FIELD OF THE INVENTIONThis invention relates to integrated circuit design and, more particularly, to output circuits (also referred to as buffer circuits) used in CMOS applications.
BACKGROUND OF THE INVENTIONFIG. 1 depicts a conventional output driver circuit. Given the assumption that the gate voltages on both transistor Q1 and transistor Q2 are at ground potential and that the output node O is low, reflections at a mismatched interface may cause output node O to fall below ground to, say, -1.0 volts. Under these conditions, the gate of transistor Q1 becomes positive with respect to the source. Consequently, the channel of transistor Q1 begins to conduct, and the source region of Q1 begins to generate free electron carriers. In an insulated-gate field effect transistor (hereinafter also FET), such as Q1, electric field intensity is greatest near the silicon-silicon dioxide interface where the drain junction is directly under the gate edge. As the free electron carriers from the source region pass through the high-field region near the drain, they can acquire energy far in excess of that which would be attributable solely to ambient temperature. In such a state, those electrons are considered "hot" carriers, and are capable of causing a number of "hot-carrier" effects. For a field effect transistor, the worst case scenario for hot electron generation is generally regarded to be a condition where gate-to-source voltage (VGS) is about one-half drain-to-source voltage (VDS).
In MOS memory circuits, "hot-carrier effects" can disturb circuit operation both by directly altering stored data values, and by permanently altering device performance. Although the vast majority of hot electron carriers are collected by the drain region, some leave the channel and travel to the gate through the gate oxide layer. Some of the electrons inevitably become trapped in the gate oxide layer, thereby shifting the threshold voltage of the device. Other electrons are injected into the substrate, through which they can migrate to the memory array where they are attracted by cells in which a logic value of "1" (i.e., a positive charge) is stored. Through this mechanism, data may become corrupted if the refresh cycle is not shortened to compensate for the charge loss. Electron injection into the substrate can also precipitate a latch-up condition in CMOS circuits.
The very structure required to fabricate bulk CMOS circuitry makes it susceptible to latchup. To have both N-channel and P-channel field effect transistors, it is necessary to have both P-type and N-type background material. Typically, the CMOS fabrication process begins with a silicon wafer of a single conductivity type. Regions of the opposite conductivity type, known as wells or tubs, are created it by diffusing or implanting dopant species, which overwhelm the original dopant. For circuitry constructed on a p-type wafer, P-channel FETs are built in an N-well, while N-channel FETs are built directly in the P-type wafer substrate. Unfortunately, the FETs are not the only structures fabricated. PNPN devices consisting of parasitic bipolar transistors are also created. Under certain operational conditions, these PNPN devices can generate a VCC (power supply voltage) to ground short that can destroy the circuitry.
Some designers have addressed the electron injection problem in output driver circuits by replacing FET Q1 of FIG. 1 with a pair of FETS, Q3 and Q4. Such a circuit is depicted in FIG. 2. Such an approach is effective in reducing electron injection when the output node O drops below ground potential, as transistors Q3 and Q4 act to divide the voltage drop between VCC and the output node. However, the area required for both FETs Q3 and Q4 is approximately four times that required for transistor Q1 of FIG. 1. Thus, this solution for reducing electron injection has its costs, which for a typical memory circuit can be significant.
What is needed is a new, space-efficient driver circuit that will reduce electron injection into the substrate.
SUMMARY OF THE INVENTIONA new inverting output driver circuit is disclosed that reduces electron injection into the substrate by the drain of the circuit's pull-up field effect transistor. This is accomplished by adding additional circuitry that allows the gate voltage of the pull-up transistor to track the source voltage. The output circuit makes use of a tri-state inverter having an output node (hereinafter the intermediate node) coupled to VCC through a first P-channel FET, and to ground through first and second series coupled N-channel FETs, respectively. The gates of the P-channel FET and the first N-channel FET are coupled to and controlled by an input node. The intermediate node controls the gate of third N-channel FET, through which a final output node is coupled to VCC. The intermediate node is coupled to the final output node through a fourth N-channel FET, the gate of which is held at ground potential. The gate of the second N-channel FET is coupled to VCC through a second P-channel FET and to the final output node through a fifth N-channel FET which has much greater drive than the second P-channel FET; the gates of both the second P-channel FET and the fifth N-channel FET also being held at ground potential. When the final output is greater than ground potential, the gate of the second N-channel FET is at VCC. Thus the channel of the second N-channel is conductive. However, when the final output node drops below ground potential, gate voltage is greater than source voltage for both the fourth and the fifth N-channel FETs, thus causing both FETs to conduct. This results in the gate of the second N-channel FET being pulled to below ground potential, cutting off current flow through that FET. Simultaneously, the intermediate node is directly coupled to the final output node through the fourth N-channel FET, so that the voltage on the gate of the third N-channel FET tracks the source voltage thereon. Thus, current flow through the third N-channel FET is cut off and hot electron injection is mitigated. Certain variations of the circuit are possible. For example, the function of the first and second N-channel FETs may be reversed. In addition, the second P-channel FET functions as a resistor, and may be replaced with any device which functions as a resistor, including a sixth N-channel FET having its gate tied to VCC or a doped or undoped polycrystalline silicon resistor.
BRIEF DESCRIPTION OF THE DRAWINGSFIG. 1 is a circuit diagram of a conventional, double N-channel output driver;
FIG. 2 is a circuit diagram of a prior-art triple N-channel output driver which reduces hot electron injection;
FIG. 3 is a circuit diagram of a first embodiment of the new, space-efficient output driver circuit which reduces hot electron injection;
FIG. 4 is a circuit diagram of a second embodiment of the new, space-efficient output driver circuit which reduces hot electron injection;
FIG. 5 is a circuit diagram of the first embodiment of the new, space-efficient output driver circuit, but with the second P-channel FET replaced by a resistor.
FIG. 6 is a circuit diagram of the second embodiment of the new, space-efficient output driver circuit, but with the second P-channel FET replaced by an N-channel FET having its gate coupled to VCC.
FIG. 7 is a plot of gate voltage on FET QN3 vs. final output node voltage.
PREFERRED EMBODIMENT OF THE INVENTIONReferring now to FIG. 3, a first embodiment of the new, space-efficient output driver circuit has an intermediate node NM that is coupled to a power supply voltage (VCC) through a first P-channel FET QP1, and to ground through a first N-channel FET QN1 and a second N-channel FET QN2, QN1 and QN2 being series coupled, with QN1 being electrically nearer node NM. The gates of FET QP1 and FET QN1 are coupled to and controlled by an input node NI. It should be understood that FETs QP1, QN1 and QN2 can be operated as a tri-state inverter. The intermediate node NM is coupled to and controls the gate of a third N-channel FET QN3, (also referred to herein as the output node pull-up transistor) through which a final output node NO is coupled to VCC. The intermediate node NM is coupled to the node NO through a fourth N-channel FET QN4, the gate of which is permanently held at ground potential. The gate of FET QN2 is coupled to VCC through a second P-channel FET QP2, and to the final output node NO through a fifth N-channel FET QN5, which has much greater drive than FET QP2. The gates of both the FET QP2 and FET QN5 are also permanently held at ground potential. When the potential on the final output node NO is greater than ground potential, the gate of FET QN2 is at VCC. Thus the channel of FET QN2 is conductive. However, when the final output node NO drops below ground potential, gate voltage is greater than source voltage for both FET QN4 and FET QN5, thus causing the channels of both of these FETs to conduct. This results in the gate of FET QN2 being pulled to below ground potential, which reduces current flow through that FET (If the magnitude of the drop below ground potential is sufficient, current flow through FET QN2 will be cut off entirely). Simultaneous with the drop in current flow through FET QN2, the intermediate node NM is directly coupled to the final output node NO through FET QN4, so that the voltage on the gate of the FET QN3 tracks the source voltage thereon. Thus, current flow through FET QN3 is reduced or cut off, and hot electron injection into the substrate is mitigated.
Referring now to FIG. 4, a second embodiment of the new output driver circuit similar to the first embodiment, with the exception that FET QN2 is coupled to the input node NI, and FET QN1 is coupled to VCC through FET QP2 and to the final output node NO through FET QN5.
It will be noted that the second P-channel FET QP2 functions as a resistor. Hence, it may be replaced by any device which functions as a resistor, including a low-drive N-channel FET having its gate coupled to VCC or a strip of doped or undoped polycrystalline silicon which provides the desired current flow. FIG. 5 depicts the embodiment of FIG. 3 with FET QP2 replaced by a resistor R1. Likewise, FIG. 6 depicts the embodiment of FIG. 4, but with FET QP2 replaced by an N-channel FET QN6 having its gate coupled to VCC.
The new output driver circuit has a definite space-savings advantage over the circuit depicted in FIG. 2. Although the area required for both FETs Q3 and Q4 of FIG. 2 is approximately four times that required for transistor Q1 of FIG. 1, the total space required for FETs QP1, QP2, QN2, QN3, QN4, and QN5 is approximately one half that required for FETs Q3 and Q4 of FIG. 2.
Referring now to FIG. 7, gate voltage on FET QN3 and voltage on the final output node NO are both plotted as a function of time. It will be observed that when the voltage on the final output node NO drops more than a threshold voltage below ground potential, both FET QN4 and FET QN5 begin to turn on, with the result that after a transition period T, the gate voltage is clamped to the final output node. As a practical matter, the output should only drop below ground potential when there is a high logic level at the input, and the intermediate node and the final output node are at low logic levels.
Although only several embodiments of the new space-efficient, hot electron injection mitigating driver circuit is depicted, it will be obvious to those having ordinary skill in the art of integrated circuit design that changes and modifications may be made thereto without departing from the spirit and the scope of the invention as hereinafter claimed.