Constant temperature coefficient self-regulating CMOS current source

[0002] This application is related to the following concurrently filed and commonly assigned U.S. patent applications: U.S. patent application Ser. No. 10/402,653, entitled “Digitizing Temperature Measurement System,” of Peter R. Holloway et al.; U.S. patent application Ser. No. 10/401,835, entitled “Low Noise Correlated Double Sampling Modulation System,” of Peter R. Holloway et al.; and U.S. patent application Ser. No. 10/402,080, entitled “A Constant RON Switch Circuit with Low Distortion and Reduction of Pedestal Errors,” of Peter R. Holloway. The aforementioned patent applications are incorporated herein by reference in their entireties.

[0003] The invention relates to constant temperature coefficient current sources and, in particular, to a constant positive temperature coefficient current source having improved power supply rejection ratio.

[0004] FIG. 1 is a circuit diagram of a constant temperature coefficient self-regulating current source according to one embodiment of the present invention.

[0005] FIG. 2 is a plot of output current versus Vdd voltage illustrating the output current characteristics for the current source in FIG. 1 over a range of Vdd voltage and temperature values.

[0006] FIG. 3 is a plot of the output current versus temperature illustrating the linear temperature dependent characteristics of the current source of the present invention.

[0007] FIG. 4 is a circuit diagram of a constant temperature coefficient self-regulating current source according to a second embodiment of the present invention.

[0008] FIG. 5 is a circuit diagram of a constant temperature coefficient self-regulating current source according to a third embodiment of the present invention where the excitation current is bootstrapped from the current source itself for increased power supply rejection.

[0009] In accordance with the principles of the present invention, a current source for generating an output current having constant temperature coefficient and high power supply rejection ratio (PSRR) uses a pair of diode-connected MOS transistors in parallel with a pair of inverter connected MOS transistors. The diode-connected MOS transistors are sized in a non-nominal manner such that when operated at a given current level, the common gate voltage of the transistor pair is at a first voltage. In contrast, the inverter connected MOS transistors are sized in a non-nominal manner such that the threshold voltage of the inverter pair at the same current level is at a voltage greater than the first voltage. The diode-connected MOS transistors and the inverter connected MOS transistors are biased in the subthreshold region to ensure that a reference voltage generated within the current source is a voltage proportional to absolute temperature (PTAT). This PTAT voltage is impressed across a resistor to set the output current. The output current of the current source exhibits a constant temperature coefficient over a wide temperature range and exhibits nearly zero power supply dependency over a wide power supply voltage range.

[0010] Linear analog circuits generally require DC biasing circuits to provide the bias currents and voltages needed for proper operation. While DC biasing circuits can be implemented using a voltage reference, reference currents are typically used as the basis for the bias circuits. In some applications, a reference current can be made to exhibit a known linear change with temperature. In that case, the reference current can be used simultaneously to bias analog circuitry and to control temperature sensing circuitry. In temperature sensing applications, it is desirable that any change in the reference current be derived from temperature change only and that changes in the power supply voltage do not induce changes to the reference current. The current source of the present invention provides an output current with constant positive temperature coefficient and near zero supply voltage dependence and is thus ideal for use in DC biasing circuits, especially in temperature sensing applications.

[0011] FIG. 1 is a circuit diagram of a constant temperature coefficient self-regulating current source according to one embodiment of the present invention. Referring to FIG. 1, current source 100 includes a voltage source V1 supplying a Vdd voltage on a node 102 and a current source I1 coupled between the Vdd voltage and a regulated voltage V_Regat a node 106. Current source I1, powered by the Vdd voltage, supplies a current (also referred herein as current I1) to circuit branches connected to the regulated voltage V_Reg. In the present embodiment, current source 100 includes a first circuit branch of diode-connected MOS transistors, a second circuit branch of inverter-connected MOS transistors and a third circuit branch including a shunt regulator. As will be explained in more detail below, the circuit branches draw current from current source I1 in a manner so as to regulate the voltage V_Regat node 106 despite variations in the Vdd voltage. In the present embodiment, current source I1 is a stand alone current source and supplies a fixed current I1. In other embodiments, current source I1 can derive its operating current value from the current source itself, as will be explained below. In the present embodiment, voltage source V1 provides a Vdd voltage of 2.4 Volts to 5.5 Volts and current source I1 provides a current value of 20 μA.

[0012] The first circuit branch of current source 100 includes a PMOS transistor M3 and an NMOS transistor M4 connected in series with a resistor R_Refat the drain terminals. Specifically, PMOS transistor M3 has a source terminal connected to the regulated voltage node 106, a drain terminal connected to a node 110 and a gate terminal connected to a node 108. Transistor M3 is diode-connected by connecting the gate and the drain terminals of the transistor together through resistor R_Ref. NMOS transistor M4 has a source terminal connected to the ground potential (node 104), and a drain terminal and a gate terminal both connected to node 108. Thus, transistor M4 is diode-connected. PMOS transistor M3 and NMOS transistor M4 are not nominally sized. In a nominally sized transistor pair, the PMOS transistor is sized two times larger than the NMOS transistor to account for the lower hole mobility. However, in the present embodiment, PMOS transistor M3 has a width to length ratio of 30:1 while NMOS transistor M4 has a width to length ratio of 60:1. Thus, when carrier mobility is taken into consideration, NMOS transistor M4 is two times as strong as PMOS transistor M3. In the present embodiment, resistor R_Refis a polysilicon resistor and has a nominal resistance of 36.5 kΩ.

[0013] Transistors M3 and M4 generate a voltage V_topat node 110 and a voltage V_botat node 108 across resistor R_Ref. When the transistors in current source 100 are biased in the subthreshold region, the difference between voltage V_topand voltage V_bot, referred herein as the reference voltage V_Ref, is a voltage proportional to absolute temperature (a PTAT voltage), as will be described in more detail below. By impressing the reference voltage V_Refacross resistor R_Ref, a nearly PTAT current is generated. In the present description, the nearly PTAT current flowing in resistor R_Refis also referred to as the reference current I_Ref.

[0014] The second circuit branch of current source 100 includes a PMOS transistor M1 and an NMOS transistor M2 configured as an inverting amplifier. Specifically, the gate terminals of transistors M1 and M2 are connected together and to node 110. Thus, voltage V_topis the input voltage to the inverter. The output voltage INV_OUT of the inverter is provided at the drain terminals of transistors M1 and M2 (node 112). PMOS transistor M1 and NMOS transistor M2 are not nominally sized. In the present embodiment, PMOS transistor M1 has a width to length ratio of 160:1 while NMOS transistor M2 has a width to length ratio of 10:1. Thus, when carrier mobility is taken into consideration, PMOS transistor M1 is at least eight times as strong as NMOS transistor M2. Thus, the threshold voltage of the inverter of transistors M1 and M2 is higher than the threshold voltage of a nominally sized inverter pair.

[0015] Output voltage INV_OUT of transistors M1 and M2 drives the third circuit branch of current source 100 which, in the present embodiment, includes a PMOS transistor M5 functioning as a shunt regulator. In operation, transistor M5 is biased by voltage INV_OUT so as to divert sufficient current from current source I1 to keep voltage V_Regregulated at the desired voltage level.

[0016] In the present embodiment, regulation of voltage V_Regis provided by a shunt regulator in the form of a PMOS transistor M5. However, the use of transistor M5 as a shunt regulator is illustrative only and in other embodiments, other means for regulating voltage V_Regusing the INV_OUT signal from the second circuit branch can be used. For instance, transistor M5 can be replaced by an operational amplifier or a vertical PNP bipolar transistor.

[0017] The “sunk” output current Iout of current source 100 is provided at output terminal 114. In the present illustration, a resistor Rload is connected from Vdd to the output terminal to represent a resistive load driven by current source 100. Output current Iout is generated by serially connected NMOS transistors M6 and M7 whose current is defined by a current mirror formed by transistor M4 and transistor M6. For reasonable values of Rload, defined as values of Rload such that the voltage at the output drain terminal of transistor M7 (node 114) is within the voltage compliance range of the current source, the output current Iout is independent of the value of Rload. Transistors M4 and M6 can be equally sized devices or may have another fixed size ratio. Because they have equal gate-to-source (V_GS) voltages, the nearly PTAT current (current I_Ref) flowing in transistor M4 is mirrored to transistor M6 to provide an output current having constant temperature coefficient. NMOS transistor M7, formed as a cascode output buffer device, is connected between output terminal 114 of the current source and the drain terminal of transistor M6. Cascode transistor M7 is biased by the regulated voltage V_Regand functions to increase the output impedance of current source 100 and further decrease the supply voltage dependence of output current Iout. Specifically, transistor M7 isolates the drain voltage at transistor M6 from any change in the Vdd voltage. In this manner, the output current generated by current source 100 remains independent of variation in the supply voltage Vdd.

[0018] Current source 100 of the present invention generates a constant temperature coefficient current using only MOS transistors and a resistor. MOS transistors can be used to generate a voltage with linear temperature dependence by biasing the MOS transistors in the subthreshold region. Specifically, when MOS devices are operated at low current density levels and are thus biased in the subthreshold or weak inversion region, the I-V characteristics of the MOS transistor are exponential and exhibit strong temperature dependence. On the contrary, when the MOS devices are operated at reasonably high current density (strong inversion region), the MOS devices operate in the square law region and the I-V characteristics are much less temperature dependent.

[0019] When a MOS device is operated at low current density, the device operation is dominated by minority carriers in the channel and the minority carrier behavior is equivalent to that of bipolar transistors. Thus, by operating a MOS device at near subthreshold, which is a point where the carrier behavior changes from square law to bipolar operation, the MOS device can be used to generate a voltage with temperature dependent characteristics. In the case of bipolar transistors, the ΔV_BEvoltage of two bipolar transistors having a fixed current density ratio is virtually perfectly proportional to temperature. When two MOS devices operating at subthreshold region with low current density are used, the gate to source voltages (V_GS) of the MOS transistors are dominated by minority carrier action and exhibits bipolar-like operation. Thus, the difference between two V_GSvoltages of two MOS transistors having a fixed current ratio will also be perfectly proportional to temperature. The current source of the present invention exploits the subthreshold operation of MOS transistors to obtain voltage behavior that approaches the temperature dependent quality of the ΔV_BEvoltage of bipolar transistors.

[0020] In current source 100 of the present invention, the voltage difference between the common gate voltage of the diode-connected pair (transistors M3 and M4) and the threshold voltage of the inverter pair (transistors M1 and M2) are impressed across resistor R_Refto generate a current nearly proportional to absolute temperature. The primary source of error in the resulting PTAT current is the temperature coefficient of resistor R_Ref. In the present embodiment, resistor R_Refis formed as a polysilicon resistor. A polysilicon resistor typically exhibits a constant temperature coefficient of +636 ppm/°C. The positive temperature coefficient of the resistance of the resistor reduces the amount of change in current caused by the PTAT voltage. As a result, the temperature coefficient at the current output of the current source is less than fully proportional to temperature. For instance, the output current can change by about 76% as much as a fully PTAT current source would. Because the reference current I_Refis inversely proportional to the resistance of resistor R_Ref, the positive temperature coefficient of the resistance results in a small amount of hyperbolic nonlinearity (less than 2%) in the reference current which nonlinearity also appears in the output current of the current source.

[0021] In other embodiments, a metal resistor, such as a nichrome resistor, of low temperature coefficient or other composite resistor of low temperature coefficient can be used in order to allow the output current of the current source to have a linear temperature coefficient closer to 100% PTAT.

[0022] The behavior of various node voltages and currents within current source 100 after it settles to steady state operation is mathematically described by the equations below. The equations assume that the four MOS transistors M1 through M4 operate in the weak inversion region. The reference entitled “CMOS analog integrated circuits based on weak inversion operation,” by E. Vittoz and L. Fellrath, IEEE J. Solid-State Circuits,vol. SC-12, pp. 662-673, December 1977, provides a description of MOS transistors operating in the weak inversion region. Specifically, the reference models the drain current of a MOS transistor as follows:

[0023] I_D=SI_D0e^V/nV(e^−(V/V)−e^−(−V/V)),  Equation (1)

[0024] where

[0025] S=W/L=size factor=effective width/effective length of the transistor;

[0026] I_D0=I_Dn0for NMOS, I_D0=I_Dp0for PMOS

[0027] I_Dn0=NMOS characteristic current;

[0028] I_Dp0=PMOS characteristic current;

[0029] V_G,V_S,V_D=gate, source, and drain voltages compared to the substrate voltage;

[0030] V_t=kT/q=device thermal voltage;

[0031] k=Boltzmann's constant;

[0032] T=absolute temperature in Kelvin;

[0033] q=charge of an electron; and

[0034] n=slope factor (here assumed equal for NMOS and PMOS).

[0035] When the substrate is connected to the source and when the transistor is biased with enough drain-source voltage to be fully in saturation (i.e. V_S=0 and V_D−V_S≧4V_t), Equation (1) reduces to Equation (2) as follows:

[0036] I_D=SI_D0e^V/nV.  Equation (2)

[0037] Applying Equation (2) at transistor M1, the drain current I₁of transistor M1 can be defined by: $\begin{array}{cc}{I}_1=I_{\mathrm{left}}=S_1{I}_{\mathrm{Dp0}}\exp \left(\frac{V_{\mathrm{reg}}-V_{\mathrm{top}}}{{\mathrm{nV}}_t}\right).& \mathrm{Equation}\left(3\right)\end{array}$

[0038] Similarly the drain current I₂at transistor M2 is given by: $\begin{array}{cc}{I}_2=I_{\mathrm{left}}=S_2{I}_{\mathrm{Dn0}}\exp \left(\frac{V_{\mathrm{top}}}{{\mathrm{nV}}_t}\right).& \mathrm{Equation}\left(4\right)\end{array}$

[0039] Because the same current (current I_left) flows through both transistors M1 and M2, Equations (3) and (4) above can be equated to solve for voltage V_Regas follows: $\begin{array}{cc}{V}_{\mathrm{Reg}}=2V_{\mathrm{top}}-{\mathrm{nV}}_t\ln \left(\frac{S_1{I}_{\mathrm{Dp0}}}{S_2{I}_{\mathrm{Dn0}}}\right).& \mathrm{Equation}\left(5\right)\end{array}$

[0040] At the first circuit branch, Equation (2) can be applied to transistors M3 and M4 to define the respective drain currents. Then, because the same current (current I_right) flows through transistors M3 and M4, the drain current equations can be equated to solve for voltage V_Reg. Voltage V_Regcan thus be given as: $\begin{array}{cc}{V}_{\mathrm{Reg}}=2V_{\mathrm{bot}}-{\mathrm{nV}}_t\ln \left(\frac{S_3{I}_{\mathrm{Dp0}}}{S_4{I}_{\mathrm{Dn0}}}\right).& \mathrm{Equation}\left(6\right)\end{array}$

[0041] Because voltage V_Regis common at both circuit branches, Equations (5) and (6) can be equated to solve for reference voltage V_Refand reference current I_Refat resistor R_Refas follows: $\begin{array}{cc}{V}_{\mathrm{ref}}=V_{\mathrm{top}}-V_{\mathrm{bot}}=\frac{{\mathrm{nV}}_t}{2}\ln \left(\frac{S_1{S}_4}{S_2{S}_3}\right);\mathrm{and}& \mathrm{Equation}\left(7\right)\\ I_{\mathrm{ref}}=\frac{V_{\mathrm{top}}-V_{\mathrm{bot}}}{R_{\mathrm{ref}}}=\frac{{\mathrm{nV}}_t}{2R_{\mathrm{ref}}}\ln \left(\frac{S_1{S}_4}{S_2{S}_3}\right).& \mathrm{Equation}\left(8\right)\end{array}$

[0042] It can be seen from the above expressions for voltage V_Refthat the reference voltage is linearly proportional to the thermal voltage V_tas long as the process slope factor n does not change significantly with temperature. Note that the thermal voltage V_tchanges in direct proportion to absolute temperature. Similarly, the current I_Refis linearly proportional to V_tif the slope factor n and the resistance of R_Refis constant with temperature.

[0043] The circuit topology also constrains the current through the inverter pair (current I_left) to be a constant ratio of the current through the diode-connected pair (current I_rightwhich is equal to current I_Ref) which ratio is dependent only upon the relative sizing of the transistors. This circuit topology constraint can be shown mathematically by applying the equation for drain current (Equation (2)) at NMOS transistors M2 and M4 and using Equation (9) below:

[0044] V_top=V_bot+I_RefR_Ref.  Equation (9)

[0045] Accordingly: $\begin{array}{cc}{I}_{\mathrm{ref}}=I_{\mathrm{right}}=S_4{I}_{\mathrm{Dn0}}\exp \left(\frac{V_{\mathrm{bot}}}{{\mathrm{nV}}_t}\right),\mathrm{and}{I}_{\mathrm{left}}=S_2{I}_{\mathrm{Dn0}}\exp \left(\frac{V_{\mathrm{top}}}{{\mathrm{nV}}_t}\right)& \mathrm{Equation}\left(10\right)\\ I_{\mathrm{left}}=S_2{I}_{\mathrm{Dn0}}\exp \left(\frac{V_{\mathrm{bot}}}{{\mathrm{nV}}_t}\right)\exp \left(\frac{I_{\mathrm{ref}}{R}_{\mathrm{ref}}}{{\mathrm{nV}}_t}\right).& \mathrm{Equation}\left(11\right)\end{array}$

[0046] Defining the ratio of currents I_left/I_Ref: $\begin{array}{cc}\frac{I_{\mathrm{left}}}{I_{\mathrm{ref}}}=\frac{S_2}{S_4}\exp \left(\frac{I_{\mathrm{ref}}{R}_{\mathrm{ref}}}{{\mathrm{nV}}_t}\right).& \mathrm{Equation}\left(12\right)\end{array}$

[0047] Substituting equation (8) for I_Refand collecting terms the ratio of currents I_left/I_Refis given by: $\begin{array}{cc}\frac{I_{\mathrm{left}}}{I_{\mathrm{ref}}}={\left(\frac{S_1{S}_2}{S_3{S}_4}\right)}^{1/2}.& \mathrm{Equation}\left(13\right)\end{array}$

[0048] Thus, the current I_leftfollows the temperature dependence of current I_Refexactly and has an amplitude which is a constant fraction of the amplitude of current I_Refwhich fraction is dictated by the relative size factors of the transistors.

[0049] Referring to FIG. 1, transistors M1 and M2 are connected as an inverting pair and are used as a self-referencing inverting linear amplifier. The self referential operation above refers to the inverter threshold voltage, and is given by that voltage at the inverter input which causes the same voltage at the inverter output. Applied input voltages above the inverter threshold voltage force the inverter output to a voltage less than the threshold voltage, and applied input voltages below the threshold voltage force the inverter output to a voltage above the threshold voltage.

[0050] In steady state weak inversion operation the relative size of transistors M1 and M2 defines the threshold voltage. Referring to FIG. 1, V_topis given by: $\begin{array}{cc}{V}_{\mathrm{top}}=\frac{V_{\mathrm{reg}}}{2}+\frac{{\mathrm{nV}}_t}{2}\ln \left(\frac{S_1{I}_{\mathrm{Dp0}}}{S_2{I}_{\mathrm{Dn0}}}\right),& \mathrm{Equation}\left(14\right)\end{array}$

[0051] where I_Dn0and I_Dp0are the process dependent characteristic currents of PMOS and NMOS devices and n is assumed equal for NMOS and PMOS. If the ratio inside the natural logarithm is greater than one, then voltage V_topis a voltage greater than voltage V_Reg/2. As the temperature increases, the thermal voltage V_tincreases, and voltage V_topbecomes larger with respect to voltage V_Reg.

[0052] Similarly, in steady state operation, when the operating current is set low enough that the devices M1 through M4 operate in the weak inversion region, diode connected transistors M3 and M4 define a voltage V_botwhich is set by the relative size of transistors M3 and M4. Specifically: $\begin{array}{cc}{V}_{\mathrm{bot}}=\frac{V_{\mathrm{reg}}}{2}-\frac{{\mathrm{nV}}_t}{2}\ln \left(\frac{S_4{I}_{\mathrm{Dn0}}}{S_3{I}_{\mathrm{Dp0}}}\right).& \mathrm{Equation}\left(15\right)\end{array}$

[0053] If the ratio inside the natural logarithm is more than one, then V_botis a voltage below V_Reg/2. It can also be seen that as the thermal voltage V_tincreases with temperature, the voltage at V_botbecomes smaller with respect to V_Reg.

[0054] By sizing transistors M1 through M4 appropriately, voltage V_topcan thus be made larger than voltage V_bot. Both voltage V_topand voltage V_botchange with temperature due to the presence of the thermal voltage term, V_t, in the equations above. The voltage difference between voltage V_topand voltage V_botcan then be used as a temperature dependent reference voltage. When this voltage difference is impressed across a resistor, a constant temperature coefficient reference current can be generated.

[0055] The generation of a constant temperature coefficient current by current source 100 while exhibiting a high power supply rejection ratio (PSRR) is described below.

[0056] When the circuit of 100 is powered up, it is assumed here that the current source 11 increases from zero current to some fixed current value. After the current reaches its fixed value, and after transients generated by the startup transition have decayed, the various circuit nodes will be at their steady state closed loop operating points.

[0057] The fixed current I1 is supplied to the V_Regnode 106. From node 106 the current will divide through three possible circuit paths to flow to ground. The first circuit path is through transistors M3 and M4. The second path is through transistors M1 and M2. The third path is through transistor M5.

[0058] In steady state operation transistors M3 and M4 establish a voltage V_botat node 108 according to Equation (15). Also, transistors M1 and M2 establish a voltage V_topat node 110 according to Equation (14). This voltage V_topis the threshold voltage of the inverting amplifier formed by transistors M1 and M2. In steady state operation the voltage difference between voltages V_topand V_botis impressed across resistor R_Ref, which causes a reference current I_Refto flow through transistors M3 and M4 in the first circuit path.

[0059] The steady state ratio of currents between the second circuit branch (transistors M1-M2) and the first circuit branch (transistors M3-M4) is set by the temperature-independent relative size constraint of Equation (13). If less of the fixed current from I1 is diverted to ground through the third circuit path at shunt regulator M5, the current through both of the first two branches will increase, while remaining in relative proportion. Because the two currents in the first and second circuit branches are constrained to operate at a fixed ratio in steady state, by monitoring and servoing one current to its desired value the other will also be servoed to its desired value.

[0060] The reference current I_Refthrough transistors M3-M4 is monitored by a feedback loop which continuously sets the amount of current diverted through transistor M5 such that the desired amount of current flows both at the first circuit branch (transistors M3-M4) and at the second circuit branch (transistors M1-M2). Diverting the required amount of current is equivalent to regulating the voltage V_Regat node 106 to the desired voltage.

[0061] The feedback loop consists of the inverting amplifier formed by transistors M1-M2 and a transconducting shunt regulator implemented as PMOS source follower M5 in FIG. 1. The inverting amplifier, operated with both devices in the linear region and with an input voltage close to its threshold voltage, supplies a large voltage gain to the error signal. Its output is applied to a large width-to-length PMOS device (transistors M5) in order to obtain high transconductance. The purpose of the shunt regulator device M5 is to divert just enough current from current source I1 such that the correct reference current flows through resistor R_Ref. In operation current source I1 must provide sufficient current to supply the first and second circuit branches (transistors M1 to M4) plus some additional current to be diverted through shunt regulator transistor M5.

[0062] The loop gain from V_topto the shunted current at the source terminal of M5 is given by the voltage gain of the inverter multiplied by the transconductance of transistor M5. The loop gain can be very large, which assures that the feedback loop will be very accurate and will maintain voltage V_Regand current I_Refto high accuracy even if power supply changes or other events disturb the amount of current flowing from I1.

[0063] In closed loop operation, the negative feedback loop from V_topto voltage INV_OUT to voltage V_Regto voltage V_topoperates in a manner such that the voltage V_topis servoed to a voltage very, very close to the inverter threshold voltage given by Equation (14) above. If the current I1 were to momentarily increase, and cause voltage V_Regto increase to a voltage V_Reg+ΔV_Reg, the voltage V_botat node 108 would increase by ΔV_Reg/2 and the inverter threshold voltage would also increase by ΔV_Reg/2. However, the voltage V_topis I_Ref*R_Refabove voltage V_botand would increase by ΔV_Reg/2+Δ(I_Ref*R_Ref). Thus the difference between voltage V_topand the inverter threshold voltage is an error signal proportional to the amount of deviation of I_Reffrom the desired value. The feedback loop seeks to minimize this error signal and would divert more current through transistor M5 until the error signal returned to nearly zero.

[0064] Due to the presence of the feedback control system and its shunt regulator, current source I1 does not need to be precise or to exhibit any required temperature coefficient. Current source I1 must supply only a current large enough for the temperature dependent requirements of transistors M1-M4 and a minimal amount of excess current in order to keep the bandwidth at transistors M5 at the desired level. However, high frequency disturbances beyond the bandwidth of the feedback loop will not be fully suppressed, and a finite loop gain can only reduce disturbances by an amount approximated by (1/loop gain). It is thus desirable to implement I1 in a manner which itself exhibits low wideband noise and high power supply rejection. In one embodiment, current source I1 is implemented as a bootstrapped current mirrored from the reference current itself for achieving extremely good power supply rejection and low noise. This configuration is illustrated in FIG. 5 and described in more detail below.

[0065] As described above, current source 100 generates an output current with constant temperature coefficient and near zero power supply dependency. In operation, the quiescent operating point of current source 100 is established by the following three constraints.

[0066] The first constraint concerns the generation of a temperature dependent current in the first circuit branch (transistors M3 and M4) of current source 100. By using non-nominal transistor size ratios in the first circuit branch and the second circuit branch, voltages V_topand V_botare established such that voltage V_topminus voltage V_botis equal to a reference voltage V_Ref. The V_Refvoltage is impressed across resistor R_Refto generate the temperature dependent reference current I_Refin current source 100, which current flows through the diode-connected transistors M3 and M4.

[0067] Equation (7) above shows that the reference voltage V_Refgenerated across resistor R_Refis exactly proportional to the absolute temperature of the isothermal transistors M1 through M4. Equation (8) above shows that current I_Refwould be exactly proportional to absolute temperature if the resistance of resistor R_Refcould be held constant with temperature. The linear resistance temperature coefficient of +636 ppm/°C. of the polysilicon resistor used in the present embodiment decreases the amount of change with temperature such that a change that is proportional to 76% of the change of an exact PTAT current source results.

[0068] The second constraint concerns forcing the currents in the first and second circuit branches to operate at the same temperature coefficient. Equation (13) above shows that the current I_left, through transistors M1 and M2, is a constant fraction of the reference current I_Ref, through transistors M3 and M4. Moreover Equation (13) shows that the fraction can be chosen by the sizing of the transistors M1 through M4 and that the fraction is independent of temperature. Thus currents I_leftand I_Refexhibit exactly the same temperature coefficient.

[0069] The final constraint concerns the regulation of voltage V_Regto ensure that the correct temperature dependent current is generated at the first circuit branch and that this current is immune to perturbations in the power supply voltage Vdd.

[0070] By considering constraints one and two, the currents in the first and second circuit branches are known to be currents with a temperature change of 76% PTAT in the present embodiment. The amount of current flowing through the inverter (transistors M1 and M2) is known to be some size-determined fraction of the current through the diode-connected transistors M3 and M4. It is therefore possible to monitor only the reference current with a feedback loop and to simultaneously set both the current through transistors M3 and M4 and the current through transistors M1 and M2 by controlling voltage V_Regon node 106.

[0071] To maintain current source 100 at the desired operating point, where there is just enough current in transistors M3 and M4 to generate the V_Refvoltage across resistor R_Ref, a control loop using transistors M1 and M2 as an error amplifier and transistor M5 as a shunt regulator is used.

[0072] In operation, there is one voltage at which transistors M1 and M2 can drive transistor M5 such that transistor M5 drains the excess current from current source I1 and the right amount of current flows through transistors M3 and M4. If voltage V_Regincreases, perhaps due to a supply voltage increase, the current through the first circuit branch will increase and voltage V_topwill increase accordingly. When voltage V_topincreases, the inverting linear amplifier at transistors M1 and M2 will cause the output voltage INV_OUT to decrease. The drop in inverter output voltage INV_OUT causes PMOS transistor M5 to turn on and draw more current away from current source I1. As a result, voltage V_Regreturns to the expected regulated voltage level. In effect, transistor M5 acts a shunt regulator to keep voltage V_Regat the desired voltage level. Transistor M5 is biased by the closed loop control system in such a way as to divert sufficient current from current source I1 to ensure that voltage V_Regremains regulated.

[0073] In current source 100, a large loop gain is provided to achieve the desired regulation. For instance, the full voltage gain of the inverter pair is employed. Transistor M5 is bootstrapped to voltage V_Regsuch that only a very small change in the gate-to-source voltage of the transistor is needed to cause a large change in current being drawn away. Thus, a compound gain composed of the voltage gain of the inverter multiplied by the transconductance of transistor M5 is realized which provides for a very large loop gain for the voltage regulator. A very tight voltage regulation can be realized.

[0074] In summary, current source 100 includes a first circuit branch and a second circuit branch of transistors connected in an “H” configuration. Both of the circuit branches are supplied by a current source I1 powered by the supply voltage Vdd. The first circuit branch includes a pair of diode-connected transistors connected in series with a resistor R_Refinterposed at the drain terminals. Resistor R_Ref, imposed between voltage V_topand voltage V_bot, defines the operating current of current source 100. The second circuit branch consists of an inverter pair functioning as an error amplifier providing an “error signal” to facilitate regulation of the supply voltage. In the present embodiment, the error signal is a function of the V_topvoltage which, when increased to a certain voltage level, will cause the error amplifier to react to return V_topto the desired operating point. The regulation is realized by a third circuit branch, which receives the “error signal” and regulates voltage V_Regaccordingly to maintain voltage V_Regat the desired regulated level. In the present embodiment, the third circuit branch includes PMOS transistor M5 acting as a shunt regulator. Transistor M5 receives the “error signal” from the second circuit branch and provides regulation of the supply voltage by draining excess current from current source I1.

[0075] Voltages V_topand V_botare generated by operating the transistors M1 to M4 in the subthreshold region and by setting the transistors effective strength by sizing the transistors in a non-nominal, asymmetric fashion. That is, in the first circuit branch, the PMOS transistor is made to be a lot weaker than the NMOS transistor, and in the second circuit branch, the PMOS transistor is made to be a lot stronger than the NMOS transistor. This non-nominal sizing of the transistors creates common gate voltages at the first circuit branch and inverter threshold voltages at the second circuit branch that are different from nominal voltage levels.

[0076] The operation of current source 100 can be divided into two parts. First, voltages VTOP and VBOT are established by using asymmetrical sizing of transistors in the first circuit branch and the second circuit branch. The difference of voltages V_topand V_botdefines the V_Refvoltage which is imposed across resistor R_Reffor defining a current in the first circuit branch. Furthermore, the transistors in the first and second circuit branches are operated in the subthreshold region so that the V_Refvoltage exhibits temperature dependent behavior. In this manner, the first circuit branch yields a nearly PTAT current which can be used as the basis for the output current for current source 100. In the present embodiment, the output current is a 76% PTAT current because of the use of a polysilicon resistor for resistor R_Ref.

[0077] Second, the second circuit branch in conjunction with the third circuit branch provides regulation of the regulated voltage V_Regto ensure that the nearly PTAT current generated by the first circuit branch is isolated from variations in the power supply voltage Vdd. Specifically, the inverter pair of the second circuit branch detects whether the current source has reached the ideal operating point and provides an “error signal” to adjust the effective current flowing in the first circuit branch so as to keep the circuit at the ideal operating point. When current source 100 has reached equilibrium, voltage INV_OUT will be near the midpoint between voltage V_Regand ground. When the current flowing in the first circuit branch becomes too high, voltage INV_OUT will decrease to allow the third circuit branch to drain off the excess current, thereby regulating voltage V_Reg. Thus, the current though the first circuit branch can remain stable despite variations in the supply voltage.

[0078] Current source 100 will operate to define an “ideal” voltage difference V_Refbetween voltage V_topand voltage V_botand resistor R_Refdefines the nearly ideal current V_Ref/R_Ref. The regulation provided by the second circuit branch and the third circuit branch is applied to keep the operating point of the circuit as close as possible to the ideal current condition.

[0079] The voltage difference between voltage V_topand voltage V_botcan be made as large as desired and in one embodiment, the voltage difference is 60 to 100 mV.

[0080] FIG. 2 is a plot of output current Iout versus Vdd voltage illustrating the output current characteristics for current source 100 over a range of temperature values. As shown in FIG. 2, after sufficient Vdd voltage is applied to power up the circuit (about 2 volts), output current Iout of current source 100 is constant over Vdd voltages from 2 volts to 5 volts. Furthermore, the output current increases linearly with temperature as illustrated by the stepwise increase of each curve representing increasing temperature. The constant temperature coefficient of output current Iout is also illustrated in FIG. 3 which is a plot of the output current versus temperature. In the present illustration, current source 100 exhibits a linear increase in current with temperature from 50° C. to 150° C.

[0081] Besides being capable of generating a nearly PTAT current with high PSRR, the current source of the present invention realizes other advantages. For instance, because the current source of the present invention does not rely on bipolar transistors to generate a ΔV_BEvoltage, the current source can be implemented entirely using CMOS transistors. Thus, the current source can be manufactured using CMOS processes and does not require complex BiCMOS fabrication processes. Furthermore, the current source of the present invention can be implemented using minimal silicon area, thereby reducing manufacturing cost.

[0082] In another embodiment, current source 100 may include components for providing AC compensation to ensure stability at high frequency. In the present embodiment, current source 100 includes a capacitor C1 coupled between node 110 (voltage VTOP) and node 112 (voltage INV_OUT) to filter out noise at the output terminal of the inverter pair. In the present embodiment, current source 100 further includes a capacitor Ccomp and resistor Rcomp connected in series for providing pole and zero compensation for frequency stability. Capacitor Ccomp rolls off at high frequency and resistor Rcomp provides a zero which allows for limited stable high frequency operation. In the present embodiment, capacitor C1 has a capacitance value of 0.1 pF, capacitor Ccomp has a capacitance value of 1.6 pf, and resistor Rcomp has a resistance value of 50 kΩ.

[0083] According to yet another embodiment, current source 100 may include switches for facilitating power up and reset operations. Referring to FIG. 1, in the present embodiment, current source 100 includes a switch S1 coupled between node 110 and the ground node 104 and a switch S2 coupled between node 112 and the ground node 104. In operation, switches S1 and S2 are closed temporarily during power up to discharge capacitors C1 and Ccomp and to ensure that the current source will start from a defined state after the switches are opened. Switches S1 and S2 may also be closed during a reset operation. In this manner, capacitors C1 and Ccomp are discharged and current source 100 is ensured to be able to start up.

[0084] In the embodiment shown in FIG. 1, the output current lout is sunk by transistors M6 and M7 through the use of a current mirror at transistors M4 and M6. In other embodiments, the output current of the current source can be sourced from the regulator voltage node as shown in FIG. 4. FIG. 4 is a circuit diagram of a constant temperature coefficient self-regulating current source according to a second embodiment of the present invention. Like elements in FIGS. 1 and 4 are given like reference numerals to simplify the discussion.

[0085] Referring to FIG. 4, current source 200 provides an output current lout that is sourced from voltage V_Reg(node 106). Specifically, the output current lout is generated by a PMOS transistor M12. The source terminal of transistor M12 is connected to voltage V_Regand the drain terminal is connected to the output terminal 216. The gate terminal of transistor M12 is driven by the voltage V_topat node 110. The sourced output current Iout is available at the drain terminal of transistor M12. In operation, voltage V_topis forced to a value which will provide the proper current through transistor M1. If transistor M12 is sized the same as transistor M1, then a current identical to that flowing through transistor M1 will be sourced from the PMOS output device transistor M12 into a reasonable value of output load resistance Rload to ground. Intentionally sizing transistor M12 greater or larger than transistor M1 can be used to select output currents proportionally greater or larger than the current chosen to flow through transistor M1. In this embodiment, the sourced output current Iout must be provided by current source I1, which also supplies the operating current to the circuit. The amount of current supplied by current I1 may need to be increased when the PMOS current output device is used in order to maintain enough current for both the external load and for proper loop operation.

[0086] In the embodiment shown in FIG. 1, current source I1 is a stand-alone current source providing a fixed current value. In other embodiments, current source I1 can be implemented as a current source having its current magnitude derived in a bootstrapped manner from the constant temperature coefficient reference current source itself which has the advantage of increased power supply rejection. FIG. 5 is a circuit diagram of a constant temperature coefficient self-regulating current source according to a third embodiment of the present invention. Like elements in FIGS. 1 and 5 are given like reference numerals to simplify the discussion.

[0087] Referring to FIG. 5, current source 300 is configured such that the excitation current I1 for the circuit is bootstrapped from the current source itself. Specifically, the excitation current I1 is derived from the constant temperature coefficient reference current I_Refgenerated at resistor R_Refof current source circuit. In current source 300, current source I1 is implemented as a PMOS current mirror (transistors M10 and M11) whose amplitude is derived from a replica of the constant temperature coefficient reference current I_Ref. As shown in FIG. 5, current source 300 includes NMOS transistor M6 for mirroring a first replica of reference current I_Refto provide as the output current Iout. Current source 300 further includes a NMOS transistor M8. The gate to source voltage of transistor M8 is equal to that of transistor M4. Therefore, the drain current of transistor M8 will also mirror that of transistor M4. Thus a second replica of reference current I_Refis caused to flow through transistor MB. This second replica reference current is used to excite the PMOS mirror at transistors M10 and M11 and thus to generate the excitation current I1. An NMOS transistor M9 is coupled to transistor M8 as a cascode device to buffer the drain terminal of transistor M8 from voltage variations in the supply voltage Vdd. Specifically, transistor M9 is connected in the same manner as transistor M7 and has its gate terminal connected to the regulated voltage node 106.

[0088] The current flowing through transistor M9 is applied to the current mirror formed by transistors M10 and M11. PMOS transistor M10 is diode-connected and has its source terminal connected to the positive supply voltage Vdd (node 102). The drain and gate terminals of transistor M10 are applied to the gate terminal of PMOS transistor M11 in a manner to mirror the current through transistor M10 at transistor M11. The relative sizes of transistors M10 and M11 are adjusted to obtain the desired magnitude of current at the drain terminal of transistor M11 which current is coupled to the regulated voltage node 106 for supplying current to the circuit branches of current source 300. In particular, the current mirror of transistors M10 and M11 can implement a ratioed gain to generate current I1 having sufficient magnitude to supply current source 300. For example, if the reference current through R_Refis 2 uA, and transistor M8 was sized identically to transistor M4, then the current through transistor M10 will also be 2 uA. If 20 uA is the desired output current I1 at transistor M11, then transistor M11 can be sized ten times wider than transistor M10 while using the same length. In this manner, current I1 is a multiple of the reference current and is guaranteed to be in excess of the current needed to supply transistors M1 and M3 which excess current is diverted by the shunt regulator (transistor M5).

[0089] At initial power-up of current source 300, switches S1 through S3 are closed. Switches S1 and S2 are used to set known startup voltages for the bootstrapped loop and to dissipate any charge which might have been held in capacitors C1 and Ccomp at their respective nodes. The current source I2, switchably connected through switch S3, is used to apply a small, known startup current to the PMOS current mirror transistors M10 and M11 to insure that the PMOS current mirror turns on. Current source I2 can be generated by a resistor to ground or by other means. After the supply voltage V1 has been applied, switches S1 through S3 are opened, and the current source circuit bootstraps itself to its proper operating point.

[0090] In an alternate embodiment of current source 300, a small resistor can be connected between the source terminal of transistor M8 and the ground potential (node 104) while the substrate connection of transistor M8 remains connected directly to ground. It is known that the operation of the bootstrapped current source implementation of current I1 through transistors M8 to M11 uses positive feedback. This feedback can detract from the dynamic recovery performance of the negative feedback loop through transistor M5 used to regulate the reference current during steady state operation. However, this positive feedback is useful at powerup to speedily push the circuit to its desired steady state operating points. The resistor degenerated connection of transistor M8 supplies larger positive feedback at powerup, insuring speedy settling to steady state, but the amount of positive feedback is reduced once appreciable current starts to flow through transistor M8, such as at steady state.

[0091] The above detailed descriptions are provided to illustrate specific embodiments of the present invention and are not intended to be limiting. Numerous modifications and variations within the scope of the present invention are possible. For example, cascode transistor M7 is included to provide additional isolation against power supply variations and to increase output impedance. In other embodiments, other buffering devices may be used in place of transistor M7 or the output current can be taken directly from the drain terminal of transistor M6. The present invention is defined by the appended claims.