Reconfigurable algorithm processing adds support to SDR commercialization

So, why has SDR failed up until now? The simple answer is one of initial cost versus end reward: SDR needs massive computing performance — not just billions of operations per second but trillions. Imagine a digital, programmable filter that can operate at IF frequencies. Sampling at 100MHz, a 64-tap filter wo uld require 6.4B MAC operations per second. Today, no DSP can do that although several companies have announced roadmaps. And that's just a simple filter.

If the problem is extrapolated to 3G, it becomes doubly difficult. 3G devices have two of these, one on I, one on Q. So far, it's been too hard to be cheap. Governments can use it simply because cost is not an issue. But for the average consumer, without a multi-billion dollar budget, the cost has been high. And the returns have been limited — how many radio standards do you see in a day? GSM and FM radio for most of us, nothing more. FM isn't even digital, so why bother using SDR?

But the world is changing. The consumer is now looking at more radio systems as technology becomes more complex and more easily affordable to the man in the street. For example a PDA might want a WLAN connection. What standard; 802.11a or 802.11b? And what about Bluetooth, using the same frequencies as 802.11b but completely different proces sing? GSM or GPRS would be nice and we might as well have that FM capability too. U.S. CDMA standards available for roaming internationally and using one global product. And, we won't mention 3G. There is a consumer class device that will need many different wireless standards.

Today each of these is a chip — which is a disaster as products begin to be combined more often. Each chip takes board space and power, not to mention cost. Suppose there was a way of doing this that did not need a new chip for each standard: a programmable chip that could adapt to new standards as they become available, or fix the standardization hiccups, Bluetooth being a now-famous example of this.

That sounds like the dream of SDR. The rewards are there — but so too are cost/power constraints. The SDR approach has to deliver Bluetooth, WLAN and cost less than the separate chips, in power/dollars/size. You can't do that with a DSP processor yet.

So, if we can deliver SDR at a now-identi fied cost target the consumer will reap huge benefits. It will be a massively successful product. The device could cost a bit more than one of the components, but probably not much more.

That's just one example. Suppose today's set top box market adopted SDR. A single device could reconfigure between digital terrestrial TV, analog TV, satellite TV and so forth. The manufacturer can truly build one product that does anything. A cable modem uses all the same technology as an SDR system. So too does ADSL. A single box can access every broadband standard.

The only problem is that the SDR for WLAN or GSM would need a huge DSP engine. The big DSP players today (TigerSHARC, C64x, Starcore) are all having serious end-cost issues with SDR. It's not that they can't do it, there's just no way they can deliver what's needed for the price at the moment.

Software transition
The problem is that SDR is not something that works on processors. Hardware performance, with t he ability to redefine the hardware under software control, is essential. It also needs to be done fast, allowing easy switching between radio standards. One technology that may be applicable here is D-Fabrix. It is based on a reconfigurable algorithm processor (RAP) that can be configured for virtual hardware — hardware defined and set up using software, but once created, operating at the speed of a hardware implementation. The technology's fine grained architecture means that resource can be matched to the problem; with 4-bit ALU's that can be programmed in VHDL or Verilog just like silicon for hardware performance. In the near future, C is expected to be added to the programming languages supported, giving even more versatility. The ability to reconfigure a D-Fabrix array in 10's of microseconds from WLAN to Bluetooth, for example, is quicker than the time needed for the user to decide which one he/she wants.

Basically, RAP technology allows an SDR to be reconfigured between multiple radio st andards. A good example is a PDA with wireless capability, such a device currently requires different baseband chips for the different wireless standards (Bluetooth & 802.11b). Employing RSP technology allows OEMs to manufacture the device using a single chip that will support both standards.This capability extends to other communications-based devices that can be built using SDR such as cable TV boxes. If this device could also operate as a satellite decoder or digital terrestrial and handle all the different cable and digital TV standards, the set top box players would be able to offer one device to both the European and US markets.

Flexibility is the key feature in the RAP platform . If we look at the example of a PDA; when it's not in use as SDR, it could do MPEG decoding for fast video calls, or JPEG for picture viewing, or graphics acceleration. This provides true functionality and the ability to cram a number of different consumer products into the same box, without losing performance or worrying about price or power consumption.

How about adding an MPEG encoder to the set-top box? Add a hard disk and turn it into a recorder too. The possibilities are virtually endless, and if nothing else, the ability for the consumer to save space is huge. No more need to have a satellite box, a tuner, a video recorder and an ADSL router all separately.

Put it all together in one box, perhaps costing a little more than one would expect for a single item normally, but offering unparalleled performance, coupled with low power, savings on space and the ability for it to automatically update the standards as and when needed. It would be connected to a high-speed Internet connection after all. It's safe to say that driven by the emerging consumer need for SDR, many companies are expecting huge demand for their respective technologies.