CMOS Buffer With Controlled Slew Rate

This invention generally relates to a method and apparatus for providing an improved controlled slew rate output circuit for use in CMOS input/output and CMOS output buffers wherein it is desirable that the rate of transition from a high/low state to a low/high state at the output be optimally controlled. An optimally controlled slew rate output circuit enables reduction of signal line reflection and ringing at the output and in the cable or connectors coupled to the output.

Without limiting the scope of the invention, its background is described in connection with a system which implements a system interface bus which couples devices in parallel, and in connection with an output buffer circuit for use on devices coupled to a bus. The system is further described using a SCSI (Small Computer System Interface) bus as an example.

Interfacing digital systems has become increasingly difficult for the designer due to the emphasis on high-speed performance. The high-speed performance of most digital systems is now limited by distributed discontinuities and transmission line effects of packaging and interconnects, and not by the switching speed of semiconductor devices. It is becoming increasingly common for digital systems to be interfaced with electrically long cables comprised of bundled-wire or fiat-wire connections. These applications are typically found in personal computers and workstations where connection to external peripheral equipment, such as disk drives, scanners, or printers, is implemented with user installed cables.

Problems are encountered when signal line reflections and high frequency ringing cause data read errors on the interface bus. The primary cause of this poor data quality is the rapid slew-rate of the I/O port output driver circuits. Typical prior art output circuits exhibit signal transition slew rates of 3 V/nsec. This rapid rate of change on the signal line will excite the resonant frequency of the bus cable as well as contribute to the strength of signal reflections. The dilemma for the prior art system designer has been that the exact characteristics of the bus signal noise cannot be predicted due to variations in physical cable length and connector impedance mismatch, and hence, signal line tuning cannot be applied to solve the problem. When the rise and fall times of the signal pulses are less than the propagation delay of the cable, transmission line effects become a serious problem to contend with. Proper design of cable terminations will eliminate some of the signal noise, however, the most dramatic reduction of these transmission line effects is achieved by making the rise and fall times of the signal pulses much slower. Controlling the slew rate of the signal driver circuit is therefore critically important in minimizing transmission line noise.

Current technologies used for digital systems bus interface include circuits designed using digital and linear CMOS processes, and digital Bi-CMOS processes. Controlled slew-rate drivers have been developed in linear CMOS and digital Bi-CMOS technologies, however these technologies are cost prohibitive, and the resulting designs typically exhibit undesirably complex circuitry as well as measurable standby power dissipation. The digital CMOS technology is attractive for bus interface driver designs due to its cost effectiveness, leakage-level standby current flow, and popularity in digital subsystems design. However, controlled slew-rate designs are very difficult to develop in digital CMOS technologies. Prior art controlled slew-rate designs utilize some form of output feed-back, and digital CMOS technology lacks the commonly used features of either the bipolar technology's junction transistors, or the linear technology's precision resistors and capacitors. Controlled slew-rate driver designs implemented without these elements exhibit widely varying performance parameters over the range of operating conditions, and hence have not been viable solutions for practical bus interface performance requirements.

Bus interface methodologies that incorporate active pull-down "open-drain" wire-ANDed driver outputs offer a partial solution by using the bus termination circuitry to passively control the low-to-high signal transition slew-rate. An example of this application is the SCSI (Small Computer System Interface) bus. This type of bus implementation, however, still exhibits undesirable transmission line noise produced by the high rate of change of the high-to-low signal transition of the prior art bus driver circuits. Accordingly, a bus driver circuit design that can be implemented in a standard digital CMOS technology as an element of an integrated circuit, and having improvements that overcome any or all of the problems of the prior art solutions, is presently desirable.

Generally, and in one form of the invention, a circuit is described which implements a controlled slew rate output buffer circuit in a standard digital CMOS process. A first embodiment is described wherein an open-dram driver circuit is coupled to a slew rate control circuit comprising P and N type MOS transistors and a capacitor. A second embodiment is described wherein the open drain circuit and the slew rate control circuit are further coupled to a pull-up circuit comprised of either active components, or passive components, to provide full push-pull output driving capacity, bus termination compensation, or both. A third preferred embodiment is described wherein the open drain circuit and the slew-rate control circuit are expanded to reflect an enhanced implementation comprising several controlling transistors for further improved output transition control and is implemented as a bidirectional input/output buffer.

An integrated circuit is described for use in a multiple chip bus system, wherein user specified application logic, either chosen from a standard ASIC or standard cell library, or specifically designed by the user, is coupled to a plurality of the input/output buffers containing the slew rate control circuitry. A system is described wherein each IC in the system contains the slew rate control circuitry and the IC's are coupled together to communicate on a multiple driver bus, such as the SCSI bus. The system thus created will exhibit superior bus noise reduction characteristics with leakage-level standby power dissipation.

In the drawings:

FIG. 1 is depicts a schematic diagram of a typical prior art CMOS output buffer circuitry;

FIG. 1A is a plot of the high-to-low output transition of the prior art buffer depicted in FIG. 1 for three different load capacitances;

FIG. 2 is depicts a schematic diagram of an open drain output buffer embodying the slew rate control circuitry of the invention;

FIG. 2A is a plot of the output high-to-low transition of the output buffer depicted in FIG. 2 for three different load capacitances;

FIG. 3 is a schematic of a second preferred embodiment comprising an output buffer having the open drain driver and a pullup circuit;

FIG. 4 is a schematic which illustrates an alternate design for the transistor M5 depicted in FIGS. 2 and 3;

FIG. 5 is a schematic which illustrates an alternate design for the transistor M8 depicted in FIGS. 2 and 3;

FIG. 6 is a schematic illustrating an input/output buffer having an open-drain output driver and embodying the controlled slew rate circuitry of the invention;

FIG. 7 is a block diagram of an integrated circuit embodying buffers which are implemented using the input/output buffer depicted in FIG. 6.

Corresponding numerals and symbols in the different figures refer to corresponding parts unless otherwise indicated.

In prior art output buffer circuits, typical output drivers exhibit slew rates of 3 V/nanosecond. When used in a multiple driver bus, such as a SCSI bus, this rapid change of voltage on the signal line will excite the resonant frequency of the bus cable, as well as disadvantageously contributing to the strength of signal reflections. Because the system designer cannot predict the exact physical characteristics of the bus, due to variations in the physical length of the cable and connector impedance mismatches, signal line tuning and impedance matching techniques are not viable solutions to the problem.

When signal line reflections and high frequency ringing on the bus caused by the rapid slew rate of the drivers on the bus become severe, data on the bus becomes unreadable and is lost. One object of this invention is to provide a circuit for input/output buffers which can drive a bus while exhibiting load independent slew rates which are controlled and limited to values chosen by the designer, for a SCSI bus example operating at a frequency of 5 Mhz typical values are in the range of 0.20 V/nanosecond. A slew rate in this range will substantially reduce signal distortions due to reflections and ringing on the signal lines.

FIG. 1 depicts a typical prior art CMOS output buffer which is implemented in a digital CMOS technology. Driving transistor M1 is coupled to inverter 1, which is coupled to the input. The output is coupled to the conductance path of transistor M1. Optional pull-up circuit 3 is provided in some applications, and omitted in open-drain applications. Pull-up circuit 3 may be passive, as a resistance or RC network, or active, in which case the dashed line shown in FIG. 1 is necessary to control the pull-up circuit 3. The output is coupled to the system bus which is represented as load 5, having a resistive pullup component and a capacitive load component.

Prior art designs, such as the one shown in FIG. 1, feature di/dt rate control which offer a modest degree of slew-rate control by reducing the rate of change of the turn-on current available in transistor M1. These designs incorporate techniques that weaken the drive available to the gate node of transistor M1. This technique does not compensate for the variations in signal line load capacitance and is primarily intended as a method of reducing internal on-chip ground noise. This design technique does little or nothing to reduce external signal bus noise.

FIG. 1A depicts the high-to-low output characteristic of the buffer depicted in FIG. 1 for three typical capacitive loads. The circuit does not compensate for different loads, as is clearly indicated by the widely varying fall time for the different cases shown in the plot.

FIG. 2 depicts the open-drain embodiment of the circuit embodying the invention. Inverters 1 and 3 are coupled to the input. Transistor M9 provides a pull down for the output pin. Transistor M8 provides a clamping action to maintain transistor M9 in the `off` state when the circuit is not driving the output. Transistor M2 operates to charge capacitor C1, while transistor M5 operates to discharge capacitor C1. Capacitor C1 is implemented in a typical digital CMOS process as an oxide dielectric capacitor, however other implementations of the capacitor are possible. Transistors M1, M3, M4, M7, M6, M5 and capacitor C1 operate to control the turn-on voltage of transistor M9 when the input switches, so as to control the rate at which the output node is pulled down to its low voltage level.

In operation, the circuit output (node "OUT") transitions through 4 states: a stable logic one or logic high state, a high to low transition phase, a stable logic low state, and a low to high transition phase. These phases are discussed each in turn in the paragraphs that follow.

When node "IN" is in the stable logic high state (greater that 90% of VDD), node "A" is at a logic low voltage and node "B" is at a logic high voltage. This causes the transistors M1, M3, M4, M5, and M9 to be `off`, and transistors M2, M6, M7, and M8 to be `on`. In this state, transistor M9 in the `off` condition results in node "OUT" being biased at the external load bias voltage, V.sub.LOAD. Transistor M9 is held in the `off` state by the clamping action of transistor M8, which is in the `on` state. Because transistor M2 is `on` and transistor M9 is `off`, the voltage present across capacitor C1 at node "C" charges the capacitor C1 up to a voltage which is equal to the voltage at node "OUT" minus the source-drain `on` voltage drop across transistor M2. Transistors M6 and M7 are `on`, however no current flows through them since transistors M1, M3, and M4 are `off`.

After node "IN" switches to the logic low state (less than 10% of V.sub.DD), node "A" is at a logic high voltage and node "B" is at logic low voltage. As a result, transistors M1 and M5 turn `on` and M2 and M8 turn `off`. The voltage at node "C" begins to decrease as the voltage across capacitor Ct discharges through transistor M5. The resistive `on` characteristics of transistor M5 and the magnitude of capacitor C1 are designed to yield a desired RC time constant for the discharge rate of node "C". This RC time constant will determine the high-to-low transition slew rate of the output at node "OUT".

As the voltage at node "C" decreases, transistor M3 begins to turn `on`. This allows current to begin to flow through transistors M1, M3, and M6, and causes the voltage at node "D" to begin to increase. As the voltage at node "D" increases, transistor M9 begins to turn `on` and the voltage at the output, node "OUT" begins to decrease. Node "OUT" connects to the gate of transistor M4 which forms feedback to the differential pair comprised of transistors M3 and M4. If the voltage at node "OUT" decreases more rapidly than the voltage discharge rate of node "C", then a situation is quickly established where node "OUT" is at a lower voltage than node "C". This will cause transistor M4 to begin to turn `on` and current that was flowing through transistor M3 will be reduced by the amount now flowing through transistor M4. The reduction of current flowing through M3 causes a reduction in the rate of increase of the voltage at node "D". This in turn modulates the rate at which M9 is turning `on` and thus reduces the rate at which the voltage at the output, the voltage at node "OUT" decreases. This negative feedback mechanism ensures that the rate of decrease of the voltage at the output is directly proportional to the rate of decrease of the voltage discharging across capacitor C1.

An important feature of the circuit design depicted in FIG. 2 is that during the stable state prior to the transition phase, the voltage at node "C" is very nearly equal to the output voltage at node "OUT" due to the coupling of transistor M2. This feature ensures that the response time of the feedback path is kept independent of the output load bias voltage, V.sub.LOAD.

Another important feature of the circuit design is that the gates of transistors M6 and M7 are tied to the gate of transistor M3 at node "C". The traditional approach used in differential-pair designs would be to connect the gates of M6 and M7 to a controlled voltage source. This technique, however, would require circuitry contributing to costly DC power dissipation. Connecting the gates of M6 and M7 to node "C" provides active biasing only during the transition phase, and avoids any DC power dissipation. In addition, as the node "C" voltage decreases, the voltage increase at node "D" is dynamically increased during the critical initial turn `on` period of transistor M3. This effect minimizes circuit propagation time variations due to processing variations.

In the final low voltage stable state, after the output high-to-low transition phase, the voltage at node "C" has discharged below the threshold voltage of transistors M6 and M7, and hence transistors M6 and M7 are `off`. Transistors M1, M3, and M4 remain in the `on` state and node "D" has risen to a voltage nearly equal to V.sub.DD. Transistor M9 is `on` in the non-saturated state with the voltage at node "OUT" being less than 0.5 volts. Since transistors M6 and M7 are clamped off by transistor M5, no current flows through the differential pair transistors M3 and M4, resulting in minimal standby power dissipation.

When node "IN" switches back to the high state, node "A" returns to logic low and node "B' returns to logic high. Transistor M1 turns `off`, immediately disabling the differential pair, M3 and M4, by shutting off their supply of source current. Transistor M8 turns `on` which in turn begins to turn `off` M9. The voltage at node "OUT" begins to rise at a rate dominated by the RC time constant of the external load. Transistor M5 has turned `off` and M2 has turned `on`, allowing node "C" to begin to rise, the voltage at node "C" will thus track the voltage at the node OUT.

Transistor M5 provides a mechanism for controlling the discharge rate of capacitor C1. By choosing the size and threshold voltage of transistor M5, the designer can determine how rapidly capacitor C1 discharges in response to a transition at the input. Alternative embodiments wherein transistor M5 and capacitor C1 are replaced with a multiplicity of resistive and capacitive elements can be used to further optimize the discharge rate of C1, and thereby reduce the propagation delay time from the input transition to the turn-on of transistor M9.

FIG. 2A depicts a plot of the high-to-low output transition characteristic of the circuit illustrated in FIG. 2 for three typical output loads. Note that the circuit compensates for the load in that the transition time for the three cases is basically the same. The transition time shown, approximately 20 nanoseconds, was chosen for an application in a system designed around the Small Computer System Interface or SCSI bus. The user may choose an appropriate transition time for any arbitrary system by designing the transistor sizes and the capacitance of capacitor C1 to get the required response using well known design techniques.

FIG. 3 depicts an output buffer embodying the invention and implementing a push-pull or full range output buffer. The circuit of FIG. 3 comprises all of the elements of the circuit of FIG. 2, along with the addition of pull-up circuitry 19. In applications where the open-drain approach is not used, this embodiment provides a circuit which will drive the bus low, with controlled slew rate fall time, and drive the bus high, with the rise time determined by the load on the bus and the pull-up circuit 19.

FIG. 4 depicts alternative designs for the transistors M5 shown in FIGS. 2 and 3. Transistor M5 of FIGS. 2 or 3 may be replaced with the transistors M5, M5A, M5B, M5C and M5D as coupled as shown in FIG. 4 for improved performance.

Transistor M5A connected in series with transistor M5 causes the discharge target voltage of node C to be elevated by the `on` voltage of transistor M5A, approximately 0.9 V in a typical CMOS process. The selection of this voltage for the voltage of node C ensures that transistors M6 and M7 (see FIG. 2 or 3) do not turn off during the critical period where transistor M9 is being turned on. Transistor M5C is coupled between transistor M5B and node "D" and operates to cause transistor M5B to turn on when node "D" is 0.9 V above the turn on voltage of transistor M9. This arrangement provides a means for turning off transistors M6 and M7 after the turn on period for transistor M9 is completed.

FIG. 5 depicts alternative designs for the transistors M8 shown in FIGS. 2 and 3. Transistor M8 may be further coupled to transistors M8A and M8B as shown in FIG. 5 for improved performance of the circuit.

Transistor M8A is coupled in series with transistor M8B. In operation this provides a means of quickly transitioning the voltage at node "D" from a high voltage level of V.sub.DD minus the turn-on voltage drops of transistors M1 and M3 (see FIG.2), to a voltage level that approaches the threshold voltage of transistor M9 in FIG. 2. The key feature of transistor M8A is that its on resistance increases as the voltage at node "D" approaches the threshold voltage of transistor M9 due to the body effect on transistor M8A. Transistor M8 is coupled in parallel with the series configuration of M8A and M8B to provide a weak secondary pull-down mechanism for node "D". Together transistors M8, M8A and M8B, provide a means of shaping the high-to-low transition of the voltage at node "D" so that the voltage at the gate of transistor M9 is controlled in a way that causes the transistor M9 to turn off slowly, but the shaping of the voltage at node "D" is performed in a way so that node "D" approaches the threshold voltage of transistor M9 quickly, but slowly transitions through the turn off period of transistor M9. This circuitry enables the designer to control the low-to-high transition of the output node ("OUT" in FIG. 2) so that it is slowed similarly to the high-to-low transition.

FIG. 6 depicts a third embodiment of the invention implementing a input/output buffer for use in an open drain bus system, such as a SCSI system. Input node "IN" is coupled to inverter 21, which is coupled to inverter 23, to provide the switching waveforms within the circuit. Transistors M3 and M4 again form a differential pair coupled to the input/output node "I/O`. Transistor M9 provides the pull down path for the output node. Inverters 29 and 31 are coupled to fore a non-inverting output equal to the value of the "I/O" node for use inside the integrated circuit. Gates 25 and 27 are arranged to form an "SR" latch which is coupled to the node "I/O". Transistors M8A, M8B, M8 and M8C provide the turn-off and clamping action on M9. Transistors M6, M7 and M5A, M5B, M5C, M5D are coupled to capacitor C1 to control the turn-on voltage at the gate of transistor M9. The bus load is represented by the load circuit 33, which consists of a resistance R.sub.LOAD, a capacitance C.sub.LOAD, and a bus termination voltage V.sub.LOAD.

In operation, the circuit of FIG. 6 operates exactly as the circuit of FIG. 2, with the enhanced circuitry of FIGS. 4 and 5 added to improve the control of the turn on and turn off characteristics of transistor M9. The SR-latch formed by the OR gates 25 and 27 coupled to the transistor M8C provides additional noise immunity on the gate of transistor M9 by a means of feeding back the high voltage at the output node "I/O" to the gate of transistor M9, the SR-latch designed so that when the high voltage at the output node reaches a strong high value, the gate of transistor M9 is further clamped down so that noise at the output cannot cause the transistor M9 to come out of the off state.

FIG. 7 depicts an integrated circuit comprising application logic 41 and output buffers 43. The application logic is specified by the user, and may be created using ASIC design tools, standard cell libraries, or specifically designed by the user. The I/O buffers 43 are each implemented using the circuit depicted in FIG. 6. This IC may be coupled to other IC's using a multiple driver bus, for example a SCSI bus, and the resulting system will exhibit excellent signal noise reduction characteristics while maintaining a leakage-level standby power for a low overall power usage for the system.

A few preferred embodiments have been described in detail hereinabove. It is to be understood that the scope of the invention also comprehends embodiments different from those described, yet within the scope of the claims.

Internal and external connections can be ohmic, capacitive, direct or indirect, via intervening circuits or otherwise. Implementation is contemplated in discrete components or fully integrated circuits in silicon, gallium arsenide, or other electronic materials families, as well as in optical-based or other technology-based forms and embodiments. It should be understood that various embodiments of the invention can employ or be embodied in hardware, software or microcoded firmware. Process diagrams are also representative of flow diagrams for microcoded and software based embodiments.

While this invention has been described with reference to illustrative embodiments, this description is not intended to be construed in a limiting sense. Various modifications and combinations of the illustrative embodiments, as well as other embodiments of the invention, will be apparent to persons skilled in the art upon reference to the description. It is therefore intended that the appended claims encompass any such modifications or embodiments.