•  
• EDN Europe MAGAZINE
•  
• Free SUBSCRIPTION
• Subscription Renewal
• View Media Kit

• EDN
• EDN Australia
• EDN Asia
• EDN China
• EDN Japan

FROM EDN EUROPE: Changing the balance of mixed-signal design

src="http://www.e-insite.net/articles/images/EDNE/20011206/pdfbutton.gif">

There is never a shortage of ingenious ideas in this industry, and it is perhaps inevitable that many of these ideas never see the light of day. But you can't discount anything that is associated with VLSI guru Carver Mead, and that endorsement comes with a technology announcement from start-up venture Impinj (www.impinj.com), which
he co-founded. The underlying concept of Impinj's technology is to produce
high-performance analogue and RF devices on standard, "logic-workhorse" CMOS
processes—which, conventionally, make poor linear circuits. Although you can
design linear and mixed-signal blocks on CMOS, the techniques you have to use
imply (among other things) that the linear blocks don't scale the same way that
the logic gates do. So, as Moore's Law proceeds, the analogue sections of your
design occupy more and more of the silicon area.

Impinj says it can fix that problem; the company's technology uses a floating-gate cell to store a voltage that calibrates the performance of an analogue-signal path, altering bias and threshold values, for example. You can make such adjustments once, after fabrication; every time your system powers up; or even on the fly in a constant measure/calibrate cycle. Impinj makes great use of differential structures and balanced circuit configurations, using the trimming facility to refine the operating point of the balanced structures. By forcing the kind of structures that work well in CMOS to work well in linear circuits, Impinj hopes to create analogue chips that can reap the same process benefits as mega-gate logic. Expect to see products from the company next year.

The question that prompts me to write about Impinj's technology in this column is: Where does it take us? For some time, conventional wisdom has said that for complex IC designs, the best philosophy is to get real-world, analogue signals into a digitized form as soon as possible. Put that tricky linear stuff as close to the front end of your system or as far up the frequency spectrum as possible. Turn it into nice, well-behaved gates. And if doing so leaves you with a great deal of digital processing, they're just gates, aren't they? Next year, those gates will occupy less area than they do now, so, digitizing is obviously the best route, right? That sort of reasoning can leave you using a lot of digital processing to carry out a function that you handled with relatively few components in the analogue domain. So, if we really can have a technology that produces good analogue performance and that scales, we might find ourselves revisiting some analogue-signal-processing techniques and reversing that seemingly inexorable trend to digitize everything as soon as possible. Now, where have all those analogue designers gone?

---

Author Information

hspace=5 src="http://www.e-insite.net/articles/images/EDNE/20011206/Xxgphead.jpg" style="HEIGHT: 102px; WIDTH: 72px" vspace=5>Contact me at graham.prophet@cahnerseurope.com.