Patent Application
20080204090
1. Field of the Invention
This invention relates generally to a circuit and method for using a stable asynchronous clock source to perform glitch-free regeneration of a clock for a synchronous interface that is not free from glitches.
2. Description of Related Art
The fields of application specific integrated circuits (ASIC) using field programmable gate arrays (FPGA), the design of such circuits, synchronous interfaces between circuit boards, telecommunications, and digital subscriber line access multiplexers (DSLAM), including redundant DSLAMs, are known. Likewise, the use of network termination (NT) cards and Utopia level two (L2) interfaces between two redundant NT cards is similarly known.
One known use for standard Utopia L2 interfaces between two redundant NT cards is to pass asynchronous transfer mode (ATM) traffic between two redundant NT cards. Such systems establish a synchronous interface between the two redundant NT cards. An example of such a system is an NT card redundant internet protocol (IP)/ATM DSLAM. Using the example of the foregoing systems, it is desirable to incorporate a glitch free clock regenerator.
Any of the foregoing objects and advantages are illustrative of those that can be achieved by various exemplary embodiments. They are not intended to be interpreted as required of all possible embodiments, and they are not intended to be exhaustive or limiting of the possible advantages which can be realized. Thus, these and other objects and advantages of various exemplary embodiments will be apparent from the description herein or can be learned from practicing various exemplary embodiments, both as embodied herein or as modified in view of any variation which may be apparent to those skilled in the art. Accordingly, the present invention resides in the novel methods, arrangements, combinations and improvements herein shown and described in various exemplary embodiments.
In light of the present need for a glitch-free clock regeneration circuit, a brief summary of various exemplary embodiments is presented. Some simplifications and omission may be made in the following summary, which is intended to highlight and introduce some aspects of various exemplary embodiments, but not intended to limit the scope of the invention.
When an NT card is physically removed from a redundant IP/ATM DSLAM system, the clock source it was providing is often subjected to an undesirable glitching that affects the input of the NT remaining in the system. As referred to herein, a “glitch” or “glitching” means false or spurious events in an electronic signal. Sometimes interface logic exists that is designed to address protocol violations and/or data errors on the synchronous interface. Typically, this interface logic assumes that the clock is healthy or otherwise free of glitches. Consequently, the presence of a glitch in the clock input can result in unreliable behavior in the NT card or cards that were not removed physically from the system. This undesirable behavior of the remaining NT card(s) may even rise to the level of an unrecoverable lock up condition in the system. An unrecoverable lock up condition in the system often requires the NT to be reset in order for proper operation to resume.
A similar problem exists when two blocks with independent resets interface synchronously. Sometimes these two blocks are on the same card. If the reset of one of the blocks asynchronously affects the clock to the system, the problems described above can also result for these reasons.
The problems described above are issues of unreliable clock sources. In various exemplary embodiments, unreliable clock source issues are resolved by passing an incoming clock signal through a phase lock loop (PLL). Passing an incoming clock through a PLL results in a “clean” clock signal on the receiving side of the synchronous interface. However, if the clock source is lost, typically the PLL will lose its lock. An indication in such a system that the PLL lost its lock is used in various exemplary embodiments to reset the logic that is running on the clock from the PLL. In various exemplary embodiments, when the clock source is re-applied, the PLL correspondingly re-achieves a lock. In various exemplary embodiments, when the PLL re-achieves a lock, a reset is removed and the circuit continues normal operation.
The embodiments described above have associated problems. For example, FPGA vendors typically specify that once a PLL loses lock, a corresponding output clock is no longer reliable unless the PLL is reset prior to an attempt to re-achieve a lock. In various exemplary embodiments, the reset activity described above requires software intervention. A need for such software intervention is undesirable.
Another problem associated with the various embodiments described above is that PLL resources in FPGAs are not limitless. Thus, in some instances, the use of a PLL will not be available because all of the resources on a device are already in use. This problem is believed to be particularly prevalent in FPGA devices from low cost FPGA families.
Another problem associated with the various exemplary embodiments described above is that PLLs often have a minimum input frequency. Such a requirement creates problems when attempting to implement an embodiment with a low-speed clock.
In various exemplary embodiments, an asynchronous stable clock source is used to perform regeneration of an incoming clock wherein the incoming clock contains glitches and the regenerated clock does not contain glitches. In various exemplary embodiments, such an asynchronous stable clock source is used in a synchronous interface. In various exemplary embodiments, an asynchronous stable clock source is used to regenerate an input clock to an FPGA or ASIC wherein the clock is sourced off-board or from an otherwise unreliable clock source. Thus, in various exemplary embodiments, an asynchronous stable clock source is implemented in a system where ATM traffic passes between two redundant NTs and a DSLAM using a Utopia L2 interface.
Accordingly, a clock regeneration circuit is provided that is free from glitches. The implementation of a synchronous clock for external interfaces subject to clock instability is beneficial when the synchronous clock does not contain glitches. Thus, in various exemplary embodiments, a clock is provided that is synchronous to an incoming clock source that contains glitches.
In various exemplary embodiments, an output clock source and synchronous interface signals are directly fed into an off-the-shelf interface IP core. In various exemplary embodiments, this is achieved because an interface protocol is maintained.
In various exemplary embodiments, the implementation of the clock regeneration circuit that is free of glitches is advantageous over other DSLAMs because it facilitates the use of synchronous interfaces across inter-PCB (Printed Circuit Board) interfaces or interfaces straddling reset boundaries. In such applications, the synchronous interface's clock often becomes unreliable. In various exemplary embodiments, a clock regeneration circuit free of glitches is combined with various embodiments of zero clock delay meta-stability filters. In various exemplary embodiments, such circuits provide a complete meta-stability protection solution for inter-PCB synchronous interfaces.
In various exemplary embodiments, where DSLAMs that do not include clock regeneration protection that is free of glitches on synchronous interfaces that cross PCB boundaries suffer from problems because a reset, extraction or insertion occurs on the block or PCB containing the clock source for the synchronous interface, the other block or PCB on the synchronous interface also requires a reset before proper operation is able to resume.
It should be apparent to those having an ordinary skill in the art that various exemplary embodiments are not implemented in DSLAMs. Rather, the subject matter described above and below is implemented in various exemplary embodiments in any form of a product that includes a synchronous interface.
Accordingly, various exemplary embodiments include clock regeneration circuits that do not contain glitches. Likewise, various exemplary embodiments include a method to prevent a meta-stability event on a synchronous interface where the event results from a manual extraction or insertion of a PCB in a system resulting in a clock instability.
In order to better understand various exemplary embodiments, reference is made to the accompanying drawings, wherein:
Referring now to the drawings, in which like numerals refer to like components or steps, there are disclosed broad aspects of various exemplary embodiments.
Exemplary clock regeneration circuit 100 includes exemplary flip flop F1, exemplary flip flop F2, and exemplary flip flop F3. Hereinafter flip flops F1, F2, F3 and other flip flops will be referred to in shorthand as “flops.” Collectively flops F1 and F2 constitute a filtering circuit wherein synchronous clock signal T1 is filtered.
In various exemplary embodiments, asynchronous clock signal T2 has a frequency much higher than synchronous clock signal T1. In this initial stage of exemplary clock regeneration circuit 100, synchronous clock signal T1, which is associated with the interface, is over-sampled and rate adapted through the meta-stability filtering circuit consisting of flops F1, F2 to asynchronous clock signal T2. This over-sampling of synchronous clock T1 is achieved by taking multiple samples of synchronous clock signal T1 before synchronous clock signal T1 is able to transition and repeat. In various exemplary embodiments, the rate adaption of synchronous clock T1 eliminates meta-stability events from synchronous clock T1 by filtering such meta-stability events out of incoming synchronous clock signal T1.
Following the meta-stability filtering circuit with flops F1, F2, flop F3 in combination with inverter I1 and and-gate A1 edge detect the rate adapted clock signal output from the meta-stability filtering circuit. In the embodiment of exemplary clock regeneration circuit 100, the rising edge of the rate adapted clock signal is edge detected by the edge detection circuit consisting of flop F3, inverter I1 and and-gate A1.
The resulting signal is then delayed by subsequent clock cycles of the T2 clock to generate delayed rising edge detect signals L1, L2 and L3 output respectively from flops D1, D2 and D3. In various exemplary embodiments, flops D1, D2, D3 are omitted. In the exemplary embodiment of clock regeneration circuit 100, delayed rising edge detected signals L1, L2, L3 illustrate an embodiment wherein exemplary clock regeneration circuit 100 is implemented in connection with a zero clock delay meta-stability filter. In various exemplary embodiments, additional flops are included in series following flop D3 with corresponding additionally delayed edge detect signals following delayed edge detect signal L3.
According to the foregoing, a top row of electrical components depicted in exemplary schematic of clock regeneration circuit 100 consists of multiple stages. A first stage is a meta-stability filtering circuit. A second stage is a rising edge detector circuit. A third stage is a series of delay circuits that delay an output of the rising edge detect circuit.
Exemplary clock regeneration circuit 100 further includes a clock regeneration stage at and-gate A2. In exemplary clock regeneration circuit 100, and-gate A2 has three inputs coming respectively from the outputs of flops D3, D4 (inverted at inverter I2) and flop D5 (inverted at inverter I3). The output of and-gate A2 thus constitutes a regenerated clock signal that is free of glitches that may be present in input clock signal T1.
As is evident in exemplary clock regeneration circuit 100, the higher frequency asynchronous clock signal T2 is also used as the clock signal for flops D4, D5, D6, D7. The inverted output signals from flops D4 and D5 that are input to and-gate A2 prevent the output from and-gate A2 from becoming active for two clock cycles of higher frequency asynchronous clock signal T2 after it has already gone active. Consequently, in various exemplary embodiments, at least three clock cycles of higher frequency asynchronous clock T2 must pass for each cycle of the regenerated clock signal T3.
Incidentally, the output of the and-gate A2 is an advanced version of regenerated clock signal T3. Under normal operation, when synchronous clock signal T1 is a valid clock signal, the period of regenerated clock signal T3 will follow the period of synchronous clock signal T1.
The ultimate output of exemplary clock regeneration circuit 100 is regenerated clock signal T3 that is free from glitches and is synchronous to the incoming interface with the timing that is known relative to the interface and guaranteed to having an absolute maximum frequency of one-third the frequency of asynchronous clock signal T2. Regenerated clock signal T3 free of glitches is used in various exemplary embodiments to clock logic using input signals from the synchronous interface and to clock signals towards the synchronous interface without fear of having an unnecessarily short clock cycle that results in violating internal logic timing. Such violations of internal logic timing often result in functional failures of the device. Such meta-stability events are filtered by the regenerated clock signal T3 that is free of glitches.
Flops D4 and D5 have no effect on the output of and-gate A2 except to filter undesired glitches in synchronous clock signal T1. Absent the presence of undesirable glitches in synchronous clock signal T1, the output of and-gate A2 follows clock signal T1. The output of and-gate A2 does not have the same duty cycle as asynchronous clock signal T2. In other words, the output of and-gate A2 is only high for one clock cycle of asynchronous clock signal T2 and is low until the next rising edge of synchronous clock signal T1 causes and-gate A2 to again become high for a single clock cycle of asynchronous clock signal T2.
In various exemplary embodiments, exemplary clock regeneration circuit 100 is combined with a zero clock delay meta-stability circuit by including flops D1, D2, D3, D6, D7 and D8. Such embodiments achieve a synchronous interface that is protected against meta-stability events caused either by changes on the synchronous inputs or by glitching of the input clock.
According to the foregoing, flop D8 is the first register that runs on the re-generated clock. Flop D8 runs on the re-generated clock synchronously. Thus, S1_T3 is an output signal synchronous to T3. Similarly, input signal S1_T1 to flop D6 is an input signal synchronous to clock signal T1. In various exemplary embodiments, S1_T1 represents a bus of synchronous signals. Flops D6 and D7 are clock enabled (CE).
In various exemplary embodiments, the purpose and operation of flops D6, D7 and D8 are as follows. Flops D6 and D7 form a meta-stability filter for a synchronous bus of signals S1_T1. In various exemplary embodiments, flop D6 is omitted and flop D8 provides the second half of the meta-stability filter. In various exemplary embodiments, the purpose of flop D8 is to finally re-time the synchronous signals to the new regenerated clock domain.
In various exemplary embodiments, the clock regeneration circuit 100 is implemented within a device containing downstream receiving logic. In various exemplary embodiments, clock regeneration circuit 100 is implemented external to a device containing downstream receiving logic. It is believed that embodiments where clock regeneration circuit 100 is implemented external to a device containing downstream receiving logic are beneficial in cases where the receiving logic is implemented in a device that is off-the-shelf such as off-the-shelf ASIC.
The use of flop stages D4 and D5, along with inverters I2 and I3 control the maximum frequency for the re-generated clock, T3, in the event of glitching on the input clock T1. Specifically, the maximum frequency of regenerated clock T3 equals (1/(N+1))* the frequency of T2, where N is the number of such stages and is greater than zero. Thus, according to the exemplary embodiment depicted in
In another example, T2 is a 100 MHz clock signal and operational specifications for a receiving circuit recommend that the receiving circuit operate at a frequency equal to, or less than, 30 MHz. In this example, it would be desirable to use at least 3 stages (N=3) in an embodiment designed for such a use. This is true because such an embodiment would have a frequency for T3 of no greater than 25 MHz, which is equal to or less than the 30 MHz desired for the end receiving logic in this exemplary use.
The ability to control the maximum frequency of T3 is believed to be a beneficial design characteristic to consider when choosing between various exemplary embodiments when it is desirable to prevent the frequency of T3 from exceeding a maximum operating frequency of a device receiving the regenerated clock and the signals from the interface synchronous with T1, such as S1_T1, that have been re-synchronized to T3, such as S1_T3.
Based on the foregoing, it should be apparent that, in various exemplary embodiments, any additional combination of flop and inverter stages which feedback into A2 in a manner similar to the structure disclosed
Exemplary method 200 then passes to exemplary step 206. In exemplary step 206, the rate adapted clock signal is edge detected. In various exemplary embodiments, the rising edge of the rate adapted clock signal is edge detected in exemplary step 206. Exemplary method 200 then passes to exemplary step 208. However, it should be noted that various exemplary embodiments exclude step 208.
In exemplary step 208, the edge detected signal is delayed. As described above in connection with exemplary clock regeneration circuit 100, one or more cycles of the faster asynchronous clock signal is used to delay the edge detected signal in various exemplary embodiments. For example, in the exemplary embodiment of clock regeneration circuit 100, the edge detected signal is delayed by three cycles of faster asynchronous clock signal T2.
Exemplary method 200 then passes to exemplary step 210. In exemplary step 210, a regenerated clock signal is generated that is free of glitches. In exemplary step 210 the regenerated clock signal is generated from the delayed signal (or the edge detected signal) using one or more inverted outputs of one or more flops that run on the asynchronous clock signal. Exemplary method 200 then passes to exemplary step 212.
In exemplary step 212 generation of the next rising edge of the regenerated clock signal is suppressed. In various exemplary embodiments, the suppression lasts for a predetermined number of clock cycles of the asynchronous clock. In various exemplary embodiments the suppression is implemented by using the predetermined number of inverted outputs of a predetermined number of flops that also run on the asynchronous clock signal. Some such embodiments ensure the maximum operating frequency of the receiving logic is not exceeded as the result of a “glitch” on T1 resulting in a short period clock, also known as a runt period clock, event on T3. Greater detail of a particular exemplary embodiment of a circuit implementing one or more of the foregoing features is described herein in connection with
According to the foregoing, various exemplary embodiments of a method of generating a regenerated clock signal are achieved using any of the various exemplary embodiments of a clock regeneration circuit described herein. It should be apparent to those having an ordinary level of skill in the art that various exemplary embodiments of the method of generating a regenerated clock signal vary based on the variances described herein in connection with the various exemplary embodiments of a clock regeneration circuit.
Although the various exemplary embodiments have been described in detail with particular reference to certain exemplary aspects thereof, it should be understood that other embodiments exist including combinations of various aspects described in connection with different embodiments or figures, and the details are capable of modifications in various obvious respects. As is readily apparent to those skilled in the art, variations and modifications can be affected while remaining within the spirit and scope of the disclosure. Accordingly, the foregoing disclosure, description, and figures are for illustrative purposes only, and do not in any way limit the invention, which is defined only by the claims.
