BACKGROUND OF THE INVENTION
Field of the Invention
This invention relates to an output buffer circuit, and specifically to an output buffer circuit having switching noise reduced by reducing the difference in the propagation delays between a pull up signal path and a pull down signal path through the buffer circuit.
DESCRIPTION OF THE PRIOR ART
Output buffer circuits are well known; one is shown schematically in FIG. 1. Such buffer circuits are typically used in integrated circuits to provide the output signals from the integrated circuit die to the integrated circuit pad, from which the signal is provided outside the integrated circuit. Low switching noise for such buffer circuits is especially desirable for integrated circuits having a large number of outputs that may be switched simultaneously. However, in the conventional output buffer circuit of FIG. 1 where both the pull up transistor Q1 and the pull down transistor Q2 are N-channel MOS transistors, there is an intrinsic skew in the signal paths associated with each of these transistors that gives rises to a substantial "crowbar" (shorting) current during output switching.
A large crowbar current undesirably increases the electrical noise in the power supply of the integrated circuit. This intrinsic skew or propagation delay differential is due to a one gate delay difference between the pull up transistor Q1 signal path and the pull down transistor Q2 signal path. This is seen in FIG. 1, where the pull up transistor Q1 includes in its signal path the output enable (OE) input terminal, NAND gate 10 and inverter 12 producing output A. However, the pull down transistor Q2 includes in its signal path the output data (D) input terminal D, inverter 14, NAND gate 16, and inverter 18. Thus the signal path of pull down transistor Q2 includes an extra logic gate (i.e. inverter 14), causing extra propagation delay.
This is illustrated in the timing diagrams of FIG. 2 where the top line represents the data signal D showing this signal going low for a particular time. The middle portion of FIG. 2 shows the signals at points A and B (i.e. at the outputs of respectively inverter 12 and inverter 18), showing that the signal at A goes low before the signal at B goes high. This creates a very small amount of noise (waveform i.sub.CB when signal D is falling. However when signal D is rising, it can be seen that signal A rises before signal B falls, thus causing a substantial amount of noise in waveform i.sub.CB at the time of the D signal rising input. This is because both of transistors Q1 and Q2 are on at this time, shorting the power supply to ground. The large "crowbar" current i.sub.CB thereby undesirably increases the electrical noise in the power supply of the chip during output switching. That is, the crowbar current affects the internal chip power supply by undesirably shorting together the positive voltage power supply and ground.
SUMMARY OF THE INVENTION
In accordance with the invention an output buffer circuit uses a single non-inverting CMOS pass gate in the output data signal path, instead of two inverting gates, to better equalize the pull up and pull down transistor signal paths, thus reducing the undesirable crowbar switching current. Thus the signal paths associated with both the pull up and pull down output transistors include equal numbers of series connected logic gates, reducing the delay skew present in the prior art. This provision of balanced length signal paths evens out the rise and fall of the crowbar switching currents, minimizing the electrical noise produced therefrom.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows a prior art output buffer with NMOS output transistors Q1, Q2.
FIG. 2 shows timing diagrams for the output buffer in FIG. 1.
FIG. 3 shows an output buffer in accordance with the invention.
FIG. 4 shows timing diagrams for the output buffer of FIG. 3.
FIGS. 5 and 6 show variations on the circuit of FIG. 3.
FIGS. 7A through 7C show variations on the transmission gate of FIG. 6.
FIGS. 8 and 9 show an output buffer in accordance with the invention in which only a single input signal is provided.
DETAILED DESCRIPTION OF THE INVENTION
An output buffer in accordance with the invention shown schematically in FIG. 3 includes the two conventional inverters 12 and 18 connected respectively to output transistors Q1 and Q2, Q1 being the pull up transistor and Q2 being the pull down transistor. Inverter 12 and transistor Q1 are the first output stage, and inverter 18 and transistor Q2 are the second output stage, both output stages being conventionally connected to pad 24.
The output enable OE signal and inverse output data signal DN are respectively the same and inverse of signals OE and D of FIG. 1. The lower illustrated portion of the output buffer of FIG. 3 includes conventional NAND gate 32, the output terminal of which is connected to the input terminal of output inverter 18. The input signals to NAND gate 32 are the output enable signal OE and the inverse output data signal DN.
In the upper portion of FIG. 3 the second data path connecting to transistor Q1 is shown including inverter 34 for logic control connecting to the gate of the P-channel MOS transistor 36. P-channel MOS transistor 36 is paired with complementary N-channel transistor 38 to form a CMOS pass gate. The output enable signal OE is provided via inverter 34 to the gate of transistor 36 and is provided directly to the gate of transistor 38. Thus the CMOS pass gate (which includes transistors 36 and 38) is controlled by the output enable signal OE and passes the inverse output data signal DN as shown to the input terminal of inverter 12.
Transistor 40 is a P-channel MOS transistor connected between a voltage source and the input terminal of inverter 12, and having its gate connected to the output enable signal OE terminal. Thus the CMOS pass gate which includes transistors 36 and 38 passes the inverse output data signal DN to the input terminal of inverter 12 at the same time transistor 40 is providing the voltage connected to its source terminal to the input terminal of inverter 12 whenever the output enable signal OE is low. Thus the delay skew from the DN signal terminal to points A and B is reduced, because both the upper (Q1) and the lower (Q2) signal paths have approximately the same number of gates, i.e. two gates in the logic path. In the case of the lower signal path these are NAND gate 32 and inverter 18, and in the case of the upper signal path they are the CMOS pass gate including transistors 36 and 38 and inverter 12. Note that transistors 36, 38, and 40 and inverter 34 together form a NAND gate having one inverted input signal.
FIG. 4 shows a timing diagram for the circuit of FIG. 3. The top waveform labelled DN is the inverse of the D waveform of FIG. 1, going high for a particular period as shown (and hence equivalent to data signal D going low in FIG. 2). The central portion of FIG. 4 shows the waveforms at points A and B of the circuit of FIG. 3, i.e. at the output terminals of inverters 12 and 18 respectively, showing the switching occurring so that the A signal goes low at the same time the B signal goes high and vice versa. This produces the resulting crowbar current labelled i.sub.CB in the lower part of FIG. 4, with minimum switching noise at both of the switching points where A and B cross over in the central portion of FIG. 4. Thus the relatively balanced signal paths for transistors Q1 and transistor Q2 even out the rise and fall crowbar current and minimize switching noise.
FIGS. 5 and 6 show other variations on the invention. The circuit of FIG. 5 does not include inverters 12 and 18 of FIG. 3, and therefore NOR gate 52 replaces NAND gate 32, and transistor 50 is tied to ground rather than VCC. Also the output enable signal OE is of the opposite polarity from the output enable signal OE of FIG. 3. FIG. 6 differs from FIG. 3 by providing a positive DATA signal, and using the conventional NAND gate 62 to drive the pull-up transistor Q1 and the transmission gate 66 along with inverter 64 and pull-up transistor 60 to drive pull-down transistor Q2.
FIGS. 7A, 7B, and 7C show variations on the transmission gate 66 which may alternatively be used. Transistors 72 (FIG. 7A) and 74 (FIG. 7B) replace CMOS transmission gate 66 of FIG. 7C and FIG. 6.
FIGS. 8 and 9 show other embodiments in which no enable signal OE is provided. The VREF signal may be equal to VCC or may be higher or lower than VCC in order to control delay, optimizing the switching characteristics.
The above description of the invention is illustrative and not limiting; it is to be appreciated that other logic elements may be substituted for those disclosed herein, and it is intended that all modifications of the above invention fall within the scope of the appended claims.