The present invention relates to circuits for generating digital clock signals. Many different circuits require a generation of a digital clock signal having a nearly 50% duty cycle with sharp rise and fall times. Such clock signals are typically generated using a crystal oscillator and level shifting buffer circuits that convert the crystal oscillator output to a signal having the voltage and current levels expected by logic circuits.
Once such circuit is described in a paper by Riley, et al. entitled “Techniques for In-Band Phase Noise Reduction in ΔΣSynthesizers”, IEEE Transactions on Circuits and Systems—II, Analog and Digital Signal Processing, Vol. 50, No. 11, November 2003, at pages 800-802. In that circuit, a self-biasing crystal oscillator provides a sinusoidal output signal via a single ended, subthreshold output drive transistor pair. The output signal is a sinusoid with an amplitude of approximately eight hundred millivolts (mV) peak to peak.
A Complimentary Metal Oxide Semiconductor (CMOS) level shifter then converts this sinusoid to a digital signal having full CMOS swing logic levels of approximately 3 volts. The level shifter has a first stage similar to a low noise amplifier (LNA) and a second stage that acts as a latch.
The first stage of the level shifter is provided by an Alternating Current (AC) bipolar, differential transistor pair. The first stage may use resistive loads; Direct Current (DC) bias levels are provided by a pair of diodes connected between the input stages and a voltage reference.
The second stage use cross coupled transistors and resistive load transistors to mirror a push-pull CMOS output stage, to generate complimentary outputs with the required voltage swing.
There are several shortcomings with the clock circuit design described in the Riley article. For one, the use of a simple linear differential amplifier as a first stage, although designed as a low noise amplifier, will actually add noise to the resulting output. This is due to the fact that it operates largely in the linear region.
In addition, the single latch stage provided by the second stage will still exhibit some level of jitter.
The present invention is a differential clock input buffer circuit that converts single and/or differential input signals into a pair of complimentary, square wave digital outputs. The input signal(s) may have a sine wave, square wave or other shape; indeed the operation of the circuit is relatively independent of the exact input wave form shape, its amplitude, or even its common mode voltage.
The circuit can be used to produce a nearly 50% duty cycle complimentary digital output. It exhibits improved noise immunity and less jitter as compared to previous clock circuit designs.
In brief summary, the invention uses a two stage approach to level shifting, adapting concepts of low-noise oscillator design to provide two stages of regeneration in a clock buffer circuit. In particular, a first stage latching circuit consists of a pair of cross coupled bi-stable transistors with resistive loads to provide gain, limiting, hysteresis, and latching functions. The transistors in the first stage are designed to be of a relatively small size in the available circuit technology, to provide as fast as possible switching performance.
These transistors operate in a linear region, as controlled by the cross-coupled latch arrangement which creates positive feedback, to restrict operation to a small range of input voltages. A voltage to current converter may be used to drive the first stage.
A second stage latching circuit, which may use circuits adapted to mirror currents in the first stage, is also provided as a pair of cross coupled transistors with resistive loads. The second stage latch provides further gain, limiting, hysteresis, and latching, and is preferably biased at the center of an output buffer's range, to provide better duty cycle control over a wider range of signal levels.
The circuit provides improved input hysteresis for greater noise immunity, and limits jitter production by ensuring that the edges of the resulting clock signals are as fast as possible, while minimizing the time that the clock buffer spends in the linear region.
The foregoing and other objects, features and advantages of the invention will be apparent from the following more particular description of preferred embodiments of the invention, as illustrated in the accompanying drawings in which like reference characters refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention.
A description of preferred embodiments of the invention follows.
The input signal(s) may be single or double ended. If a single ended signal, it is provided at terminal IN (or just at INN, with the other input terminal biased somewhere near vdd/2); if double ended, they are provided to inputs IN and INN. The voltage to current converter 30 converts these input voltage(s) to differential currents at nodes VdiffA and VdiffB. The voltage to current converter 30 consists of a pair of CMOS transistors 33, 34 arranged as a differential amplifier, to provide current from a source Vss (and to reject common-mode signals and noise).
A reference/biasing circuit 20 provides a reference current to differential transistor pair 33, 34. The reference/biasing circuit may consist of adjustable resistive load 21 and multiple drive transistors 22, 23, 24. Circuit 20 could also be implemented as cascade current source/mirror for somewhat improved performance.
First stage latch 40 consists of a pair of bi-stable, cross coupled transistors 42, 43. Nodes VdiffA and VdiffB provide source terminal inputs to transistors 42, 43, respectively. Transistors 42, 43 are diode-connected field effect transistors (FETs) by resistive load transistors 44, 45. The ratio of the gate widths for each FET with respect to its respective load (i.e., the ratio of the gate width for transistor 44 to that of transistor 42 and of transistor 45 to that of transistor 43) are selected to control the desired switching and hysteresis characteristics. Also, the ratio of the input differential pair 42, 43 gate width to the load transistor 44, 45 gate width determines the first stage linear gain (i.e., the ratio of the input transconductance to load transconductance.) If speed of operation is important, the transistors 42, 43 used in the first stage latch 40 are selected to be of a relatively small size. For example, they may have a gate width of only six microns or less. By providing the first stage 40 with transistors 42, 43 that are as small as possible, they will switch between the two bi-stable states as fast as fast as possible. The size selected for these transistors 42,43 does involve a tradeoff, however, between the desired amount of balanced drive current versus the desired hysteresis.
Turning attention briefly to
Nodes VdiffA and VdiffB are also coupled to drive respective inputs of the mirroring circuit 50. Mirroring circuit itself consists of transistors 52, 53 and respective resistive loads 54, 55. The mirroring transistors 52, 53 are set to be the same size as the transistors 42, 43 in the first stage. These transitors 52, 53 serve to mirror currents in the first stage at at nodes Pout and PoutN.
Nodes Pout and PoutN are used as nodes for the second stage. Transistors 54, 55 are arranged to set the bias at nodes Pout and PoutN so as to be centered at approximately the threshold of output buffers 70. This provides for better control over duty cycle over a wide range of input levels.
The second stage latch 60 is similar in configuration to the first stage latch 30. It is again provided by a pair of bi-stable, cross coupled differential pair transistors 62, 63 and their respective resistive loads provided by transistors 64, 65. The loads in the second stage are referenced to Vss. As with the first stage, ratio of the gate widths of the differential pair 62, 63 versus the loads 64, 65 are chosen to control hysteresis and gain. The second stage latch 60 is designed to provide an even sharper response over an even smaller variation of input voltage due to its gain, limiting, and hysteresis.
Nodes Pout and PoutN are also coupled to buffers 70. Buffers 70, in the illustrated embodiment are provided by inverters 71, 72, 73, 74. Series connected inverters 71, 72 are coupled to the Pout node to produce output signal OutN; and series connected inverters 73, 74 operate on the PoutN node to provide output signal Out. The inverters 70 are implemented in the desired Complimentary Metal Oxide Semiconductor (CMOS) technology.
It is understood now that clock input buffer 10 can be used to convert single ended or differential sign wave or square wave inputs to complimentary square wave digital outputs with low jitter and nearly 50% duty cycle. These outputs are maintained over a wide range of input power levels, exhibiting very little change in duty cycle.
The top trace in these figures shows an input triangle waveform for the “plus” and “minus” inputs, Vin and Vinn. ranging from 700 millivolts (mv) to 1.1 volt (v).
The second trace illustrates a typical output of the first stage 40 latch, at nodes VdiffA and VdiffB. Note the effect of hysteresis is to delay the crossover point “A”; but the change in response amplitude is now much more rapid.
The third trace in
The plots of
While this invention has been particularly shown and described with references to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the invention encompassed by the appended claims.
This application claims the benefit of U.S. Provisional Application No. 60/585,682, filed on Jul. 6, 2004. The entire teachings of the above application(s) are incorporated herein by reference.
Number | Date | Country | |
---|---|---|---|
60585682 | Jul 2004 | US |