Adaptive silicon comes to RF

Adaptive silicon comes to RF
By Clive Maxfield, EEdesign
November 1, 2002 (6:28 p.m. EST)
URL: http://www.eetimes.com/story/OEG20021101S0069

If you cast your mind way back into the mists of time - well, about 4 weeks (two columns) ago as I pen these words - you may recall my ponderings entitled "Reconfiguring Chip Design." In this column, I unveiled some newly released details on the cunning architecture of a new class of digital integrated circuit called an Adaptive Computing Machine from One point that I neglected to mention is that the ACM is not currently being positioned as an individually packaged device, but rather as a core function on a System-on-Chip (SoC). The r eason I mention this now is that it becomes interesting in the context of today's topic.

But what about the RF?
I'm always delighted to receive feedback on my miscellaneous musings - not the least that experience has shown that such feedback can often prompt me to learn something and form the basis for a new column. In this case, a reader sent the following query:

"Hi there,

It seems like there are several ways to optimize the Reconfigurable Digital Baseband Solutions. But, has anyone worked on the RF-Frontend and the Interface between the Digital Baseband and the RF issues? These two systems are the most power consuming and complex. In mobile-chip design, the above two factors will play more important role than the Digital Baseband. I think Intel and other Fab companies are going the better way where they are trying to implement Analog and Digital functions on the same silicon which will provide power efficiency and cost."

And indeed this is correct. Thus far, technology considerations have typically required the analog and RF portions of the design to use relatively exotic materials (compared to the standard processes used to create the digital CMOS portions of the design). In addition to being expensive, these separately packaged devices increase the size and power consumption of the system. In particular, the chip-to-chip interfacing between the analog/RF and digital devices can be a pain in the nether regions.

Impinj's self-adaptive silicon
And so we come to a relatively new company called Impinj's founders are the world-famous Dr. Carver Mead and Dr. Chris Diorio. This immediately starts flags waving in the old gray brain cells saying "there's something interesting to be found here," and indeed there is. What Impinj has come up with is a floating gate device called a pFET (s ee figure).

Figure 1 -- The basic structure of Impinj's pFET device

The idea here is that you can modify the device's characteristics with extreme precision by adding electrons to the floating gate using impact-ionized hot electron injection (the transistor on the left), or by removing them using Fowler-Nordheim electron tunneling (the transistor on the right).

Of particular interest is the fact that hot electron injection is a relatively new phenomenon in the case of PMOS transistors (most folks do it with NMOS devices). However, it turns out that there are significant advantages to using PMOS transistors, not the least that they require only a minute amount of current -- in the order of nanoamps - for the hot electron injection to take place. By comparison, achieving hot electron injection in NMOS devices requires the use of charge pumps to supply the necessary current, which can be hundreds of micro amps or even milliamps.

In addition to offering benefits in terms of reliability, the low currents required by the PMOS devices mean that you can be injecting electrons while the circuit is in operation, which allows these devices to support active tuning as discussed below.

This really is a cool beans technology
When you create an analog circuit, it's often critical to have pairs of matched transistors operating together. Any mismatches arising from the manufacturing process can be disastrous in terms of signal quality and power consumption. This is of course why analog components are often created on exotic materials in the first place - to allow for much finer control of the transistor's characteristics. The point is that, using Impinj's pFET devices, you can modify the characteristics of the transistors after they've been manufactured, which means you can fine-tune the circuits so as to achieve optimal results.

One key point that is very important to understand that these pF ETs can be implemented using any standard (low cost) CMOS process, and are compatible with the digital CMOS technologies offered by TSMC, UMC, or Chartered. In turn, this facilitates combining the analog/RF and digital baseband functions on the same SoC.

It's also important not fall into the same trap I did. When I first saw material on Impinj's technology, I mistakenly thought that the entire analog/RF portion of the design was to be created out of pFETs. In fact this is very far from the case. For example, Impinj's first "proof-of-concept" product is a 14-bit DAC, which contains several thousand transistors, of which only 31 are pFETs.

The point is that, even using only this small quantity of pFETs, the ability to fine-tune the device post-manufacturing means that it only consumes 50 milliwatts of power while running at 300 mega-samples per second (Impinj told me that comparable devices consume 500 to 700 milliwatts). Furthermore, the Impinj device provides a SPDR (spur-free dynamic range ... what ever that is) of 80dB, which is apparently 6dB better than its closest competitor.

The really cool thing is that Impinj's technology doesn't change the core functionality of the analog/RF circuits - an amplifier is still an amplifier, a D/A converter is still a D/A converter - and so forth. What Impinj brings to the party is the ability to continuously calibrate and configure the functions forming an analog device. This provides for new ways of compensating for environmental (voltage and temperature) fluctuations, and also offers significant benefits in terns of design portability when migrating an existing design, or portion thereof, to a new technology node.

It's well worth noting that Impinj was called one of "the most original and promising companies formed in the last few years" by the MIT Technology Review in September 2001. It looks like Impinj's self-adaptive silicon could well be a "disruptive technology," which always means more fun and excitement for designers and consumers ... so I think that this has to receive 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.