Switched-Capacitor Circuit With Large Time Constant

The invention relates to the field of switched-capacitor circuits in which the operating frequencies of interest are small in comparison to the clock speed operation of the circuits and in which, therefore, large capacitance ratios are needed to realize the necessary large time constants required for operation.

The realization of very large time constant circuits (circuits which have a very large ratio between the operating clock frequency and the pole/zero frequencies) is a bottleneck in the design of many analog integrated circuits. Examples of this in voice-band circuits are power supply rejection filters, such as 60 Hz/180 Hz notch filters and automatic gain control loop filters which have pole frequencies of a few Hz. The clock frequencies used for these filters are usually much higher, for example, 128 KHz, as determined by the rest of the switched-capacitor circuitry in the system.

The preferred technology for realizing switched capacitor circuits has been and is presently metal oxide semiconductor (MOS) technology. There have been a number of solutions to the problem of designing such circuits using the MOS technology. A paper entitled "Switched-Capacitor Circuit Design" by R. Gregorian et al., Proceedings of the IEEE, Vol. 71, August 1983, pp. 941-966, surveys the field. In general, the realization of a MOS switched capacitor circuit having a large integrating time constant is achieved by replacing the input resistor of a conventional active RC integrator with a capacitive network that simulates a resistance. One such network consists of two transistor switches serially connected to an input of an operational amplifier and a capacitor connected from the common point between the switches to ground. The two switches are operated in different clock phases. The integrating time constant thus obtained depends inversely on the operating clock frequency and proportionally on the ratio of the integrating capacitance in shunt with the amplifier and the effective switched input capacitance.

Lowering the clock frequency to achieve larger time constants in the above circuit is not a viable alternative in many applications. Thus, the ratio of integrating capacitance to input capacitance must be increased. Decreasing the input capacitance to increase this ratio is not desirable because parasitic capacitances then begin to play a dominant role. Thus, the integrating capacitance must be increased. This solution, however, requires large amounts of chip area.

The problems of the above-described arrangement led to the use of the now standard input T-capacitor network. This arrangement is discussed in a paper entitled "An Area-Efficient Approach to the Design of Very Large Time Constants in Switched-Capacitor Integrators" by Sansen et al., IEEE Journal of Solid-State Circuits, Vol. SC-19, October 1984, pp. 772-779. The T-network approach has the advantage that it can easily be mixed with conventional switched-capacitor circuitry. However, it suffers from a strong sensitivity to parasitic capacitances, thus causing significant deviations in the time constant of the integrator. The high parasitic sensitivity places constraints on the design and layout of circuits using these integrators. This partially nullifies the area advantage of the technique.

Other known approaches for realizing very large time constants suffer from more serious limitations. A method proposed by M. Yamamoto, "Large Time-Constant SC Circuits and Uni-Value Capacitor SC Circuits Using Multi-Phase Clock", Proceedings of the IEEE, International Symposium on Circuits and Systems, May 1982, is parasitic-sensitive, apart from requiring a multi-phase clock. An approach given by Viswanathan et al., "Increasing the Clock Frequency of Switched-Capacitor Filters", Electronic Letters, Vol. 16, 1980, relies on switching in more than two clock phases and does not yield simple circuits. Also, the way the switching is performed between the separate functions of input attenuation and integration results in an undesirable slewing of its output.

The above problems are solved, and an advance in the art is achieved in an analog circuit comprising amplifying means, a switched capacitor network connected between a signal input and an input of the amplifying means and a feedback capacitance connected between an output of the amplifying means and the input to the amplifying means. The switched capacitor network includes a sampling capacitance which is switched during a first clock phase to sample an input signal and to transfer the resulting sample charge to the feedback capacitance. During a second clock phase, the sampling capacitance is switched to withdraw the same charge from the feedback capacitance. The circuit further includes a third capacitance which is switched in one of the clock phases to acquire a charge from the output of the amplifying means and which is switched in the other clock phase to distribute any charge acquired during the one clock phase with the feedback capacitance.

In the preferred embodiment, the sampling capacitance has one connection to the input of the amplifying means and a second connection switched in the first clock phase to the signal input and switched in the second clock phase to a reference potential. The third capacitance is connected between the output of the amplifying means and the reference potential during one of the clock phases and is connected in parallel with the feedback capacitance during the other of the clock phases.

FIG. 1 shows a prior art circuit using the T-network type of input capacitance;

FIG. 2 shows an inverting integrating circuit built in accordance with the principles of the invention;

FIG. 3 shows a non-inverting version of the circuit of FIG. 2;

FIG. 4 shows an example of a lossy integrator using the principles of the circuits of FIG. 2 and FIG. 3;

FIG. 5 shows an example of a notch filter using the principles of the circuits of FIG. 2 and FIG. 3;

FIG. 6 shows a typical frequency response curve of the notch filter of FIG. 5; and

FIGS. 7, 8 and 9 show examples of a low-pass biquadratic filter stage, a band-pass biquadratic filter stage and a stage of a ladder circuit, respectively, each using the principles of the circuits of FIG. 2 and FIG. 3.

A conventional T-network 10 switched capacitor integrator is shown in FIG. 1. Magnification of the ratio of the integrating capacitor C.sub.2 to the effective input capacitance in this circuit is achieved by attenuating the input to the integrator by means of the T-network comprising the capacitors C.sub.1, C.sub.3 and C.sub.4. The designations .phi..sub.1 and .phi..sub.2 beside switches shown in this circuit indicate that the respective switch is operated (closed) during the indicated clock phase .phi..sub.1 or .phi..sub.2. .phi..sub.1 and .phi..sub.2 are mutually exclusive. That is, while .phi..sub.1 is on, .phi..sub.2 is off, and vice versa. These same clock phase designations are used consistently in the remaining figures.

It can be shown that the time constant of the integrator of FIG. 1 is approximately given by: ##EQU1##

It is seen that the time constant is determined by the product of two capacitor ratios C.sub.4 /C.sub.3 and C.sub.2 /C.sub.1. This enables the realization of large time constants with moderate capacitor ratios. However, this circuit suffers from a strong sensitivity to parasitic capacitances. A parasitic capacitance C.sub.5 at the T-junction directly affects the effective value of C.sub.4. A more serious problem arises if there is any wiring parasitic capacitance between the bottom plates of C.sub.3 and C.sub.1. This parasitic appears as a capacitor C.sub.6 in shunt with the small effective capacitance generated by the T-network, thus causing significant deviations in the time constant of the integrator.

The circuit schematic of an inverting integrator designed in accordance with the principles of the invention is shown in FIG. 2. The operation of this circuit is as follows. During a clock phase .phi..sub.1, switch S.sub.1 is operated, an input signal V.sub.in is sampled by capacitor C.sub.3, and its resulting charge C.sub.3 *V.sub.in is transferred to a feedback capacitor C.sub.2. Also, switch S.sub.3 is operated and the resulting intermediate output voltage at the output of operational amplifier A1 is sampled by capacitor C.sub.1. Thereafter, during clock phase .phi..sub.2, switches S.sub.2 and S.sub.4 are operated, and S.sub.1 is opened causing C.sub.3 to withdraw the charge C.sub.3 *V.sub.in from C.sub.2. During .phi..sub.2, C.sub.1 distributes its charge with C.sub.2. Physically, this entire operation is approximately equivalent to attenuating the input voltage V.sub.in by a factor C.sub.3 /C.sub.2, and then integrating it onto C.sub.2 through the capacitor C.sub.1. Thus, the capacitor C.sub.2 performs the dual functions of input attenuation and integration without having to break the feedback loop around the amplifier A1. It can be shown that the time constant of this integrator is approximately given by: ##EQU2##

Equation 2 shows that we can realize large time constants with modest values for the capacitor ratios C.sub.2 /C.sub.1 and C.sub.2 /C.sub.3. Moreover, this integrator is insensitive to all parasitics to ground because every capacitor node point is held at a nonfloating potential during every clock phase, as opposed to the T-network of FIG. 1, for example, where the common node between C.sub.1, C.sub.3 and C.sub.4 is at a floating potential during .phi..sub.1. Also, since the capacitor C.sub.3 of FIG. 2 is significantly larger than the small effective input capacitance of the T-network, the circuit of FIG. 2 is less sensitive to wiring capacitance. The circuit is simpler than the T-network integrator in terms of the number of capacitors and switches More important, it is seen that both the ratios C.sub.2 /C.sub.1 and C.sub.2 /C.sub.3 share the same large capacitor C.sub.2, unlike in the T-network approach. This can result in a saving of as much as 50% in total required capacitance.

The circuit schematic of a lossless non-inverting integrator, based on the circuit of FIG. 2, is shown in FIG. 3. This circuit differs from the inverting integrator of FIG. 2 only in the clock phases used to operate switches S.sub.3 and S.sub.4.

The invention can be extended to realize lossy integrators, as shown in FIG. 4. A lossy inverting integrator is obtained by using the clock phases which are shown outside the parentheses in FIG. 4. A non-inverting lossy integrator is obtained by using the clock phases shown inside the parentheses. In addition to magnifying the ratio C.sub.2 /C.sub.3 in accordance with the principles already discussed, the ratio C.sub.2 /C.sub.4 is also enhanced in accordance with the same inventive principles, which is desirable in a lossy type of circuit. This is achieved by virtue of the fact that the amplifier output is an input to C.sub.4 in the same way that the input signal V.sub.in is an input to C.sub.3. The transfer function of this integrator can be shown to be: ##EQU3##

Equation 3 shows that a very low pole frequency can be achieved without the need for very large capacitor ratios.

The integrators of FIGS. 2 through 4 can be used as generic building blocks to realize a variety of switched-capacitor filters including first-order sections, biquads and ladder realizations. The circuit schematic of one possible implementation of a notch filter, using the principles of the invention, is shown in FIG. 5. Such a filter is generally employed as a power supply rejection filter in analog systems. This circuit is based on the well-known E-type biquadratic section. The E-type quadratic section is discussed in detail in an article entitled "A Family of Active Switched-Capacitor Biquad Building Blocks" by P. Fleisher et al. in the Bell System Technical Journal, Vol. 58, December 1979, pp. 2235-2269. A differential configuration has been chosen here because of the particular application for which this filter was designed. A single-ended version can be realized along the same lines. It is seen that the clock phases for the two integrators I1 and I2 in the biquad are staggered. This is required because the output of each integrator is valid only during one of the two clock phases. Similarly, in ladder structures using these integrators, the clock phases of the integrators which form loops must be staggered.

A 60-Hz notch filter based on the configuration of FIG. 5 has been designed to work with a clock frequency of 126 KHz. This filter has been implemented in 1.5 micron CMOS. The filter uses 250 units (one unit=0.25 pf) of capacitance, a significant saving over the T-network implementation which needs about 325 units of capacitance. The chip area saving, in practice, is higher because special design and layout constraints do not have to be used to overcome the parasitic sensitivity of the T-network. The frequency response of this notch filter is shown in FIG. 6. Measurements on several samples showed a standard deviation of 0.5% in the notch frequency. In contrast, T-network implementations have often shown deviations in excess of 5%.

Other filter examples using the integrators of FIGS. 2 and 3 are shown in FIGS. 7 through 9. Specifically, FIG. 7 shows a low-pass biquadratic filter stage. This stage is based on the well-known E-type biquadratic topology described in the above-cited Bell System Technical Journal. In particular, this stage realizes the z-domain transfer function of the type ##EQU4## where k, .alpha. and .beta. are constants.

FIG. 8 shows a band-pass biquadratic stage. This stage is also based on the E-type biquadratic topology. It realizes the transfer function of the type ##EQU5##

FIG. 9 shows an integrating loop stage for a ladder filter. The loop contains two integrators consisting of an inverting lossless integrator I1 and a noninverting lossless integrator I2. Several such integrator loops may be interlaced to form ladder filters of the types described in an article by R. Gregorian, "Switched-Capacitor Circuit Design," Proc. IEEE, Vol. 71, pp. 941-966, August 1983.

It is to be understood that the above-described arrangements are merely illustrative of the application of principles of the invention and that other arrangements may be devised by workers skilled in the art without departing from the spirit and scope of the inventions.