News & Analysis
4G/LTE chips coming . . . but first, a little chaos
Junko Yoshida
3/22/2010 9:04 AM EDT
As was the case with last month's World Mobile Congress in Barcelona, expect to find LTE everywhere at the CTIA. But don't hold your breath for commercial LTE chip solutions for mobile handsets -- anytime soon.
Forward Concept's Strauss said, "Dozens of companies claim to have LTE chips, but upon closer inspection, you find that they are 'almost' solutions."
Strauss said he does not expect LTE chips for handsets before late 2012. "That's because the first rollouts are for data only, meaning for USB data dongles and PC cards," he said.
Almost all of LTE solutions are FPGA implementations of their upcoming LTE-only product (except for Icera Semiconductor), targeted only for dongles and modules.
The Linley Group's Linley Gwennap agreed. After the Mobile World Congress last month, he said, "Not counting companies like Samsung, LG, and NTT DoCoMo that have demonstrated LTE technology but are not planning to sell it commercially, the vendors that are sampling working LTE silicon right now are Qualcomm, ST-Ericsson, and Altair," Gwennap observed. "These initial solutions are single-chip LTE basebands that require an external RF chip. In most cases, additional chips are needed for compatibility with 2G/3G networks."
Open core
From a licensable IP standpoint, Strauss said that there are LTE terminal solutions (in FPGA form), which have been demonstrated by CEVA and Tensilica (both using mimoOn PHY software), Blue Wonder Communications (using 4M Wireless' LTE PHY), and Cognovo (an ARM spinout staffed by ex-TTPCom 3G stack experts).
Meanwhile, Strauss singled out Icera Semiconductor as the only company that has publicly demonstrated both LTE and 3G on their current product. "They simply re-flashed one of its OEM customer's HSPA/5MHz dongles to run the LTE stack," he said. It's "a true illustration of a software-defined modem," as it is to deliver multimode LTE technology in software at no additional silicon cost.
CEVA, too, is going for a similar SDR option. The difference is, said Wertheizer, "Icera's solution is not based on an open core." As CEVA being a DSP core IP company, he added, "We are the only choice" allowing semiconductor companies to go for SDR by using an open core and commercially available tools.
The truth is that all 3G basebands are programmable . . . albeit only by vendors themselves, said Strauss. "The key to CEVA's future success is to continue expanding their ecosystem of development tools and third-party support, mostly software," he added.
CEVA claims that CEVA-X cores are picking up interest among more chip vendors, as WiMax houses are heading to LTE. Sequans, the number two WiMax terminal chip vendor, last year announced that it is going with CEVA-X cores for LTE. Strauss said that Beceem, the number one in WiMAX terminal chips last year, told him that they were going with CEVA-X cores for their upcoming LTE product.


rick.merritt
3/22/2010 10:36 AM EDT
IMHO, CEVA has a good opportunity selling programmable chips in this transition when no what quite knows what a 4G handset should be, but by 2011-2012 when the system design gets more clearly defined dedicated chips will emerge and win the day as most cost effective.
Sign in to Reply
junko.yoshida
3/22/2010 11:51 AM EDT
That may be so; but the point here is that there are already so many permutations of 2G/3G implementation -- depending on cellular network providers and depending on different parts of the world. LTE being the first truly global standard, SDR may be just the way to go, not just as an interim solution, but as a long-term solution. At least, that was the sense I got talking to CEVA's CEO.
Sign in to Reply
Polarisgt
3/22/2010 1:34 PM EDT
True that SDR will reduce R&D cost, time to market, and tapeout risk. If you want a fast product and functional unit, SDR may be the way to go. But for very few companies who know how to do it right, hardwired solutions will yield astonishing low-cost low-power solutions with surprising high performance which is not possible given the current power of DSPs. The market will be lose or get-it-all.
Sign in to Reply
DaveKelf
3/22/2010 5:47 PM EDT
It is clear that these new vector processors have greatly narrowed the gap between custom hardware and programmable devices in terms of power consumption. The real question is the efficiency of device programming, as an effective software implementation can make an order of magnitude difference in performance. Leading companies are finding that this is the significant issue to solve, and when they do, custom hardware basebands may well become obsolete.
Sign in to Reply
tinker_dude
3/22/2010 8:09 PM EDT
Well, Rick Merritt got it right, its only a matter of time. Until that time CEVA and others can make their noise.
And to DaveKelf:
"The real question is the efficiency of device programming, as an effective software implementation can make an order of magnitude difference in performance."
How will make c-like code to compete with AND/OR/XOR gates? Well as you said that's the real question and I personally don't think today's solutions have an answer!
Sign in to Reply
Sudarshan1970
3/30/2010 1:46 AM EDT
Hi,
SDR looks interesting.
I have been working on SDR for quite sometime now. I would like to work on 4G LTE,anyone interested for this ?
contact me: sudar_ns@yahoo.com
sudarshan@spacomp.com
Sign in to Reply
xzczvxc
6/20/2010 11:08 PM EDT
what about Sequans, isn't their wimax chip in the HTC 4G phone? Also, these chip suppliers will be in CMCC TD-LTE dongles: 1. Samsung 2. Innofidei 3. Sequans 4. ST-E (T3G)
Sign in to Reply
Timothy.OBrien
11/22/2010 12:32 PM EST
"Strauss said he does not expect LTE chips for handsets before late 2012."
What about the Samsung Craft which I read somewhere uses a Qualcom chipset?
Sign in to Reply