Everything has to have a name, Peter Rickert noted wryly, and at Texas Instruments Inc. the program is called 1000X.
The goal is to cut transistor leakage current, so cell phone ICs can operate at higher frequencies and still stay within the power budget of battery-operated mobiles, said Rickert, who as platform development manager at TI (Dallas) coordinates the efforts of library developers, circuit designers and process technologists.
As TI moves to 90-nanometer design rules and then to 65 nm and beyond, "The biggest change is power. About a year and a half ago we saw the need going forward to use transistors with two or three orders [of magnitude] higher leakage at the transistor level," said Rickert. "We knew we could dial that back a little bit if we would give up performance, but our customers were telling us that for the 2.5G and 3G [third-generation] phones, they needed DSPs and microcontrollers that are running at 300, 400 and 500 MHz, to run all of these multimedia applications."
TI called on Synopsys Inc. to join the 1000X program by working on power estimation tools.
"In the past, we simulated in the time domain, but now we need to estimate power in both the standby and active domains," said Rickert.
Back-gate bias techniques can reduce MOSFET standby leakage by using the substrate, or back gate, as a fourth terminal. Also, in short-channel devices, the threshold voltage can be varied by modulating the drain voltage, taking advantage of the drain-induced barrier-lowering effect.
While some competitors are adopting triple-gate oxides to reduce power, TI worries about the extra cost and may adopt creative uses of a dual-gate oxide approach, he said.
The 90-nm macros and libraries include more support for selectively putting circuits in standby mode, and powering up only the circuits needed.
Keeping TI's baseband processor within a standby power budget of 200 to 300 microamps won't be easy, Rickert said. The challenge grows as RF functionality is added to the same CMOS die at 90-nm design rules for GSM phones. "That's something we have to be able to do in CMOS, or as a company we die," Rickert said.
If TI can move the RF function onto a single-chip cell phone IC, keep standby power consumption under control and get its 300-mm fab, DMOS 6, humming at peak efficiencies, the company will drive its per-chip costs down in the highest-volume market going: cellular handsets.
Wow. Talk about grand challenges.
Please send feedback to dlammers@cmp.com.