FET circuit with resistor-programmable transconductance

Abstract

A circuit is presented which sets the transconductance of a FET using a resistor. The circuit comprises a resistor and first and second FETs series-connected in sequence between a supply voltage and a circuit common point, and third and fourth FETs and a bias current source series-connected in sequence between the supply voltage and the circuit common point. The drain and gate of the fourth FET are connected to the gate of the second FET and the gates of the first and third FETs are cross-coupled to the drains of the third and first FETs, respectively. The bias current source provides a starting current for the circuit. When so arranged, and with the threshold voltages of the first and second FETs matched, the transconductance of the second FET is directly proportional to 1/R1. The circuit can in turn be used to bias other transistors in a reproducible way to fix the transconductance of an amplifier according to the selected resistor value.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates to the field of transconductance amplifiers, and particularly to methods of controlling the transconductance of MOS transistors.

2. Description of the Related Art

A useful element of electronic circuit design is the transconductance amplifier; i.e., a voltage-to-current amplifier having an associated transconductance g m (gain) which is expressed in units of A/V. Such amplifiers are constructed from transistors, which exhibit transconductance. However, the transconductance of a transistor is influenced by several factors which can adversely affect amplifier performance. For example, the transconductance of a MOS transistor depends on a number of factors, including transistor size, operating current, and temperature. These last two dependencies are subject to broad variations due to manufacturing tolerances, making it difficult to accurately set the transconductance and maintain it against temperature variation.

One approach to minimizing transconductance variation is described in “Understanding Wide-Band MOS Transistors”, J. Steininger, IEEE Circuits and Devices, Vol. 6, No. 3, pp. 26-31 (May 1990). Steininger proposes a circuit which is a MOS version of a self-biasing Widlar current source, comprising two PMOS and two NMOS transistors (two of which are diode-connected) and a resistor R b . When the transistors are properly sized, the transconductance becomes directly proportional to the resistance of R b .

However, this circuit suffers from a couple of weaknesses. First, two of the circuit's transistors operate at zero gate-drain potential, while the other two do not. This subjects the second two transistors to the channel length modulation effect. Since these transistors experience the effects of variations in the supply voltage, their characteristics may be different than those of the first two transistors, and a biasing error may result.

A second problem is that the circuit has a second equilibrium state, when all currents are zero. Overcoming this “start-up” problem requires a start-up circuit, which typically introduces a small current into one of the diode-connected transistors. Unfortunately, this current inevitably corrupts the circuit by ensuring that one of the diode currents is larger than the other, which worsens the impact of the channel length modulation effect problem discussed above.

Another approach is described in Johns and Martin, “Analog Integrated Circuit Design”, pp. 248-251 (1997). Here, a bias circuit is made from two PMOS and four NMOS transistors and a resistor. A cascading arrangement reduces the aforementioned channel length modulation effect somewhat, but not completely. In addition, the starting problem mentioned above also afflicts this circuit.

SUMMARY OF THE INVENTION

A FET circuit is presented which overcomes the problems noted above. The four-transistor circuit sets the transconductance of a FET using a resistor. The channel length modulation effect is practically eliminated, there is no starting problem, and all transistors are of like polarity.

The circuit comprises a resistor R 1 and first and second FETs series-connected in sequence between a supply voltage and a circuit common point, and third and fourth FETs and a bias current source series-connected in sequence between the supply voltage and the circuit common point. The drain and gate of the fourth FET are connected to the gate of the second FET and the gates of the first and third FETs are cross-coupled to the drains of the third and first FETs, respectively. The bias current source provides a starting current for the circuit. All FETs are of like polarity.

When the circuit is arranged as described, and the threshold voltages of the first and second FETs are matched, the transconductance of the second FET is directly proportional to 1/R 1 . Since resistors can be made much more predictably than transistors, the invention makes the transconductance of the second FET much more robust against temperature change and manufacturing variability than it might be otherwise. The current and/or operating voltages of the FET can in turn be used to bias other FETs in a very reproducible way to fix the transconductance of an amplifier according to the selected resistor value.

Further features and advantages of the invention will be apparent to those skilled in the art from the following detailed description, taken together with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram of a FET circuit with resistor-programmable transconductance per the present invention.

FIG. 2 is a schematic diagram of an alternative embodiment of a FET circuit with resistor-programmable transconductance per the present invention.

FIG. 3 is a schematic diagram illustrating a first method by which the present invention can be used to bias a differential input stage.

FIG. 4 is a schematic diagram illustrating another method by which the present invention can be used to bias a differential input stage, and/or additional circuitry.

FIG. 5 is a schematic diagram illustrating a method by which the present invention can be used to bias the input and output stages of a differential amplifier.

DETAILED DESCRIPTION OF THE INVENTION

An embodiment of a FET circuit with resistor-programmable transconductance per the present invention is shown in FIG. 1 . The circuit includes a first FET MP 1 connected between a supply voltage V+ and a node 10 , a resistor R 1 connected to V+ at one terminal, and a second FET MP 2 connected between R 1 's second terminal and a node 20 . A third FET MP 3 is connected between node 10 and a bias current source I 1 , and a fourth FET MP 4 is connected between node 20 and a circuit common point V−. The gates of MP 1 and MP 2 are cross-coupled to their opposite drains, such that MP 1 's gate is connected to node 20 and MP 2 's gate is connected to node 10 . MP 3 is diode-connected, and its gate/drain are connected to the gate of MP 4 .

It is seen by inspection that:

V gs1 +V gs4 =I 2 *R 1 +V gs2 +V gs3

where V gsx is the gate-source voltage for FET MPx, and I 2 is the current in R 1 . Since MP 1 and MP 3 are in series, each carries an equal current. If MP 1 and MP 3 are matched:

V gs1 =V gs3

and therefore:

V gs4 −V gs2 =I 2 *R 1.

If the threshold voltages are matched such that:

V t4 =V t2 =V t ,

where V tx is the threshold voltage of transistor MPx, then:

( V gs4 −V t )−( V gs2 −V t )= I 2 *R 1.

Since MP 1 and MP 3 are in series with R 1 :

I 2 =ID 2 =ID 4 =ID,

where IDx is the drain current in transistor MPx.

V gs −V t is given by: $$ V gs - V t = 2 * ID B

where B is the KP*W/L “beta” factor, with KP=μC ox and W/L being the FET's ratio of width to length.

Thus: $$ 2 * ID B4 - 2 * ID B2 = I2 * R1 ,

where B 4 and B 2 are the beta factors of MP 4 and MP 2 , respectively.

The transconductance G m of a FET is given by:

G m ={square root over (2 *B*ID )},

so, by substitution: $$ G m4 = 2 * [ 1 - 2 * B4 B2 ] R1 (Eq.  1)

where G m4 is the transconductance of FET MP 4 .

This means that the transconductance of MP 4 is stabilized at a value which is directly proportional to 1/R 1 and is a function of the ratio of the sizes of MP 4 and MP 2 .

The transconductance of a transistor depends in part on its operating current. The configuration of the present FET circuit automatically forces an operating current (I 2 ) to flow in MP 4 which results in G m4 being directly proportional to 1/R 1 . The operating current necessary to achieve this relationship may vary from FET to FET or from temperature to temperature, but the present circuit always forces the necessary MP 4 operating current regardless.

This “resulting operating current” can be put into an identical transistor (i.e., made at the same time, on the same chip, and at the same temperature as MP 4 ), which will then have the same transconductance as MP 4 . Alternatively, G m4 can be scaled up or down by putting the resulting operating current into a FET having a known size ratio to MP 4 . The resulting operating current or a current derived from it could then be used to, for example, bias the FET transistors making up a differential input stage of an amplifier.

Rearranging one of the equations above gives:

( V gs2 −V t )+ I 2 *R 1=( V gs4 −V t )

That is, the voltage from V+ to node 10 is equal to the gate-source voltage of MP 4 . Thus, connecting the gate of a FET MP 5 to node 10 and its source to V+ results in a drain current which is directly proportional to I 2 . Therefore, the node 10 voltage can be used to bias MP 5 , as well as other FETs MP 5 a , Mp 5 b , etc., connected in parallel with MP 5 , to provide respective currents equal to or a controlled multiple of I 2 .

Equation 1 is a correct expression for G m4 when MP 1 and MP 3 are matched. However, it is not essential that MP 1 and MP 3 be matched. Even if MP 1 and MP 3 are not matched, G m4 is still directly proportional to 1/R 1 ; however, the size ratio between MP 1 and MP 3 will now appear in the expression for G m4 .

The drain-to-gate voltages of FETs MP 1 , MP 2 and MP 3 are shielded from the supply voltage, and therefore do not suffer from the channel length modulation effect discussed above. MP 4 , however, is not shielded, and thus supply voltage variations might adversely affect MP 4 's drain current-to-gate voltage relationship. One way in which this can be mitigated is with the use of a cascoding arrangement as shown in FIG. 2 . Here, a diode-connected PMOS FET MP 5 is connected between the drain of MP 3 and I 1 , and a PMOS FET MP 6 is connected between the drain of MP 4 and circuit common, with the gates of MP 5 and MP 6 connected together. When so arranged, the drain-to-gate voltage of MP 4 is given by the difference between V gs5 and V gs6 . This voltage is much less than the supply voltage, and as a result, channel length modulation errors due to variations in supply voltage are greatly reduced. When so arranged, the “resulting operating current” used to bias another FET or circuit would be taken at the drain of MP 6 , rather than at MP 4 's drain as in FIG. 1 .

Note that, though the schematics shown in FIGS. 1 and 2 show all the FETs being PMOS, the present circuit can also be realized with NMOS FETs. Each FET is preferably fabricated with an isolation well, which is connected to the FET's source to eliminate any body effect.

The circuit is started by bias current I 1 . The cross-coupled gates of MP 1 and MP 2 ensure that over a broad range, operating current I 2 is essentially insensitive to the magnitude of the bias/start-up current. Thus, the present circuit can be started and run with almost any available current, which can even be allowed to vary while the circuit is in operation.

One application for the present FET circuit is shown in FIG. 3 . Here, the circuit provides bias current to a FET differential input stage connected between MP 4 and circuit common point V−. The input stage is made from PMOS FETs MP 6 and MP 7 connected as a differential pair, the outputs of which are combined using NMOS current mirror transistors MN 1 and MN 2 to produce a single-ended output which would typically be used to drive an output stage. The present circuit can be used to set the overall transconductance of the differential input stage. Assume, for example, that MP 7 and MP 8 are sized such that each has a W/L twice as large as that of MP 4 . Each of MP 7 and MP 8 receives half the current from MP 4 , so that when their respective outputs are combined by MN 1 and MN 2 , the input stage will have the same overall G m as MP 4 —i.e., the G m for the input stage will be directly proportional to 1/R 1 . In this way, the transconductance of a differential input stage can be set to a desired value by simply properly sizing R 1 .

Another method by which a differential input stage can be biased and its transconductance set to a desired value is shown in FIG. 4 . Here, the gate of MP 5 is connected to node 10 , thereby producing a bias current which is directly proportional to I 2 . A differential input stage is connected between MP 5 and V−, and is made from PMOS FETs MP 9 and MP 10 , the outputs of which are combined using NMOS current mirror transistors MN 3 and MN 4 to produce a single-ended output. This arrangement also sets the transconductance of the differential input stage. Assume, for example, that MP 5 's W/L is twice that of MP 4 , and that the W/L of each of MP 9 and MP 10 is equal to that of MP 4 . When so arranged, the bias current delivered to the input stage by MP 5 is equal to 2*I 2 , which is split by MP 9 and MP 10 such that each operates at nominally the same current as MP 4 . When the outputs of MP 9 and MP 10 are combined by MN 3 and MN 4 , the differential input stage will have the same overall G m as MP 4 . The same result could be achieved by, for example, making MP 5 the same size as MP 4 , and making MP 9 and MP 10 twice the size of MP 4 . In either case, the transconductance of the differential input stage is directly proportional to 1/R 1 . And, as noted above, additional circuits can be similarly biased by connecting additional PMOS FETs such as MP 5 a and MP 5 b in parallel with MP 5 .

Another application of the invention is shown in FIG. 5 , which employs the present FET circuit to establish the G m of the differential input stage of an operational amplifier, which in turn serves to control the bandwidth and frequency stability of the op amp. Here, the present FET circuit is configured as in FIG. 4 , with FET MP 5 connected to node 10 to provide a bias current which varies with 1/R 1 to the MP 9 /MP 10 differential input stage. The output produced by the input stage is fed to the gate of a second stage NMOS FET MN 5 , which has a compensation capacitor C c connected between its gate and drain. The output of MN 5 is fed to an output buffer made from a pair of NMOS FETs MN 6 and MN 7 ; the amplifier's output V out is taken at the junction of MN 6 and MN 7 . A PMOS FET MP 11 is connected in parallel with MP 5 to provide a bias current which varies with 1/R 1 to the second stage.

The unity gain bandwidth of an op amp as shown in FIG. 5 is set by the ratio of the input stage transconductance to compensation capacitance; i.e., the unity gain frequency f u is given by g m /2πC c . Thus, the invention enables the op amp's bandwidth and frequency stability to be easily set by properly selecting the device sizes and the resistance R 1 .

While particular embodiments of the invention have been shown and described, numerous variations and alternate embodiments will occur to those skilled in the art. Accordingly, it is intended that the invention be limited only in terms of the appended claims.

Claims

1. A field-effect transistor (FET) circuit having a resistor-programmable transconductance, comprising:a resistance having a value R1, first and second FETs, said resistance and said first and second FETs series-connected in sequence between a supply voltage and a circuit common point, third and fourth FETs, and a bias current source, said third and fourth FETs and said bias current source series-connected in sequence between said supply voltage and said circuit common point, the drain and gate of said fourth FET connected to the gate of said second FET and the gates of said first and third FETs cross-coupled to the drains of said third and first FETs, respectively, said bias current source arranged to provide a starting current for said FET circuit, said first, second, third and fourth FETs being of a first polarity and the threshold voltages of said first and second FETs matched such that the transconductance Gm2 of said second FET is directly proportional to 1/R1.

2. The FET circuit of claim 1, said third and fourth FETs connected together at a first node, further comprising a fifth FET having its gate connected to said first node and its source referred to said supply voltage such that said fifth FET conducts a current and has a transconductance which are directly proportional to the current in and the transconductance of said second FET, respectively.

3. The FET circuit of claim 1, wherein said third and fourth FETs are matched such that the transconductance Gm2 of said second FET is given by: Gm2=2*(1-B2B1)R1wherein B2=μ⁢ ⁢Cox*W2L2⁢ ⁢and⁢ ⁢B1=μ⁢ ⁢Cox*W1L1,where W2 and L2 are the width and length, respectively, of said second FET, W1 and L1 are the width and length, respectively, of said first FET, and μ and Cox are the channel mobility and oxide capacitance, respectively, for said second and first FETs.

4. A field-effect transistor (FET) circuit having a resistor-programmable transconductance, comprising:a resistance having a value R1, a first FET, said resistance connected between a supply voltage and said first FET's source, a second FET having its current circuit connected between said first FET and a circuit common point, said second FET's source connected to said first FET's drain at a first node, a third FET having its current circuit connected between said supply voltage and a second node and its gate connected to said first node, the gate of said first FET connected to said second node, a fourth FET which has its source connected to said second node and its gate and drain connected to the gate of said second FET, and a bias current source connected between said fourth FET's drain and said circuit common point which provides a starting current for said FET circuit that forces a current I1 to flow in said resistance and said first and second FETs, said first, second, third and fourth FETs being of a first polarity and the threshold voltages of said first and second FETs matched such that said current I1 gives a transconductance Gm2 for said second FET which is directly proportional to 1/R1.

5. The FET circuit of claim 4, wherein said third and fourth FETs are matched such that the transconductance Gm2 of said second FET is given by: Gm2=2*(1-B2B1)R1wherein B2=μ⁢ ⁢Cox*W2L2⁢ ⁢and⁢ ⁢B1=μ⁢ ⁢Cox*W1L1,where W2 and L2 are the width and length, respectively, of said second FET, W1 and L1 are the width and length, respectively, of said first FET, and μ and Cox are the channel mobility and oxide capacitance, respectively, for said second and first FETs.

6. The FET circuit of claim 4, further comprising fifth and sixth FETs of said first polarity, said fifth FET connected in series between said second FET and said circuit common point and said sixth FET connected in series between said fourth FET and said bias current source, the drain and gate of said sixth FET connected to the gate of said fifth FET.

7. The FET circuit of claim 4, further comprising a FET differential input stage, said input stage connected between said second FET and said circuit common point such that said input stage is biased with current I1 and the transconductance of said input stage is directly proportional to the transconductance of said second FET.

8. The FET circuit of claim 4, further comprising a fifth FET having its gate connected to said second node and its source referred to said supply voltage such that said fifth FET conducts a current and has a transconductance which are directly proportional to the current in and the transconductance of said second FET, respectively.

9. The FET circuit of claim 8, further comprising a FET differential input stage, said input stage connected to said fifth FET such that said input stage is biased with said current conducted by said fifth FET and the transconductance of said input stage is directly proportional to the transconductance of said second FET.

10. A field-effect transistor (FET) operational amplifier having a resistor-programmable bandwidth, said amplifier comprising:a FET circuit with resistor-programmable transconductance, comprising: a resistance having a value R1, a first FET, said resistance connected between a supply voltage and said first FET's source, a second FET having its current circuit connected between said first FET and a circuit common point, said second FET's source connected to said first FET's drain at a first node, a third FET having its current circuit connected between said supply voltage and a second node and its gate connected to said first node, the gate of said first FET connected to said second node, a fourth FET which has its source connected to said second node and its gate and drain connected to the gate of said second FET, and a bias current source connected between said fourth FET's drain and said circuit common point which provides a starting current for said FET circuit that forces a current I1 to flow in said resistance and said first and second FETs, said first, second, third and fourth FETs being of a first polarity and the threshold voltages of said first and second FETs matched such that said current I1 gives a transconductance Gm2 for said second FET which is directly proportional to 1/R1; a fifth FET having its gate connected to said second node and its source referred to said supply voltage such that said fifth FET conducts a current and has a transconductance which are directly proportional to the current in and the transconductance of said second FET, respectively; a FET differential input stage comprising a FET differential pair and a current mirror connected to said differential pair to produce a single-ended output, said input stage connected to said fifth FET such that said differential pair is biased with said current conducted by said fifth FET and the transconductance of said input stage is directly proportional to the transconductance of said second FET; a sixth FET having its gate connected to said second node and its source referred to said supply voltage such that said sixth FET conducts a current and has a transconductance which are directly proportional to the current in and the transconductance of said second FET, respectively; a FET second stage which is biased with said current conducted by said sixth FET, said second stage comprising a seventh FET connected to conduct a current which varies with said single-ended output and a compensation capacitor having a capacitance Cc connected between said seventh FET's gate and source; said operational amplifier arranged such that its unity gain frequency fu is given by gm/2πCc, where gm is the transconductance of said input stage, said FET circuit with resistor-programmable transconductance and said fifth and sixth FETs arranged such that fu is directly proportional to 1/R1.