Back to the future
More bandwidth means more power from basestations, but tricks from the 1930s are stepping in to help, notes E&T.
Cellular operators are turning green. They are demanding that basestation and core network equipment consume less power. This is not just to satisfy a fashionable sustainability agenda but to deal with the waves of data they are now pushing out to the latest generations of smartphones. In some cases, they have been going back through the annals of RF circuit design to the 1930s to come up with new (old) ways to eke out their power budgets.
On the basestation side, operators are becoming concerned about the increasing use of data leading to higher power consumption. As they move to higher-bandwidth forms of 3G and then to Long Term Evolution (LTE) and WiMax, the modulation schemes used to deliver the higher data rates have a knock-on effect on RF power, and not in a good way.
"Pushing for higher data rates means the power in the basestation expands massively and the electricity and the cost of ownership goes up dramatically. You have more complex modulation and bandwidth, which are pushing towards architectures such as [the] Doherty [amplifier architecture]," says Mark Murphy, director of marketing for RF power products at NXP Semiconductor.
In the shift from 2G protocols such as GSM to the code-division multiple access (CDMA) schemes of 3G, the modulation techniques became more complex. GSM uses the Gaussian minimum-shift keying technique, which adjusts the phase of a constant-amplitude sine wave backwards and forwards to relay data. This technique lets a linear power amplifier operate in its most efficient, saturated region.
With 3G, the demand for high data rates has pushed the protocol designers into adopting more complex modulation schemes, such as quadrature amplitude modulation. These alter both the phase and amplitude of a signal to represent the data. Unfortunately, to improve the accuracy of the signal output, the power amplifier has to operate in its linear region, which is less efficient than working in the distinctly non-linear saturated region. Digital pre-distortion can improve the situation by making it possible to get closer to the saturated region, but it's not a complete solution.
As the modulation schemes get more complex the ratio of peak to average power increases, which means that the amplifiers spend a greater proportion of their time working at lower efficiencies. "If you look at LTE, the peak to average ratios are increasing, getting up to 8 to 9.5," says Murphy. "You have to have more efficient amplifiers."
One way around the problem is to do a subtle redesign to the traditional Class AB amplifier.
Hardly a new circuit - Bell Labs researcher William Doherty demonstrated it using vacuum tubes in May 1936 - the Doherty architecture puts two amplifiers in parallel, with one supplying the bulk of the power at a constant level and a second providing additional energy for modulation peaks. Because the main amplifier can be run closer to saturation, the overall efficiency increases.
There is no reason to stop at two. NXP is considering a three-way design. Engineers from Pohang University of Science and Technology in Korea have suggested that a five-way design could be useful, although their most recent proposal was a three-way implementation for wideband-CDMA protocols.
A second concern among basestation makers is flexibility, as new procotols such as LTE and WiMax arrive and park themselves alongside existing protocols such as wideband-CDMA and CDMA2000.
Julian Hildersley, vice president of strategy and handset products at RF amplifier start-up Nujira, says: "One of the issues across the industry is the allocation of frequency bands. There are some 14 bands proposed today for LTE and others are coming up fairly regularly. We see a lot of regulatory authorities that want to reallocate UHF [analogue TV] spectrum. It is unlikely that they will all pick the same sub-bands. There is a risk of regional fragmentation both for handsets and infrastructure.
"Most amplifiers are narrow-band by nature," he added. "One of the benefits of envelope tracking is that it enables wide-band amplifiers to be designed. It is possible to think of one amplifier design that can cover, say, WiMax and LTE.
"Wide-band is not essential but if a major infrastructure supplier has to design a power amplifier for each sub-band, that is a headache."
Nujira's weapon for improving efficiency and to create wider bandwidth amplifiers is another blast from the past, first put forward a year after Doherty's proposal. Envelope tracking continually adjusts the supply voltage to the power amplifier so that it gets just enough to deliver the required signal. This reduces the amount of energy from the notional over-voltage in a conventional amplifier design that is simply dissipated as heat.
Nujira is not alone in working on envelope tracking. Murphy of NXP says: "We do [work with] Doherty but we cover other architectures as well, such as envelope tracking. All the different architectures are being looked at, but we see Doherty as the real workhorse for the moment. And for early versions of LTE and WiMax, we still see Doherty taking up 80 to 90 per cent of the market."
According to Murphy, the issues with envelope tracking are commercial: he says the designs that are possible today rely largely on discrete components, which pushes up cost. At NXP, Doherty provides greater opportunities for integration. This has been demonstrated in the launch of integrated Doherty amplifiers, built on an LDMOS process.
"With envelope tracking, the solutions that people have on the market are discrete and are expensive. You have to make it integrated to make it cost effective. And IC design becomes a challenge in terms of the speed and frequency at which the modulator has to track the envelope," Murphy explains. "On the handset side, you see a lot of envelope tracking being integrated. But the handset is a different ballpark: two different worlds of cost and power levels."
However, Hildersley says that envelope tracking is working: "Getting it all in the same box, that is the innovation that Nujira introduced. The modulator today is built from 90 per cent distributor-available parts but we have some innovative steps.
"We can do a lot with integration. I don't think it will come down to a single-chip device that delivers all that power. But we can improve performance and reduce cost and physical size."
Murphy says: "As peak-to-average ratios get higher, Doherty gives you the highest efficiency that is available in a mature architecture. We started off three years ago with symmetrical Doherty and for LTE we are looking at three-way Doherty. We see that as the next solution."
But he does not rule out a shift towards envelope tracking: "We will support both Doherty and envelope tracking. And long-term we have the switched-mode power amplifier. For those we create our own modulator ICs for the amplifier."
In the same way that switching power supplies have supplanted linear designs in most computer and telecommunication power-delivery systems, the same could happen to RF power amplifiers.
Murphy reckons switched-mode will offer greater flexibility to suit the 'one size fits all' basestation market, whereas Doherty and envelope tracking will be used in designs that have to satisfy a narrower range of frequencies and protocols.
However, there is a reason why switching architectures are not yet used in RF power amplifiers despite their much higher theoretical efficiency. "The quest for efficiency leads you to DC/DC architectures but they are slow, unless you turn the clock rate up and then you lose the efficiency through switching losses," Hildersley explains.
To move to switching architectures will demand a lot more work to overcome the switching losses or make slow conversion architectures respond much better to the rapid changes implied by broadband modulation schemes. In the meantime, Doherty and envelope tracking should help control power consumption in basestations.