Buffer interface architecture

This application claims priority from U.S. provisional application Ser. No. 60/287,674, filed May 1, 2001 and entitled “Buffer Interface Architecture.”

The present invention relates to integrated circuits, generally, and more specifically to a buffer for interfacing a low-voltage technology with a relatively high-voltage technology.

Advances in the semiconductor arts have driven devices to decreasing sizes operating at increasing speeds. This continuous effort to maximize the performance of integrated circuits (“ICs”) has produced several additional benefits, including decreased operating voltages and reductions in power consumption.

As MOS technology scales below 0.2 μm, acceptable supply voltages have lowered below the previous 3.3V and 5V standards. As lower and lower operating voltage IC technology is developed and commercialized, however, a distinct problem has arisen. Mostly because of economic reasons, electronic systems often use ICs that span several technology generations, each generation having different supply voltage requirements. The ability to interface newer low power ICs with their predecessors where each IC has a different range of operating voltages is of concern, particularly as it relates to metal oxide semiconductors (“MOS”). Interfacing an older higher operating voltage IC with a lower operating voltage technology may cause reliability issues and/or temporary or even permanent damage. For example, the buffer circuits of a 1.5V IC can neither provide nor sustain (when in a high impedance state) a 3.3V drive.

To overcome this interface problem, several solutions have been proposed. One approach entails the development of MOS devices capable of handling both low and high voltages on the same semiconductor substrate. While this “dual supply” approach is simple in circuit implementation, presently, it is substantially more expensive than the traditionally known MOS technology because of the additional processing steps required to fabricate the high-voltage devices. Currently, many 0.2 μm technologies utilize this “dual supply” approach.

Alternatively, several buffer interface architectures are also known in the art for providing high voltage drive capability using low voltage MOS technology. Using this methodology, the incremental costs associated with the additional circuitry required to realize an interface having high voltage drive capability while implemented in low voltage MOS technology are negligible.

Prior approaches to high-voltage drive buffers with low-voltage transistors (HVB/LVT) can be classified into two basic groups. FIG. 1A illustrates a circuit with both high-voltage tolerance and high-voltage drive. Such a circuit is proposed in U.S. Pat. No. 5,663,917 to Oka et al., the entirety of which is hereby incorporated by reference herein. FIG. 1B illustrates a circuit with high-voltage tolerance and low-voltage drive, such as may be found in M. Pelgrom and E. Dijkmans, “A ⅗V compatible I/O Buffer,” IEEE J. of Solid-State Circuits, vol. 30, No. 7, p.p. 823-825, July, 1995, the entirety of which is hereby incorporated by reference herein.

For purposes of circuit 10 of FIG. 1A, it is assumed that the breakdown voltage of the transistors used in the circuit is only slightly higher than ½ V HIGH —the voltage swing of the input signal. The circuit 10 of FIG. 1A includes a pad driver 12 which includes p-channel and n-channel cascode stacks, which include MOS devices P 1 , P 2 and N 1 , N 2 , respectively. The cascode transistors P 2 and N 2 allow the output at pad node 14 to traverse between 0V and V HIGH while the V GS 's (voltage gate to source) and V GD 's (voltage gate to drain) of all four transistors P 1 , P 2 , N 1 , N 2 remain lower than ½ V HIGH , and thus lower than the breakdown voltage of the transistors. The voltage capability of the pad driver 12 , therefore, is two times larger than the voltage capability of the MOSFETs used in the driver. Such a circuit may be referred to as a “2× driver.”

For proper operation, the cascode pad driver 12 requires two in-phase input signals at nodes 18 and 20 . Both signals must have a voltage swing that does not exceed ½ V HIGH in order to avoid exceeding the voltage capability of the transistors used therein. These signals are provided from the level shifter 16 to the driver 12 through two conventional inverter chains. The level shifter 16 takes a 0 to ½ V HIGH swing input data signal and produces a data signal that swings between ½ V HIGH and V HIGH at node 18 . Naturally, the level shifter 16 should be implemented in such a way that none of its transistors experience voltage overstress.

Unlike the circuit 10 of FIG. 1A, the circuit 20 of FIG. 1B is a high voltage buffer with low voltage transistors that is biased from a lower supply voltage ½ V HIGH and is characterized by high voltage tolerance but low voltage drive. As a result, its output drive is only between 0 and ½ V HIGH . The structure, however, allows the pad voltage to exceed the supply voltage when the buffer is in the tristate mode, i.e., the circuit can be driven by a voltage of approximately V HIGH without damaging the components. The circuit, therefore, may be characterized as having a “2× tolerance.” The circuit 10 of FIG. 1A may also be characterized as a “2× tolerance” circuit.

Three problems are eliminated to achieve the 2× tolerance of the circuit 20 : (a) V DG (voltage drain to gate) overstress of the n-channel transistor N 1 ; (b) conduction of the p-channel transistor P 1 in tristate mode when the output node exceeds the supply voltage by approximately a threshold voltage; and (c) forward biasing of the drain-bulk p-n junction of the p-channel transistor P 1 when the output sufficiently exceeds the supply voltage. The first problem is resolved by using an n-channel cascode—N 2 —while the second and the third problems are eliminated by using dynamic gate and bulk biasing (conceptually illustrated using two pairs of switches).

Recently, two HVB/LVT's with beyond-2× voltage capabilities have been reported. A first circuit has a 3.3V drive and 5V tolerance using 2V transistors and is proposed in L. Clark, “High-Voltage Output Buffer Fabricated on a 2V CMOS Technology,” Digest of Technical Papers, 1999 VLSI Symposium, p.p. 61-62. A circuit that extends the stress free range of a cascode stack beyond the difference between supply and ground by approximately one threshold voltage is proposed in G. Singh and R. Salem, “High-Voltage-Tolerant I/O Buffers with Low-Voltage CMOS Process,” IEEE J. of Solid-State Circuits, vol. 34, No. 11, p.p. 1512-1525, November 1999. Both circuits use dynamic gate biasing.

While the above referenced circuits address some of the issues involved with interfacing an older higher operating voltage IC with a lower operating voltage technology, the circuits possess significant long term shortcomings. Presently, there is a movement within the semiconductor industry to migrate to sub-0.2 μm sizes towards 0.16 μm, and even 0.13 μm technology powered by sub-1.5V sources. It is expected that within the next four years, the supply voltages may even be in the sub-1V range. As the industry moves below the sub-0.2 μm area and the technologies is powered by sub-1.5V sources, interface buffers will be required to handle greater than the 2× multiples of the known art in order to function with older 0.24-0.35 μm powered devices. Thus, the known art is limited as a long term solution due to the migration towards increasingly smaller MOS transistor technologies in view of the continuing commercial viability of older IC components operating at voltages more than twice that of the breakdown voltages of the smaller devices.

As such, there is a need for an improved output buffer capable of interfacing at least two ICs having operating voltages which are multiples equal to or greater than 2× and which provides no gate-to-source, gate-to-drain, and drain-to-source stresses while providing at least 2× tolerance. Still further, there is a need for a tristate capable high voltage buffer implemented with low voltage transistors that approaches 3× voltage capabilities or better.

An integrated circuit includes an output buffer having a maximum voltage that approximates the highest voltage V MAX applicable across at least one pair of nodes of a transistor. The output buffer is capable of delivering an output signal signal having a voltage swing V HIGH of up to about three times the magnitude of V MAX . The output buffer includes at least a first and a second transistor cascode stack, each of the stacks having a driver transistor and at least one cascode transistor. The output buffer also includes a biasing circuit for biasing at least one of the cascode transistors of each of the cascode stacks in response to said output signal such that the magnitude of the voltage applicable across each pair of nodes of each transistor in each cascode stack is less than or equal to V MAX .

The buffer may be utilized to provide a tristate capable buffer circuit with up to 3× voltage capabilities, including 3× drive and 3× tolerance.

The above and other features of the present invention will be better understood from the following detailed description of the preferred embodiments of the invention that is provided in connection with the accompanying drawings.

The accompanying drawings illustrate preferred embodiments of the invention, as well as other information pertinent to the disclosure, in which:

FIG. 1A is a circuit diagram of a known high voltage tolerance and high voltage driver buffer interface circuit;

FIG. 1B is a circuit diagram of a known high voltage tolerance and low voltage driver buffer interface circuit;

FIGS. 2A-2C are a schematic representation of a tristate capable high voltage buffer and accompanying active mode and tristate mode waveforms;

FIGS. 3A-3F are circuit diagrams of an exemplary driver for a 3× tristate capable buffer;

FIG. 4 is a circuit diagram of an exemplary driver for a 2× tristate capable buffer;

FIGS. 5A-5C are circuit diagrams of an exemplary level shifter circuit for a 3× tristate capable buffer;

FIG. 6 is a circuit diagram of an exemplary 3× tristate capable output buffer; and

FIG. 7 is a circuit diagram of a 3× tristate capable input buffer.

FIG. 2A is a conceptual diagram of a tristate capable CMOS buffer 30 for interfacing a low supply voltage IC chip and a high supply voltage IC chip. The buffer 30 receives an input data signal from the low voltage supply at an input and produces an output data signal which corresponds to the input data signal but has a voltage swing between ground and the higher supply voltage V DD . The buffer 30 includes a level shifter 32 and a driver stage 34 . The level shifter produces data waveforms V p and V n shown in FIGS. 2B and 2C, depending upon whether the circuit is operated in the “Active Mode” or the “Tristate Mode.” As is conventional, the tristate mode is enabled via the OEN (output enable) input. In the “Active Mode,” V n is typically the input data signal and V p corresponds to V n with a DC offset as shown in FIG. 2 B. It is assumed, but not required, that the input data signal received from the low operating voltage IC has a voltage swing with a peak of V MAX , V max approaching the maximum voltage a transistor in the buffer circuitry can withstand, i.e. /V GS /, /V GD /, and /V DS /≦V MAX .

FIG. 3A is a conceptual schematic of an exemplary embodiment of a pad driver 100 having stress-free range of 3×, i.e., the driver 100 is capable of driving an output signal V pad at output node 102 having a voltage swing V HIGH which is up to approximately three times the break-down voltage of any transistor used in the driver 100 . Driver 100 includes a p-channel cascode stack and an n-channel cascode stack connected in series at output node 102 . The p-channel cascode stack includes at least three p-channel MOS transistors P 1 , P 2 , P 3 connected in series between the supply voltage V HIGH and the output node 102 . The n-channel cascode stack includes at least three n-channel MOS transistors N 1 , N 2 , N 3 connected in series between the output node 102 and a relative ground V GROUND .

The gate terminal of N 1 is coupled to an input data signal V n (FIG. 2A) and the gate terminal of P 1 is coupled to a level shifted data signal V p (FIG. 2 A). These signals may be supplied by a level shifter circuit. The gate terminal of n-channel transistor N 2 is coupled to a first constant voltage such that the difference between the first constant voltage and V GROUND does not exceed V MAX . For a true 3× buffer, this first constant voltage value equals approximately V MAX as shown in FIG. 3 A. The gate terminal of p-channel transistor P 2 is coupled to a second constant voltage such that the difference between V HIGH and the second constant voltage does not exceed V MAX . For a true 3× buffer, the second constant voltage is set to 2V max as shown in FIG. 3 A. The first constant voltage may be obtained from the low voltage supply IC. The second constant voltage may be generated internally within the buffer by conventional techniques.

In an exemplary driver 100 , the pad voltage at output node 102 controls switches S 1 and S 2 , which thereby provide dynamic gate biasing for cascode transistors P 3 and N 3 . None of the six transistors N 1 , N 2 , N 3 , P 1 , P 2 , and P 3 experience V GS or V DG voltage overstress if the following conditions are satisfied: (1) S 1 is closed when V pad falls below the first constant voltage (e.g., V MAX ); (2) S 2 is closed when V pad rises above the second constant voltage (e.g., 2V max ); and (3) S 1 and S 2 maintain the gate voltage of N 3 and P 3 at or between the first or second constant voltages when V pad is at or between the first and second constant voltages. Of course, both S 1 and S 2 are never both closed.

The fact that S 1 must be closed or “on” when its control voltage (i.e., V pad ) is lower than the switch terminal voltages suggests that S 1 should be implemented using a p-channel transistor. S 2 on the other hand should be implemented using n-channel transistor because it must be closed or “on” when its control voltage is higher than its terminal voltages. This configuration is shown in FIG. 3B with the addition of n-channel transistor N 4 and p-channel transistor P 4 . The drain terminals of each transistor are coupled at node 104 to the gate terminals of transistors N 3 and P 3 .

The gates of both switch transistors N 4 and P 4 should be controlled by V pad , but preferably are not directly connected to the pad node 102 . A direct connection would result in voltage overstress of the switch transistors N 4 and P 4 . For stress-free operation, the gate voltage of P 4 should follow V pad , but it should not exceed ⅔ V HIGH or 2V max . Similarly, the gate voltage of N 4 should follow V pad , but it should not go below ⅓ V HIGH or V MAX . When the driver 100 of FIG. 3A is in tristate mode, nodes 110 and 112 have the required voltage excursions. The voltage at node 112 follows V pad down to V GROUND , but it would not increase much beyond ⅔ V HIGH −Vtn. The voltage at node 110 , on the other hand, follows V pad to the supply rail, but it would not decrease significantly below ⅓ V HIGH +Vtp. Note that “Vtn” is the threshold voltage of the n-channel devices, and Vtp is the threshold voltage of the p-channel devices.

A driver 100 as shown in FIG. 3B but having the gate terminals of N 4 and P 4 coupled directly to node 110 and 112 , respectively, was simulated using 0.25 μm 2.5V breakdown voltage technology and a V HIGH of 7.5V. The driver was placed in tristate mode (i.e., both N 1 and P 1 were OFF) and the pad voltage was varied between 0 and 7.5 V. As expected, the gate-source and gate-drain voltages of all eight transistors (N 1 -N 4 and P 1 -P 4 ) remained bounded to ±2.5V (i.e., the voltages did not exceed V MAX ). In active mode, however, the voltage at node 112 is not only a function of both V out (i.e., V pad ) but it is also a function of the gate voltage of transistor N 1 . Similarly, the voltage at node 110 is a function of both V pad and the gate voltage of transistor P 1 . As a result, immediately after each input transition, both P 4 (S 1 ) and N 4 (S 2 ) are ON and conducting a large “shot-trough” current. More importantly, during the same time frame, the gate oxides of both N 3 and P 3 are subjected to voltage overstress. This issue is addressed in the circuit of FIG. 3 C.

In FIG. 3C, the triple cascode is split into two separate circuits, one of which is always operated in tristate mode, and the switching transistors are shared. The controls for the switching transistors are derived from the “always-tristate” circuit. Switch transistors P 4 and N 4 respond only to changes in V pad at node 102 and provide dynamic protection for added transistors N 5 and P 5 . Transistors N 5 and P 5 have gate terminals coupled to node 104 , and drain terminals coupled to output node 102 . The source terminal of P 5 is coupled to the gate terminal of N 4 , and the source terminal of N 5 is coupled to the gate terminal of P 4 . Since the drain and gate nodes of these two transistors P 5 , N 5 are coupled respectively to the gate and drain nodes of N 3 and P 3 (also referred to as nodes 104 and 102 , respectively), the switches N 4 , P 4 also provide protection for P 5 and N 5 . An exemplary high voltage buffer, low voltage transistor circuit is therefore achieved.

The gate-source and gate-drain voltages of all ten transistors (N 1 -N 5 , P 1 -P 5 ) are always limited to ±⅓ V HIGH or V MAX . Drain-source (DS) voltages of transistors N 5 , P 5 , N 3 and P 3 can, however, exceed V MAX by at least one threshold voltage (Vtn). This may be addressed by extending the length of devices in the driver, but this may not be a viable option with increasing area constraints on ICs.

In order to keep V DS of N 5 ≦⅓ V HIGH or V MAX , node 106 may be pulled up to ⅔ V HIGH . This is accomplished in the exemplary circuit configurations of FIGS. 3D and 3E via transistor N 6 . Similarly, to keep V DS of P 5 ≦⅓ V HIGH or V MAX node 108 may be pulled down to ⅓ V HIGH , which is accomplished via transistor P 6 . Note that when activated, N 6 does not connect node 106 directly to ⅔ V HIGH ; instead, it connects it to node 104 , which for high pad voltages acquires the desired ⅔ V HIGH value. Similarly, node 108 is brought down to ⅓ V HIGH indirectly via node 104 . This feature guarantees that transistors N 6 and P 6 are not overstressed.

The final issue that should be addressed is the potential drain-source overstress of transistors N 3 and P 3 . Below is a brief description of the causes of the drain-source overstress of transistor N 3 , and a circuit approach to resolve this issue is shown in FIG. 3 F. The cause of overstress of P 3 is analogous to that of N 3 and is not described.

When the pad voltage at 102 equals V HIGH , the drain-source voltages of transistors N 2 and N 3 are as follows: V DS of N 2 =⅓ V HIGH +(V GS of N 2 −V GS of N 3 ) and V DS of N 3 =⅓ V HIGH +V GS of N 3 , respectively. Immediately after input transition of the input data signal at the gate terminal of N 1 , both V GS of N 2 and V GS of N 3 can increase so that the cascode transistors N 2 and N 3 can carry the current conducted by N 1 . These changes alter V DS of N 2 and V DS of N 3 . According to the first equation, the change in V DS of N 2 can be kept low at approximately constant and equal to ⅓ V HIGH by making N 2 and N 3 identical in size. The second equation reveals that N 3 would experience a drain-source overstress. This overstress could be prevented by connecting an additional cascode transistor N 7 between the drain of N 3 and node 102 as shown in FIG. 3 F. With the addition of transistor N 7 in FIG. 3F to the n-channel cascode stack, the drain-source voltage of N 3 becomes V DS =⅓ V HIGH +(V GS of N 3 −V GS of N 7 ). This drain-source voltage can now be kept nearly constant and equal to ⅓ V HIGH by simply making N 4 and N 3 approximately equal in size. This drain-source overstress protection requires the gate terminal of N 7 to have a potential of V HIGH whenever the pad node has potential V HIGH . The gate terminal potential of transistor N 7 should, however, be lowered to ⅓ V HIGH as pad node 102 traverses toward V GROUD , i.e. N 7 requires a V HIGH -to-⅓V HIGH dynamic gate biasing. Such biasing is readily available at node 108 .

Similarly, the drain-source overstress of P 3 is eliminated by the addition of the P 7 cascode transistor coupled between the drain terminal of P 3 and output node 102 . As shown in FIG. 3F, required dynamic biasing (0-to-⅔V HIGH ) is obtained by connecting the gate terminal of transistor P 7 to node 106 .

FIG. 4 is provided to show that the same basic approach as described above in achieving a 3× driver with no gate-source, gate-drain, and drain-source overstress may be utilized to provide a 2× driver 200 that exhibits no gate-source, gate-drain, and drain-source overstress. Note that the driver 200 of FIG. 4 includes n-channel and p-channel cascode stacks connected at an output node 202 as shown in FIGS. 3A-3E. The supply rail is set to 2V max in the 2× driver 200 instead of 3V max . Also, the gate terminals of P 2 and N 2 are coupled to a single constant voltage at node 204 (shown as V MAX ) such that the difference between V HIGH and node 204 is not greater than V MAX and the difference between node 204 and V GROUND is not greater than V MAX . The input data signal at the gate terminal of N 3 may have a maximum voltage swing of V MAX , and the input data signal at the gate terminal of P 3 is the input data signal of N 3 level shifted by the DC value at node 204 , i.e., the data signal traverses between V MAX and 2V max . Control circuits 206 and 208 provide dynamic gate biasing signals Gn and Gp to n-channel transistor N 1 and p-channel transistor P 1 , respectively.

Both signals Gn and Gp are in-phase with the output signal produced at node 202 and have a voltage swing of V MAX , with Gn traversing between V MAX and V HIGH and Gp traversing between V GROUND and V MAX . To understand how the circuit 200 of FIG. 4 provides stress-free operation, the behavior of the circuit before and after input transition may be considered. Assuming that the initial condition is as follows: input is low (the gate of N 1 is 0V and the gate of P 1 is V MAX ) and the output is V HIGH . Due to the action of the biasing circuits 206 , 208 , Gn and Gp respectively are V HIGH and V MAX . Under these stated conditions, it can be shown that the potential difference between any two transistor terminals does not exceed V MAX .

As the input signal at the gate terminal of N 1 goes high, the current carried by N 1 increases. In order for this current to be conducted by N 2 and N 3 , the transistor source potentials of N 2 and N 3 decrease from their initial values of V MAX −Vtn and 2V max −Vtn. If N 2 and N 3 are matched, the source decrement for both transistors is the same and V DS of N 2 remains initially constant and approximately equal to V MAX . The drain-source voltage of N 3 also is less than V MAX . As the load capacitance is being discharged, the drain voltage of N 3 decreases. At some point, the biasing circuit 206 is activated and it lowers the gate voltage of N 3 to prevent V GD of transistor N 3 from becoming to large. The output node 202 keeps discharging until it reaches V GROUND where it settles.

It can be shown that for output “low” or V GROUND , and input of V MAX , no terminal-to-terminal voltages exceed V MAX . The circuit 200 exhibits similar behavior when the input transitions from “high” to “low” (and output transitions from “low” to “high”). During this transition, the presence of P 3 prevents transistor P 2 from developing large drain-source voltages while the control circuit 208 provides gate-source and gate-drain overstress protection for P 3 .

Biasing circuits 206 and 208 may be implemented as shown in FIG. 4 with p-channel transistors P 4 and P 5 and n-channel transistors N 4 and N 5 . P-channel transistor P 4 has its gate terminal coupled to V pad at output node 202 , a drain terminal coupled to the constant voltage at node 204 , and a source terminal coupled to the gate terminal of N 3 . Transistor P 5 has its gate terminal coupled to node 204 , a drain terminal coupled to output node 202 , and its source terminal coupled to the gate terminal of N 3 . Likewise, biasing circuit 208 may be implemented with n-channel transistor N 4 having its gate terminal coupled to output node 202 , its source terminal coupled to the gate terminal of P 3 , and its drain terminal coupled to the constant voltage at 204 . N-channel transistor N 5 has its gate terminal coupled to the constant voltage at node 204 , its source terminal coupled to the gate terminal of P 3 and its drain terminal coupled to output node 202 . The biasing circuits provide the advantage of not dissipating static power.

FIGS. 5A-5C are circuit schematics of an exemplary level shifter circuit 300 for use in a 3× tristate capable buffer. This circuit may be used to provide signal V p to driver 100 . A level shifter circuit comprises an input n-channel transistor, a cascode stack for protecting the input transistor from voltage overstress and including at least one n-channel cascode transistor, and a load transistor coupled to the cascode stack at an output node. As shown in FIG. 5A, the level shifter 300 for providing a signal to a driver as described above may generally be illustrated as a modified N-MOS inverter comprising at least four n-channel transistors N 8 , N 9 , N 10 and N 11 coupled in series. A source terminal of input transistor N 8 is coupled to V GROUND , a source terminal of cascode transistor N 9 is coupled to a drain terminal of N 8 , a source terminal cascode transistor N 10 is coupled to a drain terminal of N 9 and a drain and gate terminals of load transistor N 11 are both coupled to supply V HIGH or 3V max . Transistors N 11 and N 10 are coupled together at first output node 302 . A gate terminal of N 8 is coupled to an input data signal with a maximum voltage swing which does not exceed V MAX . This input data signal may be provided by the low voltage IC. The gate terminals of N 9 and N 10 are coupled to a first constant voltage and a second constant voltages, respectively, such that a difference between V HIGH and the second constant voltage does not exceed V MAX and a difference between the first constant voltage and V GROUND does not exceed V MAX . For a true 3× level shifter stage, the first and second constant voltages are V MAX (⅓ V HIGH ) and 2V max (⅔ V HIGH ), respectively. Through the level shifter 300 , a data signal V p is produced at output node 302 which corresponds to the input data signal but has a DC offset of 2V max , i.e. the signal traverses between 2V max and 3V max as needed for the driver circuits shown in FIGS. 3A-3F. It should be apparent that higher voltage shifted data signals may be achieved with the circuit approach of FIG. 5A by simply adding additional n-channel cascode with appropriate constant gate voltage biasing and an increased supply voltage V HIGH .

The circuit 300 provides a very robust output signal at node 302 , but can dissipate static power. To reduce this static power dissipation, the duty cycle of the input data signal can be reduced. If the input data signal duty cycle is reduced, the output duty cycle must still be preserved. Therefore, an exemplary level shifter may further include a second inverter stage driven out-of-phase with the input inverter by one half cycle and an RS (reset-set) latch connecting the two inverter stages. This exemplary embodiment of a level shifter is shown in FIG. 5 B.

The level-shifter of FIG. 5B includes inverter 300 a and inverter 300 b (N 12 , N 13 , N 14 , and N 15 ) coupled to an RS latch 306 . The level shifter also optionally includes cross-coupled p-channel transistors P 8 and P 9 and added series n-channel transistors N 16 and N 17 . These devices (P 8 , P 9 , N 16 , N 17 ) are not required but help the level shifter produce smooth output waveforms at outputs 308 and 310 . Each inverter 300 a , 300 b comprises an input transistor coupled to input signals IN 1 and IN 2 at their gate terminals, a cascode stack and load transistor. Input signals IN 1 and IN 2 are shown as having reduced duty cycles below 50%. The “dashed” wave forms illustrate input waveforms with 50% duty cycles. The cascode transistors provide overstress protection for all of the devices. For effective overstress protection all devices preferably are the same size. With equally sized input and load transistors, the level shifter of FIG. 5B exhibits a gain of near-unity for large signals. Inverter gain is on the first order insensitive to process and temperature variations. The level-shifter of FIG. 5B, however, dissipates static power if input signals IN 1 and IN 2 have normal duty cycles of 50% (as shown by the dashed input signals of FIG. 5 B).

Since the inverter stages 300 a , 300 b dissipate static power whenever their inputs are “high,” the static power dissipation could be reduced if the two inverters are impulse driven. Impulse duration, however, should be sufficiently large so that the latch 306 can change its state. If the two inverter structure are pulse driven, both inverter outputs are “high” most of the time. In order to be able to retain its state, the RS latch 306 should be implemented using NAND gates (as opposed to NOR gates).

The circuit of FIG. 5C illustrates one means of generating the modified pulse signals IN 1 and IN 2 . The circuit of FIG. 5C also provides the additional advantage of producing two additional signals at outputs 316 and 318 having voltage swings between 0 and V MAX . One of these signals may be used to provide signal V n to drive the n-channel transistor N 1 of the driver stage 100 .

The desired impulse drive of FIG. 5C is realized using a “one shot” circuit. This circuit employs three MOS inverters within input circuitry 314 , two NAND gates within input circuitry 314 , and a NAND-based RS latch 312 . Transistors N 8 , N 9 , N 12 and N 13 are also part of the one-shot circuit. P-channel transistors P 10 and P 11 may be included for smoother waveform generation. A pulse is produced at the gate of the transistors N 8 whenever there is a positive input transition and at the gate of input transistor N 12 whenever a there is a negative input transistor. The duration of the produced pulse is approximately equal to τ mos +τ N8/N9 +τ RS where τ mos is the delay of the MOS inverter, τ N8/N9 is the delay of the N 8 /N 9 inverter and τ RS is the switching delay of the RS latch 312 . As long as RS latch 308 and RS latch 312 are equally loaded and present minor loading to their corresponding driving circuits, the duration of the generated driving pulses would be sufficient to guarantee switching of RS latch 308 .

FIG. 6 is an exemplary embodiment of a 3× tristate capable output buffer circuit 400 including a level shifter circuit 300 described above coupled to a driver stage 200 described above. The tristate-capable 7V output buffer 400 was fabricated with a 0.25 μm 2.5V CMOS process. The circuit was designed to drive a 10 pF load capacitance at 200MHz. The transistor sizes (width/length) in μm)) were as follows:

The above designed circuit was tested, and “on-wafer” probing was performed successfully. To verify high-voltage capability, internal nodes 104 , 106 and 108 were monitored while the buffer was operated in a “package-like environment.” This was done by bonding a bare die directly on a PCB (printed circuit board) and using active probes. The obtained waveforms were then compared to the output waveform generated at node 102 . The potential differences V 102 −V 106 , V 102 −V 108 and V 102 −V 104 are indicative of the presence or absence of GS and GD voltage overstress. These differences remained bounded to approximately ±2.5 V. Therefore, the voltage drive and tolerance of the developed buffer circuit is nearly three times larger than the breakdown voltage of the MOS devices used in the circuit.

FIG. 7 illustrates that the circuit approach described in FIGS. 3D-3F may be used to form an input buffer circuit which receives an input signal at node 102 which swings between 0 and 3Vmax and produces an output data which swings between 0 and V MAX at output node 506 . Of course the output signal can be amplified by an inverter 510 or chain of inverters as is conventional. The n-channel and p-channel cascode stacks of FIGS. 3D-3F are not required for the function of an exemplary input driver. This circuitry is shown within the dashed box 508 . These cascode stacks may be disenabled by coupling the gate of P 1 to the supply voltage and the gate of N 1 to V GROUND .

An input stage 502 may be designed as shown in FIGS. 3D-3F from transistors N 4 -N 6 and P 4 -P 6 particularly those circuit configuration introduced in FIGS. 3D and 3E. Likewise, the biasing circuit 208 of FIG. 4 including n-channel transistors N 5a and N 4a may be coupled to the input stage 502 as shown. An input data signal from a high voltage IC presented at node 102 and having a voltage swing between 0 and 3V max produces a data signal at node 506 which is appropriate for safely driving a low voltage IC. The data signal at node 504 has a voltage swing between 0 and 2 V MAX , and the data signal at node 506 has a voltage swing of 0 to V MAX .

Although the invention has been described in terms of exemplary embodiments, it is not limited thereto. Rather, the appended claims should be construed broadly to include other variants and embodiments of the invention which may be made by those skilled in the art without departing from the scope and range of equivalents of the invention.