Transient IR-drop waveform measurement system and method for high speed integrated circuit
A transient IR-drop waveform measurement system and method for a high speed integrated circuit are provided. The system includes all-digital elements and is based on a ring oscillator in GHz. Through oscillation with a Fast Ring Oscillator, sampling with an Edge Detector and counting with a Ripple Counter, a width and a peak of an IR-drop waveform are obtained. Moreover, a power supply network is adapted during a clock cycle through sending an adaptation signal to a connected dynamic voltage frequency scaling (DVFS) system. The measurement method includes 11 steps. The measurement system has following features: 1) IR-drop peak/width measurement ability; 2) low fabrication and test cost; 3) high accuracy and sensitivity; 4) early adaptation ability. Therefore, the measurement system can be used alone for chip monitoring or testing, in order to reduce a power supply noise disturbance to a chip.
1. A transient IR-drop waveform measurement method for a high speed integrated circuit, comprising steps of:
(1) selecting regions with a large IR-drop, wherein the regions with the large IR-drop represent regions with a plurality of gates switching simultaneously, and are selected and marked by applying functional or structural test patterns through an electronic design automation (EDA) tool named IC Compiler;
(2) integrating a measurement system, wherein: a Control Register and a Decision Logic Module are placed in spare areas of an integrated circuit chip, while transient IR-drop waveform measurement modules, respectively 2 A, 2 B . . . 2 L, are inserted into the marked regions; each transient IR-drop waveform measurement module is connected to a power supply network corresponding to the respective region, and a system clock is routed to an input end of a Timer ( 20 C) automatically by the EDA tool;
(3) determining a measurement window length and a sampling window length, wherein: for a single-clock system with only positive-edge devices, a measurement window is as long as one system clock cycle; for a region comprising both the positive-edge devices and negative-edge devices, double IR-drop peaks occur in each clock cycle, thus the measurement window is equal to half of the system clock cycle; for a multiple-clock region, the measurement window is the same as the shortest clock cycle; and meanwhile, an optimal sampling window length is determined according to an amount of buffers in an Edge Detector ( 20 E);
(4) generating lookup tables, wherein: besides an overall counter value within the measurement window and signatures of the Edge Detector ( 20 E) obtained directly from each transient IR-drop waveform measurement module, the IR-drop peak is required to obtain an IR-drop waveform, and thus a fast three-dimensional lookup table is generated for showing a relationship among the IR-drop peak, an IR-drop width and a normalized counter value; moreover, as process and temperature variations may affect an operation of the measurement system, several lookup tables are generated for showing the relationship among the IR-drop peak, the IR-drop width and the normalized counter value under different temperatures; the step (4) of generating the lookup tables further comprising steps of:
4a) generating an N 0 -K lookup table suitable for a manufactured Fast Ring Oscillator (RO), wherein the N 0 -K lookup table correlates a noise-free counter value N 0 and a temperature K ranging from −40° C.-120° C. ; and depicting the N 0 -K lookup table by an automatic test equipment (ATE) during a production test;
4b) generating a K-
∂
T
∂
VDD
lookup table suitable for a specific Fast RO during the production test through stepping VDD and measuring a variation of a Fast RO ( 20 D) oscillation period under the various temperatures K ranging from −40° C.-120° C., wherein the K-
∂
T
∂
VDD
lookup table correlates the temperature K and the
∂
T
∂
VDD
;
and the
∂
T
∂
VDD
represents a delay-VDD sensitivity;
4c) repeating above two steps during the production test, and storing obtained lookup tables either on-chip or off-chip; and
4d) obtaining a three-dimensional V p -t w -N norm lookup table through simulating by Simulation Program with Integrated Circuit Emphasis (SPICE), wherein: the three-dimensional V p -t w -N norm lookup table correlates a width t w of the IR-drop waveform, a peak V p of the IR-drop waveform, and the normalized counter value N norm at the various
∂
T
∂
VDD
;
and the various
∂
T
∂
V
D
D
are easily obtained by changing a load or a cell type of an arbitrary Fast RO;
(5) determining an adaptation threshold, wherein: to minimize an adaptation reaction time, the adaptation threshold is determined according to an output of a Differentiator ( 20 B), which is obtained within one adaptation grid; during an ATE test, an intensity of the structural or functional patterns is increased until a first failure occurs, and the corresponding output is assigned as the adaptation threshold;
(6) configuring the Control Register, wherein: before measurement, the Control Register is configured; the step (6) of configuring the Control Register further comprising steps of: writing a start clock cycle, the measurement window length, a branch configuration of the Edge Detector ( 20 E), the adaptation threshold, and a length of the Fast RO ( 20 D) into the Control Register;
(7) calibrating the Edge Detector ( 20 E), wherein: inside each branch of the Edge Detector ( 20 E), lengths of two sub-branches are calibrated to be close to each other; the two sub-branches are respectively a strong sub-branch and a weak sub-branch; the weak sub-branch has a longer delay then the strong sub-branch; by increasing a length of the strong sub-branch, the Decision Logic Module checks an m-bit signature value Edge_Indicator[m-1:0] all the time; if no IR-drop exists, every bit of the signature is logic “0”; and, once a specific bit flips during checking, a calibration for the corresponding branch ends;
(8) measuring the temperature, wherein: in the step (7), a three-step lookup table checking process is provided; since different delay-VDD sensitivities correspond to different three-dimensional planes, a proper plane for obtaining a noise peak is determined before in-field detecting; to obtain a local temperature, the Fast RO oscillates for a time length of the measurement window, and the counter value is recorded; and finally, according to the already generated lookup tables, an in-field temperature is obtained;
(9) monitoring in-field, wherein: during in-field monitoring, the system starts at a predefined time, and works for a time that equals the measurement window length; and, the obtained counter values within the measurement window and each adaptation grid, and the m-bit signature value Edge_Indicator[m-1:0] of the Edge Detector ( 20 E) are stored in an on-chip flash;
(10) adapting in-field, wherein: the Decision Logic Module compares the output of the Differentiator with the predefined adaptation threshold; if an IR-drop is over the threshold, an output of the Decision Logic Module switches to logic “1” to adapt a dynamic voltage frequency scaling (DVFS) system and compensates the power supply network; and
(11) reconstructing the IR-drop waveform, wherein: after in-field measurement, the recorded N 0 , the counter value N and values of the m-bit signature value Edge_Indicator[m-1:0] are read out; then, the width and the peak of the IR-drop are calculated according to the above read out values and the lookup tables; and finally, a triangular IR-drop is reconstructed.