To: pat mudge who wrote (4042 ) 10/1/1998 10:45:00 PM From: Danny Hayden Respond to of 6180
TI preps Merced-compliant power supply controller By Stephan Ohr DALLAS — Texas Instruments Inc. will take the wraps off a power-supply controller next week that will support Intel Corp.'s upcoming Merced processor. The TPS5210 breaks new ground in transient response and current-handling capability, and is being released as details emerge of Merced's power requirements and architecture. No published information exists on the power requirements of the IA-64 architecture, which will be first implemented in the Merced processor scheduled to appear in mid-2000. But discussions with industry sources suggest a continuation of the trend toward lower core voltages, higher output current and faster transient response times exhibited with the advanced Pentium parts. The 64-bit processor will not represent a drastic change over the Pentium in voltage-regulator-module (VRM) requirements. The biggest change — one that will make things easier for VRM designers but actually suggest a sea change on motherboard design — is that the base system voltage will be 12 V, not 5 V. Like the Pentium VRM, the Merced will demand a precise core voltage set by a 5-bit voltage identification (VID) code. The voltage can be anywhere between 3.6 and 1.8 V, though a 1-V core is likely on Intel's road map. The current requirement, up to 11 A on the Pentium parts, will be up to 19 or 20 A on Merced. Like the Pentium, the Merced will toggle between sleep and awake states. Tied to a 400-MHz clock, the transition from quiescent states to a full 20-A current draw can go as fast as 250 ns — and motherboard designers will like not having to leave a lot of space for a capacitor reservoir. Rather than being called a Merced controller, TI's TPS5210 might be more appropriately characterized as "the most advanced Pentium VRM controller," said Steve Goacher, TI's analog section manager. "Customers will find this part meets the needs of future PCs based on the Merced," he said. The part claims an ability to deliver up 30 A, and can effect a 30-A/microsecond current ramp. The device has a programmable output voltage range of 1.3 V to 3.5 V, determined by an integrated 5-bit D/A converter. The controller is essentially a synchronous dc-to-dc converter that uses a hysteretic control loop to improve the switching regulator's current response to transient requirements. In operation, a trip comparator looks at the hysteretic loop voltages and compares these with a precision reference. The comparator will be triggered the instant the two voltages are not symmetrical. That gives the controller a 250-ns response time. An additional droop compensation circuit improves regulation. On the test bench, said Goacher, TI's engineers put the part through a 0.1-A to 20.4-A load step, similar to what a Merced-based system might demand in a desktop. With a 12-V input and 2-V output, the droop on the output from the current surge was only 111 mV. The droop compensation also eliminates the need for extra external capacitors — a savings of 60 cents, which can become substantial in very high volumes, said Goacher. The part includes two 2-A internal MOSFETs that drive external n-channel power MOSFETs. The entire VRM can come together for less than $9, TI said. TI may be trying to leapfrog the market for VRM controllers for Intel CPUs. The company does not now appear to be a significant player in the market for Pentium VRM controllers, an arena dominated by Harris Semiconductor, Raytheon (now part of Fairchild Semiconductor) and, to a much smaller extent, Cherry Semiconductor. Unitrode also made an offering for Pentium VRM control. Analog Devices' late entry into the market this year felt like someone shouldering his way onto a crowded bus. On the other hand, manufacturers with acknowledged strengths in voltage regulators — National Semiconductor, Maxim Integrated Products and Linear Technology Corp. — did not appear to be significant factors in the Pentium VRM market. It is possible that design wins were obtained by these companies in the less-visible market for mobile Pentium power regulators. In any case, whether they claim to support the Merced processors or not, many voltage-regulator manufacturers are likely to participate in a growing trend toward double- and triple-output regulators. In some cases, these triple-output devices will show an odd combination of output voltages.