Method for transferring data between a first digital domain and a second digital domain, and corresponding system-on-chip.

The interface circuit in a system on chip prevents data corruption during transfers by using synchronized internal resets and controlled data handling, addressing metastability issues in asynchronous resets across power domains with different supply voltages and frequencies.

FR3134462B1Active Publication Date: 2024-11-08STMICROELECTRONICS (ROUSSET) SAS
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
FR2022003375
Authority / Receiving Office
FR · FR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2022-04-12
Publication Date
2024-11-08
Estimated Expiration
2042-04-12

AI Technical Summary

Technical Problem

Data corruption due to metastability in registers during asynchronous resets occurs when transferring critical information between power domains in a system on chip with different supply voltages and operating frequencies, leading to system malfunctions.

Method used

Implement an interface circuit that includes transfer elements and synchronization circuits configured to avoid resets from asynchronous external signals, ensuring data transfer completion with uncorrupted data by using synchronized internal resets and controlled data handling.

Benefits of technology

Prevents data corruption by asynchronous resets, maintaining data integrity during transfers between power domains with different supply voltages and frequencies, thereby ensuring reliable system operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000018_0000
    Figure 00000018_0000
  • Figure 00000019_0000
    Figure 00000019_0000
Patent Text Reader

Abstract

The system-on-a-chip includes at least one first digital domain (V11) configured to be reset by a first reset signal (PADRST), a second digital domain (V22), and an interface circuit (INTF). The interface circuit includes a start register (RDEP) in the first domain, a destination register (RDEST) in the second domain, and a synchronization circuit (FFSYNC) in the first domain. The interface circuit is configured to transfer data (PWDAT, PADD) from the start register (RDEP) to the destination register (RDEST) upon command of a control signal (W_pls_sync) transmitted by the synchronization circuit (FFSYNC). The start register (RDEP) and the synchronization circuit (FFSYNC) are configured not to be reset by the first reset signal (PADRST). Figure for the abstract: Fig 1
Need to check novelty before this filing date? Find Prior Art

Description

Title of the invention: Method for transferring data between a first digital domain and a second digital domain, and corresponding system on chip.

[0001] Embodiments and implementations relate to data transfers between distinct digital domains of a system-on-chip, particularly digital domains having different supply voltages and operating frequencies.

[0002] Conventionally, in systems on chip comprising distinct power domains, i.e. in particular digital domains powered and clocked by different voltage-frequency pairs, certain power domains can be powered intermittently while other power domains can be powered continuously.

[0003] Intermittently powered power domains are typically high-performance domains that can be turned on and off depending on the system-on-chip usage. These high-performance domains are power-intensive because they are powered at relatively high frequencies in the system-on-chip.

[0004] Permanently powered domains are typically energy efficient and in particular allow critical data or critical functions to be maintained when the system is operating in a low power mode. Consequently, permanently powered domains are typically clocked at low frequencies.

[0005] However, critical information to be kept in the permanently supplied domain can be written by high-frequency transaction initiators, for example from the high-performance domain.

[0006] On the one hand, critical information must be received and recorded, typically in registers in the low frequency domain, regardless of the data transmission frequency relative to the register writing frequency.

[0007] On the other hand, a reset of the system-on-chip may occur randomly with respect to the writing of the registers, typically upon actuation of an external and asynchronous reset signal.

[0008] This type of external and asynchronous reset can cause metastability on the registers containing the critical information to be transferred, leading to a malfunction of the system on chip, if the reset occurs at the same time as the writing of the registers.

[0009] Metastability of a register, such as a conventional and per se known "D" flip-flop, typically occurs when the data pins of the register flip during a transition of its clock signal or when the clock signal is interrupted, in which case the output of the register remains between low and high states for some time and converges to an unknown state.

[0010] Thus, metastability of the low-frequency domain registers containing critical data may be caused when the inputs of this register are reset asynchronously with respect to the low-frequency clock. For example, the external reset event during a data transfer between the two domains may corrupt the critical contents of the register.

[0011] There is therefore a need to avoid corruption of critical data during a transfer between two power domains of a system-on-chip.

[0012] In this regard, according to one aspect there is provided a system-on-chip comprising at least a first digital domain configured to be reset by a first reset signal, a second digital domain and an interface circuit comprising at least one transfer element in the first domain. The interface circuit is configured to transfer data from said at least one transfer element to the second domain. Said at least one transfer element is configured not to be reset by the first reset signal.

[0013] The transfer element of the interface circuit is for example a register containing data to be transferred, or a digital signal state to be transferred.

[0014] Since the transfer element of the interface circuit is not affected by the first reset signal, which is for example an asynchronous external reset signal, the risk of incurring corrupted data due to metastability resulting from an asynchronous reset by the first reset signal during the transfer is greatly limited. Indeed, this allows the interface circuit to react in a controlled manner to a reset by the first reset signal, the transfer element thus being able to complete a transfer with uncorrupted data, despite a reset occurring at the time of the transfer.

[0015] According to one embodiment, the second domain is configured to be reset by a second reset signal, said at least one transfer element being configured to be reset by the second reset signal.

[0016] For example, the second reset signal can be an internal power-on reset signal, usually called a "power-on reset" in English, used during a total shutdown of the system-on-chip or during an intermittent shutdown of the domain, which occur synchronously.

[0017] According to one embodiment, the first domain is configured to be clocked by a first clock signal at a first frequency and the second domain is configured to be clocked by a second clock signal at a second frequency lower than the first frequency.

[0018] This corresponds to an example where the first domain is intended for high-performance numerical executions, while the second domain is intended to be kept in operation, for example, to record critical data from the first domain.

[0019] According to one embodiment, the first domain is configured to receive the first reset signal from outside the system-on-chip asynchronously with the first clock signal, and the second domain is configured to generate the second reset signal during a synchronous reset process upon startup of the system-on-chip.

[0020] According to one embodiment, the interface circuit comprises a starting register among said at least one transfer element, a destination register in the second domain and a synchronization circuit in the first domain, the interface circuit being configured to transfer data from the starting register to the destination register upon command of a control signal transmitted by the synchronization circuit, in which the synchronization circuit is configured not to be reset by the first reset signal.

[0021] This also helps to limit the risk of corrupted data being involved in the transfer. Indeed, here again the interface circuit can react in a controlled manner to a reset by the first reset signal, since the synchronization circuit is thus able to avoid initiating a transfer command with corrupted data due to metastability of the starting register, caused by a reset occurring at the moment the starting register is loaded.

[0022] According to one embodiment, the first domain is configured to communicate the data to be transferred on an input of the start register and to communicate the control signal to the synchronization circuit at least one cycle of the first clock signal after said data, and the start register is configured to capture the data to be transferred in a single cycle of the first clock signal and then retain said data independently of said input.

[0023] This embodiment makes it possible on the one hand to guarantee that the data to be transferred, when they are retained by the starting register, are not corrupted by a possible reset occurring after said single cycle used to enter the data. Furthermore, this embodiment makes it possible on the other hand to guarantee not to transfer data corrupted by a reset occurring during said single cycle used to enter the data, given that in such a case, the control signal cannot be communicated to the synchronization circuit one cycle after said single cycle used to enter data, since the reset occurs during this cycle.

[0024] In other words, this embodiment makes it possible not to suffer data corruption occurring after the cycle used for data entry, and not to transfer corrupted data during the cycle used for entry.

[0025] According to one embodiment, the interface circuit comprises a signal adaptation element among said at least one transfer element, configured to transfer, between the first domain and the second domain, a link establishment signal.

[0026] This again makes it possible to limit the risk of engaging said signal contributing to said transfer, corrupted due to a reset during the transfer.

[0027] According to one embodiment, said at least one signal adaptation element comprises a second start register configured to capture the value of a handshake conditioning signal in a single cycle of said first clock signal and then retain said value.

[0028] This embodiment also makes it possible to avoid corruption of said signal occurring after the cycle used for capturing this signal.

[0029] According to one embodiment, a control logic circuit is configured to generate a transfer status signal during the first cycle of the first clock signal of the transfer of data from the first domain to the second domain, the start register(s) being configured to capture said data or said handshake conditioning signal upon command of the transfer status signal, such that said one cycle of the first clock signal is this first cycle.

[0030] According to one embodiment, the control logic circuit is configured to provide the transfer state signal shifted by at least one cycle of the first clock signal to the synchronization circuit triggering the transmission of said control signal.

[0031] According to another aspect, there is provided a method of transferring data between at least a first digital domain, and a second digital domain of a system on chip, comprising transferring the data from at least one transfer element in the first domain to the second domain, wherein a first reset signal causes a reset of the first domain but does not cause a reset of said at least one transfer element.

[0032] According to one embodiment, a second reset signal resets the second domain, and said at least one transfer element.

[0033] According to one embodiment, the first domain is clocked by a first clock signal at a first frequency and the second domain is clocked by a second clock signal at a second frequency lower than the first frequency.

[0034] According to one embodiment, the first reset signal originates from outside the system-on-chip asynchronously with the clock signal of the first domain, and the second reset signal is generated during a synchronous reset process at startup of the system-on-chip.

[0035] According to one embodiment, said transfer comprises a data transfer made from a starting register among said at least one transfer element, to a destination register in the second domain, on command of a control signal transmitted by a synchronization circuit of the first domain, the first reset signal not causing a reset of a synchronization circuit.

[0036] According to one embodiment, the method comprises a communication of the data to be transferred on an input of the starting register and a communication of the control signal to the synchronization circuit at least one cycle of the first clock signal after said data, and in which the transfer of the data comprises an entry, by the starting register, of the data to be transferred in a single cycle of the first clock signal and a conservation of said data thereafter, independently of said input.

[0037] According to one embodiment, said transfer comprises a signal adaptation from a signal adaptation element among said at least one transfer element, comprising a transfer of a handshake signal between the first domain and the second domain.

[0038] According to one embodiment, said at least one signal adaptation comprises an entry, in a second starting register, of the value of a link establishment conditioning signal in a single cycle of said first clock signal and a retention of said value thereafter.

[0039] According to one embodiment, the method comprises generating a transfer status signal during the first cycle of the first clock signal of the transfer of data from the first domain to the second domain, said entry into the start register(s) being controlled by the transfer status signal, such that said single cycle of the first clock signal is this first cycle.

[0040] According to one embodiment, the transfer state signal is provided, shifted by at least one cycle of the first clock signal, to the synchronization circuit and triggering the transmission of said control signal.

[0041] Other advantages and characteristics of the invention will appear on examining the detailed description of embodiments and implementations, which are in no way limiting, and the appended drawings in which:

[0042] [Fig.l] ;

[0043] [Fig.2] illustrate embodiments and implementations of the invention.

[0044] [Fig.l] illustrates an example of a system on chip SOC comprising at least two digital domains Vil, V22 having different supply voltages and different operating frequencies.

[0045] The first digital domain Vil is supplied by a first supply voltage VDD11, for example 1.1 volts, generated by a supply stage SUPP11. The first digital domain Vil is clocked by a first clock signal CLK11 at a first frequency, generated by a clock generation stage CLKGEN11.

[0046] The second digital domain V22 is supplied by a second supply voltage VDD22, for example 3.3 volts, generated by a supply stage SUPP22. The second digital domain V22 is clocked by a second clock signal CLK22 at a second frequency, generated by a clock generation stage CLKGEN22.

[0047] In this example, the second frequency is lower than the first frequency, and the second supply voltage VDD22 is higher than the first supply voltage VDD11.

[0048] In the context of this example, the first domain V11 may for example be intended to carry out operations with high performance and consequently high energy consumption, while the second domain V22 may for example be intended to be always active, so as to ensure critical operations, and with low energy consumption.

[0049] On the other hand, the first domain Vil is configured to be reset by a first reset signal PADRST, while the second domain V22 is configured to be reset by a second reset signal POR distinct from the first reset signal PADRST. The respective reset signals PADRST, POR are used to place the respective domain Vil, V22 in a virgin and mastered state (typically to restart the domain), and have the effect in particular of interrupting the actions in progress and erasing the data contained in the registers of the reset domain.

[0050] The first PADRST reset signal can, in particular, originate from outside the system-on-chip (SOC) asynchronously with the first CLK11 clock signal. For example, the first PADRST reset signal can be a forced reset signal activated by a user independently of the actions being performed by the first VIL domain.

[0051] The second domain V22 can be configured to generate the second reset signal POR during a synchronous reset process at the start of the system on chip SOC, usually designated by the English terms “power-on reset”.

[0052] Critical information to be stored in the second domain V22 can be written by initiators from the first domain VI1, and the system on chip SOC comprises in this regard an interface circuit INTF configured to implement a transfer of data and signals between the two domains V11, V22.

[0053] The interface circuit INTF makes it possible to make compatible, in particular according to the voltage-frequency pairs respective to each domain, the signals from the first domain Vil for the second domain V22, and possibly the signals from the second domain V22 for the first domain Vil.

[0054] In addition, the INTF interface circuit makes it possible to protect potentially critical data against corruption caused by an asynchronous reset occurring during the transfer.

[0055] Indeed, if for example an asynchronous reset causes a switch of the register input signals during a transition window of the clock signal timing the registers (usually "setup-hold window" in English), then the outputs of the registers can be in metastable states. In the metastable state, an output signal oscillates between the high state and the low state for a certain time then converges towards an unknown state independent of the input signal.

[0056] From the point of view of compatibility between the first supply voltage VDD11 and the second supply voltage VDD22, the interface circuit INTF comprises voltage level shift circuits LS configured to convert the voltages of the logic levels of one of the two domains Vil, V22 to the corresponding voltage of the other domain V22, VIL

[0057] From the point of view of compatibility between the first frequency CLK11 and the second frequency CLK22, the interface circuit INTF comprises in particular a start register RDEP in the first domain VI1, a synchronization circuit FFSYNC in the first domain VI1, and a destination register RDEST in the second domain VIL

[0058] By "a circuit in the first domain (respectively second domain)" is meant that this circuit belongs to the first domain (respectively to the second domain), that it is supplied by the first supply voltage VDD11 (respectively the second supply voltage VDD22) and that its operation is clocked by the first clock signal CLK11 (respectively the second clock signal CLK22).

[0059] The interface circuit INTF comprises transfer elements RDEP, ADPT1, (see also [Fig.2] ADPT2, RDEP2) in the first domain Vil, in order to transfer data PWDAT, PADD, or signals such as a handshake signal CDC11, to the second domain V22.

[0060] In the examples of Figures 1 and 2, the transfer elements comprise registers, in particular starting registers RDEP, RDEP2, intended for data transfers, and signal adaptation elements ADPT1, ADPT2, intended for transfers of functional signals, such as handshake signals.

[0061] The transfer elements RDEP, ADPT1, ADPT2, RDEP2, are advantageously configured not to be reset by the first reset signal PADRST of the first domain Vil.

[0062] To transfer data PWDAT, PADD, from the first domain V11 to the second domain V22, the interface circuit INTF is configured to load the data to be transferred PWDAT, PADD into the start register RDEP, in a manner clocked by the first clock signal CLK11 in the first domain V11. The data accDAT, accADD loaded into the start register RDEP are then transferred to the destination register RDEST on command of a control signal W_pls_sync transmitted by the synchronization circuit FFSYNC adapted to clock the destination register of the second domain V22.

[0063] For example, the data to be transferred is intended to be written into a memory of the second domain V22 and may in this respect comprise PWDAT content data and a respective PADD memory address. The data and addresses are transmitted in N-bit words (with N a natural integer) for example 16 bits, 32 bits or 64 bits, on respective buses. The starting register RDEP consequently comprises a register of respectively 16, 32 or 64 bits to contain the PWDAT data word to be transferred and a register of respectively 16, 32 or 64 bits to contain the corresponding PADD address word.

[0064] From the point of view of protection against corruption caused by an asynchronous reset during the transfer, the interface circuit INTF is configured to implement measures 100, 200, 300 to ensure an absence of risk of corruption.

[0065] In the first measurement 100, the transfer elements RDEP, ADPT1 and the synchronization circuit FFSYNC are configured not to be reset by the first reset signal PADRST, unlike the other circuits of the first domain VIL

[0066] The transfer elements RDEP, ADPT1 and the synchronization circuit FFSYNC can nevertheless be configured to be reset by the second reset signal POR. This does not present a risk from the point of view of data corruption since the second reset signal POR is generated synchronously by the system on chip SOC.

[0067] Thus, due to the first measurement 100, if the first reset signal PADRST occurs during the transfer of the accDAT, accADD data correctly loaded into the start register RDEP, these accDAT, accADD data will not be erased at the output of said start register RDEP. In particular the accDAT, accADD loaded in the starting register RDEP will not be able to switch during the transition window of the destination register RDEST. Similarly, due to the first measurement 100, the control signal W_pls_sync transmitted to the destination register RDEST by the synchronization circuit FFSYNC will not be able to switch unexpectedly during a data capture by the destination register RDEST, thus ensuring a correctly performed capture of said data accDAT, accADD in the second domain V22.

[0068] Furthermore, the data to be transferred PWDAT, PADD are communicated by general purpose circuits (not shown) of the first domain Vil, on input channels of the starting register RDEP.

[0069] In the second measurement 200, the start register RDEP is configured to capture the data to be transferred PWDAT, PADD in a single cycle of the first clock signal CLK11, and then store said data accDAT, accADD, for example in a D-type flip-flop, independently of the input channels PWDAT, PADD.

[0070] The starting register RDEP may in this respect comprise a multiplexer element providing the data to be stored in the starting register. The multiplexer element is configured in this respect to select either the data PWDAT, PADD present on the input channels, or the data accDAT, accADD contained in the starting register RDEP, looped back from the output of the starting register RDEP to an input of the multiplexer.

[0071] The multiplexer element is thus controlled to select (on "1") the input channels PWDAT, PADD during a single cycle to enter the data PWDAT, PADD in the starting register RDEP, then to select (on "0") the looping back of the output of the starting register RDEP then, in order to keep the entered data accDAT, accADD.

[0072] Advantageously in this respect, the multiplexer element can be controlled by a transfer state signal PRDY_fe, generated by a control logic circuit FSM, also called a “state machine” or “finite state machine”. The transfer state signal PRDY_fe is generated during the first cycle of the first clock signal CLK11 of the data transfer.

[0073] The multiplexer element is thus controlled to select the input channel PWDAT, PADD by the transfer status signal PRDY_fe. The start register RDEP is thus configured to capture said data PWDAT, PADD on command of the transfer status signal PRDY_fe, so that said single cycle of the first clock signal CLK11 is the first cycle of the clock signal CLK11 of the implementation of the transfer.

[0074] This ensures that the data to be transferred accDAT, accADD, cannot be corrupted on the input channel PWDAT, PADD by a possible reset. PADRST occurring after the first cycle of the transfer, since the data accDAT, accADD are retained by the start register RDEP regardless of the state of the data signals on said input channel PWDAT, PADD.

[0075] As an alternative to the multiplexer element, a clock gating circuit, usually referred to as “clock gating”, may be provided on the clock signal clocking the D flip-flop of the start register in order to capture the data on the input channels in a single cycle. A conventional clock gating circuit, advantageously without deviation (usually “glitch free” in English), may be obtained by mounting a D flip-flop clocked by the inverted clock signal CLK11, and receiving the transfer status signal PRDY_fe as input. An AND gate between the output of the D flip-flop and the clock signal CLK11 makes it possible to transmit a clock pulse conditioned by the transfer status signal PRDY_fe.

[0076] In practice, and in an example following a communication protocol of the “APB” type (acronym for the usual English terms “Advanced Peripheral Bus”), the transfer status signal PRDY_fe can come from a PRDY signal provided in the APB protocol, and generated by the FSM control logic circuit. The PRDY signal provided in the APB protocol is generated at a first logic level, for example 0, during any implementation of the transfer, and at a second logic level, for example 1, outside the implementation of the transfer. Consequently, a falling edge occurs on this PRDY signal at the start of the first cycle of the transfer.

[0077] From this falling edge, the transfer status signal PRDY_fe controlling the multiplexer element and as defined above can be generated, i.e. the generation of the instant of capture of the data PWDAT, PADD during the first cycle of the transfer.

[0078] Furthermore, from the signal PRDY provided in the APB protocol, the offset transfer status signal PRDY_fe_d can be generated, defining the instant at which writing is commanded in the second domain V22, i.e. after and only after this first cycle, as defined below in relation to the third measurement 300.

[0079] For example, and in a manner adapted to the context of the “APB” protocol, the FSM control logic circuit is also configured to receive control signals PWRITE, PSEL, PENABLE, from the first domain Vil, controlling the transfer.

[0080] In the third measurement 300, the synchronization circuit FFSYNC is configured to transmit the control signal W_pls_sync to the destination register RDEST after the first cycle of the clock signal CLK11 timing the implementation of the transfer in the first domain VI1. In this regard, the control logic circuit FSM is advantageously configured to provide the shifted transfer status signal PRDY_fe_d by one (or more) cycle(s) of the first clock signal CLK11, in order to trigger the generation and transmission of the control signal W_pls_sync to the destination register RDEST by the synchronization circuit FFSYNC.

[0081] The control signal W_pls_sync is for example generated by a pulse stretching circuit STRCHR, configured to generate a pulse W_pls having a duration long enough to pass through the voltage level shifting circuit LS, from a pulse generated in the first domain VI1 and having a duration corresponding to the first frequency CLK11.

[0082] In this regard, a trigger event, for example generated by combinational logic between two signals, for example an AND between an access validity signal AccVld generated by the control logic circuit FSM and the falling edge of the transfer status signal shifted by one cycle PRDY_fe_d, is captured by a D flip-flop. The D flip-flop being clocked by the first clock signal CLK11, the output of the D flip-flop provides a pulse with a duration of one cycle of the first clock signal CLK11 and triggered by the transfer status signal shifted by one cycle PRDY_fe_d. The pulse stretching circuit STRCHR may typically comprise a series of flip-flops and logic conditions on intermediate signals in the series to extend the duration of the pulse over several cycles of the first clock signal CLK11.The number of series flip-flops and cycles of the first clock signal is chosen to be adapted to the crossing time of the LS voltage level shifting circuit.

[0083] In other words, the first domain Vil is configured to communicate the control signal W_pls to the synchronization circuit FFSYNC, at least one cycle of the first clock signal CLK11 after the provision of said data PWDAT, PADD.

[0084] This ensures that data held accDAT, accADD by the starting register RDEP, possibly corrupted by a PADRST reset occurring during the first cycle of the transfer, is not captured by the destination register RDEST, since the PADRST reset prevents the generation of the transfer status signal shifted by one cycle PRDY_fe_d and therefore prevents the triggering of the transmission of the control signal W_pls.

[0085] Furthermore, the synchronization circuit FFSYNC, for transmitting the control signal W_pls_sync, advantageously comprises a series of flip-flops, for example of the D flip-flop type, in sufficient number to guarantee a risk of propagation of a metastability lower than a specification of mean time between failures of the system on chip SOC (usually designated by the acronym “MTBF” from the English terms “Mean Time Between Failures”).

[0086] Therefore, if a PADRST reset occurs after the first cycle of the transfer and after the generation of the trigger command PRDY_fe_d, a possible metastability will not be transmitted in the control signal W_pls through the FFSYNC synchronization circuit, given its configuration in this regard and given that the FFSYNC synchronization circuit is not impacted by the first reset signal PADRST.

[0087] The INTF interface circuit thus described therefore makes it possible not to communicate to the destination register RDEST data corrupted by a PADRST reset occurring during the first cycle of the transfer thanks to the third measure 300, nor of data corrupted by a PADRST reset occurring in subsequent cycles during the transfer thanks to the second measure 200, nor of corrupting data being entered by the destination register RDEST thanks to the first measure 100.

[0088] The same effects are respectively obtained on the control signal W_pls_sync, thanks to the third measurement 300 and the first measurement acting on the generation of the control signal W_pls and the transmission of the control signal W_pls_sync by the synchronization circuit FFSYNC.

[0089] Furthermore, other signals exchanged between the first domain V11 and the second domain V22, in particular handshake signals CDC11, CDC22, contributing for example to the transfer of data, can benefit from said first measurement 100, second measurement 200, and / or third measurement 300.

[0090] For example, a CDC22 signal can be generated in the second domain V22 to communicate a state ready for transfer of the destination register RDEST. The CDC22 signal can undergo a "handshake" procedure, and in this respect pass through an ADPT1 signal adaptation element of the first domain V11, and then be retransmitted as CDC11 to the second domain V22. In the second domain V22, the CDC11 handshake signal resulting from the handshake procedure can, for example, enable the activation of an activation terminal E of the destination register RDEST, for example in combination with other signals via a logic circuit conditioning the activation.

[0091] In this example, the signal adaptation element ADPT1 may comprise a series of D flip-flops clocked by the first clock signal CLK11, so as to provide the handshake signal CDC11 contributing to the capture of the destination register RDEST, synchronously with the other signals accDAT, accADD, W_pls_sync coming from the first domain and controlling the capture of the destination register RDEST.

[0092] Such a signal adaptation element ADPT1 is advantageously configured not to be reset by the first reset signal PADRST, and thus benefit from the effects of the first measurement 100. Furthermore, the signal adaptation element ADPT1 can be reset by the second POR reset signal.

[0093] [Fig.2] illustrates another example of a signal adaptation element ADPT2, belonging to the first domain Vil, configured to transfer, between the first domain Vil and the second domain V22, an activation signal Enbl22 contributing to said data transfer. The activation signal Enbl22 is for example supplied to the logic circuit conditioning the activation cond.

[0094] This further example of an ADPT2 signal adaptation element includes a suitable means for a "handshake" method, as described above in relation to [Fig. 1], providing a synchronized handshake signal CDC11 in the first domain VI1 from a handshake signal CDC22 from the second domain CDC22. The CDC11 signal is transmitted to an input of a PL logic gate, for example an OR gate, conditioning the transmission of the Enbl22 signal to the second domain VI22 with a handshake conditioning signal Fig. 1, from the first domain VI1

[0095] And, the signal adaptation element ADPT2 advantageously comprises a second start register RDEP2 configured to capture the value, i.e. the logic state, of the link establishment conditioning signal flgl 1, in a single cycle of said first clock signal CLK11, i.e. according to the second measurement 200 described previously in relation to [Fig.l].

[0096] This second starting register RDEP2 is further configured not to be reset by the first reset signal PADRST, i.e. according to the first measure 100 described previously in relation to [Fig.1].

[0097] This again makes it possible to limit the risk of engaging said signal contributing to said transfer in a corrupted state by a reset occurring during the transfer, after the cycle used for the capture of the digital signal.

[0098] Furthermore, the invention is not limited to the embodiments and implementations described in relation to Figures 1 and 2, but embraces all variants thereof; for example, the invention embraces a system on chip provided with a signal adaptation element ADPT1, ADPT2, as described in relation to Figures 1 and 2, but considered in isolation, i.e. without the start registers RDEP1, RDEP2 or the synchronization circuit FFSYNC; or a system on chip provided with a second start register RDEP2 configured to transfer a conditional signal flgl 1 as described in relation to [Fig. 2], but considered in isolation, i.e. without the signal adaptation circuit ADPT2, in combination or not with the circuit described in relation to [Fig. 1]; or a system on chip provided with an ADPT2 signal adaptation element as described in relation to [Fig.2], but considered in isolation, that is to say without the starting register RDEP1, nor the synchronization circuit FFSYNC described in relation to [Fig.l].

Claims

Claims

1. A system-on-chip comprising at least one first digital domain (Vil) configured to be reset by a first reset signal (PADRST), a second digital domain (V22) and an interface circuit (INTF) comprising at least one transfer element (RDEP, ADPT1, ADPT2, RDEP2) in the first domain, the interface circuit being configured to transfer data (PWDAT, PADD, CDC11) from said at least one transfer element (RDEP, ADPT1, ADPT2, RDEP2) to the second domain (V22), wherein said at least one transfer element (RDEP, ADPT1, ADPT2, RDEP2) is configured not to be reset by the first reset signal (PADRST); wherein the interface circuit (INTF) comprises a signal adaptation element (ADPT1, ADPT2), among said at least one transfer element, configured to transfer, between the first domain (Vil) and the second domain (V22), a link establishment signal (CDC11, CDC22);and wherein said at least one signal adaptation element (ADPT2) comprises a second start register (RDEP2) configured to capture the value of a handshake conditioning signal (flg 11) in a single cycle of said first clock signal (CLK11) and then retain said value (acc_flg).;

2. The system-on-chip of claim 1, wherein the second domain (V22) is configured to be reset by a second reset signal (POR), said at least one transfer element (RDEP, ADPT1, ADPT2, RDEP2) being configured to be reset by the second reset signal (POR).

3. System on chip according to one of claims 1 or 2, wherein the first domain (Vil) is configured to be clocked by a first clock signal (CLK11) at a first frequency and the second domain (V22) is configured to be clocked by a second clock signal (CLK22) at a second frequency lower than the first frequency.

4. The system-on-chip of claims 2 and 3, wherein the first domain (Vil) is configured to receive the first reset signal (PADRST) from outside the system-on-chip (SOC) asynchronously with the first clock signal (CLK11), and the second domain (V22) is configured to generate the second reset signal (POR) during a synchronous reset process at system-on-chip (SOC) startup.

5. System on chip according to one of claims 1 to 4, wherein the interface circuit (INTF) comprises a start register (RDEP) among said at least one transfer element, a destination register (RDEST) in the second domain and a synchronization circuit (FFSYNC) in the first domain, the interface circuit being configured to transfer data (PWDAT, PADD) from the start register (RDEP) to the destination register (RDEST) on command of a control signal (W_pls_sync) transmitted by the synchronization circuit (FFSYNC), wherein the synchronization circuit (FFSYNC) is configured not to be reset by the first reset signal (PADRST).

6. System on chip according to claim 5 taken in combination with claim 3, wherein the first domain (Vil) is configured to communicate the data to be transferred (PWDAT, PADD) on an input of the start register (RDEP) and to communicate the control signal (W_pls) to the synchronization circuit (FFSYNC) at least one cycle of the first clock signal (CLK11) after said data, and wherein the start register (RDEP) is configured to capture the data to be transferred (PWDAT, PADD) in a single cycle of the first clock signal (CLK11) and then retain said data (accDAT, accADD) independently of said input.

7. System on chip according to one of the preceding claims, wherein a control logic circuit (FSM) is configured to generate a transfer status signal (PRDY) during the first cycle of the first clock signal (CLK11) of the transfer of data from the first domain to the second domain, the start register(s) (RDEP, RDEP2) being configured to capture said data (PWDAT, PADD) or said handshake conditioning signal (flg 11) on command of the transfer status signal (PRDY_fe), so that said one cycle of the first clock signal (CLK11) is this first cycle.

8. System on chip according to claim 7, wherein the control logic circuit (FSM) is configured to provide the shifted transfer state signal (PRDY_fe_d) by at least one cycle of the first clock signal (CLK11) to the synchronization circuit (FFSYNC) and adapted to trigger the transmission of said control signal (W_pls, W_pls_sync).

9. A method of transferring data between at least a first digital domain (Vil) and a second digital domain (V22) of a system on chip (SOC), comprising transferring data (PWDAT, PADD, CDC11) from at least one transfer element (RDEP, ADPT1, ADPT2, RDEP2) in the first domain to the second domain, wherein a first reset signal (PADRST) causes a reset of the first domain (Vil) but does not cause a reset of said at least one transfer element (RDEP, ADPT1, ADPT2, RDEP2); wherein said transfer comprises signal adaptation (ADPT1) from a signal adaptation element (ADPT1) among said at least one transfer element, comprising transferring a handshake signal (CDC11, CDC22) between the first domain and the second domain;and wherein said at least one signal adaptation (ADPT2) comprises entering, in a second start register (RDEP2), the value of a handshake conditioning signal (flgll) in a single cycle of said first clock signal (CLK11) and retaining said value thereafter (acc_flg).;

10. Method according to claim 9, wherein a second reset signal (POR) resets the second domain (V22), and said at least one transfer element (RDEP, ADPT1, ADPT2, RDEP2).

11. Method according to one of claims 9 or 10, in which the first domain (Vil) is clocked by a first clock signal at a first frequency (CLK11) and the second domain (V22) is clocked by a second clock signal at a second frequency (CLK22) lower than the first frequency.

12. The method of claims 10 and 11, wherein the first reset signal (PADRST) originates from outside the system on chip (SOC) asynchronously with the first domain clock signal (CLK11), and the second reset signal (POR) is generated during a synchronous reset process at the start of the system on chip (SOC).

13. Method according to one of claims 9 to 12, wherein said transfer comprises a data transfer (PWDAT, PADD) made from a starting register (RDEP) among said at least one transfer element, to a destination register (RDEST) in the second domain, on command of a control signal (W_pls_sync) transmitted by a synchronization circuit (FFSYNC) of the first domain, the first reset signal (PADRST) not causing a reset of a synchronization circuit (FFSYNC).

14. Method according to claim 13 taken in combination with claim 11, comprising a communication of the data to be transferred (PWDAT, PADD) on an input of the starting register (RDEP) and a communication of the control signal (W_pls) to the synchronization circuit (FFSYNC) at least one cycle of the first clock signal (CLK11) after said data, and in which the transfer of the data (PWDAT, PADD) comprises a capture, by the starting register (RDEP), of the data to be transferred (PWDAT, PADD) in a single cycle of the first clock signal (CLK11) and a conservation of said data thereafter (accDAT, accADD), independently of said input.

15. Method according to one of claims 9 to 14, comprising a generation of a transfer status signal (PRDY) during the first cycle of the first clock signal (CLK11) of the transfer of the data from the first domain to the second domain, said entry in the start register(s) (RDEP, RDEP2) being controlled by the transfer status signal (PRDY_fe), so that said one cycle of the first clock signal (CLK11) is this first cycle.

16. Method according to claim 15, in which the shifted transfer state signal (PRDY_fe_d) is supplied by at least one cycle of the first clock signal (CLK11) to the synchronization circuit (FFSYNC) and triggering the transmission of said control signal (W_pls, W_pls_sync).