# US 2013/0154583 A1 · Dynamic voltage scaling system based on on-chip monitoring and voltage prediction

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2013/0154583 A1
- Document ID: US20130154583A1
- Status: Published application
- Kind: A1
- Application number: 13/700,426
- Filed: 2011-10-17
- Published: 2013-06-20
- Earliest claimed: 2011-08-08
- Assignees: Southeast University
- Inventors: Longxing Shi; Weiwei Shan; Jun Yang; Haolin Gu; Xinning Liu; Yang Zhang
- Web page: https://patentgrep.com/patent/US20130154583A1

## CPC

- G05F 1/10: Regulating voltage or current (G05F1/02 takes precedence)
- G06F 1/26: Power supply means, e.g. regulation thereof (for memories G11C)
- G06F 11/008: {Reliability or availability analysis}
- G06F 11/3058: {Monitoring arrangements for monitoring environmental properties or parameters of the computing system or of the computing system component, e.g. monitoring of power, currents, temperature, humidity, position, vibrations (thermal management in cooling arrangements of a computing system G06F1/206)}
- G05F: SYSTEMS FOR REGULATING ELECTRIC OR MAGNETIC VARIABLES
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

The present invention discloses a dynamic voltage scaling system based on on-chip monitoring and voltage prediction, comprising a main circuit that has integrated on-chip monitoring circuits, a supply voltage scaling module, and voltage converters, wherein, the supply voltage scaling module comprises a sampling and statistics module designed to calculate the error rate of the main circuit in the current time slice, a state recording module designed to record the error rate and the corresponding supply voltage, an error prediction module, and a state transition probability generation module; the error prediction module predicts the error trend of the main circuit in a future time slice according to the state recording module and the state transition probability generation module, and generates regulation signals and sends to the corresponding voltage converters, so as to generate the voltage required for operation of the entire main circuit. The present invention utilizes the Markov theory to predict the “future” timing violation level of the circuit according to the “past” working condition and “current” working condition of the main circuit, and reserves some time for voltage scaling in the voltage converters; as a result, the dynamic voltage scaling has high directivity and purposiveness.

## Claims

### Claim 1

A dynamic voltage scaling system based on on-chip monitoring and voltage prediction, comprising a main circuit that has integrated on-chip monitoring circuits, a supply voltage scaling module, and a voltage converter connected to the output terminal of the supply voltage scaling module, with the output terminal of the voltage converter connected to the input terminal of the main circuit, wherein, the supply voltage scaling module comprises a state transition probability generation module and an error prediction module designed according to the main circuit, and a sampling and statistics module and a state recording module connected in sequence to the output terminals of the on-chip monitoring circuits, the output terminals of the state recording module and state transition probability generation module are connected to the input terminals of the error prediction module; the on-chip monitoring circuit send s the monitored error signals to the sampling and statistics module which calculates the error rate of the main circuit in the current time slice, and outputs to the state recording module, and the state recording module records the error rate and the corresponding supply voltage; the error prediction module utilizes the Markov theory to predict the error rate of the main circuit in a future time slice according to the probability generated by the state recording module and the state transition probability generation module, and generate regulation signals to control the output voltage of the voltage converter.

### Claim 2 (depends on 1)

The dynamic voltage scaling system based on on-chip monitoring and voltage prediction according to claim 1, wherein, the state transition probability generation module employs a state transition probability look-up table.

### Claim 3 (depends on 2)

The dynamic voltage scaling system based on on-chip monitoring and voltage prediction according to claim 2, wherein, the state transition probability look-up table is created as follows: after the main circuit is designed completely, the main circuit is subjected to run a large-size program at different voltages, and perform statistics in a large quantity, to obtain the state transition probability of the main circuit from the current state skip to the next state at different voltages, and store in the form of a look-up table.

### Claim 4 (depends on 1)

The dynamic voltage scaling system based on on-chip monitoring and voltage prediction according to claim 1, wherein, the processing procedure of the sampling and statistics module comprises the following steps: first, dividing the operation time of the main circuit into N time slices in equal size; next, performing statistics on the total number of error signals Nerror in the main circuit sent from the on-chip monitoring circuits in n−1th time slice; finally, calculating the error rate Rerror of the main circuit in the n−1th time slice: Rerror=Nerror/Ntotal, where, Ntotal is the total number of sampled signals of the main circuit in the n−1th time slice; wherein, 2≦n≦N−1, N≧3, and both N and n are integral number.

### Claim 5 (depends on 4)

The dynamic voltage scaling system based on on-chip monitoring and voltage prediction according to claim 4, wherein, the length of the time slice is not shorter than the maximum duration required for performing voltage scaling once by the voltage converter.