27 March 2012
Could a true worldphone be built in the near future?
Apple's new iPad went on sale earlier this month sporting a modem capable of handling the 4G Long Term Evolution (LTE) mobile protocol, but with no 4G network to which it could connect in the UK.
And there was little clarity as to which networks around the world – other than the US – it might be able to deal with. But this is not going to be a problem isolated to Apple, which traditionally has invested heavily in global roaming for its reassuringly expensive iPhones.
"Apple has high enough margins to be willing to ship the same handset everywhere in the world. Most of the other guys would take out some power amplifiers and filters to make region specific handsets," says Christopher Taylor, director of rf and wireless components at Strategy Analytics.
Later this year, the UK telecom regulator Ofcom will auction the spectrum needed by mobile operators to deploy LTE and the country will add its preferred frequency bands to a list that spans of no less than 43 from which governments can allocate to the 4G standard.
What makes an LTE 'worldphone' worthy of the term? The China Mobile led team behind an initiative launched by the Next Generation Mobile Networks Alliance (NGMN) to investigate the LTE worldphone's feasibility, is looking at a device that can support at least 10, possibly 14, individual bands.
"If the phone makes it to market, it will be interesting," says Rodd Novak, chief marketing officer at Peregrine Semiconductor. "But it's the goal of the NGMN to get a global phone to market."
Francis Sideco, senior principal analyst at IHS iSuppli, says: "Ultimately anything is feasible, especially given enough battery and enough space. With the way Samsung has been putting out 10in megadevices that do voice calls, maybe we have enough battery. But it's a little early for a true commercial device."
Gradually, mobile handset makers have moved towards software defined radio designs, where a high performance, parallelised dsp engine handles not just international variations on a communications protocol, but also different standards.
Eran Briman, vp of marketing at mobile dsp specialist Ceva, says the company's XC architecture was developed primarily as a communications engine for protocols based on the orthogonal frequency division multiplexing (OFDM) family of algorithms. "The communications world has been marching towards OFDM. So the XC was designed to be generic enough for LTE, HSPA and Wimax, as well as Wi-Fi."
Software defined radio streamlines production for chipset makers and OEMs because it cuts down the bill of materials (BOM) and the number of different stock keeping units (SKUs) they need to maintain. But that is not currently where the pressure lies.
Sideco says pressure for a worldphone is coming from the operator side and these considerations are going into the requests for proposals (RFPs) and requests for quotes (RFQs). "From the operator side, it's about capturing roaming revenue," said Sideco. "And it's better to have phones roam using the higher generation technology from a cost per megabit and available services standpoint. When you think about the balance of the subscribers, there are not many roaming, but they are responsible for a lot of the revenue.
"Although the manufacturers want to do SKU optimisation, they are driven more by the RFPs and RFQs coming from the operators. Manufacturers are doing what's required of them," Sideco adds.
Unfortunately, the push for global roaming from operators and the flexibility of the dsp in the baseband processor only solve part of the problem for the LTE worldphone. The analogue and rf sections in front of the baseband are where all the challenges lie.
A true worldphone has to cover a frequency range – though not contiguous – from 700MHz to almost 3GHz. To handle each specific band, a handset needs to meet stringent criteria. The specifications for mobile phones demand they can pick up a weak signal adjacent to a very strong blocking signal – and the phone can, potentially, block itself with the incoming and outgoing signals carried on nearby, if not the same, band with time division duplex versions of LTE.
Finding the buried signal
The software workaround would be to sample a much wider band and use digital filters to extract the signal – but that would demand not just high sampling frequencies, but also very high dynamic ranges to let the computer find the buried signal. That demands excessively high resolution from the a/d converter and, in turn, more space and power.
A collection of dedicated passive filters – either surface acoustic wave (SAW) or bulk acoustic wave – that can be switched on and off dynamically works out to be more economic. Ideally, the filters and duplexers – used to filter the transmitter's output from the receive chain – could be tuned to cover different bands.
Taylor says: "People talk about tunable duplexers. I have spoken to experts and they say it will not happen soon. If the bands were laid out and spaced uniformly, the problem would not be so bad. But they aren't. For some of the bands it's easy; for others, it's virtually impossible. A tunable duplexer could maybe hit two bands."
Novak says: "There is work being done on tunable SAWs, but the techniques being used result in huge insertion losses. No one has accepted those. We think we have an approach that can bring this type of device in much sooner. It's in research and, for the moment, we are keeping this as quiet as we can."
The power amplifier (PA) does not represent as large an issue as the filter and duplexers, but the tradeoffs involved are not good news for handset makers. Ideally, the 2GHz or so of spectrum which a handset might need to address would be covered by a single wide bandwidth amplifier. Taylor says while PA bandwidth is likely to increase, it will not cover the entire spectrum.
More likely, says Ben Thomas, director of marketing for advanced cellular platforms at RFMD, is a collection of PAs grouped into low, mid and high ranges. "The high range will be around 2.5GHz; I don't see that consolidating into the mid for a long time. I think any consolidation will happen in the low and the mid range. The high will serve very few bands."
Thomas adds that, while increasing the bandwidth of one amplifier reduces the total number needed inside a phone, this would reduce efficiency and, with it, battery life. The options open to PA designers to claw back efficiency are now very limited.
"The next step is to move to envelope tracking," says Thomas. Envelope tracking adjusts the supply voltage to the power amplifier continually, so it gets just enough to deliver the required signal. This reduces the amount of energy compared to that delivered by a conventional amplifier and dissipated as heat. The efficiency improvement from envelope tracking tends to increase the further the amplifier operates from its frequency sweet spot, lending itself to wideband amplifier designs.
Thomas says envelope tracking still faces problems. "The problem is how to get high total efficiency while keeping noise low. People have shown they can hit the noise target better at lower efficiency. And last, but not least, folks are underestimating how the handset manufacturers will implement this in the factory."
With envelope tracking, the amplifier section will call for a greater degree of calibration at assembly, which could be complex to implement for companies who have become accustomed to building standard consumer products. Chipsets that support envelope tracking will probably have to include a number of self calibration functions to streamline this process.
The increase in filters and amplifiers will see designers paying greater attention to the switches that connect the modules together and direct the signal to the right path between the baseband and the right part of the antenna.
To try to reduce the amount of space taken up by these switches, companies such as Peregrine are trying to integrate as many as possible onto one die. Peregrine recently launched a 16 throw switch as handset developers begin to look at designs that may need to include as many as 32 throws.
Because of the massive increase in component count using existing front end devices, manufacturers cannot realistically, in the short term, build a 10 band LTE device without increasing its size or restricting what can go into the applications processing subsystem – unthinkable, given the focus on the user interface in today's high end handsets. They have to make choices, largely based on what operators want, but there is little consensus.
The result, says Novak is that 'the design of the LTE handset is in a state of flux'. "There is no single network operator with a lot of clout anymore," he claimed. "The days of a network operator saying 'we are going to make this phone with all these bands' are gone. The phone makers try to make a phone that satisfies enough of the network operators, while companies such as Apple and Samsung come back to the operators and say 'this is the phone I have'."
The rf issues raised by LTE and its many bands could see a renewed emphasis on call quality. Sideco says infrastructure exists within operators to determine how well vendors' designs perform once they are out in the field.
"On the operator side, there are a lot of things happening on trying to understand higher performance devices," Sideco says. "How clean the rf reception is has implications on the battery life and, as a result, a happier consumer. And a chief network officer will be happier when the noise floor is kept at a minimum as that's more profitable across the board.
"Amongst the metrics that operators use are International Mobile Equipment Identity specific data on performance. They can see if a group of handsets behaves in a certain way versus a different set. My feeling is they go back to the OEMs and say 'Your handset, while it meets specification, does this to my network. It costs more to run: you can't ask me to subsidise it as much'."
Despite the rf performance, other considerations will continue to influence which As with the days of different 2G standards, designing a phone which can roam on 4G is going to remain difficult for a number of years and the current focus on application features, rather than rf, is not going to help.