Alan Gatherer is the Chief Technical Officer for the Wireless Infrastructure Business at Texas Instruments, based in Dallas, Texas, USA, and is responsible for all strategic development of TI’s digital baseband modems for 3G wireless infrastructure.

He joined TI in 1993 and has worked on various digital modem technologies including cable modem, ADSL and 3G handset and basestation modems. He was elected to Senior Member of Technical Staff in 1996 and then to Distinguished Member of Technical Staff in 2000. Alan also served as the manager of the wireless cellular communications group in research and development.

He has authored seven journal papers and 20 conference papers and holds 33 awarded patents and is author of The Application of Programmable DSPs in Mobile Communications. He holds a bachelor of engineering in the area of microprocessor engineering from Strathclyde University in Scotland. He also attended Stanford University in California where he received a master’s in electrical engineering in 1989 and his doctorate in electrical engineering in 1993.

Nick Flaherty: Texas Instruments is a major supplier of semiconductors to the wireless infrastructure business, so you are in a good place to see how realistic the next generation Long Term Evolution (LTE) wireless technology is. It is very much the buzz word at the moment but how real is it?

Alan Gatherer: We provide key components to these systems, and typical infrastructure takes 15 to 18 months to develop products so we need to be two to three years ahead.

We think towards the end of 2008 there will be field trials and then after that systems will start rolling out. It needs to be standards compliant so its just these field trials at the moment.

Our feeling is as a standard such as LTE comes along there’s a rush to field trails then a period of calm when people work out when it has to be rolled out.

There are two distinct phases, with the technical trial and the business decision of whether you can make money out of the technology and the service, so the technical is decoupled from the business decision and the success of HSPA (with downloads to 3G mobile phones from 14.4Mbit/s to 50Mbit/s) could delay LTE but by how much we don’t know.

DoCoMo in Japan for example is looking at services (for LTE) in 2010.

In many ways the LTE physical layer borrows a lot from WiMax and HSPA so it will evolve from the systems that are rolling out today.

NF: Multicore processors have been highlighted as the way forward, particularly for the higher performance requirements of LTE. When will we see, say 10, digital signal processor (DSP) cores on a chip from TI?

AG: I think the whole multicore thing is somewhat of a marketing play among many people.

We have been doing multicore for many years as our customers put multiple chips on a board, and we can put a lot of DSPs down on a board.

We find a lot of use for general purpose single cores. But we also have devices such as the 6488 with the MAC and PHY and the 6487 for GSM and EDGE. These are multi-core devices.

NF: Will dedicated multi-core processors overtake DSPs for those next generation LTE systems that will have to offer bandwidth of up to 100Mbit/s to an individual phone?

AG: Customers are very interested at this point in reuse of W-CDMA systems and we continue to do well in the more deployed systems.

If you look at the past we started with generic devices then applied an accelerator for Viterbi, for the things that mattered most.

So we have devices such as the 6488 and 6487 for GSM and EDGE. It takes time to evolve to an optimum solution.

For example, an operator such as

T-Mobil’s has the goal to reuse the hardware from W-CDMA for LTE.

As we move forwards, we are looking towards chips that drive not only LTE enhancements of the MIPS for symbol rate and bandwidth related signal processing

On W-CDMA it was the bandwidth that was the struggle to achieve. With LTE there’s an increase in bandwidth and therefore an increase in symbol but when you look at the true signal processing we require it’s linked to the increase in bandwidth – as an operator I need 5MHz, 10MHz is optimal but it has to be capable of supporting 20MHz.

But MIMO is also coming in to give more bandwidth and that is increasing the complexity. But no one knows how MIMO will be deployed and how systems will use it.

It’s the cost of the antennas that gates MIMO. If you become too dependent on MIMO just to get your functionality you are dead in the water.

In the US you rent pole space by the antenna and so with multiple antennas, (16in a 4 x4 MIMO array, each) with its own power amplifier, that becomes expensive.

As we go forwards we will be looking to provide cost optimisation for W-CMDA as well as performance for LTE.

NF: Picocells and femtocells, providing 3G basestations in the home and office, are getting popular, but there ha been no sign of TI in this market which is potentially very high volume. What is happening here – is Texas Instruments behind companies such as picoChip?

AG: Femtocells are still undergoing standardisation and they are specifically looking at the effects of interference and that’s a non understood issue for femtocells.

The (coming) field trials will throw some light on this so there’s a lot of work that still needs to be done and it would be a mistake to put a lot of effort into claiming we had a solution. It’s not time to be claiming victory.

Where we are today in femtocells if we talk about the cost of producing thousands of femtocells we have the difference between the NRE (Non Recurring Engineering cost) and the bill of material (BoM).

The problem with femtocells at the moment is that it could be a mass market but its not today.

Reuse is tremendously important in a market that has no volume yet so our general philosophy is to get as much reuse out of the basestations that exist today.

We already have devices that can do femtocells. Maybe they don’t hit the BoM but they have a much lower NRE and that’s vitally important.

Software is the dominant cost for our customers and the key issue in the time to market. It’s the thing that they do well and get reuse from.

Reuse from the macro cell to the micro cell to the pico cell has been tremendously important so they don’t have to start from scratch for the femto cell.

If you are late you lose a lot of money, and this applies not just to femtocells or LTE but to W-CDMA and to HSPA, and this will continue to happen to silicon companies as well as equipment companies. It’s better to put a little more money up front and perhaps even spend a little more on unit cost to hit the market window. A lower NRE beats a slightly higher unit price and the shared NRE beats a slightly higher unit price.

We build these devices not to destroy our business but to have a path to optimisation, so we want to build the right thing at the right time on the right process – our business is always cutting edge. But my finance director will not let me build these parts until there is a clear market.

I think the business model also needs to be tested in field trials. Femtocells at the end of the day is an unknown market, and we can provide the BoM optimisation, or the macro basestation reuse approach. TI is the only one to provide both models. But don’t over optimise for femtocells as you will lose on the NRE.

NF: How do you see the future for TI in wireless infrastructure?

AG: Moving to OFDM for LTE is more back to our roots, more so than W-CDMA. There, chip rate processing was a challenge and we did have to come up with some accelerators.

With OFDM it’s the complexity on the number of different ways you need to run the modem that’s the issue. It's quite a dynamic system so we need to design for more than the worst case.

I think the DSPs will have more impact on MIMO arrays – this is good stuff, we like it!