As designers turn their attention to mobile WiMAX devices, they are quickly learning that there are some specific design challenges regarding power amplifiers. For Wave 2 mobile WiMAX products, the mobile device needs to efficiently deliver +23 dBm output power with high linearity from a 3.3 VDC supply.
Managing power in mobile WiMAX is quickly shaping up to be vitally important as first-generation designs are tested and deployed. One of the challenges of designing for mobile WiMAX is its long range, since WiMAX networks typically achieve coverage of about 1 km per cell.
To achieve these ranges, WiMAX must have an optimized power profile—from the base station right down to the components in the mobile device. High transmit power, then, is important. But how high can WiMAX go and what are the limitations imposed by regulatory bodies, technological limits, and usage models?
Designers of the power amplifier (PA) and those selecting PAs need to find the optimal balance between high power and high efficiency in order to ensure robust links, high data rates, and good range for their WiMAX services.
The nature of WiMAXWhat makes WiMAX challenging for designers is that it is an access technology with a unique set of constraints. As a result, power amplification circuits that were used for cellular or Wi-Fi applications cannot simply be dropped into WiMAX designs and tweaked to perform adequately.
In many ways, WiMAX can be considered a hybrid technology because it shares aspects of both cellular and Wi-Fi networks. Mobile WiMAX is very similar to cellular; it is meant to be used in highly mobile devices and it uses licensed frequency bands (so users expect high reliability). It also employs transmit power control techniques, much like CDMA cellular does.
However, it differs from cellular because it operates at much higher data rates (resulting in more stringent linearity requirements) and must simultaneously handle voice over Internet Protocol (VoIP), data, and video transmissions. Managing the bandwidth and priority of transmission for these types of services requires a quality of service (QoS) component that is not required for mobile voice alone.
On the other hand, WiMAX is also similar to Wi-Fi. For instance, it offers high data rates, uses orthogonal frequency division multiplexing (OFDM) with modulations from BPSK to 64-QAM, and is an all-IP-based network.
However, it differs from Wi-Fi because it uses a fully-scheduled service, unlike the collision-based carrier sense multiple access (CSMA) technique used by Wi-Fi. This gives WiMAX a significant advantage over Wi-Fi.
As the number of users increases in a CSMA network, overall capacity drops dramatically since each collision requires a subsequent retransmission. With a scheduled service, overall network capacity is unaffected as the number of users increases, since the basestation manages each user's access to the network efficiently.
Simply put, RF power translates directly into range, so higher power equals longer range. To achieve long range, WiMAX basestations transmit at power levels of approximately +43dBm (20W), as compared to Wi-Fi APs, which typically transmit at +18 dBm (60 mW).
A WiMAX mobile station (MS) typically transmits at +23 dBm (200mW), as compared to +18 dBm (60 mW) for Wi-Fi. Cellular (CDMA) transmit powers for both the BS and MS are similar to those used in WiMAX.
However, because WiMAX uses much higher modulation orders to achieve higher throughput, WiMAX requires a much better SNR than cellular. For the mobile transmitter, high modulation orders require a PA with much better linearity and greatly complicates PA design compared to GSM or CDMA.
You might notice that there is a large difference (approximately 20 dB) between downlink power (from the BS to the MS) and uplink power (from the MS to the BS), so mobile WiMAX networks are severely uplink limited (this is also the case for cellular networks, of course).
This means that, while a mobile can easily receive transmissions from a BS, the mobile's relatively low transmit power makes it difficult for the BS to hear it.
One way to combat this mismatch is by using a technique called subchannelization, where only a subset of all of the available subchannels is used for any particular user.
In effect, each mobile concentrates its power over a smaller range of frequencies, and the net signal gain is 10*log(Ntotal/Nused), where Nused is the number of subcarriers assigned to the user, and Ntotal is the total number of subcarriers available.
For example, if a user is assigned one subchannel made up of 24 subcarriers, the net gain that is achieved relative to the BS that is transmitting on all 841 allocated subcarriers is 10*log(841/24)=15.4 dB. The other subcarriers are made available to other users, and they can use these simultaneously.
Another technique to address the link imbalance is adaptive modulation. In this case, the mobile transmits using a lower order modulation compared to the BS. For example, the mobile could transmit QPSK or 16QAM signals, while the BS transmits using 64QAM.
Because the SNR required to receive QPSK or 16QAM is lower than 64QAM, using a lower order modulation allows the MS to communicate with the BS using less transmit power (although uplink throughput is reduced, since fewer bits are transmitted per subcarrier with lower order modulation).
For example, the SNR required for QPSK-1/2 is 5 dB as compared to 10.5 dB for 16QAM-1/2 and 20 dB for 64QAM-3/4 modulation1. If the MS transmits with QPSK, the BS can tolerate 5.5 dB more link loss than with 16QAM.
When sub-channelization and adaptive modulation are combined, a network operator can effectively balance the uplink and downlink budgets, and the network will operate bi-directionally.
The downside is that when these techniques are used, the uplink throughput will be lower than the downlink throughput; subchannelization limits the number of subcarriers available for mobile transmission, and lower order modulation means that fewer bits are transmitted on each available subcarrier.
To understand why this is the case, consider a mobile device moving from the edge of the cell directly towards the BS. When it is at the extreme cell edge, path loss will be very large, so the mobile device will be transmitting at maximum power with the most robust modulation.
As a result, uplink data rates will be relatively low. However, with the high MS transmit power and robust modulation, the BS will be able to receive transmissions from the MS, and the link is sound.
As the mobile moves closer to the BS, path loss decreases. The signal level at the BS increases, and the SNR improves, since the received signal is now farther above the noise floor.
In response, the BS may instruct the mobile to start reducing power (to minimize potential for interference between different mobile stations). However, as soon as the signal level supports a higher order modulation, the BS will instruct the mobile to switch modulations in order to increase overall network capacity.
Going back to our example comparing QPSK and 16QAM, suppose a transmitter operates at +23 dBm and it just achieves the 5 dB SNR required for QPSK when it is at the edge of the cell. As is moves closer to the BS, path loss drops, and the BS may ask the MS to reduce its transmit power.
However, as soon as the path loss has decreased by 5.5 dB, the BS will instruct the MS to switch to 16QAM-1/2, and will increase transmit power back to +23 dBm, since the MS will now be able to achieve a 10.5 dB SNR. Therefore, a mobile will typically transmit at higher powers until it is close enough to the BS to achieve 16QAM operation (or even 64QAM in many instances), at which point power is reduced. This is shown in Figure 1.
Click here for Figure 1.Figure 1: Achievable modulation versus distance with +23 dBm transmit power.
Figure 1 was derived using parameters from a WiMAX Forum whitepaper2. It shows the modulation that is achievable as a function of distance from the BS. We use the parameters in the whitepaper, so, for example, maximum available path loss is calculated assuming a 10 MHz channel bandwidth at 2.5 GHz, with 3 subchannels, and 10 dB penetration loss.
In calculating the path loss, we have assumed a COST231 suburban model at 2.5 GHz with 32 m BS height and 1.2m MS height. This analysis has assumed the presence of slow (lognormal) fading, but is somewhat simplified, since we assume a fixed 5.5 dB fade margin.
In reality, of course, fading is a random process, and closed loop power control will be used to help mitigate its effects. However, for the sake of this analysis, the conclusions are valid, as fading will simply blur the boundaries between the different modulations.
Note that the red ring, labeled QPSK-1/8 represents QPSK-1/2 modulation with a repetition factor of 4. This is the most robust modulation scheme, and it can be seen that it is indeed required at maximum range.
In our analysis, we calculate that with +23 dBm transmit power, an MS must use QPSK-1/8 for mobiles from 0.9 km to 1.35km from the BS. At closer distances, the MS is able to use higher order modulations, and network capacity is therefore increased.
For example, the MS is able to use 16QAM-1/2 modulation at distances from 0.45 to 0.6 km from the BS. Since 16QAM-1/2 modulation transmits 2 bits per symbol, while QPSK-1/8 transmits only 0.5 bits per symbol, one can see that the throughput in the green ring is 4 times higher than in the red ring.
We can also estimate the required transmit power as a function of range. At the edge of each of the zones in Figure 1, the MS will be transmitting at maximum power. It will decrease its transmit power as it moves towards the BS, until it has sufficient power to achieve the next modulation order.
At this time, it will increase transmit power again to maximize capacity.
Figure 2 shows the expected transmit power as a function of distance, showing the impact of adaptive modulation. It can be seen that transmit power is significantly reduced only once the maximum modulation order has been achieved, which in this case is 64QAM-3/4.
Click here for Figure 2.Figure 2:
If the maximum modulation order was instead 16QAM-3/4, then the transmit power would be monotonically reduced once the 16QAM-3/4 rate was achieved.
It should be noted that the presence of fading will result in significant changes to this curve. In a real-life fading environment, additional margin may be required to counteract fading effects, and one would expect that transmitting at maximum power would occur less frequently.
However, the overall trend shown in Figure 2 is correct, and shows that mobile stations will be required to transmit at high powers not only at the cell edges, but also at much closer distances in order to achieve higher-order modulation.
There are two benefits to transmitting at higher power. First, transmitting at this higher output power increases the maximum range. Using parameters from the WiMAX Forum 3, maximum mobile to BS distance is increased from 1.35 to 1.5 km when the output power is increased from 23 to 24.5 dBm, so that the overall coverage area increases by 23.5%.
In principle, one might expect that a network operator could deploy 23 percent fewer base stations, and realize a cost savings. However, this effect may be of only limited benefit, since many networks will have been designed with cell sizes assuming +23 dBm uplink transmit power, so cell sizes may already be fixed.
The second benefit is more significant, however. If an MS is able to transmit at higher power, then it can achieve the SNR required for higher order modulation when it is further from the BS. This improves overall network capacity, so increases overall spectral efficiency.
Figure 3 shows the modulation that is achievable as a function of distance from the BS with +24.5 dBm transmit power.
Click here for Figure 3.Figure 3:
In this figure, we again plot achievable modulation as a function of distance from the BS (and the dashed lines show the ranges for +23 dBm from Figure 1 for reference). Note that the maximum distance has increased from 1.35 to 1.5 km, as discussed above.
However, it is more important to note that users can now achieve higher order modulations over a wider range. For example, for 16QAM-1/2 modulation the maximum range is now 0.7 km, versus 0.6 km for +23 dBm.
As a result, each user will achieve higher throughput over a wider range, and the network aggregate capacity will be increased accordingly. With every additional user who can transmit at a higher power level, overall network capacity increases.
It is important to understand that all users would need to transmit at a higher transmit power in order to allow cell sizes to expand. However, each and every higher power user added to the network increases overall network capacity.
Finally, it is relatively straightforward to calculate the capacity increase seen by increasing transmit power from +23 to +24.5 dBm. We know how many bits per symbol can be transmitted for each modulation scheme, and we know the relative areas that can be covered for each modulation scheme, for both power levels.
When this information is used to calculate relative capacity, we find that it increases by 24% when transmit power is increased from +23 dBm to +24.5 dBm.
Even if the maximum cell size remains fixed at 1.35km when the transmit power is increased to +24.5 dBm (as would be the case if networks were rolled out assuming +23 dBm devices) the capacity still increases by 18% when devices are able to transmit at higher power.
So why not transmit even more power? There are three important factors that limit our ability to transmit at higher power: PA efficiency, available supply voltage, and regulatory requirements.
PA efficiencyIn PAs, efficiency is the measure of the RF power out versus the DC power in. For example, if a PA has a 10 percent efficiency, it would consume 3.55 W to transmit at +25.5dBm (355 mW). If the PA efficiency could be doubled to 20 percent, then the peak power consumption drops to 1.7W.
Today's state-of-the-art WiMAX PAs, like SiGe Semiconductor's SE7262, operate with >20 percent efficiency (See sidebar Why is PA efficiency so low for WiMAX?.)
The PA efficiency has a direct impact on battery life for mobile devices. Of course, the PA is not working all of the time, so the average power consumption will be considerably lower than the peak power consumption quoted above.
For instance, transmit duty cycles for WiMAX devices are typically about 40 percent when the MS has data to transmit. Therefore the average power consumption for a 20 percent efficiency PA will be about 680 mW if the PA is transmitting at maximum power.
Furthermore, often there will be no data to transmit, and in this case, the device will transmit very infrequently (essentially, it transmits only ranging messages to let the BS know that it is still in the cell).
In the end, however, the PA power consumption can have a significant impact on battery life, and it is important that PA efficiency is as high as possible.
Available supply voltageMobile WiMAX devices will be powered directly from the mobile station's battery, and battery supply voltages vary significantly during use. When freshly charged, the battery will operate at about 4.8V.
The supply voltage drops as the battery discharges, and the minimum practical supply voltage before the device shuts down is typically 2.7V. Most manufacturers want to use the battery for as much of this range as possible, and therefore specify that the power amplifier must faithfully deliver fully rated power at 3.3V (and occasionally 3.0 V).
Delivering high power under these conditions imposes some significant challenges. As most circuit designers know, a low supply voltage requires a high current, which implies a very low output impedance. Consequently, matching the low impedance PA output to a 50 Ohm antenna is difficult to achieve.
If higher output powers are required, the impedance becomes even lower, and it becomes increasingly difficult to achieve a good broadband match between the PA and the antenna.
Regulatory requirementsRegulatory requirements also place a serious constraint on how much power a PA can deliver. An ideal linear PA produces only the original frequency from the input signal. In real-world implementations, PA non-linearities introduce new frequencies through intermodulation distortion (IMD), and these out-of-band signals can interfere with users in adjacent channels (referred to as spectral regrowth or spectral leakage).
Regulatory bodies have imposed strict regulations on the amount of power that can be emitted out of band. For example, for mobile devices in the 2.5GHz band, the FCC specifies4 that the emissions must be below -25 dBm/MHz, measured 5.5MHz outside the device's assigned band.
Since this limit is an absolute power measurement, as output power is increased, more and more rejection of out-of-band emissions is required, and the power amplifier must be made more and more linear.
For example, when transmitting at +23 dBm with a 10 MHz channel bandwidth, achieving -25 dBm/MHz requires a net rejection of 23-10log(10)+25=38 dB rejection. Transmitting at 24.5 dBm requires 39.5 dB rejection.
Therefore, it becomes increasingly difficult to meet regulatory requirements as output power is increased. To reduce IMD distortion, the PA must operate more linearly, and the result is that PA efficiency will drop as the output power target is increased.
Recognizing the Tradeoffs Undoubtedly, higher transmit power is important for mobile WiMAX networks. Networks are currently being deployed specifying that the minimum transmit power is +23 dBm.
Each user who enters a network transmitting at powers greater than +23 dBm increases overall network efficiency. However, delivering higher transmit powers comes at a cost to power consumption. As a result, power amplifier efficiency becomes more important as higher output powers are used.