference from a Bluetooth transmitter. The relative Bluetooth signal level, which produces a BER of 0.001 in the 802.11g link, is shown against the frequency offset of the Bluetooth and 802.11g signal center frequencies. At 0-MHz offset, an −11-dB Bluetooth signal produces 0.001 BER. Between 0-MHz and 5-MHz offset, the Bluetooth signal power that's required to produce 0.001 BER is nearly constant—except at 2-MHz offset. When the Bluetooth and 802.11g signals are offset by 2 MHz, the Bluetooth interference—which is 22.5 dB below the 802.11g signal level—produces 0.001 BER in the received 802.11g signal. At 2-MHz offset, the 802.11g network is 11 dB more sensitive to degradation than it is at either 1-MHz or 3-MHz offset. At 7-MHz offset, the Bluetooth power drops to 13 dB. It then increases to about 3 dB/MHz from 7-to-10-MHz offset.
At 2-MHz offset, the Bluetooth signal interferes with an 802.11g pilot subcarrier that's 2.19 MHz from the 802.11g center frequency. This interference causes the rise in 802.11g network degradation at this offset. Although increases in degradation also occur at 6-MHz and 7-MHz offset, they're not as strong. The pilot subcarrier at 6.56-MHz offset is between these 6- and 7-MHz offset channels. Thus, the edges of the 1-MHz wide Bluetooth signals fall on the pilot subcarrier.
The simulated EVM of each OFDM subcarrier was analyzed using the software-based VSA (FIG. 4). For the subcarriers numbered from −26 through 0, the EVM is between 2% and 5% rms. The subcarriers that are nearer to the Bluetooth interfering signal have larger EVM. Subcarrier number 12, for example, has an EVM greater than 100%. The power of the Bluetooth signal is 11 dB less than the total 802.11g power. It is 6 dB greater than the power of the individual OFDM subcarriers. As a result, the EVMs are expected to be very large for subcarriers that have frequencies very close to the frequency of the Bluetooth signal.
In an OFDM signal, pilot subcarriers are used as phase and amplitude references for demodulating data subcarriers. Bluetooth interference on a pilot subcarrier can therefore cause errors in the demodulation of data subcarriers. Figure 5 and Figure 6 show the 802.11g constellations with the interfering Bluetooth signal offset 2 and 4 MHz, respectively. In both cases, the power of the interfering Bluetooth signal is 21 dB below the 802.11g signal's power. The interference at 2-MHz offset causes errors in the reference phase, which is used for the demodulation of all subcarriers. In Figure 5, this phase error shows up as a circular smearing of the 64-QAM constellation. The interference does not fall directly on a pilot in Figure 6, however. There, the constellation errors are more randomly scattered.
To approximate the interference from 802.11b/g signals on Bluetooth networks, determine the Bluetooth degradation that will be produced by broadband noise interference. Be sure to include models of 802.11b/g transmit filters in simulations of broadband noise interference with Bluetooth networks. It's then possible to determine performance degradation as a function of frequency offset.
The degradation of an 802.11g network by an interfering Bluetooth signal is much more severe when the Bluetooth frequency is very near an OFDM pilot subcarrier. Say an 802.11g network is degraded by an interfering Bluetooth signal, which falls directly on the pilot signal in the seventh OFDM subcarrier. The degradation will be the same as the degradation produced by the 10-dB stronger Bluetooth signals falling on OFDM data subcarriers. To determine 802.11g network performance as a function of frequency offset, simulation can include the effects of 802.11g modulation, Bluetooth modulation, and system filters.
The simulations reveal that interference between 802.11b/g and Bluetooth networks can be significant when they're used simultaneously, like Joe Engineer uses them. As he downloads files on his WLAN while there's a large amount of activity on his WPAN, Joe's experience will be determined by the relative signal power of the 802.11b/g and Bluetooth signals at each receiver. In some cases, Joe Engineer may be frustrated by his simultaneous wireless networking experience. Luckily, other cases will render him unaware of any interference.
The authors would like to acknowledge the support of Innovative Wireless Technologies (IWT) and Agilent Technologies in this joint effort to demonstrate wireless-networking issues and analysis techniques.