Apparatus and a method of detecting errors on registers

[0001] The present invention relates to semiconductor memory devices. More particularly, the present invention relates to an apparatus and a method of detecting errors on registers on a semiconductor chip.

[0002] ASICs (application specific integrated circuits) are becoming more and more complex with multi-million gates. Due to low power requirements, chips are reduced in size, making it prone to soft errors due to interface from alpha particles or neutrons emitted by packaging materials and cosmic rays. Soft errors are glitches in devices and occur when a logic state from “1” as initially written to “0” or vice versa. Soft errors cause no permanent damage. However, soft errors can limit the reliability of hardware devices.

[0003] The technology trend has been to go towards even more reduced geometry (20 nm, 14 nm, etc., integrated circuits), which makes soft error problems worse. Conventionally, memories on chip have used techniques such as ECC (error correction code), CRC (cyclic redundancy check) and LDPC (low density parity check) to detect and recover from such errors. However, apart from memories, ASICs contain a large amount of programming state in flip-flops. This state, in control registers, is programmed initially by software and controls how devices operate. Unlike pipelines and dynamic content, this state rarely changes and stays static, which makes this state even more vulnerable to soft errors as once soft error occurs, no corrective actions can be taken. Soft errors are extremely dangerous as they can alter the way a device is operating.

[0004] An error detection circuit on a semiconductor chip detects whether soft errors have affected flip-flop implemented registers on the semiconductor chip. A signature of these flip-flop implemented registers on the semiconductor chip is periodically captured. The signature allows for the integrity of the flip-flop implemented registers to be constantly monitored. A soft error occurring on any of the flip-flop implemented registers can be immediately detected. In response to the detection, an interrupt is raised to notify software to take action.

[0005] In one aspect, a semiconductor chip is provided. The semiconductor chip includes a plurality of control registers implemented as flip-flops, wherein each in the plurality of control registers stores control data, and an initial signature captured across all of the control registers.

[0006] The semiconductor chip also includes an error detection circuit configured to receive output signals from the plurality of control registers, periodically calculate a subsequent signature based on values of the received output signals, compare the subsequent signature with the initial signature, and based on the comparison, raise an interrupt.

[0007] In some embodiments, the semiconductor chip also includes a global register, wherein the initial signature is stored in the global register. Alternatively, the initial signature is stored in a field.

[0008] In some embodiments, the initial signature and the subsequent are calculated according to an error detection algorithm. The error detection algorithm is CRC scheme, a party bit scheme, a checksum scheme. Other error detection algorithms are contemplated.

[0009] In some embodiments, the plurality of control registers is associated with a slave stage. Alternatively, the plurality of control registers is associated with a master stage. Alternatively, a first portion of the plurality of control registers is associated with the slave stage, and a second portion of the plurality of control registers is associated with the master stage.

[0010] In another aspect, a system for detecting a soft error in a semiconductor chip is provided. The system includes memory storing an error detection algorithm, a plurality of flip-flop implemented registers, wherein each in the plurality of flip-flop implemented registers is preprogrammed to store substantially static data, and an initial signature captured across all of the flip-flop implemented registers, wherein the initial signature is calculated based on the error detection algorithm. The system also includes an error detection circuit configured to periodically calculate a subsequent signature according to the error detection algorithm, compare the subsequent signature with the initial signature, and based on the comparison, raise an interrupt.

[0011] In some embodiments, the substantially static data includes control data. In some embodiments, the substantially static data includes configuration data.

[0012] In some embodiments, the initial signature is stored in the memory.

[0013] In some embodiments, the plurality of flip-flop implemented registers is associated with a slave stage. Alternatively, the plurality of flip-flop implemented registers is associated with a master stage. Alternatively, a first portion of the plurality of flip-flop implemented registers is associated with the slave stage, and a second portion of the plurality of flip-flop implemented registers is associated with a master stage.

[0014] In some embodiments, the system further includes a global register, wherein the initial signature is stored in the global register.

[0015] In some embodiments, the system further includes a register that is a concatenation of the plurality of flip-flop implemented registers.

[0016] In some embodiments, the system further includes a protection bit such that enabling the protection bit activates the error detection circuit.

[0017] In some embodiments, the system further includes a check bit such that enabling the check bit activates a check feature of the error detection circuit, wherein when the check feature is enabled, the error detection circuit checks the subsequent signature against the initial signature.

[0018] In yet another aspect, a method of detecting a software error is provided. The method includes maintaining an initial signature captured across all flip-flop implemented registers of a semiconductor chip, periodically calculating, by an error detection circuit, a subsequent signature across all of the flip-flop implemented registers, comparing, by the error detection circuit, the subsequent signature with the initial signature, and raising, by the error detection circuit, an interrupt when the subsequent signature does not match the initial signature.

[0019] In some embodiments, the method further includes initializing all of the flip-flop implemented registers, and calculating the initial signature.

[0020] In some embodiments, the initial signature is stored in software. Alternatively, the initial signature is stored in hardware.

[0021] In some embodiments, the method further includes deactivating the error detection circuit via software prior to initializing all of the flip-flop implemented registers, and activating the error detection circuit via software after calculating the initial signature.

[0022] In some embodiments, the method further includes enabling a check feature of the error detection circuit via software prior to the comparison step.

[0023] In some embodiments, the method further includes, in response to the interrupt, suspending operation of the semiconductor chip until one or more of the flip-flop implemented registers are reprogrammed.

[0024] In some embodiments, the flip-flop implemented registers are control registers. In some embodiments, the flip-flop implemented registers are configuration registers. In some embodiments, the flip-flop implemented registers store substantially constant data.

[0025] In some embodiments, the plurality of flip-flop implemented registers is associated with a slave stage. Alternatively, the plurality of flip-flop implemented registers is associated with a master stage. Alternatively, a first portion of the plurality of flip-flop implemented registers is associated with the slave stage, and a second portion of the plurality of flip-flop implemented registers is associated with a master stage.

[0029] In the following description, numerous details are set forth for purposes of explanation. However, one of ordinary skill in the art will realize that the invention can be practiced without the use of these specific details. Thus, the present invention is not intended to be limited to the embodiments shown but is to be accorded the widest scope consistent with the principles and features described herein.

[0030] An error detection circuit on a semiconductor chip detects whether soft errors have affected flip-flop implemented registers on the semiconductor chip. A signature of these flip-flop implemented registers on the semiconductor chip is periodically captured. The signature allows for the integrity of the flip-flop implemented registers to be constantly monitored. A soft error occurring on any of the flip-flop implemented registers can be immediately detected. In response to the detection, an interrupt is raised to notify software to take action.

[0031] In some embodiments, the error detection circuit can be enabled or disabled via a protection bit. The protection bit can be set via software. The error detection circuit is first disabled to allow the flip-flop implemented registers to be preprogrammed with substantially static or constant data. For example, the flip-flop implemented registers can be control registers initialized with control data. For another example, the flip-flop implemented registers can also be configuration registers initialized with configuration data. The flip-flop implemented registers are associated with a slave stage. Alternatively, the flip-flop implemented registers are associated with a master stage. Alternatively, a first portion of the flip-flop implemented registers is associated with the slave stage, and a second portion of the flip-flop implemented registers is associated with the master stage. An initial signature is calculated across all of the flip-flop implemented registers after the flip-flop implemented registers are initialized. The initial signature can be stored in a global register. Alternatively, the initial signature can be stored in a field. The error detection circuit is then enabled. When the error detection circuit is enabled, the error detection circuit starts to receive output signals from the flip-flop implemented registers and periodically calculates a subsequent signature based on values of the received output signals.

[0032] In some embodiments, a check feature of the error detection circuit can be enabled or disabled via a check bit. The check bit can be set via software. When the check feature is enabled, the error detection circuit checks a subsequent signature against the initial signature. When the subsequent signature is different from the initial signature, the error detection circuit raises or causes an interrupt to notify software to take action. An exemplary action is to reprogram one or more of the flip-flop implemented registers. In some embodiments, normal operation of the semiconductor chip is suspended until the reprogramming of the one or more of the flip-flop implemented registers is completed.

[0033] In some embodiments, the initial signature and the subsequent signatures are calculated based on an error detection and/or correction algorithm such as a checksum scheme, a parity bit scheme, an ECC (error correction code) scheme, a CRC (cyclic redundancy check) scheme or a LDPC (low density parity check) scheme. Other error detection and/or correction algorithms are contemplated. The error detection and/or correction algorithm is typically stored in memory of the semiconductor chip.

[0034] FIG. 1 illustrates a block diagram of an exemplary semiconductor chip 100 in accordance with some embodiments. The semiconductor chip 100 includes flip-flop implemented registers 105, memory 110, an error detection circuit 115, and an interrupt module 120. The error detection circuit 115 is coupled with the flip-flop implemented registers 105, the memory 110 and the interrupt module 120. The flip-flop implemented registers 105 are preprogrammed and store substantially constant data such as control data or configuration data. The memory 110 stores an error detection and/or correction algorithm, which is used to calculate an initial signature and subsequent signatures. In some embodiments, the initial signature is stored in software. In some embodiments, the initial signature is stored in hardware. The error detection circuit 115, when enabled, receive output signals from the flip-flop implemented registers and periodically calculates a subsequent signature according to the error detection and/or correction algorithm, using values of the received output signals. When the subsequent signature is different from the initial signature, the error detection circuit raises or causes an interrupt to notify software to take action.

[0035] In a hypothetical, assume Cfg1, Cfg2 and Cfg3 are configuration registers and CfgALL is a register that is a concatenation of Cfg1, Cfg2 and Cfg3 (i.e., CfgALL={Cfg1, Cfg2, Cfg3}). CfgCRC is the computed CRC value using some specific polynomial across the CfgALL (e.g., CfgCRC=CRC(CfgALL)). The CfgCRC is an initial signature, which can be stored in stored in a field or in a global register. When a subsequently calculated CRC(CfgALL) is different from the initial CfgCRC, an interrupt is raised.

[0036] FIG. 2 illustrates an exemplary method 200 of detecting a software error in accordance with some embodiments. In some embodiments, prior to the method 200, an error detection circuit is deactivated by disabling a protection bit via software. Flip-flop implemented registers on a semiconductor chip are initialized and an initial signature across all of the flip-flop implemented registers is calculated. The error detection circuit is thereafter activated by enabling the protection bit via software.

[0037] At a step 205, the initial signature captured across all of the flip-flop implemented registers of the semiconductor chip maintained. The initial signature can be maintained in a software field. Alternatively, the initial signature can be maintained in a global register in hardware.

[0038] At a step 210, the error detection circuit periodically calculates a subsequent signature across all of the flip-flop implemented registers in the semiconductor chip.

[0039] At a step 215, the error detection circuit compares the subsequent signature with the initial signature. In some embodiments, a check feature of the error detection circuit is enabled via software prior to the step 215.

[0040] At a step 220, the error detection circuit raises an interrupt when the subsequent signature does not match the initial signature. In some embodiments, in response to the interrupt, operation of the semiconductor chip is suspended until one or more of the flip-flop implemented registers are reprogrammed.

[0041] The signature of the flip-flop implemented registers on the semiconductor chip that is periodically captured allows for the integrity of the flip-flop implemented registers to be constantly monitored. Detection of an occurrence of error on the flip-flop implemented registers is automated. If an error is detected, an interrupt is raised which will assist software to reprogram the one or more flip-flop implemented registers. This implementation is low cost since only one signature is maintained across all flip-flop implemented registers on the semiconductor chip.

[0042] One of ordinary skill in the art will realize other uses and advantages also exist. While the invention has been described with reference to numerous specific details, one of ordinary skill in the art will recognize that the invention can be embodied in other specific forms without departing from the spirit of the invention. Thus, one of ordinary skill in the art will understand that the invention is not to be limited by the foregoing illustrative details, but rather is to be defined by the appended claims.