RF Simulation Improves 802.11a System Performance

As pointed out by M. Mizoguchi,², the receive signal of an OFDM system introduces a signal amplitude reduction because there is a phase rotation caused by the carrier frequency offset. Hence, OFDM signals are more sensitive to carrier frequency offset than the single carrier modulation signals. Thus, frequency synchronization is very important.

The issue of frequency synchronization becomes modified in WLAN systems. In these systems, packetized burst signals ar e transmitted without scheduling. Therefore, synchronization must be established burst by burst. The proposed burst structure based on IEEE 802.11a specification (Figure 2).³

Figure 2 shows that the OFDM burst actually has four distinct regions. The first is the short preamble (initial training sequence). This is followed by a long preamble (further training sequence) and, finally, by the signal and data symbols. Guard intervals are inserted between each burst section.

Synchronizing Frequencies
In order to estimate frequency offset, the first two short preambles are used to compute the coarse carrier frequency offset between transmitted signal and receiving signal based on a maximum likelihood algo rithm. We can assume that the received signal sequence for computing the phase offset is given by:

where K is the calculation start point, M is the number of one short preamble, with the eighth and ninth short preambles (t₈ and t₉) used for the phase offset calculation. The correlation can be expressed as follows:

After computing correlation R, the phase offset Δθ can be determined by:

and the frequency offset Δƒ can be found by using the following equation:

where T is duration of one short preamble (0.8 μs).

To estimat e the carrier frequency offset more accurately, a fine frequency offset estimation is performed using the two long preambles after the coarse frequency offset estimation. Then, a DemuxBurst model, based on both coarse and fine carrier frequency offsets, is used to detect and remove the carrier frequency offset in the receiver.

Simulation example
In order to better synchronize frequencies, designers can turn to simulation for help. To illustrate, let's simulate the performance of a power amplifier used in a WLAN system. The purpose of the simulation is to test and verify that the designed power amplifier can meet the WLAN standard.

Figure 3 shows a simulation template that has been created for simulating a power amplifier for WLAN data transmission.

In this simulation example, the signal source block generating the RF WLAN signal has a hierarchical structure. Thus, users can push into the block and see the lower level structure (Figure 4).

As can be seen from Figure 4, the second level of the hierarchy includes a baseband source and a RF modulator. A further push into the baseband source reveals the third level of the hierarchy (Figure 5). At this level it is clearly seen how the WLAN signal is generated.

To generate a WLAN signal, we will follow the IEEE 802.11a stand ard.³. Specifically, using Annex G of IEEE Std 802.11a-1999, WLAN signals can be generated by the following steps:

1. Generate the short training sequence section of the preamble by using W1, W2 and F1 in the lowest branch in Figure 5.
2. Generate the long preamble sequence section of the preamble by using W3, W5, and F2 in the third branch in Figure 5.
3. Generate the signal field bits, coding, interleaving, modulating, multiplexing with the data section by using B2, convolutional coder, interleaver, and BPSK modulator in the second branch in Figure 5.
4. Form the data, scrambling, convolutional coding, interleaving, 16-QAM modulation, and multiplexing with signal section data by using B1, data, scrambler, L1, tail, PuncCoder, interleaver, 16-QAM, and MuxSigandData in the first branch in Figure 5.
5. Map the signal and data into frequency domain, then transform from frequency to time by using MuxSym and IFFT buffer, F3 in the first branch in Figure 5.
6. F orm the short preamble, long preamble, signal, and data to the OFDM burst by using MuxBurst model.

For systems based on IEEE 802.11a, the error vector between the vector representing the transmitted signal and the vector representing the error-free modulated signal defines modulation accuracy. This magnitude is determined using error vector magnitude (EVM) measurements.

The EVM measurement helps design engineers determine whether root-mean-square (RMS) value of a specific part of the burst does not exceed limits set under the IEEE 802.11a specification.

Figure 6 shows a typical EVM measurement on a power amplifier employed in an 802.11a system.

In this figure, EVM is estimated using the following steps:

1. The start of frame is detected by using W1 (WLAN_BurstSy nc) in Figure 6.
2. The transition from short sequences to channel estimation sequences is detected and fine timing (with one sample resolution) is established by using the WLAN_BurtSync in Figure 6.
3. Frequency offsets are estimated by using the WLAN_FreqSync model in Figure 6. The packet is derotated according to estimated frequency offset by using the WLAN_DemuxBurst.
4. The complex channel response coefficients are estimated for each subcarrier by using the WLAN_PhaseEst and the WLAN_ChannelEst in Figure 6.
5. Each data OFDM symbol is transformed into subcarrier received values, the phase from the pilot subcarriers is estimated, subcarrier values are rotated according to the estimated phase, and each subcarrier value is divided with a complex estimated channel response coefficient by using the WLAN_MuxDataChEst, WLAN_PhaseTrack and WLAN_Equalizer.
6. For each data-carrying subcarrier, the closest constellation point is determined and the Euclidean distance from it is calc ulated.

The RMS average of all errors in a packet is calculated using the formula:

where:

L_p is the length of the packet;N_f is the number of frames for the measurement; (I₀(i,j,k), Q₀) denotes the ideal symbol point of the i^th frame; j^th OFDM symbol of the frame; k^th subcarrier of the OFDM symbol in the complex plane; (l(i,j,k), Q(i,j,k)) denotes the observed point of the i^th frame; j^th OFDM symbol of the frame; k^th is the subcarrier of the OFDM symbol in the complex plane; and P₀ is the average power of the constellation.

As can be seen from the above calculation procedure, the EVM represents the distance between the measured and expected carrier magnitude and phase at some point in time af ter it has been compensated in timing, amplitude, frequency, phase and DC offset.

Testing/Verifying Power Amps
For testing and verifying any power amplifier design, the basic WLAN system design shown in Figure 4 above is employed. Assume that the data is coded by 64-QAM modulator, the pilot signal coded by BPSK, which, along with 10 short preambles and 2 long preambles, are framed as the WLAN signal, as shown in Figure 2 from above. Per the 802.11a OFDM modulation scheme, the framed WLAN signal is routed through a power amplifier and sent to a receiver.

Now let's test and verify an actual WLAN power amplifier to see if it can meet the requirements given by IEEE 802.11a. In the example to follow, we will consider a power amplifier built using two cascaded stages. From circuit-level simulations, the relation between input and output power for the power amplifier is shown in Figure 7.

There are two modeling approaches that can be employed for measuring the power amplifier. Under the first approach, an RF/DSP co-simulation can be performed using the circuit envelope simulation for the PA. Under the second approach, which will be discussed here, a behavior timed RF_Gain model is employed to simulate the performance of the power amplifier at the system level.

In this modeling approach, the RF power amplifier's complex input signal V_l(t) is represented by the in-phase (I) and quadrature (Q) components surrounding its carrier frequency. The output signal is given by:

where a denotes the gain of the c omponent as set by the component parameter gain. If the input is a baseband-timed signal, then only the real part of the gain is used. g_comp denotes the gain compression factor as determined by the gain compression parameters, such as GCType, TOIout, dBc1out, PSat, GCSat and Gcomp. In this example, we will discuss the dbc1out case.

The non-linear characteristics for dBc1 are described in Figure 8.

Based on the non-linear power amplifier characteristic shown in Figure 8, designers can determine the dBc1 value for the power amplifier. So, the RF gain parameter can be specified as shown in Figure 9.

EVM Measurements on the Power Amp
Now let's turn to EVM measurements to determine the modulation accuracy of the power amplifier module. Under the 802.11a specification, power amplifiers must deliver modulation rates 6, 12 and 24 Mbps rates when EVM are performed. In a production environment, it should be necessary to measure the EVM only at the highest rate supported. This would be an EVM value of 15.8% for all modems--54 Mbps modems will need to achieve 5.6% EVM; while 36 Mbps modems will need to achieve 11.2% EVM.

Figure 10 displays some EVM test results on the power amplifier module. These EVM values were automatically compared to the IEEE 802.11a standard and the most important final result is shown.