In this paper, we describe an advanced optical diagnostic technique used for diagnosing the IBM z990 eServer microprocessor (Slegel et al., 2004). Time-to-market pressure demands quick diagnostic turnaround time and high diagnostic resolution while the ever increasing design complexity, density, cycle time, and shrinking technologies dramatically add difficulties to diagnostics. Although design-for-test (DFT) and design-for-diagnostics (DFD) features are implemented in the latest microprocessors to help easing the diagnostic efforts, they may still not be sufficient to diagnose certain fails. The well-known picosecond imaging circuit analysis (PICA) (Kash and Tsang, 1997) tool, equipped with the high quantum efficiency superconducting single-photon detector (SSPD,) shows a unique diagnostic capability for optically probing the internal nodes of a chip. Several hard-to-diagnose examples will be used to demonstrate the unique capabilities of this technique

Abstract This paper describes the analysis of a Phase-Locked Loop (PLL) internal phase detection circuit built in IBM’s 0.13 µm Silicon On Insulator (SOI) CMOS technology by using the Picosecond Imaging Circuit Analysis (PICA) [1,2] tool equipped with the high quantum efficiency Superconducting Single-Photon Detector (SSPD) [3,4]. Signals corresponding to the internal nodes of the PLL are for the first time measured and compared to circuit simulations in order to characterize the behavior of the different components of the circuit.

The clock distribution on the Power4 dual-processor chip supplies a single critical 1.5 GHz clock from one SOI-optimized PLL to 15,200 pins on a large chip with 20 ps skew and 35 ps jitter. The network contains 64 tuned trees driving a single grid, and specialized tools to achieve targets on schedule with no adjustment circuitry.

The IBM eServer z900 microprocessor is a seventh-generation zSeries™ (formerly S/390®) CMOS design which has achieved 1.3-GHz operation. This paper describes the 0.18-µm bulk CMOS, seven-level copper metal process and the high-frequency circuit, integration, and design methodologies developed to achieve this operation. The microprocessor was floorplanned to closely mimic the flow of the microarchitecture pipeline and reduce the communication delay overhead between units. Novel circuit techniques were used in the implementation of the arrays and cache hit detection logic to save power and reduce circuit complexity without sacrificing performance. A four-dimensional gate library and novel synthesis algorithms were developed to yield synthesized control implementations with the performance characteristics of a fully custom circuit design.

The use of phase-locked loops (PLLs) for clock generation in modern microprocessors has been proliferating in recent years. This is because PLLs have the advantages of allowing multiplication of the reference clock frequency and allowing phase alignment between chips. The PLL locks to a reference clock but can generate output clocks that are a multiple of the reference. It is argued that excessive “jitter”, caused primarily by power supply noise, can detract from the advantages of phase-locked loops. Moreover, in a multichip system, the accumulated phase error must be measured-not just the jitter

A phase-locked loop (PLL) in silicon over insulator (SOI) for clock generation for CMOS processors can be used on all chips in a mainframe nest which constitute up to a 10-way computer. Test site measurements show a continuous range of operation between 65 MHz and 1.3 GHz with 12 ps steady-state jitter.

The measurement of the accumulated phase error of phase-locked loops (PLLs) in microprocessor systems is discussed. A system which creates controlled power supply noise and measures the PLL response is described. Examples of the use of this technique are shown for a PLL used in a 400 MHz microprocessor.

Phase error in phase-locked loops (PLLs) in microprocessor systems is discussed. The distinction between jitter, or random phase error, and systematic, accumulated phase error is explained Accumulated phase en or is caused by disturbances of the power supply voltages. Even a short disturbance can lead to a phase error which persists and accumulates over many clock cycles. A system which creates controlled power supply noise and measures the PLL. response is described. Examples are shown using the PLL of a 400MHz S/390 microprocessor.

Phase error in phase-locked loops (PLLs) in microprocessor systems is discussed. The distinction between jitter, or random phase error, and systematic accumulated phase error is explained. Accumulated phase error is caused by disturbances of the power supply voltages. Even a short disturbance can lead to a phase error which persists and accumulates over many clock cycles. A system which creates controlled power supply noise and measures the PLL response is described. Examples are shown using the PLL of a 400 MHz S/390 microprocessor.

CMOS microprocessors operating in the hundreds of megahertz create significant current deltas due to the variation in switching activity from clock cycle to clock cycle. In addition to the high-frequency voltage variations more commonly discussed, a lower frequency noise component is also produced that lasts from 50 to 200 ns which we refer to as mid-frequency noise. In this paper, we discuss the design of IBM's CMOS S/390 computer for control of mid-frequency noise. This machine has a 10-way multiprocessor on a 127 mm by 127 mm MCM on a FR4 board. The chips on the MCM cause a current step of tens of Amps in a few cycles that can be sustained for many cycles. The power distribution and decoupling capacitors must supply that current without disturbing the voltage level at the circuits. The design of the system power distribution and modeling and verification of mid-frequency noise in this system is presented.

This paper describes the circuit design techniques used for the IBM S/390® Parallel Enterprise Server G4 microprocessor to achieve operation up to 400 MHz. A judicious choice of process technology and concurrent top-down and bottom-up design approaches reduced risk and shortened the design time. The use of timing-driven synthesis/placement methodologies improved design turnaround time and chip timing. The combined use of static, dynamic, and self-resetting CMOS (SRCMOS) circuits facilitated the balancing of design time and performance return. The use of robust PLL design, floorplanning, and clock distribution minimized clock skew. Innovative latch designs permitted performance optimization without adding risk. Microarchitecture optimization and circuit innovations improved the performance of timing-critical macros. Full custom array design with extensive use of SRCMOS circuit techniques resulted in an on-chip L1 cache having 2.0-ns cycle time.

The performance required for logic gate arrays by the IBM Enterprise System/9000TM (ES/9000TM) family of water-cooled processors was obtained by redesigning chips that previously consisted of emitter-coupled logic (ECL) circuits. Multiple bipolar logic circuit families were implemented for the first time on a single IBM chip by using a modular cell approach. In 60% of the ECL circuits, ac coupling in ECL gates reduced the maximum operating power per ECL circuit on ES/9000 chips by 50% and decreased the signal delay per loaded gate by 30%, to 150 ps. About 10-20% of the remaining ECL circuits were replaced by differential current switches (DCS) which dissipated less power and improved the overall chip performance. Circuits to communicate between ECL and DCS circuit families and to improve DCS circuit reliability were included on the ES/9000 chips without affecting logic function density.

An 1100 circuit bipolar gate array and a multi-chip high density glass-ceramic module are described. The chip features multiple logic circuit families built using a modular cell approach. Key features include use of a capacitor for faster ECL delays and the introduction of a 200 mV signal swing differential cascode current switch. The module offers a glass ceramic substrate featuring a dielectric constant of 5.2 and a total of 63 layers of metallization. The cooling capability has been increased to 30 W per chip.<>