Of course this wouldn't be particularly strange if we were talking about a simple copycat product or technique. What makes it unusual is when we're talking about concepts that may have taken numerous person-years to develop and refine.

One classic example of this phenomenon is that of the telephone. Alexander Graham Bell first conceived the idea of this device in 1874 - the same year that he met Thomas Watson, a young electrician who was to become Bell's long-time assistant. After much experimentation, Bell filed a patent application for the telephone on February 14, 1876.

What is less well known is that just a few hours after Bell had filed his patent, another inventor called Elisha Gray attempted to file a sort of pre-patent known as a "caveat" for a similar device. You can only imagine how Gray felt about this when he discovered the dreadful truth: "If only I'd got up a few hours earlier that morning," we can envision him lamenting to himself.

picoChip's parallel processors
Fast forward to 2002. A few columns ago I wrote about QuickSilver Technology and the architecture of their new ACM devices. At the time, I mentioned that a company called picoChip,based in England, appeared to be doing something similar. Unfortunately picoChip was in "secret squirrel" (stealth) mode at the time, so it was somewhat difficult to determine exactly what they were doing based on the meager crumbs of information they scattered throughout their website.

However, on December 2, 2002, picoChip flung the veils asunder and revealed their solution. This is mega-super-cunning, let me tell you, and shows that - as usual - there's more than one way to skin a cat (this is just an expression and not one I invented, so please don't take the time to write in and complain about it).

Like QuickSilver, picoChip commenced by evaluating various word-orientated and bit-orientated algorithms to determine the various types of elements that are required to implement algorithms like the wireless WCDMA standard. And like QuickSilver, picoChip also came up with a heterogeneous node-based structure, but in this case their collection of nodes involves different flavors of RISC processors.

These 16-bit devices are optimized in a variety of different ways; for example, one processor type may have lots of memory, while another will support special algorithmic instructions that can perform operations like "spread" and "de-spread" from the CDMA wireless standard using a single clock cycle (as opposed to 40 cycles using a general-purpose processor).

Each of these processor nodes is approximately equivalent (in processing capability, not in architecture) to an ARM9 for control-style applications or a TI C54xx for DSP-style applications. The trick is that each picoArray (which is what these devices are called) can contain hundreds of such nodes. The result is a truly ferocious amount of processing power.

For example, one of the absolute top-of-the-line DSPs in the world today is the TMS320C6415 from Texas Instruments. This bad boy can perform such a humongous number of calculations at such a breathtaking speed that it makes your eyes water. However, a single picoArray running at only 160 MHz can deliver almost 20x more processing power (measured in 16-bit ALU MOPS) than a TMS320C6415 running at 600MHz. Wow!

Going wireless
Although the end results of 3G (third generation) wireless are taking a little longer to hit the streets than were originally expected, this doesn't mean that 3G is dead in the water. In fact, by the end of this year, somewhere between $5 and $6 billion dollars will have been spent on 3G infrastructure in 2002 alone!

A large portion of these funds go to developing the digital baseband processing portions of wireless base stations. A typical base station may be required to handle 64 simultaneous channels, but this can rise to hundreds of channels in highly populated areas. Today, operators ideally require the cost-per-channel to be in the region of $150, but it's actually hovering around $200 (which is 1.3x higher than they want). And by 2005, operators need to get the cost per channel down to only $10. However, based on predicted advances in current technology, they are currently forecasting $50 per channel (which is 5x higher than they want).

What is required is some sort of technology step function, and it looks like the picoArray may fit the bill. A single picoArray can replace a number of today's extreme-end FPGAs along with 20 or more high-end DSPs, thereby dramatically reducing the cost per base station channel.

Single design environment
One great point about the picoArray solution is that it largely brings everything down to a single design environment (see figure 1).

Figure 1 - picoArray design environment

With today's conventional solutions (Figure 1a), ASICs are inflexible and expensive to develop, and they have with long design cycles and high fixed costs. Even worse, wireless standards are evolving so quickly that, by the time an ASIC design has actually been implemented, it's often obsolete. In the case of DSP and FPGA-based alternatives, these do offer flexibility, but wireless solutions based on these devices require a lot of components and the cost and power consumption of the resulting implementations is often unacceptable.

Of equal importance is that conventional solutions require at least three design environments - ASIC/FPGA, DSP, and RISC - which complicates development and test, slows time to market, and generally increases the risk of not getting the entire design to work in a timely manner (that is, before the need for this particular implementation has faded away into the mists of time).

By comparison, a picoArray-based implementation largely brings everything into a single design environment (Figure 1b). Also, the fact that every processor node on the picoArray is fully programmable means that each channel can be easily reconfigured to adapt to hourly changes in usage profiles, weekly enhancements and bug fixes, and monthly evolutions in wireless protocols. Thus, a base station channel based on picoArray technology will have a much longer life in the field, thereby reducing operating costs.

There's way more to discover about picoChip's new technology than I can cover here. For example, the fact that they provide a complete library of picoArray programming/configuration modules that can be hooked together to implement a fully-functioning base station (or users can "tweak" individual modules to implement their own algorithm variations, thereby gaining a competitive advantage).

Also of interest is the fact the picoChip's design environment uses a VHDL framework to describe parallel processing and connect modules together at the block level, but that the actual "internals" of each block are implemented in pure C or assembly language.

If you want to know more about picoChip and their picoArray technology, please feel free to roam wild and free around their website. For my part, I think this looks like another incredibly interesting technology that deserves an official "Cool Beans" from me. Until next time, have a good one.

Clive (Max) Maxfield is president of Techbites Interactive, a marketing consultancy firm specializing in high-tech. Author of Bebop to the Boolean Boogie (An Unconventional Guide to Electronics) and co-author of EDA: Where Electronics Begins, Max was once referred to as a "semiconductor design expert" by someone famous who wasn't prompted, coerced, or remunerated in any way.