Circuit and Method of Indicating Data Hold-Time

The present invention relates in general to digital timing circuits and, more particularly, to a data hold-time indicator circuit that produces a recurring pulse width that reflects the data hold-time of an embedded register.

Parallel-to-serial converters are commonly used in digital circuit design to convert multi-bit signals to a string of data bits that are serially transmitted one at a time. One common application involves a data register embedded within an integrated circuit that periodically receives new data sourced by external logic. Timing generation logic for loading the data register is also embedded within the integrated circuit. The timing generation logic asserts a periodic signal to the external logic requesting more data be presented to the data register.

Many applications involve high speed operation, say in the gigahertz range. The data transaction must be completed within a predetermined time period. The data must be present and valid for a setup time before the data is loaded into the register by a clock signal. In addition, the data must remain present and valid for a hold-time after the transition of the clock signal. Unfortunately at such high data rates, the propagation delay uncertainty of the digital signals from the external sourcing logic are almost as long as the entire transaction period.

When the periodic signal is asserted to request more data, the external logic begins the time-consuming process of retrieving new data. When the external logic finally presents new data to the integrated circuit, the new data typically propagates through buffer logic and eventually reaches the data register. The timing generation logic asserts a clock signal to load the data register. When the data transaction is so fast that propagation delay uncertainties consume almost the entire time period, there is no guarantee that data arrives at the data register within register setup and hold-time constraints.

Since the data register and timing logic are embedded within the integrated circuit, it is difficult to directly measure the actual setup and hold-time. That is, the setup and hold-time are not readily observable by the external logic. If the data setup and hold-time cannot be observed, the data rate of the external sourcing logic must be reduced to ensure sufficient setup and hold-time. The data rate of the external sourcing logic cannot be maximized unless the actual setup and hold-time of the register is known. Otherwise, where the propagation time uncertainty consumes a large portion of the transaction time period, the data transaction may fail to correctly time the data transfer under a worst-case timing analysis.

Hence, a need exists for a circuit that provides data setup and hold-time information for embedded circuits that are otherwise unobservable.

FIG. 1 is a block diagram illustrating a parallel-to-serial converter; and

FIG. 2 is a schematic diagram illustrating the data hold-time indicator circuit of FIG. 1.

Referring to FIG. 1, a parallel-to-serial converter 10 is shown suitable for manufacturing as an integrated circuit using conventional integrated circuit processes. A CMOS-ECL voltage translator circuit 12 receives a 32-bit DATA IN word from external sourcing logic (not shown) operating at CMOS logic levels. CMOS-ECL voltage translator circuit 12 provides a 32-bit ECL logic level output signal to 32-bit register 14. Register 14 loads data at rising edge of a REGISTER CLOCK signal. Multiplexer 16 rotates through the individual bit locations of register 14 under control of the MUX ADDRESS signal and provides serial bits to the data input of flipflop 18. Flipflop 18 transfers the serial data signal to DATA OUT at its Q-output upon receiving a FLIPFLOP CLOCK signal. Data hold-time indicator circuit 22 receives the least significant D0 bit from CMOS-ECL voltage translator circuit 12 and the REGISTER CLOCK signal for providing a DATA HOLD signal that indicates the data hold-time as the time between the rising edge of the REGISTER CLOCK signal and the transition edge of the D0 bit.

Timing logic 20 operates in response to a SYSTEM CLOCK signal, operating for example at 2.5 gigahertz, for providing the FLIPFLOP CLOCK signal and the REGISTER CLOCK signal. The FLIPFLOP CLOCK signal operates in phase and at the same frequency as the SYSTEM CLOCK. The REGISTER CLOCK signal and the WRITE are derived from dividing the SYSTEM CLOCK by value 32. The MUX ADDRESS is reset to zero with each REGISTER CLOCK and counts up with the SYSTEM CLOCK to value 32. Thus, the serialized data stream DATA OUT is sent at the SYSTEM CLOCK frequency with the duration of each serial bit time the same as the period of the SYSTEM CLOCK. Timing logic 20 further asserts a WRITE control signal every thirty-two SYSTEM CLOCK signals. ECL-CMOS translator 24 converts the WRITE control signal from ECL logic levels to CMOS logic levels for the external logic to send the next DATA IN word. Timing logic 20 includes combination logic to divide the SYSTEM CLOCK and reset the MUX ADDRESS signal. Such combinational logic can be implemented from the aforedescribed operations.

It is important for the overall circuit operation that the output signal from REGISTER 14 does not become metastable. The 32-bit data at the output of CMOS-ECL voltage translator circuit 12 must be stable for a finite "setup time" before the rising edge of REGISTER CLOCK. Likewise, the 32-bit data must remain stable for a finite "hold-time" after the rising edge of REGISTER CLOCK. Any violation of setup and hold-time may cause the REGISTER 14 output to become metastable, yielding indeterminate logic levels for an indeterminate time duration.

Accordingly, as a feature of the present invention, data hold-time indicator circuit 22 provides a DATA HOLD signal that represents the actual data hold-time as the time between the rising edge of the REGISTER CLOCK signal and the transition edge of the D0 bit. Thus, the DATA HOLD signal provides a mechanism of determining the actual data hold-time of register 14 without directly observing data transitions within the embedded circuit. Once the actual data hold-time is accurately determined, the external sourcing logic can be set to provide the DATA IN signal as fast as parallel-to-serial converter 10 can handle it given the required set-up and hold-time of register 14.

Turning to FIG. 2, further detail of data hold-time indicator circuit 22 is shown including flipflop 26 having a data input coupled for receiving the D0 bit from CMOS-ECL voltage translator circuit 12. The Q-output of flipflop 26 is coupled to a first input of OR gate 28. The D0 bit is applied to the reset input of flipflop 30. The D0 bit is further inverted by inverter 32 and applied to the reset input of flipflop 26 and to the data input of flipflop 30. The REGISTER CLOCK signal is applied to the clock inputs of flipflops 26 and 30. The Q-output of flipflop 30 is coupled to a second input of OR gate 28 for providing the DATA HOLD signal at its output.

The data hold-time is typically determined during test. The D0 bit is toggled between zero and one with successive 32-bit DATA IN test vectors to determine the data hold-time for the integrated circuit given the process technology. Assuming the D0 bit starts as logic one. When REGISTER CLOCK rises from logic zero to logic one, the Q-output of flipflop 26 goes to logic one. At the same time, the logic one D0 bit resets flipflop 30 and produces a logic zero at the second input of OR gate 28. The logic one from flipflop 26 causes the DATA HOLD signal to go to logic one. The rising transition of the DATA HOLD pulse reflects the time when the REGISTER CLOCK goes high. When the D0 bit goes to logic zero, the output of inverter 32 goes to logic one and resets the Q-output of flipflop 26 to logic zero. However, the logic one at the output of inverter 32 does not propagate through flipflop 30 until the next rising edge of the REGISTER CLOCK signal. The logic zero at the first input of OR gate 28 causes the DATA HOLD signal to go to logic zero. The trailing edge of the DATA HOLD pulse reflects the time at which the data changed at the input to register 14. Thus, the overall pulse width of DATA HOLD represents the data hold-time of register 14 as the time between the rising edge of the REGISTER CLOCK signal and the transition edge of the D0 bit

Alternately, if the D0 bit starts as logic zero, the data input receives a logic one from inverter 32. The Q-output of flipflop 30 goes to logic one at the rising edge of the REGISTER CLOCK signal. At the same time, the logic one from inverter 32 resets flipflop 26 and produces a logic zero at the first input of OR gate 28. The logic one from flipflop 30 causes the DATA HOLD signal to go to logic one. The rising transition of the DATA HOLD pulse reflects the time when the REGISTER CLOCK goes high. When the D0 bit goes to logic one, the Q-output of flipflop 30 is reset to logic zero. However, the logic one D0 bit does not propagate through flipflop 26 until the next rising edge of the REGISTER CLOCK signal. The logic zero at the second input of OR gate 32 causes the DATA HOLD signal goes to logic zero. The trailing edge of the DATA HOLD pulse reflects the time at which the data changed at the input of register 14. Again, the overall pulse width of DATA HOLD represents the data hold-time of register 14 as the time between the rising edge of the REGISTER CLOCK signal and the transition edge of the D0 bit

The data hold-time determination is useful during design and development where the DATA HOLD signal may be observed with an oscilloscope. If the DATA HOLD pulse is much narrower than one half of the total period of the SYSTEM CLOCK, then there is too little hold-time and too much setup time. A delay circuit may be added in the external sourcing logic for the WRITE control signal to delay application of DATA IN and balance the observed setup and hold-time. Likewise, for too much hold-time and too little setup time, the solution may optimize external logic to eliminate delays, or reduce the operating frequency, to balance setup and hold-time.

By now it should be appreciated that the present invention provides a circuit and method of indirectly observing the timing relationship of the data and clock applied to a register embedded within an integrated circuit. The invention generates a recurring pulse having a width that reflects the data hold-time at the input to the embedded register. By indirectly observing the timing relationship, the externally-sourced data timing can be calibrated to meet the setup and hold-time requirements of the embedded register.

While specific embodiments of the present invention have been shown and described, further modifications and improvements will occur to those skilled in the art. It is understood that the invention is not limited to the particular forms shown and it is intended for the appended claims to cover all modifications which do not depart from the spirit and scope of this invention.