Chip design – easier than semiconductor IP?

By Luke Collins |  1 Comment  |  Posted: June 28, 2012
Topics/Categories: no topics assigned  |  Tags: No tags for this post.

Software-defined modem company Cognovo has been acquired by u-blox, a Swiss fabless semiconductor company that specialises in positioning chips and modules, for $16.5m in a cash deal. The move once again calls into question the viability of the semiconductor IP licensing model for all but the most dominant players.

Cognovo was formed when ARM Holdings decided that it couldn’t give enough attention to a vector signal-processing architecture that it acquired when it bought Belgian design house Frontier Design.

ARM had commissioned researchers at the University of Michigan to define the ideal instruction set for a processor handling wireless protocols, and then fed that definition into a tool that Frontier had developed that could synthesise an architecture from an instruction-set definition. The result was the Ardbeg processor, optimised to run wireless protocols up to and including LTE-Advanced.

Rather than let this work lie fallow, ARM turned to a group of experienced wireless design people, who had cofounded Cambridge design consultancy TTPCom and later sold it to Motorola, to commercialise it. The result was Cognovo, backed by ARM and its founders, to license the architecture to companies that wanted to take control of their destinies in the cellular handset market by designing their own chips.

Cognovo made the Ardbeg processor the core of a strategy to license software-defined modem technology, on the basis that this approach would accelerate the chip design and software-integration process as well as allowing later design changes, both attractive options in the tumultuous wireless market. Underpinning this approach was an intriguing design flow, detailed here, based on defining a chip design in terms of a set of decoupled constraints.

According to Charles Sturman, vice president of sales and marketing at Cognovo, by the time of Mobile World Congress in Barcelona in February, the company had developed its design flow and implemented a test chip with Samsung, and was able to demonstrate the chip running LTE-Advanced protocols within a sensible power envelope. Tier One handset vendors were apparently showing interest, and the next step would have been to add an RF front end to the demo board and the protocol-stack software and so demonstrate a working LTE terminal.

“It is incredibly hard to convince the big players,” said Sturman. “You have to do much more than your basic product and it just gets harder and harder.” For Tier 2 handset vendors, Sturman predicted that Cognovo would have had to develop pretty much an entire handset, including the modem stack.

“It would have needed another year or two to get to that point and ARM and the founders could see where that was going and that was fine, but the other option was to be acquired, which would be OK in terms of the security of the staff and the vision of where we wanted to go.”

So Cognovo, using a processor that was developed in part by IP licensing powerhouse ARM, and staffed by highly experienced technology licensors from TTPCom, is now being absorbed into a fabless chip company focused on selling chips. The staff are expected to keep their jobs, and u-blox has already said it will start a wireless chip design activity to incorporate the Cognovo technology into future products.

What happens to the licensing activity?

According to Sturman “There’ll be a huge amount of work on hand to develop end products, so we’re not expecting to licence the IP in future.”

What happens to any existing licensees?

“There are some relationships that we are going to have to manage.”

What happens to the semiconductor IP licensing model?

“As far as the IP licensing model goes, companies like ARM are more of an exception.”

Will anyone make any money from the Cognovo acquisition? If you reckon that the company launched in 2009 with around 20 staff and had around 30 as of today, then it seems unlikely that anyone will have made much from this latest foray into semiconductor IP licensing – although a ‘liquidity event’ is apparently planned at a local pub.

As Sturman put it: “No-one’s lost anyone’s shirt but we haven’t taken [Cognovo] to the level of another TTPCom. The staff and founders feel like we have taken technology out of ARM and built a platform of proven technology, but in terms of turning it into taking cash out of the market, that’s what we’ll do within u-blox.”

One Response to Chip design – easier than semiconductor IP?

  1. Pingback: Decoupled constraint modelling – a design methodology for hard real-time systems on chip | EDA

Add Comment Register



Leave a Comment

PLATINUM SPONSORS

Mentor Graphics GLOBALFOUNDRIES Synopsys Samsung Semiconductor Cadence Design Systems
View All Sponsors