This technology relates to self-starting bandgap reference devices and methods of use thereof.
Bandgap references are widely utilized in integrated circuits to produce a fixed voltage to serve as a reference. The voltage remains constant regardless of power supply variations, temperature changes, and circuit loading from a device. Conventional bandgaps create a current source with the core bipolar devices and cross-couple this source with a current mirror.
Bandgap references are based on temperature behavior of the base-emitter voltage (VBE) of a bipolar junction transistor. When biased in a stable manner, VBE falls with increasing temperature. Thus, VBE is said to have a temperature behavior that is Complimentary to Absolute Temperature (CTAT). Two identical devices having different current densities will have VBE values that fall with temperature at different predictable slopes. These characteristic curves will both converge at zero degrees Kelvin. The difference between the VBE values of the two devices with different current densities (ΔVBE) increases with temperature. This characteristic is said to be Proportional to Absolute Temperature (PTAT). The appropriate scaling and summation of PTAT and CTAT voltages, produces a voltage that is substantially temperature independent. This serves as the basis for basic bandgap reference techniques.
In most bandgap references, ΔVBE is created by biasing identical devices with different emitter junction areas with either currents of the same magnitude or currents of a stable integer ratio. The created ΔVBE is generally impressed across a resistor to generate a current that is PTAT, which is then scaled by the required factor using a resistor of the same type and added to the VBE to create the reference output voltage. The ΔVBE generator is often referred to as the “bandgap core.” The bandgap core that produces the ΔVBE provides a PTAT current source, such that the core current is PTAT. It is convenient in bandgap implementations, to mirror this core PTAT current from one core bipolar transistor to the other. Mirroring the core PTAT current is performed using a current mirror, a feedback circuit, or both. Since the circuit is stable wherever the current-verses-voltage transfer functions of the current source and mirror cross each other, standard bandgap references have a stable state at zero current which produces zero output voltage.
More specifically, A bipolar junction transistor collector current is given by the following equation:
IC=Is*exp(VBE/Vt) (1)
wherein, IC is the collector current, Is is the scale current, VBE is the base-emitter voltage, and Vt is the thermal voltage. Vt is determined by the following equation:
Vt=KT/q (2)
wherein, K is Boltzmann's constant, T is the temperature in Kelvin, and q is the electronic charge.
Solving equation (1) for base-emitter voltage yields the following equation:
VBE=Vt*ln(IC/Is) (3)
For two devices with different base-emitter voltages, the difference between the VBE values for the two devices, i.e., ΔVBE, is given by the following equation:
ΔVBE=(VBE1−VBE2) (4)
Combining equations (3) and (4) provides:
ΔVBE=Vt*ln(IC1/Is1)−Vt*ln(IC2/Is2)=Vt*ln[(IC1/IC2)*(Is2/Is1)] (5)
Where IC1 and Is1 are the collector current and scale current for the first device and IC2 and Is2 are the collector current and scale current for the second device.
For identical devices, the scale current is proportional only to the emitter area. Therefore, equation (5) becomes:
ΔVBE=Vt*ln[(IC1/IC2)*(AE2/AE1)] (6)
wherein AE1 and AE2 are the emitter areas for the first and second devices, respectively. For practical matching considerations, area ratios are achieved in standard fashion by replication of unit devices. Ratio m results from m devices ratioed with a single device. Thus, equation (6) becomes:
ΔVBE=Vt*ln[(IC1/IC2)*(m)] (7)
In equation (7), only the collector current ratio (IC1/IC2), not its magnitude, is important. For devices driven at the same current, equation (7) becomes:
ΔVBE=Vtln(m) (8)
Equation (8) is independent of collector current. The ΔVBE expression in equation (8) has all of the VBE-characterizing parameters eliminated, leaving only a simple ratio (m) and the thermal voltage Vt. As provided in Gilbert, “Monolithic Voltage and Current References: Theme and Variations” MEAD Design Course, San Jose, Calif. (1996), equation (8) provides the basis for producing a temperature dependent voltage that is quite fundamental (not empirical). The temperature dependent voltage is proportional to temperature regardless of the absolute value of the operating currents, doping profiles, junction areas, transistor polarity (NPN or PNP), or even the material type (Si, Ge, SIGe, or even Schottky junctions). Only Boltzmann's constant and the charge on an electron and are involved.
Most bandgap references create a current source with the ΔVBE core and cross-couple this source with a current mirror that is balanced either by the mirror itself, or an amplifier servo-loop, or both. This is convenient and reduces curvature slightly, but introduces an undesired zero-current state. Since standard bandgap reference techniques have an undesired zero-current or zero-voltage state, they require a start-up circuit for operation. Although many successful start-up circuits have been developed, they are difficult to design and notoriously difficult to evaluate with complete confidence over all conditions. Start-up circuits also introduce delay in start-up for the bandgap reference circuit.
A self-starting bandgap reference circuit comprises a bias current source configured to provide a bias current. A bandgap core is coupled to the bias current source. The bandgap core includes a first device configured to receive the bias current and provide a first current output based on the bias current and a second device configured to receive the bias current and provide a second current output based on the bias current. A difference mirror is coupled to the first device and the second device to receive the first current output and the second current output. The difference mirror is configured to provide a difference current between the second current output and the first current output, wherein the difference current is a proportional-to-absolute temperature current. A voltage reference output and a current reference output are coupled to the difference mirror to receive the proportional-to-absolute temperature current and provide a voltage reference and a current reference based on the proportional-to-absolute temperature current.
A method of making a self-starting bandgap reference circuit comprises providing a bias current source configured to provide a bias current. A bandgap core is coupled to the bias current source. The bandgap core includes a first device configured to receive the bias current source and provide a first current output based on the bias current and a second device configured to receive the bias current and provide a second current output based on the bias current. A difference mirror is coupled to the first device and the second device to receive the first current output and the second current output. The difference mirror is configured to provide a difference current between the second current output and the first current output, wherein the difference current is a proportional-to-absolute temperature current. A voltage reference output and a current reference output are coupled to the difference mirror to receive the proportional-to-absolute temperature current and provide a voltage reference and a current reference based on the proportional-to-absolute temperature current.
This technology provides a circuit that provides a stable voltage reference that is independent of supply voltage and temperature. The circuit forms the basis for Analog-to-Digital conversion, as well as a basis to create stable power supply voltages and bias currents. The circuit provides a single stable state when powered with no undesired zero-current or zero-voltage state, thus eliminating the requirement for a start-up circuit. Specifically, the circuits of the present technology put the core devices in current drive with a sure-starting current source.
The circuits of the present technology advantageously provide a particular voltage value in situations where supply power is unreliable or intermittent, such as in RF energy scavenging systems. Additionally, the circuits of the present technology are not susceptible to having their operating state upset from a disturbance. This means the circuits are radiation tolerant, i.e., they have no false state that they can be tripped into, when implemented in an appropriate process. The technology may advantageously be employed in both bipolar and CMOS technologies.
As set forth in equation (8) above, the PTAT signal ΔVBE is dependent on only a fixed ratio and fundamental constants. Most bandgap references create a current source with the ΔVBE core and cross-couple this source with a current mirror which is balanced either by the mirror itself, or an amplifier servo-loop, or both. This is convenient and reduces curvature slightly, but introduces an undesired, zero-current state. By contrast, the circuits of the present technology drive the collector currents of the core bipolar junction transistors, which in turn creates an accurate value current source from a rough value current source—with the important additional feature that no undesired, zero-current state is possible as the core bias is generated by an independent current source. Therefore no start-up circuit is required. The problem that remains is to extract the ΔVBE signal so that it can be properly scaled and added to the CTAT signal (VBE) and ground referenced. This problem is solved in several different ways in the exemplary circuits below.
The present technology recognizes that the ΔVBE generator is independent of current magnitude over decades of bias current. This holds as long as the core devices carry the same current or currents with a fixed ratio relationship. In the present technology, instead of mirroring the core device currents one to the other, both bipolar junction transistors are put in current drive by independent current sources that are stable, produced in a predictable ratio, and have no zero current state. It is not necessary that this bias current be PTAT or even fixed as long as the ratio is stable and it is bound over extremes of process, temperature, power supply voltage, and mismatch to avoid low extremes, where leakage errors can be problematic and to avoid high currents where high level injection can be problematic or parasitic resistance can destroy log conformity. Errors due to decreasing beta are also avoided with moderate bias currents.
A second problem addressed by the exemplary circuits of the present technology is extracting the ΔVBE PTAT voltage so that it can be scaled and added to the VBE CTAT voltage. The standard bandgap circuit uses a core mirror so the signal of interest is simply mirrored out. The present technology provides a number of exemplary solutions for this problem. One solution applies the ΔVBE to a source-coupled pair modified by placing a resistor between the sources. In addition, simple servo amplifiers drive the gate-source of each device in the pair in order to reduce matching errors. In this case, the PTAT current is developed in the source coupling resistor. This current appears in the drain of one source-coupled pair device so that it can be mirrored out. This first method is compatible with both CMOS and BiCMOS technologies.
A second exemplary solution for extracting the ΔVBE includes creating current sources with each of the bipolar transistors that deliver a current of VBE divided by a resistor. These current sources take in the rough bias current and reliably deliver a controlled current VBE/R. These VBE based currents are then bucked against each other in a current mirror to generate the required difference current which has the PTAT signature.
A third exemplary solution uses the VBE of the core devices as inputs to an operational amplifier based voltage-to-current converter, bucks the resulting currents to generate the PTAT current, and drives this current through the output branch in the standard manner. This circuit is also CMOS compatible.
An example of a self-starting bandgap reference circuit 10(1) that may be employed with a BiCMOS integrated circuit device, by way of example, is illustrated in
Referring again to
The BiCMOS self-starting bandgap reference circuit 10(1) illustrated in
In this example, the first device 18 and the second device 20 comprise bipolar junction transistors Q1 and Q2. The first device 18 includes Q1, M10, R1, and the second device 20 includes Q2, M11, and R2 and act as current sources. The first device 18 and the second device 20 are three terminal NPN devices, although other core devices may be employed. The first device 18 and the second device 20, in this example bipolar junction transistors, are configured to turn the rough current from the bias current source 12 into a complimentary to absolute temperature (CTAT) current as an output. The first device 18 and the second device 20 of the bandgap core and difference mirror 14 portion of the circuit are configured such that the first current output and the second current output are proportional to a first base emitter voltage (VBE1) for the first device 18 and a second base emitter voltage (VBE2) for the second device 20, as described in further detail below. More specifically, in this example, the first current output is equal to the first base emitter voltage divided by the resistance of resistor R1 in the first device 18 and the second current output is equal to the second base emitter voltage divided by the resistance of resistor R2 in the second device 20. In one example, the resistors R1 and R2 are configured to provide an equal resistance.
In this example, the second device 20 (including Q2) has a greater emitter area than the first device 18 to create a current difference. In this example, the emitter areas of the second device 20 and the first device 18 have an 8:1 ratio, although other ratios may be employed. The circuit is configured such that the first current output is delivered to an input of the difference mirror 22 and the second current output is delivered to an output of the difference mirror 22 as shown in
The core devices (first device 18 and second device 20) in the bandgap core and difference mirror 14 portion are coupled to the difference mirror 22 (including M12-M13) as illustrated in
The BiCMOS self-starting bandgap reference circuit 10(1) further includes a circuit portion that provides the voltage and current reference output 16. As shown in
An exemplary operation of the self-starting bandgap reference circuit 10(1) illustrated in
Next, the second device 20 comprised of Q2, R2, and M11 produces the current output VBE2/R2 which is delivered to the input of difference mirror 22 (M12-M13). The first device 18 comprised of Q1, R1, and M10 produces the second current output VBE1/R1, which is bucked against the output of the M12-M13 difference mirror 22. In this example, Q2 of the second device 20 has factor m greater emitter area than Q1 of the first device, such that the difference of the M10, M11 currents is given by the equation:
IPTAT=(VBE1/R1−VBE2/R2) (9)
In this example, Vgs(M13) is set equal to Vgs(M18) so that current sources M10 and M11 will see the same compliance. This causes these current sources to turn on at the same supply voltage so that there is no overshoot of VR as the supply voltage comes up. This matching also reduces errors caused by finite current source output impedance. The NPN current source devices M10 and M11 will benefit from native or low Vt devices. This will lower required power supply voltage and incur no liability.
The PTAT current (IPTAT) is delivered to mirror M18-M19-M20 the voltage and current reference output 16. The PTAT current is driven into R3-Q3 to produce the bandgap output voltage in standard fashion. M20 provides a PTAT reference output current. The PTAT gain is set by the ratio R3/R1. The voltage and current references provided by the voltage and current reference output 16 may be utilized in any application in which bandgap reference circuits are employed. In this example, the references may be utilized in a BiCMOS application.
Bandgap circuits having core devices biased with PTAT currents are advantageous as compared with those biased with T-stable or CTAT currents. As provided in Gilbert, “Monolithic Voltage and Current References: Theme and Variations” MEAD Design Course, San Jose, Calif. (1996), an expression for base-emitter voltage often used in precision bipolar design appears below:
VBE(H,Ic)=EGE−H(EGE−VBEN)+VTN*H*log(Ic/In)−η*VTN*H*log(H) (10)
wherein, VBE is the base-emitter voltage as a function of temperature (H) and collector current (Ic) (H=T/Tn, where T is the temperature in Kelvin and Tn is the design center temperature), EGE is the Y-intercept (extrapolated value at zero Kelvin), VBEN is the base-emitter voltage at the design center current In and temperature Tn, and VTN is the thermal voltage at the design center temperature (K*Tn/q).
The term H(EGE−VBEN) in Equation (10) is the linear CTAT term (Complimentary to Absolute Temperature), the term VTN*H*log(Ic/In) is the log dependence on collector current, and the term H*log(H) models curvature with temperature, where η is the exponent of temperature in the scale current expression Is(T) and is the SPICE parameter XTI (2 to 4 typical).
For a conventional bandgap reference, with PTAT (Proportional to Absolute Temperature) core currents, the collector currents can be expressed as:
Ic=λ/*H*In (11)
where λ/=Ic/In at design temperature Tn. Since log(a*b)=log(a)+log(b), the third term in equation (10) can be expressed as:
VTN*H*log(λ/*H)=VTN*H*(log(λ/)+log(H)) (12)
Inserting Equation (12) into Equation (10) gives:
VBE(H,Ic)=Ege−H(Ege−Vben−Vtn log(λ/))−(η−1)VtnH(log H) (13)
In comparing Equation (13) with Equation (9) it is apparent that a benefit of PTAT bias is that it reduces the curvature factor from η to (η−1).
The simple and reliable, independent bias current source 12 used as a basis for core current drive for first device 18 and second device 20 is essentially a modified Wilson current mirror, with the diode connected device replaced by a resistor. It is a feedback circuit with a small loop-gain that makes it quite stable. Sensitivity to power supply variations are given by the equation:
S=(VDD/Iout)*(dIout/dVDD)=(Vov/2*Vgs)*Sin/VDD (14)
wherein, Vov is the mosfet overdrive voltage (Vgs−Vt), Vgs is the gate-source voltage and Sin/VDD is roughly unity for supply voltages significantly greater than Vgs as set forth in P. R. Gray, et al., “Analysis and Design of Analog Integrated Circuits, Fourth Edition” John Wiley & Sons, Inc. (2001), the disclosure of which is hereby incorporated herein by reference in its entirety. This evaluates to roughly 0.05 for typical devices. There is a significant further reduction in power supply sensitivity because the independent current source of the bias current source 12 biases the bipolar junction transistor based current sources (first device 18 and second device 20), the output currents of which in turn are subtracted to produce the voltage reference output branch current of the voltage and current reference output 16. This creates a significant improvement in the power supply rejection ratio (PSRR).
In this example, the bandgap core and difference mirror 14 includes PNP bipolar transistor junctions Q3A and Q4A that are biased at the same level as PNP bipolar transistor junctions Q3 and Q4 of the first device 18 and the second device 20. This causes their base currents to track those of the core. These base currents are cross-coupled and injected into the PTAT current differencing mirror 22 (M13-M14) such that they subtract the original base currents. In this manner this current difference mirror 22 serves both functions so additional mismatch is not introduced. In addition, the difference mirror 22 operates at collector current magnitudes rather than significantly lower base current magnitudes, so difference mirror accuracy is improved. Nearly the same accuracy can be obtained by making Q3A and Q4A 1× devices as the base currents will be nearly identical at identical collector currents.
Referring now to
In this example, the bandgap core and difference mirror 14 portion of the circuit outlines two current sources, first device 118 and second device 120, proportional to VBE1 and VBE2, along with a difference mirror 122 that creates the difference in these currents, which is a PTAT current. The second device 120, which provides a current source, is comprised of Q2, R2, A2 and M15 and produces current VBE2/R2 which is delivered to the input of difference mirror 122 (M12-M13). The first device 118, comprised of Q1, R1, A1 and M14, produces current VBE1/R1, which is bucked against the output of difference mirror 122 (M12-M13). In this example, Q2 has factor m greater emitter area than Q1. The difference of the M14, M15 currents is given by the equation:
IPTAT=(VBE1/R1−VBE2/R2) (15)
The PTAT gain in this example is set by the ratio R3/R1. Vgs(M13) is set equal to Vgs(M18) so that current sources M14 and M15 will see the same compliance. This causes these current sources to turn on at the same supply voltage so that there is no overshoot of VR as the supply voltage comes up. This matching also reduces errors due to finite current source output impedance. NMOS current source devices M14 and M15 included in first device 118 and second device 120, respectively, will benefit from native or low Vt devices. This will lower required power supply voltage and incur no liability.
In this example, the voltage and current reference output 16 provides the PTAT current delivered to mirror M18-M19-M20. The PTAT current is driven into R3-Q3 to produce the bandgap output in standard fashion. M20 provides a PTAT reference output current.
In this example, Vgs(M12) is set equal to Vgs(M13) so that current sources M14 and M15 of the first device 118 and the second device 120 will see the same compliance. This causes these current sources to turn on at the same supply voltage so that there is no overshoot of VR as the supply voltage comes up. NMOS current source devices M14 and M15 of the first device 118 and the second device 120 will benefit from native or low Vt devices. This will lower required power supply voltage and incur no liability. In the voltage and current reference output 16 illustrated in
The example illustrated in
In this example, the ballast current source 12 drives current into core devices Q1 and Q2 of first device 218 and second device 220. This current is set at an amplitude that is lower than the design target servo current. The current mirror 222 formed by amplifier A1 and M8 and M9 delivers current in parallel with the ballast source 12. It introduces additional current to bring the core to its target bias point. The loop tracks temperature and process changes. Note that in the case of a reference that requires start-up, this start-up current must be introduced to wake-up the circuit and then turned off after start-up is reached. The turn-off is important because, if left on, the start-up current will shift the operating point and introduce errors. In contrast, this circuit uses ballast currents for the bias current source 12, which are a fraction of the intended bias level and remain on constantly. The feedback loop adds the remaining fraction of current required to reach the design target level. The ΔVBE signal is impressed across R1 making its current PTAT. The VBE voltage is impressed across R2 making its current CTAT. These resistors are ratioed such that the sum of their currents is temperature stable (T-stable).
The circuit 10(6) illustrated in
In this example, both the ballast and A1 mirror currents are mirrored to the output branch. Since the sum is T-stable, the output branch resistor R3 can be chosen at any convenient value to scale the reference output voltage as needed. This includes the possibility of a fractional reference where the output voltage is set to 600 mV for example.
Ballast currents must be sufficiently small over all operating conditions, such that the amplifier always makes a contribution to the core currents. Otherwise its control of the loop is lost. (This constraint is not present in a start-up version, since these currents are shut down after start-up is achieved). These considerations determine the range of suitable ballast currents for the bias current source 12. By driving identical currents into both amplifier input nodes and sweeping this current, the above described behavior is clearly seen and a plot of this behavior informs the choice of ballast current.
Referring again to
RB=2*IB=Vt/RB (16)
This current is split by cascode devices M3 and M4, which provide current to the PMOS mirror reference and its associated cascode device. The ballast current is mirrored out by M7, M8 and M9 to provide a reliable bias for servo amplifier A1 as well as two ballast currents.
The ballast currents advantageously cannot fail to turn on, do not turn off while the circuit operates, and impart no deleterious impact as a result of the continuous operation of ballast currents. The core mirror operates consistently at a current defined by (ΔVBE/R) as desired. As shown in
When no current, or currents of small magnitude, are available from the core mirror 322 (M10 and M11), the drop across the resistor R1 will be small. Since equal ballast currents are driven into Q1 and Q2 of the first device 318 and the second device 320, the device with a ratio of “m” greater area (Q1) will have a smaller voltage drop than Q2. As a result, the positive input of the servo amplifier A1 is at a lower voltage than the negative input and the output from servo amplifier A1 drives downward. This causes PMOS mirror 322 (M10-M11) to deliver current to the core. Alternatively, when excess current flows in the core, the drop across R1 increases, and the output of servo amplifier A1 drives upward and the core currents decrease. The loop drives from any other condition to one of equal voltages at the opamp input. As a result, the core must start with continuous ballast current applied to the bipolar devices of first device 318 and second device 320 (Q1 and Q2).
When the inputs on servo amplifier A1 are driven to equal voltages, the difference in base-emitter voltages (ΔVBE) appears across R1 so that it carries a PTAT current equal to (ΔVBE/R1). This PTAT current is mirrored by the high-compliance mirror 322 comprised of M10-M11 and A1. This means the bipolar devices both carry a current equal to (ΔVBE/R1)+I_BALLAST. Thus, ΔVBE is independent of collector currents over several decades of current. In this example, the collector currents must be equal. The PTAT reference current output is mirrored out by device M12, which is cascoded by M19 for tight matching and high output impedance. This circuit advantageously eliminates the problem of double boundaries on the ballast current. Thus, the circuit illustrated in
The circuit 10(8) further includes a CTAT current source 28 prior to the output branch. The CTAT current source 28 takes the base-emitter voltage of Q3 of third device 319 (VBE3) as an input. The voltage-to-current converter comprised of amplifier A2 and PMOS device M21 drives current through R2 such that base-emitter voltage of Q3 (VBE3) appears across R2. The resulting Complimentary to Absolute Temperature (CTAT) current equal to (VBE3/R2) is generated and mirrored to R3 by M22. The PTAT current is also mirrored by M20 to R3. Current through R3 then is the sum of the PTAT and the CTAT currents. This current is scaled to be temperature-stable by the PTAT gain ratio of (R2/R1) and the resulting reference voltage appears across R3. Since the R3 current is stable with temperature, this is a fractional bandgap reference and R3 can be scaled to produce any convenient reference voltage magnitude that is within compliance of the output current mirror.
Accordingly, examples of the present technology provide self-starting bandgap reference circuits that advantageously provide a stable voltage reference that is independent of supply voltage and temperature. The circuit forms the basis for Analog-to-Digital conversion, as well as a basis to create stable power supply voltages and bias currents. The circuit provides a single stable state when powered with no undesired zero-current or zero-voltage state, thus eliminating the requirement for a start-up circuit. Specifically, the circuits of the present technology put the core devices in current drive with a sure-starting current source.
Having thus described the basic concept of the invention, it will be rather apparent to those skilled in the art that the foregoing detailed disclosure is intended to be presented by way of example only, and is not limiting. Various alterations, improvements, and modifications will occur and are intended to those skilled in the art, though not expressly stated herein. These alterations, improvements, and modifications are intended to be suggested hereby, and are within the spirit and scope of the invention. Additionally, the recited order of processing elements or sequences, or the use of numbers, letters, or other designations therefore, is not intended to limit the claimed processes to any order except as may be specified in the claims. Accordingly, the invention is limited only by the following claims and equivalents thereto.
This application claims the benefit of U.S. Provisional Patent Applications Ser. No. 62/534,726, filed Jul. 20, 2017 and Ser. No. 62/586,295, filed Nov. 15, 2017, which are hereby incorporated by reference in their entirety.
Number | Date | Country | |
---|---|---|---|
62586295 | Nov 2017 | US | |
62534726 | Jul 2017 | US |