Monthly Archives: April 2016

Models squeeze the most speed from submicron cells

Accurate circuit simulation is necessary to get the most speed from submicron chip technology. This is also true for NCR Corp’s newest cell library which include submicron-feature devices that fall and rise 40 percent quicker than larger-feature devices. Accurate model techniques are needed to simulate the gate delays of the new Application-specific integrated circuits. The new VS700 line of cells are also 40 percent denser than NCR’s VS1500 cell library. Designers can simulate gate delays caused by rise and fall times down to 0.1 ns with the precise modeling technology.

Accurate circuit simulation is essential to pulling the highest possible speed from submicron chip technology. This is no less true for the latest cell library from NCR Corp., Fort Collins, Colo., which boasts submicron-feature devices that have much faster rise and fall times than their large-feature counterparts. And because gate-delay times vary with the rise and fall time, precise model technology is needed to accurately simulate the gate delays of these new ASICs.

Specifically, NCR’s new VS700 family of cells are 40% faster and denser than the company’s VS1500 cell library. These increases promise small, fast, and powerful end products. The submicron process produces features that are 0.95 [micrometer] drawn and have a 0.7-[micrometer] effective-channel length.

Simulation becomes difficult with submicron technology because second-order effects in a large-feature process assume first-order proportions at the submicron level. With this library, workstation models calculate each cell’s delay time as a function of rise and fall time and load capacitance. Accounting for load capacitance is a feature that makes the library especially valuable for improving critical-path performance.

With the precise modeling technology, designers can simulate gate delays caused by rise and fall times down to 0.1 ns (see the figure). In contrast, its competitors model delays from 1- to 1.5-ns edge times–a significant difference for submicron chips.

Lightly loaded critical paths have fast rise and fall times, and precise models accurately represent the faster delay times that result. Models are accurate enough to unmask circuit race conditions, giving designers a chance to correct them.

Another helpful feature of the new cell family’s design system is its set of high-level macros and compilers. Memory compilers, for example, are so flexible that designers can select word width as well the number of words. For example, designers can configure a RAM compiler to an exact bit width, and not have to settle for predefined blocks that may be larger and consume more die area than needed.

Customized symbols and models for simulation are automatically generated from the designer’s workstation input; there’s a faster time to market and no waiting for the vendor to configure high-level functions. And post-layout as well as prelayout routing capacities can be simulated. As a result, chip real-estate is assigned only for the amount of memory actually needed.


Automated assembly line builds power modules singly or by the hundreds

Vicor Corp, Andover, MA, uses a robotic assembly and overall automation to produce power-converter modules on printed circuit boards. Each module consists of a printed circuit board with small passive, active and magnetic components. A metal base plate that is automatically assembled holds the TO-220-housed FETs and diodes. Some stations along the assembly line still require off-line mechanized handling, including insertion of the chokes and transformers and coil-winding. Automation of these functions will come later, increasing the number of robots to nine. When the automation is completed, the line will produce approximately two modules per minute.

Robotic assembly and overall automation have set one computer-controlled production line apart from others in the power-supply industry. Over the past few months, this line–which produces power-converter modules–was phased into production at Vicor Corp., Andover, Mass. It handles any mix of Vicor’s modular units, from lots of hundreds to “batches of one,” according to Patrizio Vinciarelli, founder, president, and CEO of the fast-growing firm.

Each of Vicor’s modular converters consists of a pc board that carries small active, passive, and magnetic components, plus a metal baseplate that holds the TO-220-housed FETs and diodes. The baseplate is automatically assembled in a final assembly cell, where one of the line’s six robots joins it to the pc board.

Some stations along the way still require off-line mechanized handling, including coil-winding and insertion of the chokes and transformers that form the module’s magnetic components. Automating the production and insertion of those components will come late this year. For that, the number of robots on the line will increase to nine.

When completed, the line will churn out most variations of the company’s products at a rate of two modules/min. A duplicate line, to be installed late this year, will boost the rate to four modules/min. (The first line, which stretches 155 ft. across the back of one building at the Andover headquarters, underwent extensive shakedown tests at a systems-integration firm before being installed.)

A dedicated HP 9000 minicomputer controls the line, but takes its cues from an HP 3000 business computer. Orders are entered on the business computer, which tells the HP 9000 the amount and type of models to produce, and what components to select for the various lots.

Vinciarelli says that Vicor’s eventual goal is to manufacture all of its modules on the automated line with little or no operator intervention. Once each step is automated, he expects the line to process orders from start to finish in about four hours. Comparably, a manual assembly line that’s now running in tandem with the automated setup takes up to a week. The manual line will eventually be phased out.

Vicor’s wide product mix and the need to turn out high volumes precluded offshore assembly. For example, the VI-200 power-conversion module alone comes in about 300 variations.

“Our long-term strategy of high-volume and high-mix lots lends itself to automation,” says Vinciarelli. That’s why he committed $5 million to implementing the line, an investment that will roughly double with the second setup.

The first station in the line swages 40- or 80-mil pins through the small circuit board. The pins supply an electrical path when the finished module is in use. A robot then positions the pc board and the pins (see the figure). Then the robot transfers the pinned board to a second station, where the board is placed on a pallet that will convey it (pins facing up) along a moving belt until the pc board is mated to its baseplate.

A solder-dispensing station is the next stop on the line. Using a video inspection system, the station checks the board’s alignment against fiducial marks. Then it dispenses solder paste onto the pads where chip components are to be placed, inspecting the dispensed solder using gray-scale techniques to assure that no bubbles are present.

Next on the line, at the surface-mounting device (SMD) station, a robot takes active and passive components from movable vehicles. The vehicles get the components from as many as 300 feeders. Each component is tested and those that pass are placed on the solder-pasted pads, as directed by the computer, using five vacuum SMD pipettes for placement. Two such SMD stations work side by side.

A second visual inspection assures that all of the components are properly placed before the pallet-borne printed-circuit module travels to the reflow-soldering station. At the reflow-soldering station, solder paste is heated to the reflow temperature to complete the component-bonding process.

After the magnetic assemblies are placed on the board, the loaded assembly passes through a cleaning station before final assembly. In the final assembly cell, the separately traveling metal baseplate is automatically fitted with alumina pads. These pads serve as thermal conductors for the TO-220-housed power devices that are also mounted to the baseplate.

The final assembly cell also trims and forms the TO-220 leads. In addition, the alumina pads are cemented to the baseplate. Finally, the pc board is carefully mated to the baseplate so that the TO-220 leads feed through the pc board. A reflow-solder step follows to secure the pc board to the baseplate.

The finished module is then ready for electrical testing, epoxy encapsulation, and final testing. These three phases are accomplished automatically in the test, encapsulation, and test (TET) machine. All of the steps before the TET machine take just 15 min., including testing the TO-220-housed power FETs and diodes, and checking the entire assembly for characteristic outputs compared to what the lot order calls for.

Most of the TET machine’s processing time is for oven-curing the encapsulation material. A second TET machine will be installed in the second quarter of this year, with each machine drawing assemblies upstream of the line as needed. A final station prints bar-code and other identification information for the module.