Advertisement

By Chris Irwin, RF Micro Devices, Inc.

Due to the line drawings and tables in this artice, we have made a PDF available for you to download.

With the high levels of integration currently available from Wireless Local Area Network (WLAN) chipset vendors, the front end architecture is one of the key remaining areas that allow WLAN product designers to differentiate their products. Parameters, such as transmitted output power, current consumption, receiver sensitivity, image and spurious rejection, Bill of Material (BOM) cost and occupied footprint are all critically affected by the choice of front end architecture. The fact that these parameters can each play a part in determining how a customer perceives the performance of a WLAN product emphasizes the importance of choosing the correct front end architecture.


There are many applications for WLAN and each has unique requirements and operating conditions that place different demands on the front end. For example, Access Point (AP) and home gateway products are powered from an AC to DC wall adapter, therefore current consumption is not a critical concern, but high output power is perhaps a strong marketing advantage. Other examples include a PDA with an integrated WLAN adapter that needs to consume the lowest amount of current possible and occupy a small board footprint, and a PC Card WLAN adapter that is sold into a more competitive market space where BOM cost is an overriding concern.


The most obvious approach

There are several companies offering highly integrated WLAN chipsets. For example, the Taurus™ chipset from RF Micro Devices is a high-performance solution that uses a minimum number of external components, minimizing the occupied footprint, and it also offers very low current consumption in both receive and transmit modes. These advantages are complemented by using a front end architecture that meets the performance goals of the product while also enhancing the key differentiating features.


The heart of the Taurus chipset is the RF2958 super-heterodyne transceiver. The RF2958 integrates the entire receiver, from the Low Noise Amplifier (LNA) to the analog quadrature demodulator, and the entire transmitter, from the analog quadrature modulator to the Power Amplifier (PA) driver. The RF2958 also features fully integrated dual synthesizers, including on-chip Voltage Controlled Oscillators (VCOs) and image reject up- and down-converters. An external IF SAW filter is used to implement the channel filtering for both receive and transmit paths. The advantage to using an external IF SAW filter is, because it is a passive device, it is much less current intensive than the baseband channel filtering required for a direct conversion architecture.


The front end implementation shown in Figure 2 (architecture 1) is referred to as the most obvious because it is arrived at by a straightforward, step-by-step analysis of the functions that the front end must perform.


First, the PA and LNA must be connected to an antenna without adversely affecting the function of either device and with the minimum insertion loss added. A PHEMT Radio Frequency (RF) switch is the ideal solution for solving this problem because it provides low insertion loss (typically 0.5 dB) and good isolation to the open contact port (greater than 20 dB). This approach is made possible by the half-duplex nature of the IEEE802.11b standard, because no unit is ever transmitting and receiving simultaneously.


The next requirement of the front end is band-pass filtering. In receive mode the LNA should be protected from strong out-of-band signals to prevent blocking of the desired signal and energy at the image frequency and other spurious responses that can jam the receiver. The RF2958 features an LNA with high input power handling capability and an image-reject downconvertor to ease the requirements on the front end filtering. An additional 25 dB of attenuation in the stop-bands and at the image frequency is recommended for robustness of reception in a range of environments. In transmit mode, the spurious mixing products from the upconvertor must be attenuated. The RF2958 features an image-reject upconvertor and excellent linearity in the signal path to ease the requirements on the front end filtering. An analysis of the output spectrum of the RF2958 shows that only the Local Oscillator (LO) leakage needs 25 dB of attenuation to ensure compliance with FCC and ETSI limits for spurious emissions.


The last function required of the front end is antenna diversity switching. The RFMD Taurus chipset offers the capability to implement spatial antenna diversity based on a measurement of the relative signal strength from two antennas. In the case where one antenna happens to be in a position where multi-path fading causes a significant reduction in signal strength, the system will select the other antenna that is receiving a stronger signal. This has the benefit of maximizing Signal to Noise Ratio (SNR) at the receiver input and avoiding severe inter-symbol interference (ISI) situations that could cause errors in the received signal. It should be noted that the proprietary equalizer in the RF3002 effectively cancels ISI over a wide range of multi-path environments, but no equalizer alone can compensate for the cases when the signal amplitude is severely attenuated by fading. This antenna diversity switching can easily be implemented using another PHEMPT RF switch.


Cascading these three functional blocks, from right to left, results in the architecture shown in Figure 2. The RF band-pass filter is effectively multiplexed between the receiver and transmitter on the right and the two antennas on the left.


Front End Metrics

Before considering alternative architectures, it is necessary to establish what metrics are important in evaluating a front end implementation. These metrics are then used to compare the suitability of different architectures for various applications.


The front end must pass the signal from the selected antenna to the LNA or from the PA to an antenna according to the mode of operation. In doing so, it should have the minimum possible adverse impact on the signal, therefore the amount of signal loss and signal distortion should be defined as key metrics of the front end. These parameters can be conveniently evaluated as insertion loss and output 1 dB compression power level, commonly referred to as OP1dB. OP1dB is defined as the output power level at which the gain has decreased by exactly 1 dB, or the signal has begun to compress or distort to the point that its amplitude is reduced by 1 dB at the output. In the architectures considered here, all devices are passive so the gain is always negative and the OP1dB point will occur where the insertion loss has increased by 1 dB.


Insertion loss has an adverse effect on transmit and receive performance. In receive mode, the front end insertion loss precedes the LNA in the signal path. This means that the insertion loss adds directly to the noise figure of the receiver and has a direct impact on sensitivity. In transmit mode, the power output from the PA will be attenuated by the insertion loss, therefore the PA must be capable of providing a higher amount of power, linearly, than will be delivered to the antenna. At the same time, it is necessary for the current consumption of the PA to be higher in order to produce the extra power dissipated in the front end. The power dissipation reduces the overall efficiency of the system.


Output compression causes spectral regrowth, which is a concern in WLAN design. This regrowth is a result of 3rd and 5th order intermodulation products of the main transmitted signal being generated and mixed by the non-linearity of the compressing device. Figure 3 illustrates the case where two unmodulated sinusoidal tones of equal amplitude are being transmitted. Figure 3 also illustrates how the second harmonic of one tone can mix with the other tone to produce an intermodulation product above or below the two original tones. The intermodulation products occur at an offset equal to the difference between the original tone frequencies. This is normally referred to as adjacent channel spectral regrowth. Similarly the fifth order intermodulation products occurs at twice the frequency spacing of the original tones and is normally referred to as alternate channel spectral regrowth. In reality, an IEEE 802.11b modulated signal is a Direct Sequence Spread Spectrum (DSSS) signal as opposed to two pure tones. However, the DSSS signal in the frequency domain can be modeled as a series of many closely spaced tones, each with the appropriate amplitude to give the same power spectral density as the modulated signal. This is illustrated in Figure 4. Any two tones can intermodulate and, if all combinations are taken, the adjacent and alternate channel spectral regrowth can be predicted.


All the functional blocks in the front end have a nominal 50 O input and output impedance. Assuming that mismatch losses are insignificant (this assumption has been proven to be valid by experimental measurement), then insertion loss of the front end is the sum of the individual functional blocks' insertion losses. According to the state of the receive/transmit and diversity switches, there are four paths of interest when evaluating insertion loss.


Table 3 shows that the insertion loss of the architecture 1 front end implementation is 3.0 dB, using insertion loss figures for the filters and switches taken from Table 2. Insertion loss is the same for each path as expected from the symmetry of the circuit.


Our objective with the front end is to not cause any significant spectral regrowth, and typically we need an OP1dB at least 6 dB higher than the maximum desired output power to satisfy this. Commercially available PHEMT RF switches typically offer minimum OP1dB of approximately +29 dBm. For the architecture shown in Figure 2, the power level at the output of the receive/transmit switch must be greater than the desired output power at the antenna by the amount of the insertion loss of the filter and diversity switch. The OP1dB of this switch therefore supports an output power level at the antenna port equal to:

Pout = OP1dB – 6 dB – insertion loss (from switch output to antenna)


Evaluating this equation with maximum expected figures for insertion loss gives a result of +20.6 dBm. Note, this does not mean that the entire system can produce this amount of output power. For instance, the PA may only be capable of driving +21 dBm before the insertion loss of the front end, which provides an output power level of +18 dBm at the antenna port. Rather, +20.6 dBm is the level at which the front end begins to compromise the overall linearity of the system.


Other metrics that should not be ignored are BOM cost and occupied footprint. We can evaluate this in terms of the number of devices in the front end. The front end implementation for architecture 1 uses one filter and two RF switches.


A Minimum Insertion Loss Architecture

Architecture 2 represents the highest performance possible, with no compromise.


In Table 1, looking first at the transmitter, the band-pass filtering and receive/transmit switching are essential functions in the transmit signal path and diversity is not needed. This suggests that the diversity switch does not need to be in the transmit path, which will reduce the insertion loss from the PA to antenna. This implies that transmission always occurs on the same antenna. Looking at the filtering requirement in more detail, the band-pass filter can be moved from the front end and be placed at the input of the PA. This is made possible because the system is linear, and the spurious products that must be filtered all originate in the upconvertor stage of the transceiver device, so moving the filter does not change the spurious emissions level. On the other hand, eliminating the transmit filtering from the front end significantly reduces the insertion loss from PA to antenna. The benefit of this is reduced output power required from the PA for a given amount of power delivered to the antenna (or conversely, a higher amount of power delivered to the antenna for a given amount of power from the PA). This allows reduced power consumption in the PA, or a higher transmitted power level. There is a possibility that the PA could regenerate some harmonics of the transmitted signal, which could require a low pass filter after the PA to provide some rejection. If this proves necessary then there are several options available all with very low cost and low insertion loss, so the benefit of moving the band-pass filter is not lost.


By considering the receiver and looking at Table 1 again, we see that antenna diversity switching, receive/transmit switching and some filtering functions are all desirable or essential in the receive signal path. As mentioned above, the receive band-pass filter and the antenna diversity switch can be removed from the shared path with the transmitter, which dictates that they are placed between the LNA and the receive/transmit switch. By looking at the receive filtering requirement in more detail and as can be seen from Table 1, LO rejection is not required. This opens the possibility that a different filter can be used in the receiver than the transmitter. A search of available filter parts shows that in general, filters without much LO rejection have less insertion loss, by about 0.5 dB, than filters with a minimum 25 dB of LO rejection, everything else being equal.


The next step is to evaluate the metrics that were reviewed in architecture 1 for this architecture.


Table 4 shows that the insertion loss in the transmit path has been reduced by 2.4 dB compared with architecture 1. It also reveals that all paths are no longer equal because the circuit is asymmetrical.


The Rx/Tx switch's OP1dB now supports an output power level at the antenna equal to:

Pout = OP1dB – 6 dB – Insertion Loss (from Switch output to Antenna)


Insertion loss to the antenna has now been reduced to zero, which supports a maximum output power level of +23 dBm.


BOM cost and occupied footprint are increased with architecture 2. Three filters are used (if a LPF is required after the PA) and two RF switches are required.


A Reduced Functionality Architecture

Architecture 3 is another approach that can be considered. This approach eliminates all functions that are not essential in the front end. While this will obviously compromise some aspects of performance, it will also give the minimum possible BOM cost, occupied footprint and complexity.


Table 1 shows that transmit filtering and receive/transmit switching are the only essential functions. We have already seen that the transmit filter can be placed at the input of the PA, which means that the front end can be reduced to simply an receive/transmit switch, as shown in Figure 6. At the same time, an interesting variant on this architecture is to replace the PA with a 50 O gain block. RF Micro Devices' RF2361 GaAs HBT 3 V PA driver amplifier can provide more than +10 dBm linear output power at an extremely economical 20 mA current consumption.


The insertion loss is obviously minimized, so we would expect very good receiver sensitivity, output power and current consumption when transmitting.


On the downside, we have now sacrificed antenna diversity and receive image and blocking rejection. This translates into less robust reception in environments with high levels of multipath or interference signals, which will be observed as throughput varying with time as the interference or multipath conditions change.


The Rx/Tx switch's OP1dB now supports an output power level at the antenna equal to:

Pout = OP1dB - 6 dB - Insertion Loss (from switch output to antenna)


Insertion loss to the antenna has been reduced to zero, so this supports a maximum output power level of +23 dBm.


The BOM cost and occupied footprint are minimized with architecture 3. Only one filter and one RF switch is used.


Other Considerations

Another consideration is the necessary Printed Circuit Board (PCB) traces that connect the signal path. The traces should be implemented as 50 O transmission lines, usually in microstrip, to minimize mismatch losses. There will still be some finite loss per unit length. Cost constraints normally dictate that FR4 substrate material be used in the PCB construction and this can have as much as 1 dB of loss per inch at 2.5 GHz, so these transmission lines must be kept as short as possible.


Choosing the right architecture

Table 6 shows a summary of features and performance of the architectures discussed in this article. The appropriate architecture to use in a given product depends on the application itself. Referring to examples given at the beginning of this article, architecture 2 is probably the most applicable for AP and home gateway products, because it allows for high output power and robust performance. An integrated WLAN adapter in a PDA could use architecture 3 with a gain block, such as the RF2361, as the PA in order to consume very low current and occupy a small footprint. Physical constraints may already dictate the use of only one antenna in that application. Architecture 1 is probably best suited for a PC Card WLAN adapter to keep to a low BOM cost while also delivering robust performance.


Conclusions

Undoubtedly, there are many other possible front end architectures that combine the benefits and features of the architectures discussed in this article in a variety of ways. The approach to conceiving and analyzing the front ends presented in this article can be applied to any WLAN product to select and implement the best architecture to deliver the right features to the intended customers.


Advertisement
Advertisement