Temperature Insensitive CMOS Precision Current Source

1. Field of the Invention

The present invention relates to a CMOS precision current source and, more particularly, to a temperature insensitive CMOS precision current source which varies only with sheet resistance and is insensitive to other process parameters.

2. Description of the Prior Art

Many parameters of analog circuits, such as slew rate and operational amplifier gain, are a strong function of bias current. Therefore, it is important to bias analog circuits with a precision temperature independent current source. One early attempt to provide such a current source is disclosed in U.S. Pat. No. 3,813,595 issued to A. C. N. Sheng on May 28, 1974. In this arrangement, an MOS transistor which has its gate electrode connected to its drain electrode is utilized as a constant current device by means of a resistive network placed between its source and substrate electrodes. A remaining problem with this circuit, however, is that the output current is still a function of the circuit bias voltage, which will drift with changes in temperature.

Circuits which provide a constant reference voltage are also difficult to design, for many of the same reasons. One prior art arrangement for providing a temperature independent voltage reference is disclosed in U.S. Pat. No. 4,158,804 issued to W. J. Butler et al on June 19, 1979. In the Butler et al arrangement, the reference voltage circuit consists of a series-connected long channel MOS transistor and short channel MOS transistor which produce, at their junction, a temperature independent voltage. A differential circuit containing three MOS transistors is then provided with one of the transistors serving as a current source that carries the current to the other two MOS transistors which are in parallel. The gates of the two parallel MOS transistors are connected between the above-defined junction and the output. Current divides between the two parallel MOS transistors in such a way as to cause a constant output voltage to be produced regardless of supply variations. This circuit, however, remains sensitive to processing variations.

An article entitled "A Highly Stable Current or Voltage Source", by L. E. MacHattie appearing in Journal of Physics E: Scientific Instruments, Vol. 5, 1972, at pp. 1016-7, describes an arrangement which may serve as either a current or voltage source. As described, a chain of FETs is utilized, where a zero temperature coefficient is obtained at a certain current value characteristic of the individual transistor. However, the dynamic operating range of this arrangement is rather limited.

The present invention relates to a CMOS precision current source and, more particularly, to a temperature insensitive CMOS precision current source which provides an output current that varies only with sheet resistance and is insensitive to other process parameters.

It is an aspect of the present invention to provide a CMOS circuit which forces both the voltage V(T) and on-chip resistance R(T) to exhibit the same temperature dependency. That is, I.sub.o =V(T)/R(T)=constant, where I.sub.o is defined as the desired constant output current.

Another aspect of the present invention is to provide, at a minimal increase in circuitry, a high quality voltage reference which is likewise insensitive to changes in ambient temperature.

Yet another aspect of the present invention is to provide a precision source which is completely self-contained, requiring no external voltage or current sources.

Other and further aspects of the present invention will become apparent during the course of the following discussion and by reference to the accompanying drawings.

Referring now to the drawings, where like numerals represent like parts in several views:

FIG. 1 illustrates an exemplary CMOS precision current source formed in accordance with the present invention;

FIG. 2 contains a detailed illustration of a portion of the circuit of FIG. 1 which comprises a temperature dependent voltage source;

FIG. 3 illustrates an alternative embodiment of a CMOS precision current source formed in accordance with the present invention which guarantees "start-up", the presence of the desired reference current output after the power supplies are activated; and

FIG. 4 illustrates an exemplary CMOS reference voltage circuit formed in accordance with the present invention.

As previously stated, many analog systems require a current source to bias their circuits in the linear region. It is known that many circuit parameters, such as slew rate and transconductance, are a function of the bias current and, therefore, circuits require a precision current source to reduce variations in performance over processing and temperature fluctuations. Additionally, a precision reference voltage is often required, where the current source and voltage reference are often realized as two separate circuits. As will be explained in greater detail hereinafter, the precision current source formed in accordance with the present invention provides a temperature insensitive current source which approaches the minimum attainable process sensitivity for a fully integratable design, and is also capable of providing a precise voltage reference with a minimum of additional components.

The present invention provides a precision current source that will not vary with temperature by forcing the on-chip circuit voltage to vary with temperature in the same manner as the on-chip resistance. That is, V(T)/R(T)=constant, where V(T) is defined as the temperature dependent circuit voltage and R(T) is defined as the temperature dependent on-chip resistance. Therefore, in accordance with the present invention,

where I.sub.o is defined as the constant current output from the precision current source.

It is well known that on-chip resistors exhibit a temperature dependency of the form ##EQU1## where T.sub.o is defined as the reference temperature, R.sub.o is defined as the value of R(T) at temperature T.sub.o, and A, is defined as the temperature coefficient in parts per million. Equation (2) which defines R(T), is of the familiar y=mx+b form, which is known to represent a linear function. In particular, ##EQU2## where m.sub.R is defined as the slope of the function R(T) and b.sub.R is defined as the y-intercept of the function R(T). In accordance with the present inention, therefore, V(T) must also be of the y=mx+b form, where in association with equation (1), ##EQU3## where m.sub.V is defined as the slope of V(T) and b.sub.V is defined as the y-intercept of V(T). Therefore, a current source formed in accordance with the present invention is designed to provide the values of m.sub.v and b.sub.v required for equation (5) to be valid. It is to be noted that R.sub.o, the value of R(T) at T.sub.o, is a process dependent parameter, known to vary .+-.15%, and variations of R.sub.o define the minimum attainable process sensitivity for a precision current source formed in accordance with the present invention.

FIG. 1 illustrates an exemplary precision current source 10 formed in accordance with the present invention, utilizing the information associated with equations (1)-(5). Additionally, a large .DELTA.V.sub.BE1 is utilized to reduce amplifier offset errors that prior art arrangements are known to exhibit. As will be explained in detail hereinafter, all of the analog computations needed to form V(T) are performed in one feedback loop, thus eliminating any current mirroring errors which would otherwise exist. Referring now to FIG. 1, a voltage generator 12 is included in current source 10, where the output of voltage generator 12, denoted V.sub.1 (T), is a known function of temperature. An exemplary voltage generator 12 which may be utilized in association with the present invention is illustrated in FIG. 2, and will be discussed in detail hereinafter. Voltage V.sub.1 (T) is applied as an input at the junction of a resistor divider network, defined as node A, formed by a first resistor 14 and a second resistor 16, where resistor 14 is connected between node A and ground. Resistor 16, as seen in FIG. 1, is connected between node A and the emitter of a PNP transistor 18. The voltage drop across resistor 14 is defined as .DELTA.V.sub.BE1 and the voltage drop across resistor 16 is defined as .DELTA.V.sub.BE2.

In accordance with the present invention, the base of PNP transistor 18 is connected to the source of a P-channel MOS transistor 20, where the interconnection is defined as node B. A resistor 22 is connected between node B and ground. As seen in FIG. 1, the gate of MOS transistor 20 is activated by a reference signal, denoted C(T), generated by voltage generator 12. Control signal C(T) functions to adjust the gate voltage in order to have the voltage drop required across resistor 22 to provide the desired constant output current I.sub.o. As seen in FIG. 1, output current I.sub.o is the drain current associated with MOS transistor 20.

Output current I.sub.o flows through MOS transistor 20 and subsequently appears at the drain of an N-channel MOS transistor 24, defined at node C. Transistor 24 is utilized as an output device and also provides a feedback signal to voltage generator 12. In particular, MOS transistor 24 is connected to a pair of N-channel MOS transistors 26 and 28 as shown in FIG. 1, where the sources of transistors 24, 26, and 28 are connected to VSS and the drains of transistors 26 and 28 are connected to voltage generator 12. Therefore, the gate-to-source voltage (V.sub.GS) of transistor 24 will activate the gates of transistors 26 and 28 to in turn provide an activating signal which is fed back to voltage generator 12. Thus, as stated above, a constant output current I.sub.o is provided by a circuit which contains only one feedback loop.

Since V(T) is defined as the voltage drop between node B and ground, ##EQU4## where .DELTA.V.sub.BE2 (T.sub.o) is defined as the value of .DELTA.V.sub.BE2 at the temperature T.sub.o, V.sub.ref is defined as .DELTA.V.sub.BE1 +V.sub.BE, and R(T) is determined for resistor 22. Therefore, in association with equations (1) and (6), ##EQU5## Applying the y=mx+b relation, ##EQU6## Further, for a given process, as stated above, both V.sub.ref and A (defined above as the temperature coefficient of the on-chip resistance) are fixed values. Therefore, only the quantity .DELTA.V.sub.BE2 need be determined to provide the desired constant output current I.sub.o. The value of .DELTA.V.sub.BE2 needed is found by applying equation (5), which may be rewritten as follows: ##EQU7## Substituting equations (3), (4), (8), and (9) into equation (10) yields ##EQU8## where simplifying and solving for .DELTA.V.sub.BE2 yields ##EQU9## Therefore, since the equation for V(T) can be solved completely, and the desired constant output current I.sub.o is known, the value of resistor 22 needed to provide I.sub.o can be found simply by solving

In other words, the quantity R.sub.o, which is defined as the value of resistor 22 at the reference temperature, will dictate the magnitude of output current I.sub.o.

FIG. 2 illustrates in detail an exemplary voltage generator 12 which may be utilized in accordance with the present invention to supply an output voltage which is proportional to temperature and, in particular, exhibits a positive temperature coefficient. To explain the operation of voltage generator 12 in general, reference is made to a first PNP transistor 30 and a second PNP transistor 32 interconnected as shown in FIG. 2, where the emitter of transistor 30 is n times larger than the emitter of transistor 32. Transistor 30, as shown in FIG. 2, is also connected in series with a plurality of base-emitter connected transistors, each of which comprises an emitter region n times larger than that of transistor 32. Similarly, transistor 32 is connected in series with a plurality of base-emitter connected transistors, all comprising the same size emitter region as transistor 32. Also, MOS transistors 26, 50, and 52 are scaled to size with respect to transistors 28, 54, and 56 such that the current I.sub.a flowing through transistor 52 is m times smaller than I.sub.b, the current through transistor 56, or ##EQU10## It is well known that the base-to-emitter voltage of a PNP transistor is defined by the following relation: ##EQU11## where V.sub.T is equal to the well-known relation kT/q. Therefore, ##EQU12## where V.sub.BEa defines the base-to-emitter voltage associated with transistor 30 and V.sub.BEb defines the base-to-emitter voltage associated with transistor 32. Ignoring the remaining transistor structure for the moment, the signal V.sub.1 (T), which is applied to node A of the structure illustrated in FIG. 1, is equal to V.sub.BEa -V.sub.BEb, which may be expressed as ##EQU13## Simplifying this equation utilizing the relation of equation (13) yields

Stacking the plurality of PNP transistors as shown in FIG. 2 simply multiplies the value of V.sub.1 (T) in equation (18) by 4, without amplifying the offset error of operational amplifier 34.

As stated above, control signal C(T) adjusts the gate voltage of MOS transistor 20 to provide the correct voltage drop across resistor 22. Referring to FIG. 2, signal C(T) is provided by the output of operational amplifier 34. The high gain of amplifier 34 forces the emitter voltages of transistors 30 and 32 to be equal (if amplifier 34 has no offset) by controlling C(T). This feedback action allows equation (17) to be valid.

Since the circuit of the present invention is completely self-contained, a problem may exist if there is no voltage V(T) present when the circuit is first connected. Given that situation, no current I.sub.C and I.sub.D would be fed back to voltage generator 12, the pluralities of MOS transistors 36 and 38 would not be turned on, and both output voltage V.sub.1 (T) and control signal C(T) could remain equal to zero. Therefore, to protect against this situation, an additional MOS transistor 40, disposed as shown in FIG. 3, may be utilized to guarantee the presence of a nonzero V(T) under any operating circumstances. As seen by reference to FIG. 3, the drain of transistor 40 is connected to node B and the source of transistor 40 is connected to node C, where the gate of transistor 40 is connected to ground. Since the gate of transistor 40 is grounded, transistor 40 will always be conducting some current and some voltage V(T) will always be present, thus allowing the circuit to "start-up", since only the slightest voltage V(T) is needed to initiate the circuit.

As previously stated, the present invention may also be utilized to provide a precise voltage reference by incorporating a few additional devices into the circuit illustrated in FIG. 1. In particular, FIG. 4 illustrates the additional circuitry required to form a voltage reference circuit in accordance with the present invention. As shown, the reference voltage V.sub.1 (T), which as defined above is a function of absolute temperature, is applied as the positive input to an operational amplifier 42. The node A is also contained in FIG. 4 to illustrate the point of interconnection between the circuitry of FIG. 4 and the circuitry of FIG. 1. A resistor 44 is connected between the negative input terminal of operational amplifier 42 and ground, where the value of resistor 44 is equal to the value of resistor 14 of currnt source 10. Therefore, the same .DELTA.V.sub.BE1, will appear across resistor 44 as appears across resistor 14. As seen in FIG. 4, the emitter of a PNP transistor 46 is also connected to the negative input of operational amplifier 42, where the collector of transistor 46 is connected to the VSS power supply and the base of transistor 46 is connected to the output of operational amplifier 42. Transistor 46 is chosen in accordance with the present invention to exhibit a V.sub.BE2 equal to V.sub.BE1 of transistor 18 included in current source 10. Therefore, a bandgap voltage reference, defined as V.sub.o, will be formed by the addition of .DELTA.V.sub.BE1 and V.sub.BE2, where the presence of V.sub.1 (T) will allow V.sub.o to remain constant in the presence of changing temperature conditions.