February 17, 2010
TI's multicore SoC: Right notes, ringing hollow
By Patrick Mannion

While the new multicore system on chip (SoC) signal-processing architecture

For sure, the requirements of next-generation basestations will push all architectures to their limits and beyond. Balancing lower power and lower cost with increasingly parallel, math-intensive processing to meet multiuser demands for high-data-rate data in 3GPP Long Term Evolution (LTE) Release 8 all-IP networks is not going to be easy, especially with the introduction of MIMO, beam forming, OFDMA and many other enhancements engineered to maximize spectral efficiency.

These issues were all at the front of Kathy Brown's mind when she gave me an overview of the new TI architecture. Kathy is manager of TI's wireless basestation infrastructure division and was refreshingly quick to get to the point.

Beyond the impressive basics, such as four-core and eight-core options to start, 1.2-GHz operation and 256 GMACS and 128 GFLOPS performance (eight cores), the use of a 40-nm TSMC process and the incorporation of coprocessors for Layer 1, 2, and 3 processing functions, there are four main architectural innovations being outlined here, each of which are groundbreaking for TI.

Along with mixing fixed-point and floating point, TI's multicore SoC architecture has impressive intentions, such as advanced memory management and a Multicore Navigator to provide core 'load balancing'. Implementation details will be worth waiting for.

The first is the integration of fixed and floating-point processing within each DSP core. The cores are based on TI's 'C64x+ high-performance fixed-point line. Fixed point comes up short with regard to precision when it comes to solving the matrix inversion math required for the likes of MIMO processing, so the combination of both fixed and floating-point processing, if done right, enhances processing efficiency and simplifies programming.

The second feature is the Multicore Navigator. According to Brown, this performs 'load balancing' similar to Apple's Grand Central Dispatch and is a codification of much if what TI has learned about multicore processing. It coordinates data movement throughout the processor and breaks up a task between peripherals and accelerators--without sucking core cycles.

The third is the Multicore Shared Memory Controller. Using prefetch mechanisms the controller 'anticipates' what each core may need and goes, "a step beyond DMA," said Brown. In effect, shared memory becomes as efficient as dedicated local memory.

Last but not least comes the TeraNet 2 2-terabit/s non-blocking on-chip switch fabric that connects all the elements, allowing them all to run simultaneously and independently.

Other features abound, but these are the heart of the new SoC architecture. The problem is that TI is unable to comment about how each of these functions is implemented and how they work. Also, while L1, L2 and L3 processing is supported, only software for L1 is available from TI. It is relying on the TI ecosystem for the L2 and L3 stacks. Furthermore, no concrete data on availability or estimated pricing is anywhere near forthcoming, beyond a timeline 'sometime in the second half of 2010'.

Granted, it's an 'architecture' announcement and not a 'product' announcement, but the lack of detail on the implementation, beyond a sketchy overview that's more akin to a statement of 'intent' vs. actual design, smacks of a rush to market to meet the Mobile World Congress deadline. This contrasts sharply with Freescale's six-core MSC8155 DSP, also for basestations.

This processor also plays host to an accelerator platform technology, in this case MAPLE-B2L, and uses the Gen 2 Serial RapidIO interconnect for inter-core communication. The chip is based on the SC3850 StarCore DSP, which earned the best ever BDTImark2000 fixed-point benchmark.

That I used the term 'chip' is important. Freescale has qualified the MSC8156 in a 45-nm technology and expects to sample to customers in Q3, in a 783-pin FC-PBGA package. I don't expect to see a chip based on TI's new architecture this year. Do you?

By the time it does materialize, will it then seem so far ahead, given what's sure to emerge between now and early 2011 from the likes of Freescale, as well as core vendors such as CEVA and heterogeneous processing vendors such as Altera and Xilinx, both of which are moving way beyond formal interpretations of FPGA and do already have relationships with basestation vendors. It'll be an interesting year. What do you think?

For more on the new SoC and how to take advantage of it, read:
Realizing Full Multicore Entitlement and Enabling LTE Development with TI's New Multicore SoC Architecture

Comment on this blog entry

November 19, 2009
Magic microphone: All smoke, no mirrors
By Patrick Mannion

To get around the dynamic range issues of current microphone pickups, Schwartz Engineering and Design has devised a laser-based pickup that detects voice-induced 'distortions' in a flowing stream of smoke and that then relies on proprietary digital signal processing to translate those distortions into audio.

To be honest, I was skeptical, as laser-based pickups have been tried before by Bell Labs, but the man behind the system is David Schwartz, who performed seminal work in audio compression in the early '80's that led to what is now MP3, so my interest was piqued.

Skepticism seemed all the more justified when I met David in our reception area holding what looked alarmingly like a really elaborate bong! The walk from reception to the meeting room did not go unnoticed by my colleagues.

Called the Particulate Flow Detection Microphone and shown in its latest rendition in the YouTube video below, the system does actually require a smoke source (regular smoke, that is.) That smoke is then pulled vertically through a tube by two fans at a rate of two to three inches per second. Near the top is a low-cost, off-the shelf laser set up such that the smoke passes between the source and the detector. At that same point, the vertical tube has an aperture through which Schwartz speaks to 'distort' the rising smoke.

As a tech editor, we see lots of demonstrations, this was one of the most bizarre. We also see lots of mishaps in even the most well-planned, pre-staged demonstrations. This was no exception. One of the fans got broken in transit so the rate of smoke flow was reduced accordingly. That was a problem.

Why? Smoke is not elastic, so it needs to be replenished in the beam path otherwise the sound pressure waves leave "trails" or after-effects in the intersection of laser and smoke. Also, reverberations in the detection chamber would affect the signal even more than they already do. The laser needs to "see" a clean, smooth slat--or at least as clean a slate as possible. The cleaner the smoke, the better. Two fans aid in that.

With one fan out of action, it became more difficult to detect Schwartz's speech, this translated to a barely audible replay on the accompanying laptop, but it definitely worked. I videoed the demonstration, but the audio is too low, but that's fine: Schwartz has posted better recordings on YouTube across each stage of development. Here's the latest:

In this version, he has figured out how to make an ultra-low noise-floor version that sets the smoke stream between the sound input aperture and the beam, instead of aiming the laser beam directly through the smoke column.

This way, when there is no sound, the beam crosses the detection chamber without any interference by smoke: it is dead quiet. As soon as any sound pressure affects the smoke, it intrudes into the beam's path, modulating the laser and thus yielding the audio signal.

The new mode does not replace the previous mode, but rather gives mic designers more options, which has the potential of leading to different applications for both modes. Combined modes using two or more lasers are disclosed in the patent.

Though demonstrated at the recent Audio Engineering Society event in New York, it clearly has a long way to go before becoming a practical system, but Schwartz is confident he can get into a more integrated and portable system. If not, I'm sure it can be adapted to 'other' purposes.

For more information, review the patent and an accompanying updated paper.

Comment on this blog entry

September 20, 2009
The path to money in medical systems design
By Patrick Mannion

Between government stimulus packages, an aging population, exploding healthcare costs and increasing need for better remote diagnostics, it's understandable that engineers looking for the next big opportunity would look up and cry: "Thars gold in them thar [medical] hills." But not so fast.

They may be right, but mining this particular mother lode has proven difficult at best, to downright stomach wrenching at worst, and the medical-device landscape is littered with the dried-up carcasses of what at one time seemed like a good idea .

Why? The reasons are many, which is why we've convened a panel of experts in the field to discuss medical Trends and Opportunities at ESC Boston this coming Tuesday, but the general issues are not so much technical as they are systemic: changing FDA rules and approval delays and lack of reimbursement models.

Take remote patient monitoring, for example. Along with implantables and diagnostic imaging, it is one of the three hottest spaces in medical electronics, yet is mired in a race to the bottom because of a lack of insurance reimbursement models. "Our system is based on people going to see a doctor [and getting billed accordingly], not on doctors reading in-coming data," said Steve Dean, medical marketing director at Texas Instruments, one of the medical panel members.

This lack of reimbursement relegates remote monitoring devices to consumer price points, while still having to meet FDA requirements for everything from connectors to firmware. Not a strong business model.

Speaking of the FDA: it has recruited 1,000 more approvers in the past year, yet the process is still bogged down in red tape as those approvers have ditched the principle of predicate technology and are looking at designs from scratch. This doesn't help anyone looking to develop implantable devices, which already has high barriers to entry, according to Mir Imran, serial medical entrepreneur, chairman and CEO of InCube Laboratories and ESC medical panelist. "Only a handful of people have the stomach for it [the approvals process]."

So yes, mining medical is indeed difficult, but the opportunities exist, given the strong drivers at play. Imaging, remote diagnostics, ultrasound, molecular sensing all have potential. To identify and explore those opportunities, Dean and Imran will be joined on the panel at ESC by Newton De Faria, business development manager, National Instruments, Charles Sodini, professor of electrical engineering at MIT, Al Wegener, CTO and founder of Samplify Systems and Steve Ohr, research director at Gartner Dataquest.

Together, the group will look at everything from where the technology and device gaps lie and how designers can go about targeting those market opportunities, to specifics such as low-power design to extend product life spans and firmware partitioning and other design techniques to secure FDA approval. More on the outcome of this discussion and where you should be putting your medical design energy in my follow-up posting.

In the meantime, join the pre-panel discussion and pose your own questions on EETimes at Where's the money in medical design? Join the conversation. See you on Tuesday!

Update: The full panel is now online, view it here.

Comment on this blog entry

September 11, 2009
Chip vendors making VCs redundant
By Patrick Mannion

The next time you have a great idea for an end product, you should look at bit more closely at your choice of IC: the vendor-as-partner model can do more to realize your idea than any venture capitalist or angel investor ever could--without eating into your long-term profits.

I've discussed the Irrelevance of Silicon and in a follow-up blog (Engineering as Art) I highlighted some great feedback on where we stand on the silicon vs. software debate (I recommend you read the comments, all very thoughtful). More recently I've bemoaned the lack of innovation in DSP architectures.

Does anyone recall the crazy days leading up to the telecom crash when new processing architectures were popping up like daffodils in Spring? I believe BOPS Inc. was one of the last of that batch to disappear, while picoChip has managed to keep the dream alive. Maybe it was that Comet Award we (EETimes) gave it way back in 2003? Let's ignore many of the other winners in the communications field that year. Out of respect for the dead.

My point is that now we're, for the most part, 'stuck' with the 'usual suspects' when it comes to DSP architectures. (As editors, we're always on the lookout for 'the next big thing'). TI and ADI still rule the roost in catalog ICs, though Ceva and Tensilica are making great strides in the IP space. Tensilica most recently with its ConnX D2 16-bit dual MAC engine.

I mentioned this to ADI a while back and they rightly took me to task on this, pointing to innovations on the Blackfin lineup in particular. But again, those didn't constitute real 'architectural' changes. It turns out, however, that the company has done a lot more to benefit its customers through partnership and ecosystem innovation, than could ever be achieved through architectural innovation. Though it took a day at ADI's facility to hammer this home.

Yes, it recently launched its Engineer Zone to help designers and customers work with each other as a community to get questions answered and its developers connected. That's extremely useful and apparently has taken off exponentially (according to ADI). But that's long overdue (just ask TI, with its very popular e2e community).

Fortunately, that's just a starting point. ADI has raised the bar on customer support in other ways, to the point that if your idea matches its own targeted application areas, you can bring your idea to its door and the ADI in[-house designers will not only 'suggest' some ICs that can meet your needs, but will design the whole system and even open up its sales channels to you, where suitable.

Now, there's a continuum here. Some startups or established companies may not want that level of support and the openness that it requires. Also, the concept is clearly not new and good IC vendors have always been good customer advocates, but in an era when ICs are no longer being differentiated by horsepower and silicon alone isn't enough, vendors like ADI are taking this customer support to a whole new level, from silicon to software to sales.

So, if you have an idea, we've discussed ad nauseum of late how venture capitalists aren't likely to fund your concept, you want to get to market quickly at reasonable cost and you have enough software expertise to add value at the top end, this vendor-as-partner model may be your best bet. It's not just about the IC anymore.

In targeted areas such as medical, automotive, industrial, instrumentation, VoIP and others, companies such as ADI have a vested interest in getting your design off the ground faster than you can say "I just burned through my Series C funding and am out of options." They want to be seen as innovators in those spaces. Take advantage of that!