TI launched a new range of chips last month targeted primarily at the larger enterprise and metro-femto femtocells, both 3G and LTE. What capabilities do they offer and how are these positioned alongside the more well known femtocell silicon providers? I asked TI for the lowdown.
Texas Instruments have been involved in mobile basestation products since their inception, providing the high performance baseband processing power that drives many of today’s outdoor macrocells. However they haven’t been associated with femtocells that much, perhaps because their silicon solutions are traditionally designed to accommodate the high capacity and more complex processing demands of larger cellsites. Their design parameters meant they were seen as being costly, power hungry and inappropriate for a femtocell application.
They haven’t been competing directly with the new femtocell chipset vendors such as Picochip and Percello (now part of Broadcom), who have designed very low cost and compact System-on-a-Chip designs that are ideal for the residential market.
New product launched in June 2011
In June, TI launched a new range of chips targeted at the high performance end of the femtocell market – suitable for the larger enterprise and metro-femto product designs including LTE. While I still don’t see TI being able to compete for the low end residential and SOHO femtocells, the raw processing capacity of the chipset will give others targeting high capacity femtocells a good run for their money. Already, Ubiquisys have publicly selected it for their larger metro-femto range (while retaining Broadcom for their smaller residential products) and plan to launch products based on it in the first half of 2012.
What’s under the hood
The 6612 and 6614 chips announced at Femtocell World Summit in June are part of the “KeyStone” product range and targeted at enterprise/pico/metro small cells. A single chip can fill both a 20MHz LTE and a 5 MHz WCDMA carrier simultaneously, supporting a dual-mode 3G/LTE femtocell. The various processor intensive functions required to handle mobility (i.e. faster moving users), interference (e.g. poor radio conditions) with headroom to add further features are all included onboard.
Another significant advance has been adding a 1.2GHz floating point core into the chip – this function was previously fixed point, together with a matrix processing algorithm that provides some 5x improvement.
TI also pairs this with their own RF front end chips so that their reference design is based on all their own silicon. (This is a similar approach to Qualcomm, while other femtocell chipsets from Picochip and Broadcom typically offer a choice of 3rd party RF front ends).
I’d compare TI’s optimisation approach to that seen by other silicon product companies, who initially embed the most demanding processing functions directly in silicon while allowing further non-standard features to run alongside on standard DSP as firmware. As new features become more popular (and standardised), later editions of the silicon can incorporate these functions, freeing up more processing power for even newer innovations.
Examples of non-standard functions are some of the MIMO processing, which could be 2x2, 4x4 etc., and some of the “secret sauce” that equipment vendors might add (such as proprietary radio optimisation algorithms).
The 6612 can support up to about 64 active users/sessions and could support either WCDMA or LTE (but not both simultaneously). The 6614, being quad core DSP, can handle some 128 users at peak rates of 150Mbit LTE and 42 Mbit WCDMA simultaneously.
Announcements don’t always mean you can buy the product today of course. The chips are planned to be sampling in 3rd quarter of 2011. A development starter kit is available which includes separate Integra scalar CPU, DSP and RF front end. Femtocell products based on the chipset should appear in the first half of 2012.
Has the market evolved?
TI tell me they see a change in market requirements for femtocells in the past 18 months. For them, it’s no longer about stripped down residential femto. They’ve seen enterprise femtocells introduced in 2010 and the term “small cells” become popular this year.
I see this as an expansion and evolution of the femtocell market. The residential and SOHO products are well served by the likes of Picochip and Broadcom, who have little serious competition for the functionality and price point they serve.
The expansion into enterprise femtocells with somewhat higher capacity has already been addressed by those same vendors, with larger/scaled up versions of their low cost System-on-a-Chip designs.
The battlefield opens up
As we see the demand growing for even higher capacity 64 or 128 user femtocells for public outdoor/high traffic areas, combined with the emerging LTE femtocell market, we are also seeing the larger silicon vendors such as Texas Instruments and Freescale enter the fray. Approaching this with a scaled down, re-engineered and cost optimised version of their macrocell products, it introduces more competition to the market.
Whether the price/performance/capability offered by the top down rather than bottom up approach will win out remains to be seen, but I do believe that once again the heavy investments being made by silicon vendors such as Texas Instruments demonstrates the strong belief in a long term market for small cells (and the silicon inside them).
You can read more details in TI's product bulletin for these chips.