Variable delay element for use in delay tuning of integrated circuits

This application is a continuation of application Ser. No. 08/989,917, filed Dec. 12, 1997 now U.S. Pat. No. 6,628,157, the contents of which are incorporated herein by reference.

1. Field of the Invention

The present invention relates to a method and apparatus that facilitates delay tuning of integrated circuits and, in particular, to a method and apparatus for providing a software implementation of delay tuning in integrated circuits.

2. Description of the Prior Art

Timing analysis is a critical element of the design of an integrated circuit. Vendors of integrated circuits often standardize the performance of their circuits to ensure that they meet a standardized timing scheme. The scheme may be the subject of a market-wide standard or may be established by the vendor itself. Often, the vendors distribute timing diagrams with their products to permit purchasers to integrate the products into larger designs. Thus, uniform, consistent timing performance is critical to market success of integrated circuits.

Consider the circuit of FIG. 1 , for example. There, a microprocessor (μP) generates an address signal on, for example, sixteen (16) parallel address lines, labeled AØ-A 15 . Due to characteristics of the circuitry that generates the address signals, the signals may be skewed: They may not be established in unison. Certain lines may become active before certain others. This performance may cause the microprocessor to deviate from the requirements of an established timing protocol.

Each address signal A 0 -A 15 may be generated by the circuit of FIG. 2 . There, a master slave flip flop 210 receives a data signal (the address data) generated from within the microprocessor. The flip flop 210 captures the data signal when triggered by a clock signal 220 and outputs the data signal thereafter. Typical trigger signals are the rising edge or the falling edge of the clock. Because the data signal may not be established at the flip flop input when the triggering signal is received from the clock, a delay circuit 240 may be added to delay propagation of the clock to the flip flop.

Excessive delays contribute to loss of data. New data may is overcome old data before the old data is clocked into the flip flop 210 . Accordingly, a latch 200 is provided between the source of the data signal and the flip flop. The latch 200 itself is clocked by a non-delayed clock signal 200 . Under this scheme, old data may be clocked into the flip flop 210 before new data is input to the latch 200 .

During circuit design, an amount of desired delay is unknown. Therefore, delay circuits are tuned. FIG. 3 illustrates the configuration of a known delay circuit 240 for a single data line. There, the circuit provides a plurality of alternate paths 242 - 246 for a data signal. Each path provides a different number of delay buffers in the path. Each delay buffer imposes an incremental amount of delay upon the signal. Thus, the selection of a path determines how much delay is imposed on the signal.

Through trial and error, technicians alternately direct the signal through each of the delay paths 242 - 246 to determine how much delay is necessary to meet desired timing requirements. The trial and error procedure is implemented in a physical prototype of the integrated circuit. To direct the signal through a particular delay path, a technician must establish a physical electrical connection to test a first path, test it, then destroy the connection and establish a second connection to test a second path. Once a preferred delay path is identified for the particular signal, the selected delay path becomes part of the design of the integrated circuit.

A single integrated circuit may have several hundred delay circuits. Every delay circuit within the integrated circuit must be “tuned” through the trial and error process described above. Delay tuning, therefore, materially increases the time and expense of integrated circuit manufacture.

Accordingly, there is a need in the art for a delay tuning process that reduces time and expense associated with delay tuning. Further, there is a need for such a process that eliminates the need for providing physical interconnections between a signal and the alternate delay paths that the signal must traverse for testing.

An embodiment of the present invention provides a delay element that receives an input signal and imposes a variable delay upon the signal. The delay element includes a plurality of delay stages interconnected in a cascaded relationship. Each stage imposes an incremental delay upon the input signal when enabled. The delay element receives a selection signal that determines how many of the delay stages are enabled.

FIG. 1 illustrates a prior art microprocessor having a plurality of address line outputs.

FIG. 2 illustrates a known latch-flip flop circuit.

FIG. 3 illustrates delay paths used in the prior art.

FIG. 4 is a block diagram of a latch-flip flop circuit according to an embodiment of the present invention.

FIG. 5 is a block diagram of a delay circuit according to an embodiment of the present invention.

FIG. 6 is a block diagram illustrating operation of the delay circuit of FIG. 5 in response to a first predetermined SELECT input.

FIG. 7 is a block diagram illustrating operation of the delay circuit of FIG. 5 in response to a second predetermined SELECT input.

FIG. 8 is a block diagram illustrating operation of the delay circuit of FIG. 5 in response to a third predetermined SELECT input.

FIG. 9 is a block diagram of a delay circuit according to a second embodiment of the present invention.

FIG. 10 is a circuit diagram of a delay element 100 according to an embodiment of the present invention.

FIG. 11 is a schematic diagram of a system for delay tuning an integrated circuit constructed in accordance with the present invention.

Turning to FIG. 4 , there is shown a data circuit 10 according to an embodiment of the present invention. The data circuit 10 may be used in an integrated circuit, such as the microprocessor of FIG. 1 , in place of the circuit of FIG. 2 . The data circuit 10 provides a conventional latch 200 and master slave flip flop 210 . A data signal generated from a source internal to the integrated circuit 10 (not shown) is input to the latch 200 . A clock signal 220 is input directly to the latch 200 . From the latch, the data is input to the master slave flip flop 210 . The clock 220 propagates through a variable delay element 100 to the flip flop 210 . The delay element 100 provides variable delay to the propagation of the clock signal 220 to the flip flop 210 . The variable delay is adjusted according to the value of a multibit SELECT signal 230 input to the delay element 100 .

FIG. 5 is a block diagram of an embodiment of the delay element 100 . The delay element 100 accepts an input signal at input 110 . The delay element 100 is populated by a plurality of “stages” 120 A-C, each imposing a predetermined same or different amount of delay upon the input signal. Each stage is populated by an inverter buffer and a delay block. The inverter buffer and delay blocks impose incremental delay upon the input signal. The delayed input signal is output from the delay element 100 at output 130 .

In one embodiment, the orientation of the inverter buffer and delay blocks of each stage may vary by stage. In a first stage 120 A, a first input of delay block 140 A is coupled to the input terminal 110 . The signal from input terminal 110 also passes through the first stage 120 A to a next stage such as 120 B. A second input of the delay block 140 A is coupled to an output of another stage 120 B. An output of delay block 140 A is input to inverter buffer 150 A. An output of buffer 150 A is output from the delay element 100 at terminal 130 .

Intermediate stages, such as stage 120 B, receive an input from a previous stage 120 A which is input to the inverter buffer 150 B. An output of buffer 150 B is input to a first input of the stage's delay block 140 B and also output from the stage 120 B to the next successive stage 120 C. A second input to the delay block 140 B is returned from the next stage 120 C. An output of the delay block 140 B is output from the stage 120 B to the previous stage 120 A.

A last stage 120 C is configured somewhat similarly to the intermediate stage 120 B. The input signal is input to an inverter buffer 150 C. The delay block 140 C of the last stage 120 C has no second input. An output from the inverter buffer 150 C is input to a delay block 140 C. An output of the delay block 140 C is returned to the previous stage 120 B.

The SELECT signal 230 is a multibit signal. In the embodiment of FIG. 5 , it provides a bit for every stage in the delay element 100 . The SELECT signal 230 is a three bit code in the illustrated embodiment. Each bit determines how the associated delay block 140 A, 140 B, 140 C is configured. When the SELECT bit is a “1,” the delay block 140 A, 140 B, 140 C switches to the signal input at terminal IN 1 . When the SELECT signal 230 is a “Ø,” the delay block 140 A, 140 B, 140 C switches to the signal input at terminal IN 2 . Of course, because delay block 140 C receives only one input, the delay block 140 C may be permanently switched to the first input; the select bit may be omitted in this case.

FIG. 5 illustrates a three stage embodiment, populated by only three delay blocks. However, the delay element 100 may provide as many delay stages as may be desired. For example, at the time of this writing, Intel Corporation, the assignee of the present invention, is considering an eight (8) stage delay element 100 for use in an integrated circuit.

FIGS. 6-8 illustrate paths taken by the input signal depending upon the value of the SELECT signal:

As is shown in FIG. 6 , when the SELECT signal 230 provides a “1” to delay block 140 A, the delay block 140 A routes the input signal from input terminal IN 1 directly to the output inverter buffer 150 A. The delay block 140 A ignores any input to terminal IN 2 . Thus, the values of the SELECT signal 230 as input to delay blocks 140 B and 140 C are immaterial to the performance of the delay element 100 . Delay block 140 A and output inverter buffer 150 A impose a minimum delay on the input signal. When the SELECT signal 230 provides a “1” to delay block 140 B and a “Ø” to delay block 140 A, the input signal traverses the path highlighted in FIG. 7 . There, the input signal is delayed by inverter buffer 150 B, delay blocks 140 B and 140 A, and output inverter buffer 150 A. When the SELECT signal 230 provides a “1” to delay block 140 C and a “Ø” to delay blocks 140 A and 140 B, the input signal traverses the path highlighted in FIG. 8 . There, the input signal is delayed by inverter buffers 150 B and 150 C, by delay blocks 140 C-A, and by inverter buffer 150 A.

As noted above, each stage incrementally increases the delay imposed upon the input signal. Additional stages permit the delay element 100 to impose additional delay upon the input signal.

In another embodiment, as is shown in FIG. 9 , the SELECT signal 230 need not provide a separate bit for each delay block. Instead, a decoder 240 receives the SELECT signal 230 and, responsive to the value of the SELECT signal, controls operation of the delay blocks of each stage. The SELECT signal 230 remains a multibit signal that possesses a unique value associated with each path available in the delay element 100 . For example, for an eight stage delay element, the SELECT signal 230 may be a three bit signal.

FIG. 10 is a circuit diagram of an embodiment of a single delay block 140 of FIG. 5 . The delay block is a tri-state multiplexer that chooses one of two inputs based on a locally decoded SELECT bit. The delay block 140 is populated by eight transistors. It provides two selection PMOS transistors, T 1 and T 2 , each coupled to a high voltage source such as +5 volts (often termed “VCC”). Additionally, the delay block 140 provides two selection NMOS transistors, T 3 and T 4 , coupled to a low voltage source, such as ground. As is known in the context of binary circuits, PMOS transistors become conductive in response to a low voltage input; NMOS transistors become conductive in response to a high voltage input.

Input IN 1 is input to a pair of transistors T 5 and T 6 . Transistor T 5 is a PMOS transistor that, in combination with transistor T 1 , couples VCC to the output terminal OUT. The leg formed by transistors T 1 and T 5 outputs a high voltage signal (“1”) when IN 1 is low (“Ø”) and the SELECT bit is high (“1”). The second transistor T 6 , in combination with transistor T 3 , couples the low voltage source to the output terminal OUT. The leg formed by transistors T 3 and T 6 drive the output terminal low (“Ø”) when IN 1 is high (“1”) and the SELECT bit is high (“1”).

Similarly, input IN 2 is input to a pair of transistors T 7 and T 8 . Transistor T 7 is a PMOS transistor that, in combination with transistor T 2 , couples VCC to the output terminal OUT. The leg formed by transistors T 2 and T 7 outputs a high voltage signal (“1”) when IN 2 is low (“Ø”) and the SELECT bit is low (“Ø”). The second transistor T 8 , in combination with transistor T 4 , couples the low voltage source to the output terminal OUT. The leg formed by transistors T 4 and T 8 drive the output terminal low (“Ø”) when IN 2 is high (“1”) and the SELECT bit is low (“Ø”).

The block 140 defines its operation based upon the value of the SELECT input signal. When SELECT=“Ø” the delay block 140 outputs a signal that represents an inversion of the IN 2 input ({overscore (IN 2 )}). When SELECT=“l,” the delay block 140 outputs IN 1 inverted ({overscore (IN 1 )}). Switching of the transistors of the delay block is demonstrated in the following table:

SELECT T1 T2 T3 T4 OUT ∅ Off On Off On {overscore (IN2)} 1 On Off On Off {overscore (IN1)}

Although PMOS and NMOS transistors are shown in FIG. 10 , the present invention finds application with transistors of other types.

The delay element of the present invention eliminates any need for manual delay tuning of integrated circuits. The delay element 100 substitutes for known delay circuits within the integrated circuit. To tune the delay element, different SELECT signals may be applied to the delay element 100 over an appropriate electrical interface. To change delay in the delay element 100 , only the value of the SELECT signal need be changed. Physical interconnections need not be changed.

In an embodiment of the present invention, an integrated circuit that includes delay elements may be tuned by a computer. Such a system is shown in FIG. 11 . Consider an integrated circuit populated by n delay elements 100 a-n , driving output lines 130 a-n . There, a computer 300 is connected to each of the delay elements 100 a-n by an electrical interface 310 . The interface 310 drives SELECT lines 230 a-n for each of the delay elements 100 a-n . The computer 300 senses the output of the delay elements 100 a-n over sensory lines 330 a-n coupled to the computer via a second interface 320 .

To delay tune a delay element 100 a , the computer 300 may vary the SELECT signal 230 a and measure delay associated with each SELECT setting via sensory line 330 a . Once each setting is tested, a preferred setting that causes the delay to fall within a predetermined window may be detected.

After testing, when the desired SELECT code is identified, the SELECT signal is hardwired into the integrated circuit. That is, within the code, the SELECT bits that are a “ 1 ” are electrically connected to VCC; the SELECT bits that are a “Ø” are tied to ground. The integrated circuit may be mass produced with this hardwired implementation of the SELECT code.

Delay tuning also may be performed dynamically within in a circuit. Rather than hardwiring the SELECT signal to predetermined voltage sources, the SELECT signal may be coupled to a controller or state machine that senses when delay falls outside a preferred operating condition. To compensate, when it is detected that the delay is less than desired, the controller may vary the SELECT signal to increase the delay. When the delay is more than desired, the controller may vary the SELECT signal to reduce the delay.

The present invention provides a mechanism for use in skew compensation. When two data signals are to be generated synchronously, one of the signal may exhibit a natural delay that does not occur in another. In such a case, the signal with less natural delay may be configured to propagate through the delay element 100 . By varying the SELECT signal, the two data signals may be brought into the desired synchronism.

The present invention has been described in the context of “data” signals and “clock” signals. It should be understood that such labels are provided to facilitate the presentation of the present invention. The principles of the present invention find application where it is desired to delay the propagation of any signal, regardless of the substantive information that the signal represents.

Several embodiments of the present invention are specifically illustrated and described herein. However, it will be appreciated that modifications and variations of the present invention are covered by the above teachings and within the preview of the appended claims without departing from the spirit and intended scope of the invention.