Ultra-low power and ultra-low voltage bandgap voltage regulator device and method thereof

Abstract

A family of bandgap embodiments are disclosed herein, capable of operating with very low currents and low power supply voltages, using neither any custom devices nor any special manufacturing technology, and fabricated on mainstream standard digital CMOS processes. As such, manufacturing cost can be kept low, manufacturing yields of digital CMOS system-on-a-chip (SOC) that require a reference can be kept optimal, and manufacturing risk can be minimized due to its flexibility with respect to fabrication process node-portability. Although the embodiments disclosed herein use novel techniques to achieve accurate operations with low power and low voltage, this family of bandgaps also uses parasitic bipolar junction transistors (BJT) available in low cost digital CMOS process to generate proportional and complementary to absolute temperature (PTAT and CTAT) voltages via the base-emitter voltage (VEB) of BJTs and scaling VEB differential pairs to generate the BJTs thermal voltage (VT).

Description

FIELD OF THE INVENTION

This invention relates to improvements in voltage regulators for integrated circuits.

BACKGROUND

Voltage regulators are fundamental functions in powering most analog or digital functions in electronic systems. Operating at low currents, low power supply voltages, and over extended temperatures are key objectives in portable electronics systems, including in system-on-chips (SOC), especially in medical or emerging wireless and battery-less applications such as power-on-a-chip (PoC), Internet-of-things (IoT), or energy harvesting, to name a few.

In light of the potentially large markets for such emerging applications, unless solution costs are low, the markets may not realize their full potential reach. As such, low cost innovations in component designs are needed to enable such prospects of high volume markets to get off the ground and take off. In the last few decades, there has been a tremendous amount of annual manufacturing investment in semiconductors, some of which is sunk-cost. As a result, there is likely ample capacity in standard complementary metal-oxide-semiconductor (CMOS) manufacturing factories that may be readily and cheaply available for exploitation by new and promising high volume markets. The enablement of such new high volume markets would greatly benefit if their solutions require neither any custom devices nor any special processes, which could impose higher costs or risks on manufacturing yields. For design solutions in such emerging applications, to optimally exploit the existing digital CMOS capacity, manufacturing process node-portability should not be impeded. While the die-size of such solutions need to be kept small to keep the costs down, optimally no performance interference on the rest of the functions of the SOC should be imposed by innovative designs when serving such new applications requirements (low voltage and ultra-low current over maximal power supply and temperature spans).

A survey of alternative voltage regulator topologies is presented in Fayomi, C J B. et al (2010) “Sub 1 V CMOS bandgap reference design techniques: a survey”, AICSP 62:141-157, DOI 10.1007/s10470-009-9352-4 discusses many techniques and their trade-offs for low power references and which is hereby incorporated by reference. Some of the trade-offs which are discussed in Fayomi et al. are briefly summarized as follows. Using Bipolar CMOS (BiCMOS) processes or depletion mode CMOS transistors frees up some operational supply headroom but adds to fabrication complexity and cost. Also, utilizing process deviations such as threshold differentials between positive channel metal-oxide semiconductors (PMOS) and negative channel metal-oxide semiconductors (NMOS), multiple depletion mode transistors, and differential thresholds via channel length variations. Alternatively, using custom devices such as a dynamic threshold metal-oxide semiconductor transistor (DTMOST), bulk driven CMOS, or lateral positive negative positive transistors (PNPs) can provide some design flexibility in extending digital power supply voltage (V_DD) and bandgap voltage (V_BG) spans, but at a price. Such deviations impose additional manufacturing costs, require special device modeling, or may demand post production control monitoring for the entire SOC that contains the reference. There are other design techniques, such as switch capacitors, that can provide extra operational headroom without the need for process or device variations. Switching techniques can help make low-noise references with stable temperature coefficients (TC) but may add cost due to capacitor extra mask or large capacitors for low droop rates. Chopping bandgap topologies can also cause high transient current consumption or may interfere with other substrate-noise sensitive analog functions within the SOC. Moreover, switch capacitor topologies can increase latch up risk and may require on chip charge-pumps in sub-1V environments. Banking only on CMOS transistors in subthreshold (which emulate pseudo-bipolar junction transistor (BJT) like behavior) to generate a V_BG helps with low-power and sub-1V objectives, but such designs generally exhibit wider V_BG variations, and likely require extra trimming and testing cost in post production. Forward biasing PMOS or NMOS body-source terminals can help increase bandgap amplifier headroom, but requires proper modeling at different process nodes. Additionally, such a design approach generally entails increased parasitic current and leakage risk at elevated temperatures, which limits the reference high temperature operations. The majority of reference topologies that impose on the SOC by requiring special processes or rely on non-standard use of transistors have not made it to high-volume applications. This is not just because of their added manufacturing cost, special device characterization, or special device modeling requirements. As noted earlier, variations on process, and deviation on devices for one function of an SOC (such as a reference) increases costs and manufacturing risk on the entire SOC and handicaps the SOC's process node portability, which ties the manufacturer's hands.

SUMMARY OF THE INVENTION

An aspect of the disclosure herein is a fractional bandgap reference circuit comprising: a voltage loop having two scaled transistors to provide a positive TC signal across a first resistor; a feedback loop having a first amplifier which generates a positive TC signal across a second resistor and a third resistor; a second amplifier controlling a plurality of current sources which conjunctively produce the residual signal with near zero TC to the feedback loop to produce a voltage reference across a fourth resistor; and wherein the voltage loop positive TC signal is supplied by the positive TC signal of the feedback loop to generate the residual signal with near zero TC.

Another aspect of the disclosure herein is a method of creating a fractional bandgap comprising: providing a positive TC signal across a first resistor from a voltage loop having two scaled transistors; generating a positive TC signal from a feedback loop having a first amplifier across a second resistor and a third resistor; controlling a plurality of current sources by a second amplifier to conjunctively produce a residual signal with near zero TC to the feedback loop to produce a voltage reference across a fourth resistor; and generating the residual signal with near zero TC by supplying the voltage loop positive TC signal from the positive TC signal of the feedback loop.

Another aspect of the disclosure herein is a fractional bandgap reference circuit comprising: a voltage loop having a first transistor and second transistor isolated by a first amplifier from a first resistor and which creates a positive TC current and a negative TC voltage; a second amplifier in conjunction with a plurality of current sources to receive the positive TC current from the voltage loop and supply a feedback loop with a positive TC signal; and the feedback loop having a scaled second resistor and third resistor which receive the positive TC current, which is fed to a second and a third resistors to produce a positive TC voltage that is combined with the negative TC voltage to generate a near zero TC voltage as a reference voltage.

Another aspect of the disclosure herein is a method of generating a fractional bandgap comprising: isolating a voltage loop having a first transistor and second transistor by a first amplifier from a first resistor, wherein the voltage loop creates a positive TC current and a negative TC voltage; receiving the positive TC current at an amplifier and conjunctively with a plurality of current sources to supply feedback loop with a positive TC signal; and receiving the positive TC signal at the feedback loop which includes a scaled second resistor and third resistor and which feeds a positive TC current to a second resistor and a third resistor to produce a positive TC voltage that is combined with the negative TC voltage to generate a near zero TC voltage as a reference voltage.

Another aspect of the disclosure herein is a fractional bandgap reference circuit comprising: a voltage loop having a first transistor, a second transistor, first amplifier and a first resistor; said first resistor connected to the base of the first transistor to isolate a quiescent current of the first transistor from the current through the first transistor to generate a positive TC current; and said first amplifier buffering a negative TC voltage between the voltage loop and a feedback loop; a second amplifier in conjunction with a plurality of current sources which receives a positive TC current that was generated by the voltage loop and supplies feedback loop with a positive TC signal; and the feedback loop having a scaled second resistor and third resistor which receive the positive TC current, which is fed into the second and the third resistors to generate a positive TC voltages that is combined with the negative TC voltage to generate a near zero TC voltage as a reference voltage.

Another aspect of the disclosure herein is a bandgap reference circuit comprising: a plurality of V_T cells connected in series to produce a V_PTAT; wherein each of said V_T cells comprise a first amplifier and scaled first transistor and second transistor operating at scaled currents which are biased by the two scaled current sources; said first amplifier generating a positive TC output voltage at a base of the first transistor so that the amplifier inputs are substantially equalized; a V_EB transistor biased by a current source to generate a V_CTAT; and wherein the V_PTAT and V_CTAT are connected in series to create a voltage reference.

Another aspect of the disclosure herein is a method of producing a bandgap reference comprising: producing a V_PTAT from a plurality of V_T cells connected in series; wherein each of said V_T cells comprise a first amplifier and scaled first transistor and second transistor operating at scaled currents which are biased by at least two scaled V_T cell current sources; said first amplifier generating a positive TC output voltage at a base of the first transistor so that the amplifier inputs are substantially equalized; biasing a V_EB transistor by a current source to generate a V_CTAT; and creating a voltage reference from the V_PTAT and the V_CTAT.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic circuit diagram of a first embodiment illustrating a bandgap (or fractional bandgap) in a voltage regulator circuit.

FIG. 2 is a schematic circuit diagram of second embodiment illustrating a bandgap (or fractional bandgap) in a voltage regulator circuit.

FIG. 3 is a schematic circuit diagram showing a third embodiment of an alternative bandgap reference to FIG. 2.

FIG. 4 is a schematic circuit diagram showing a bandgap reference according to a fourth embodiment.

FIG. 5 is a circuit schematic showing a prior art conventional bandgap reference.

DETAILED DESCRIPTION

Numerous embodiments are described in the present application and are presented for illustrative purposes only. The described embodiments are not, and are not intended to be, limiting in any sense. One of ordinary skill in the art will recognize that the disclosed embodiment(s) may be practiced with various modifications and alterations, such as structural, logical, and electrical modifications. Although particular features of the disclosed embodiments may be described with reference to one or more particular embodiments and/or drawings, it should be understood that such features are not limited to usage in the one or more particular embodiments or drawings with reference to which they are described, unless expressly specified otherwise.

The present disclosure is not a literal description of all embodiments of the invention(s). Also, the present disclosure is not a listing of features which must necessarily be present in all embodiments On the contrary, a variety of components are described to illustrate the wide variety of possible embodiments of the present invention(s).

Although process (or method) steps may be described or claimed in a particular sequential order, such processes may be configured to work in different orders. In other words, any sequence or order of steps that may be explicitly described or claimed does not necessarily indicate a requirement that the steps be performed in that order. The steps of processes described herein may be performed in any order possible. Further, some steps may be performed simultaneously despite being described or implied as occurring non-simultaneously (e.g., because one step is described after the other step). Moreover, the illustration of a process by its depiction in a drawing does not imply that the illustrated process is exclusive of other variations and modifications thereto, does not imply that the illustrated process or any of its steps are necessary to the embodiment(s), and does not imply that the illustrated process is preferred.

Although a process may be described as including a plurality of steps, that does not imply that all or any of the steps are preferred, essential or required. Various other embodiments within the scope of the described invention(s) include other processes that omit some or all of the described steps.

Although a circuit may be described as including a plurality of components, aspects, steps, qualities, characteristics and/or features, that does not indicate that any or all of the plurality are preferred, essential or required. Various other embodiments may include other circuit elements or limitations that omit some or all of the described plurality.

The following description has been presented for purposes of illustration and description and is not intended to be exhaustive or to limit the embodiments to the precise form disclosed. Many modifications and variations are possible in light of the teachings disclosed herein. The embodiments were chosen and described to explain principles of operation and their practical applications. However, the scope of the invention is to be defined by the claims.

As mentioned earlier, over the last few decades, multi-billion dollar annual investments in equipment and fabrication technologies have resulted in advancements in low voltage sub-micron CMOS processes that have yielded continued improvements in die density, tightly controlled device parameters {i.e. threshold voltage (V_TH), Boltzmann Constant (k), oxide thickness (T_ox), and other parameters}, and increased matching. Naturally, such manufacturing advancements have materially helped enhance the performance, precision, and matching of analog functions (i.e., conventional 2-stage amplifiers with lower offset, lower offset drift, and tighter matching) which is leveraged in the disclosure herein. The embodiments described in this application may be implemented using standard digital CMOS processes and parasitic vertical BJTs with inherent predictability due to the stable device physics base-emitter voltage (V_BE) and V_T parameters that should enhance manufacturing viability and ease design portability. The majority of transistors used in the bandgap family operate in the sub-threshold region to accommodate their low power objective and help with headroom without use of any custom devices. Additional analysis of relevant to low power and low cost bandgap designs are also hereby incorporated by reference: 1) Ali Far, (2013), “A 5 μW Fractional CMOS Bandgap Voltage and Current Reference”, 2013 IEEE GHTCE Conference Proceedings, China; 2) Ali Far, (2013), “A 400 nW CMOS Bandgap Voltage Reference”, 2013 IEEE ICEESE Conference Proceedings, Malaysia.; and 3) Ali Far, (2014), “A Low Supply Voltage 2 ρW Half Bandgap Reference In Standard Sub-p CMOS”, 2014 IEEE CONNECT Conference Proceedings, India.

The term “near zero” is used throughout this disclosure and hereby a brief explanation is provided as follows. A bandgap voltage, or a fractional bandgap voltage, generally is a voltage that has near zero temperature coefficient (TC). For example for a bandgap reference to have a near zero TC, the output reference voltage can be about 1.2V with variations for instance in the range of +/−10 mV over a span of temperature from negative 50 degrees centigrade to positive 150 degrees centigrade (i.e. 1.99V<VBG<1.21V with −50 C<T<+150 C). Bandgaps without (post manufacturing adjustments or) trim can have wider variations such as for example +/−50 mV or tighter variations of for example +/−1 mV with trim. For a fractional bandgap (say for a half bandgap), an example of near zero TC half bandgap reference voltage could be about 0.6V with variations for instance in the range of +/−5 mV over a span of temperature from negative 50 degrees centigrade to positive 150 degrees centigrade (i.e 0.595V<VHBG<6.05V with −50 C<T<+150 C). Half bandgaps without (post manufacturing adjustments or) trim can have wider variations such as for example +/−25 mV or tighter variations of for example +/−0.5 mV with trim.

In the first embodiment discussed herein a bandgap voltage V_BG may be generated by adding (or combining) a proportional to absolute temperature (PTAT) voltage, V_PTAT, whose voltage increases with temperature (positive TC) to a complementary to absolute temperature (CTAT) voltage, V_CTAT, whose voltage decrease with temperature (negative TC). Henceforth, summation of the V_CTAT (with a negative TC) and V_PTAT (with a positive TC) can be scaled via resistors (i.e. resistor dividers) to produce a fractional bandgap voltage, V_ra, which is substantially constant over temperature variation. This fractional bandgap reference embodiment may be implemented using standard CMOS manufacturing processes, and may not require any custom device or special process, and hence has lower production costs. One of the benefits of these embodiments is that both a fractional current reference I_ra and a fractional voltage reference V_ra are generated concurrently and low current consumption is optimized around ambient temperature (which is where some applications operate most of the time). This first embodiment includes elements such as amplifiers, resistors, BJTs, and CMOS transistors (i.e. PMOS transistors that can be configured to function as current sources).

FIG. 1 is a circuit schematic showing a bandgap reference according to the first embodiment. Throughout this description the values of R_f1a=R_f2a=R_fa and current source values which are determined as a function of PMOS W/L scale factors may be set to a predetermined value b=c=d=1. Also note that ‘b’, ‘c’, and ‘d’ scale factors may each be in a predetermined range of 1 to 10 depending on considerations such as die size and current consumption amongst others. The current through R_ia is a PTAT current where I_Ria has a positive TC, comprised of the difference of V_EB voltages (between Q_1a and Q_2a) across R_ia. One terminal of each of R_f1a and R_f2a is connected to the reference output V_ra, with near zero TC, while each of R_f1a and R_f2a's other terminal is connected to a V_EB (of Q_1a and Q_2a) potential with a negative TC. As a result, R_fa's current I_Rfa will increase with temperature. Now I_Ria+I_Rfa+I≈0 in order to meet the requirement that sum of currents at a circuit node must be zero. Since I_D of P_3a=V_ra/R_Oa=I_ra, then the I_D in P_1a through P_4a is equal to I_ra. Moreover, since I_Ria increases with temperature and I_Rfa supplies R_ia with the positive TC current it needs, then the residual I_ra is left for the current sources that would be biased with a near zero TC. Therefore, P_1a through P_4a operate at an I_D that has constant TC that is I_ra, which generates the fractional voltage reference across R_0a before it is buffered to generate V_ra.

The connections of the elements in FIG. 1 are described as follows. Transistors P_1a, P_2a, P_3a, and P_4a have their source terminals connection to node 1 which is the positive supply voltage or V_DD. Also, P_1a through P_4a have their gates connected to node 6a, whose voltage is driven by the output node of amplifier A_1a. At node 3a, drain of P_1a is connected to the emitter of BJT, Q_1a, as well as the inverting input terminal of amplifier A_1a. At node 2a, emitter of BJT, Q_2a, is connected to one terminal of R_ia. At node 4a, drain of P_2a is connected to the other terminal of resistor R_ia as well as non-inverting input terminal of A_1a. Collectors and bases of Q_1a and Q_2a are connected to V_SS, which is node 0. At node 7a, drain of P_3a is connected to one terminal of resistor R_Oa whose other terminal is connected to the ground potential. Also, node 7a is where the fractional bandgap voltage is generated. At node 5a, the inverting input of A_2a is connected to its output, which configures A_2a as a unity gain amplifier buffer, enabling node 5a which is V_ra to substantially equal the voltage at node 7a. Also, the buffered V_ra voltage at node 5a is able to drive the common terminal of R_f1a and R_f2a feedback resistors, whose other terminals are connected to nodes 3a and 4a respectively. Setting the source voltages of transistors P_1a through P_4a to the same voltage at node 1 and setting their gates to the same voltage at node 6a enables configuring transistors P_1a through P_4a as four current sources whose quiescent currents can be scaled substantially proportional to their respective transistors' width over length ratios, W/L. Node 8a is the drain of PMOS transistor P_4a which operates and provides a reference current source. For example, in the FIG. 1 circuit schematic (W/L)_P1a=b×(W/L)_P2a, and (W/L)_P3a=c×(W/L)_P2a, and (W/L)_P4a=d×(W/L)_P2a, which accommodates scaling PMOS current sources P_1a, P_3a, and P_4a quiescent currents may be substantially scaled as multiples of ‘b’, ‘c’, and ‘d’ times the quiescent current in P_2a. In the embodiment described here, b=c=d=1, which causes the quiescent currents in P_1a through P_4a to be equal (or substantially equal) and the same (or approximately the same) as I_ra. Also note that ‘b’, ‘c’, and ‘d’ scale factors may each be in a predetermined range of 1 to 10 depending on considerations such as die size and current consumption, amongst others.

The details of various functions of the circuit in FIG. 1 are described in 2 segments.

In the first segment, we begin with the voltage loop that starts (at node 0) going through V_BE of Q_2a (to node 2a) then through R_ia (to node 4a) then across A_1a inputs (to node 3a) and finally back through V_EB of Q_2a to node zero (0). In describing the operation of this voltage loop, amplifiers A_1a and A_2a have substantially minimal offset voltages. Also, note that V_EB of Q is denoted as V_EB|_Q and I_D of P is denoted as I_D|_P.V_EB|_Q1a≈I_Ria×R_ia+V_EB|_Q2a.Thus: V_EB|_Q1a−V_EB|_Q2a≈I_Ria×R_ia≈ΔV_EB.Therefore, I_Ria≈ΔVEB/R_ia≈(V_EB|_Q1a−V_EB|_Q2a)/R_ia.To generate ΔV_EB, the emitter areas of Q_2a are scaled ‘a’ times bigger than that of Q_1a, and as described earlier the quiescent currents for Q_1a and Q_2a can be scaled via P_2a and P_1a currents, whose W/Ls are scaled by ‘b’:ΔV_EB≈V_EB|_Q1a−V_EB|_Q2a≈V_T×ln(a×b),where V_T is the thermal voltage and equal to (K×T)/q and k is the Boltzmann constant, q is the electron charge, and T is the temperature. Thus:I_Ria≈I_PTAT≈ΔV_EB/R_ia≈V_T.×ln(a×b)/R_ia.Note for example that for a=5 and b=11, then 100 mV≈VT×ln(55) could be a typical value of V_T×ln(a×b) term in the above equation. Consistent with principles of conventional bandgaps (such as the ones depicted in prior art FIG. 5), note that ΔV_EB tracks V_T which has a positive TC, and this yields the PTAT signal (current) needed in the bandgap.

In the second segment, we turn our attention to the mechanism by which fractional reference current I_ra and voltage V_ra (with near zero temperature variations) are generated conjunctively. As noted earlier, in this embodiment:b=c=d=1I_D|_P1a≈I_D|_P2a≈I_D|_P3a≈I_D|_P4a≈I_ra.R_f1a=R_f2a=R_fa.The sum of currents at node 4a and 3a must equal zero. In other words:I_D|_P2a+I_Ria+I_Rfa=0.Also, I_Rfa≈(V_ra−V_EB)/R_fa I_ra=V_ra/R_Oa I_ra+[V_T.×ln(a×b)]/R_ia+I_Rfa≈0Thus, V_ra/R_Oa+[V_T. ln(a×b)]/R_ia+(V_ra−V_EB)/R_fa≈0Solving for V_ra in the above equation (with b=1), we therefore get:V_ra≈[V_EB+(R_fa/R_ia)×V_T×ln(a)]×(1+R_fa/R_0a)⁻¹

The method of operation of the FIG. 1 embodiment may be summarized as follows. An I_PTAT or I_Ria is generated in the Q_2a−R_ia−A_ia−Q_1a voltage loop via ΔV_EB of Q_2a−Q_1a placed across R_ia, where ΔV_EB≈V_EB|_Q1a−V_EB|_Q2a≈V_T×ln (a×b). Concurrently, the sum of currents at a circuit node (node 4a) are zero, meaning that I_Rf2a+I_Ria+I_D|_P2a=0. Amplifier A_2a accommodates the (unity gain) buffering and the feeding back of V_ra on to R_f1a and R_f2a. Hence, I_Rfa≈(V_ra−V_EB)/R_fa, which increases with temperature since V_ra is designed to have near zero TC while V_EB voltage decreases with temperature. Therefore, since I_RF2 which increases with temperature can provide the positive TC I_Ria, then (in order for A_1a to keep its inputs substantially equal) amplifier A_1a provides the gates of P_1a-P_4a (current sources) with a voltage such that P_1a-P_4a will have near zero TC currents. Thus, I_D|_P2a≈V_ra/R_Oa which is I_ra and has a near zero TC.

To describe some of the benefits of this embodiment, an example may be helpful. Given that V_EB has a −2 mV/° C., let us say V_EB|_T=cold=0.8V, V_EB|_T=ambient=0.6V, and V_EB|_T=hot=0.4V. The circuit is configured such that V_ra|_cold>T>hot=0.6V. Also note that current consumption of the FIG. 1 circuit embodiment can be in a predetermined range of for example 1 micro ampere to 5 micro amperes, depending on considerations such as die size and power consumption objectives amongst others.

Another benefit is that the current consumption at ambient temperature is minimized, and this is advantageous because some applications operate in ambient most of the time but need to withstand hot and cold temperature hikes sometimes. The current through R_f1a and R_f2a is nearly zero since the voltages across them are nearly zero: V_ra|_cold>T>hot=0.6V=V_EB|_T=ambient=0.6V. Note also that here the zero current crossing point can be set to the middle of hot and cold temperature.

Another benefit of the FIG. 1 embodiment is that the same amplifier A_2a may be used to generate I_ra and V_ra. Furthermore, A_2a may be configured as a unity gain buffer amplifier which has some current output drive capability and can eliminate the need for an external amplifier for driving some output loads. Besides cost, an external amplifier can contribute to additional offset and gain error variation over temperature.

Conventional bandgaps (such as the one depicted in FIG. 5) as well as this proposed embodiment would require trimming for more precise TC. A trim for TC in the embodiment disclosed in FIG. 1 goes a longer way since both I_ra and V_ra values (with A_1a to provide some voltage output drive capability) can be centered for optimal TC with the same single trim as opposed to multiple trims for each of V_ra, I_ra, and one to null-out the non-ideal effects of an external buffer amplifier.

Another advantage of the embodiment of FIG. 1 is that by programming different R_Oa, R_fa, or scaling P_1a-P_4a, different fractional bandgap voltages can be attained.

Another benefit of the FIG. 1 embodiment is that it also allows for decoupling of V_SS and ground (GND), where technically V_SS can run at a lower voltage than the ground potential. Note that V_DD may be lowered to substantially similar value by which the V_SS is lowered limited by the forward biasing of base-collectors of Q_1a and Q_2a. Such capability of this embodiment can accommodate proper operations of the Bandgap reference at lower V_DD while V_ra remains referenced to the ground potential.

Supply voltage coefficient and bandgap's TC of this embodiment should also benefit from the circuit in FIG. 1 for at least 2 reasons: (1) there is more temperature stability associated with current sources P_1a-P_4a which are biased with V_ra/R_Oa which is a current with near zero TC; and (2) transistors P_3a-P_4a drain voltages are centered near V_ra, which at the middle point of P_1a-P_2a drain voltages that shift around V_ra from 400 mV at hot temperatures to 800 mV at cold temperatures. This should minimize and help center systematic mismatch due to V_A (early voltage of CMOS transistors) of PMOS transistors, and variations of V_ra over V_DD and temperature spans.

The embodiment of FIG. 1 further allows for stacking CMOS transistors to help enhance the output impedance of current sources P_1a-P_4a. Note also that V_ra can be fed back through only one resistor R_f1a to the Q_ia side. Eliminating R_f2a can save power and area at the expense of additional bow in TC.

In the embodiments disclosed in FIGS. 2 and 3 a bandgap voltage, V_BG, is generated by adding a PTAT voltage, V_PTAT, whose voltage increases with temperature (positive TC) to a CTAT voltage, V_CTAT, whose voltage decrease with temperature (negative TC). Henceforth, summation of the V_CTAT (with a negative TC) and V_PTAT (with a positive TC) can be scaled via resistors (divider) to produce a fractional bandgap voltage, V_rb, which is substantially constant over temperature variation. To accomplish this objective in standard CMOS and in the context of low supply voltage and current, alternative embodiments are shown in FIG. 2 and another in FIG. 3. These embodiments include elements such as amplifiers, resistors, BJT, and CMOS transistors that can be configured to function as current sources. The embodiments of FIGS. 2 and 3 improve bandgap's cost-performance by maintaining low current consumption, keeping the bandgap operational at lower power supply, and maintaining stability over frequency while keeping the die size small to keep the cost down.

FIG. 2 is a circuit schematic showing a fractional bandgap reference. The embodiment depicted in FIG. 2 establishes a PTAT current (I_Rib) in P_3b via forcing a ΔV_EB across R_ib, where ΔV_EB=V_T×ln (a×b) which is a PTAT signal. Note that Q_2b is ‘a’ times the size Q_1b, and P_1b is ‘b’ times the size of P_2b. Amplifier A_1b is configured in unity gain to set the voltage at one side of R_ib to V_EB of Q_2b. Conjointly, amplifier A_2b substantially equalizes the voltage on the other side of R_ib (by providing enough voltage to the gate of P_3b) with V_EB of Q_1b. Therefore a voltage of ΔV_EB≈V_EB|Q_1b−V_EB|Q_2b is generated across R_ib. The current I_Rib=I_PTAT is properly scaled and mirrored in P_1b through P_4b, where P_4b feeds such scaled I_PFAT to the output node at V_rb. In concert, a CTAT current I_Rfb flows through R_fb, which is generated via the same V_EB of Q_2b, after it is buffered by unity gain amplifier A_1b. The aggregation of the three currents I_Rfb, I_Rib, and I_ROb at the output node 7b generates the desired V_rb. This embodiment segregates the quiescent currents of Q_1b-Q_2b-P_1b-P_2b loops from the currents that flows through R_ib, R_fb, and R_Ob. Such implementation provides a needed flexibility in biasing Q_1b-Q_2b, including operating them at advantageously lower quiescent currents, to minimize the V_EB voltage drop (without requiring prohibitively large R_ib or R_fb or R_Ob). This move, in turn can free up the scarce operational supply voltage headroom, especially in deep sub-μ CMOS sub-1V power supply environments, where every 10 s of mV counts. For example, lowering the quiescent current in the Q_1b and Q_2b BJTs from 10 ρA to 5 nanoamperes (nA) lowers V_EB by about 200 mV, which translates to gaining back about 20% of extra V_DD supply headroom in sub-1V environments. By segregating low quiescent currents in Q_1b and Q_2b BJTs from the PTAT and CTAT signal loop, very large value resistors R_ib, R_fb, R_Ob may be avoided (e.g. in the 10 MΩ+ range for currents in 10s of nA).

The connections of elements in FIG. 2 are now described. Transistors P_1b, P_2b, P_3b, and P_4b have their source terminals connection to node 1 which is the positive supply voltage or V_DD. Transistors P_1b through P_4b have their gates connected to node 6b, whose voltage is set by the output node of amplifier A_2b. At node 3b, drain of P_1b is connected to the emitter of BJT, Q_1b, as well as the inverting input terminal of amplifier A_2b. At node 2b, drain of P_2b is connected with the emitter of BJT, Q_2b, as well as the non-inverting node of amplifier A_1b. At node 4b, drain of P_3b is connected to resistor R_ib as well as non-inverting input terminal of A_2b. At node 7b, which is the output of the reference voltage V_rb, drain of P_4b is connected to resistor R_ob as well as resistor R_fb. The other side of R_Ob is connected to ground (GND), which can be at a different voltage that V_SS where base and collector terminals of Q_1b and Q_2b are connected to V_SS. At node 5b, R_ib is connected to R_fb, which are connected to the output of A_1b. Note that at node 5b, amplifier A_1b has its inverting input terminal connected to its output, which configures A_1b as a voltage unity gain buffer amplifier, where the voltage at the non-inverting input of A_1b is substantially equalized with the voltage its output at node 5b. Amplifier A_2b is utilized here to substantially equalize its negative at node 3b with its positive input at node 4b. Setting the source voltages of transistors P_1b through P_4b to the same voltage at node 1b (which is V_DD) and setting their gates to the same voltage at node 6b enables configuring transistors P_1b through P_4b as four current sources whose quiescent currents can be scaled substantially proportional to their respective transistors' width over length ratios, W/L. For example, in the FIG. 2 circuit schematic (W/L)_P1b=b×(W/L)_P2b, and (W/L)_P3b=c×(W/L)_P2b, and (W/L)_P4b=d×(W/L)_P2b, which accommodates scaling P_1b, P_3b, and P_4b quiescent currents to be substantially scaled as multiples of ‘b’, ‘c’, and ‘d’ times the quiescent current in P_2b.

The details of various methods of operation of the circuit in FIG. 2 is described in the following segment. We begin by describing the voltage loop that starts (at node 0 which is the VSS) going through Q2b (to node 2b) then across A_1b (to node 5b) then through Rib (to node 4b) then across A2b inputs (to node 3b) and finally through Q_2b back to node 0. In this voltage loop, amplifiers A_1b and A_2b have substantially minimal offset voltages (note that V_EB of Q is denoted as V_EB|_Q)V_EB|_Q1b≈I_PTAT×R_ib+V_EB|_Q2b, where I_PTAT≈I_Rib.Thus: V_EB|_Q1b−V_EB|_Q2b≈I_PTAT×R_ib≈ΔV_EB.Therefore,I_D|_P3b≈I_PTAT≈ΔV_EB/R_ib≈(V_EB|_Q1b−V_EB|_Q2b)/R_ib To generate ΔV_EB, the emitter areas of Q_2b are scaled ‘a’ times bigger than that of Q_1b, and as described earlier the quiescent currents for Q_1b and Q_2b can be scaled via P_2b and P_1b currents, whose W/Ls are scaled by ‘b’:ΔV_EB≈V_EB|_Q1b−V_EB|_Q2b≈V_T×ln(a×b).Thus:I_D|_P3b≈I_PTAT≈ΔV_EB/R_ib≈[V_T×ln(a×b)]/R_ib.Note for example that for a=5 and b=11, then 100 mW≈VT×ln(55) could be a typical value of [V_T×ln (a×b)] term in the above equation. Consistent with principles of conventional bandgaps (such as the one depicted in prior art FIG. 5), note that ΔV_EB tracks V_T which has a positive TC, and this yields the PTAT signal (current) needed in the bandgap.

Next is a summary of what has been described thus far about the FIG. 2 bandgap which pertains mostly to the operation of PTAT signal generation. Amplifier A_1b (which is configured in unity gain with its inverting input connected to its output) operates as a buffer for V_EB|_Q2b voltage to be fed to one side of R_ib. Concurrently, amplifier A_2b senses the V_EB|_Q1b at its inverting input and while the output of A_2b generates the needed gate voltage for P_3b. Configured as such, transistor P₃ drain current feeds the other side of R_ib in order for A_2b to substantially equalizes its (inverting and non-inverting) inputs it to V_EB|_Q1b. Hence, the voltage across R_ib is set to ΔV_EB=V_EB|_Q1b−V_EB|_Q2b which is a P_TAT current or I_PTAT. As noted earlier, given that A_2b drives the gates P_1b through P_4b, whose sources are all connected the V_DD, then the quiescent current of P₁ through P₄ are also scaled (by ‘b’, ‘c’, ‘d’) as a function of I_PTAT, which has a positive temperature coefficient as stated earlier.

The circuit of FIG. 2 enables Q_1b and Q_2b quiescent currents to be scaled and track I_PTAT, which has a positive TC. An important benefit of this embodiment is that it also concurrently accommodates both Q_1b and Q_2b to operate at currents that are segregated (or isolated) from the current through R_ib. Such segregation results in feasible and meaningful design flexibility to scale the ratio (‘n’) of currents in R_ib as compared to the quiescent currents in Q_1b and Q_2b, to accommodate low voltage supply operations, which will be explained shortly. Since the current through R_i is I_PTAT≈I_D|_P3, then the current through Q_1b and Q_2b can be I_PTAT/n. For example, the amount of I_PTAT current through R_ib can be typically in the 10 μA level, while because of the segregation, the amount of I_PTAT current through BJTs Q_1b and Q_2b can be scaled down in the 5 nA levels, which make n=10 ρA/5 nA=2000. In this example, by lowering quiescent currents of BJTs by 2000, then the V_EB in Q_1b and Q_2b is lowered by V_T×ln (2000) which is about 200 mV. Having 100 mV to 200 mV less voltage across V_EB in sub-1V power supply environments is about 10% to 20% of extra headroom, which is very beneficial for bandgap operations to specifications with lower power supplies. Also note that current consumption of the FIG. 2 circuit embodiment can be predetermined in the range of for example 0.5 micro ampere to 2 micro amperes, depending on considerations such as die size and power consumption objectives, amongst others.

Moreover, by segregating the Q_1b and Q_2b currents from R_ib, the circuit of FIG. 2 has another benefit which allows for small size R_ib at very low currents. To make the point more clear a comparison with (such as the one depicted in FIG. 5) bandgaps may be helpful. As in the cases with ordinary bandgaps, R_ib is generally in series with emitter of Q_2b, and in such schemes both R_ib and Q_2b quiescent currents must be the same or in the μA ranges. To reduce the V_EB of Q_2b, its quiescent current needs to be substantially reduced. For example, if one were to design for ΔV_EB=100 mV and 5 nA of quiescent current through both R_ib and Q_2b, then that would require a 20 MΩ resistor which is size prohibitive. Instead, this embodiment allows for ΔV_EB=100 mV across R_ib with current of 10 μA which makes R_ib=10 KΩ which is very feasible and economical. In this example, therefore 2000 times smaller resistor and 20% extra headroom for sub-1V power supply environment is achieved. Area and headroom benefits less than 2000 and 20% (respectively) would make using the embodiment in FIG. 2 worth the while.

Now we turn to describing the CTAT and fractional voltage generation of this embodiment. The fractional bandgap voltage or V_rb at node 7b is provided with (1) the PTAT signal, which is via the drain terminal of P_4b feeding the equivalent parallel resistors R_Ob and R_fb, and (2) the CTAT signal, which unity gain amplifier A_1b feeds the resistor divider of R_Ob and R_fb. The method to arrive at the fractional bandgap voltage is described as follows. The sum of currents at node 7b must equal zero. In other words, I_D|_P3b+I_ROb+I_Rfb=0. Note that I_D of P_3b is denoted as I_D|_P3b. Transistor P_4b is a current source, whose current is a function of I_PTAT with a positive TC. If scale factors c=d then (W/L)_P3b=(W/L)_P4b causes, I_D|_P3b=I_D|_P4b. We showed earlier that:I_D|_P4b≈I_D|_P3b≈I_PTAT≈[V_T×ln(a×b)]/R_ib Moreover, at node 7b it can be seen that I_ROb=V_rb/R_Ob and I_Rf=(V_EB−V_r)/R_f. Given that I_D|_P3+I_RO+I_Rf=0, therefore the simplified mathematical solution for V_rb can be described as:V_rb≈V_BG×(1+R_fb/R_ob)⁻¹ ≈(m×V_T+V_EB)×(1+R_fb/R_ob)⁻¹ ≈[(R_fb/R_ib)×V_T×ln(a×b)+V_EB]×(1+R_fb/R_ob)⁻¹.Note for example that for a=5 and b=11, then 100 mV≈VT×ln(55) could be a typical value of V_T×ln (a×b) term in the above equation. Note that in the equation above, V_EB generates the CTAT voltage, which is added to that PTAT voltage (m×V_T) to generate a V_BG, where ‘m’ is a constant set by designing the proper scale factors in (a×b) and R_fb/R_ib. Production yields are significantly optimized when specifications depend on constants such as ‘m’ that are derived from ratios of transistor sizes. As just described in this embodiment, manufacturing yields are not compromised here because the V_BG and V_rb terms depend on ‘a’ which is the ratio of sizes of Q_1b and Q_2b, and ‘b” which is the ratio of W/L of P_1b and P_2b, and R_fb/R_ib and R_fb/R_ob which are the ratios of 2 resistors.

An example to describe the ratio of transistor and resistors sizes for this embodiment is presented. As noted earlier, a typical value of V_EB of a particular parasitic BJT device in a CMOS process is +0.6V with TC of −2 mV/° C. The value of V_T is about +25.8 mV with TC of +0.086 mV/° C. To generate a voltage that has a near zero TC, ‘m’ needs to be set at a multiple of 23.25, which is derived from ratio of TC of V_EB (−2 mV mV/° C.) to TC of V_T (+0.086 mV/° C.). Hence, V_BG≈m.V_T+V_EB≈1.2V. Moreover, by choosing R_fb=R_Ob, then a fractional V_rb voltage of V_BG/2 or 0.6V can be produced, which is desirable in sub-1V power supply applications.

Recall that the PTAT signal is generated while current through R_ib is segregated from quiescent current through Q_2b, in order to keep V_EB low and resistor sizes feasible. Similarly, this embodiment enables generating the CTAT signal from the same V_EB of Q_2b where the quiescent current in Q_2b is segregated from the current through R_fb via the same amplifier A_1b which is configured as a unity gain buffer amplifier. As was the case in the PTAT signal path, here in the CTAT signal path, there is a material benefit in segregating Q_2b from R_fb, where normal currents (e.g. in the ρA levels) can run through R_fb in order to keep the size of R_fb small for optimal die cost, while running current through Q_2b very small (e.g. in the nA levels) to keep Q_2b V_EB low and hence improve power supply headroom for sub-1V application. Note that unity gain amplifier buffer A_1b segregates Q_2b (which operate at very low currents) from R_fb by sinking and sourcing current changes in R_fb (as V_rb is kept constant while V_EB of Q_1b decreases with temperature) where R_fb currents are ‘n’ times higher that Q_2b quiescent currents. As such one of the additional benefits of this embodiment is that the same V_EB of Q_2b plays a central part in generating the CTAT and PTAT signals which improves manufacturing reliability.

The embodiment of FIG. 2 has an additional benefit which is a more stable frequency response when compared with conventional bandgaps (such as the one depicted in FIG. 5 which require a large size R_ib in series with emitter of Q_2b). By operating the currents in the nano-ampere level, P_2b and P_1b operate in subthreshold, where PMOS transistors behave like pseudo-BJTs. Therefore, current sources P_1b and P_2b (which behave like pseudo BJTs) provide the quiescent currents for BJTs Q_1b and Q_2b. As such, small signal frequency response in P_2b-Q_2b and P_1b-Q_ib signal paths would depend on and track V_T/I_BIAS≈1/g_m, and will largely not depend on the segregated R_ib, which would have otherwise interfered with frequency response as is the case with conventional bandgaps (such as the one depicted in FIG. 5). Moreover, because I_BIAS of P_2b-Q_2b and P_1b-Q_ib transistors are scaled as a multiple of I_PTAT with a positive TC, their frequency response would also track to be more stable over temperature variations.

Another benefit of FIG. 2 embodiment is that it also allows for decoupling of V_SS and ground (GND), where technically V_SS can run at a lower voltage than the ground potential. Note that V_DD may be lowered to substantially similar value by which the V_SS is lowered limited by the forward biasing of base-collectors of Q_1b and Q_2b. Such capability of this embodiment can accommodate proper operations of the Bandgap reference at lower V_DD while V_rb remains referenced to the ground potential.

The inclusion of the second amplifier in both FIGS. 2 and 3 will negatively contribute to higher noise and higher offset. Including the offset and noise (simplifying the equation to V_OFNS˜V_OFFSET+V_NOISE as an example) of the amplifiers in the V_rb equation can be approximated as follows:V_rb≈[(R_fb/R_ib)×V_T×(ln(a))+V_EB]×(1+R_fb/R_ob)⁻¹+[R_fb/R_ib)×(V_OFNS1±V_OFNS2)−V_OFNS2]×(1+R_fb/R_ob)⁻¹.However, this impact on noise and offset is minimal and manageable by the embodiment of FIG. 2. Firstly, although Q_1b and Q_2b BJT's noise contribution increases with lowered currents, noise is generally dominated by the CMOS section (amplifiers and current sources) and not the BJTs. Secondly, the multi-billion dollar annual investments in the manufacturing of CMOS fabrication processes, over the last few decades, has significantly improved devise matching, noise, and offset performance which will also benefit this embodiment. Thirdly, the manufacturing yield impact due to the second amplifier offset is minimal. This is explained as follows. For demonstrative simplicity let us assume that identical amplifiers A_1b and A_2b have similar V_offset, and that b=c=1 and R_ob=R_fb. Statistically, the contribution of amplifiers A_1b and A_2b random inaccuracies (non-systematic and random offset voltage, noise, and drift) in the PTAT loop (Q_1b-A_1b-R_ib-A_2b-Q_1b) follow the square root of sum of squares of such random effects. As such, employing an extra amplifier, A_1b, should not statistically translate to doubling the impact of amplifier's noise and offset on the V_rb output. That is the statistical contribution of amplifier offset voltage (V_ofs) could approximately be as follows:[√2(R_fb/R_ib)−1]×V_ofs.

FIG. 3 is a circuit schematic showing a third embodiment which is an alternative bandgap reference to FIG. 2. Here, the voltage loop consisting of R_ic, Q_2c, A_1c, and Q_1c forces a difference V_EB, or ΔV_EB, voltage to be placed across R_ic. The current ΔV_EB/R_ic is the I_PTAT current that is mirrored via P_3c onto P_1c, P_2c and P_4c. Concurrently, V_EB of Q_2c feeds R_fc, which constitutes the CTAT signal. Similarly, note that Q_2c is ‘a’ times the size of Q_1c and P_1c is ‘b’ times the size of P_2c. In FIG. 3, the PTAT signal through P_4c and CTAT signal through R_fc are then summed and scaled at the V_rc node, by the ratio of R_fc and R_Oc to generate the desired fractional V_BG. With R_ic connected to base of Q_2c to ground, the PTAT current through R_ic (e.g. in the ρA level) is segregated from the quiescent current in the Q_1c and Q_2c BJTs (e.g. in the nA level). Running currents in R_1c, R_fc, R_Oc at normal currents allows for small size resistors which save die costs, while concurrently running Q_1c and Q_2c at very low currents reduces V_EB voltage of Q_1c and Q_2c which helps operating the fractional bandgap at lower power supply voltage.

The connections of elements in FIG. 3 will now be described. Given the commonalities between embodiments of FIG. 2 and FIG. 3, descriptions of circuit operations and benefits that overlap between them and that have been covered in the section for FIG. 2 are not discussed in detail. Transistors P_1c, P_2c, P_3c, and P_4c have their source terminals connection to node 1 which is the positive supply voltage or V_DD. Transistors P_1c through P_4c have their gates connected to node 6c, whose voltage is set by the output node of amplifier A_2c. At node 3c, drain of P_1c is connected to the emitter of BJT, Q_1c, as well as the inverting input terminal of amplifier A_2c. At node 2c, drain of P_2c is connected with the emitter of BJT, Q_2c, as well as the non-inverting node of amplifier A_1c plus the non-inverting node of amplifier A_2c. At node 4c, drain of P_3c is connected to resistor R_ic as well as the base terminal of BJT transistor Q_2c. At node 7c, which is the output of the reference voltage V_rc, drain of P_4c is connected to resistor R_Oc as well as resistor R_fc. The other side of R_Oc is connected to ground (GND), which can be at a different voltage than V_SS when base and collector terminals of Q_1c and Q_2c are connected to V_SS. At node 5c, R_fc is connected to the output of amplifier A_1c which is configured in unity gain (functioning as a voltage buffer) with its inverting input connected to its output. Amplifier, A_2c is utilized here to substantially equalize its negative at node 3c with its positive input at node 4c. Setting the source voltages of transistors P_1c through P_4c to the same V_DD voltage at node 1c and setting their gates to the same voltage at node 6c enables configuring transistors P_1c through P_4c as four current sources. Here, P_1c-P_4c quiescent currents can be scaled substantially proportional to their respective transistors' width over length ratios, W/L. For example, in the FIG. 3 circuit schematic (W/L)_P1c=b×(W/L)_P2c, and (W/L)_P3c=c×(W/L)_P2c, and (W/L)_P4c=d×(W/L)_P2c, which accommodates scaling P_1c, P_3c, and P_4c quiescent currents to be substantially scaled by ‘b’, ‘c’, and ‘d’ times the quiescent current in P_2c.

The details of various methods of operation of the circuit in FIG. 3 is described in the following segment. We begin by describing the voltage loop that starts (at node 0) going through R_ic (to node 4c) then over Q_2c (to node 2c) then across A_2c (to node 3c) then across Q_2c back to node 0. To describe operation of this voltage loop, amplifiers A_1c and A_2c will have substantially minimal offset voltages:I_PTAT×R_ic+V_EB|_Q2c≈V_EB|_Q1c, where I_PTAT≈I_Ric.Thus: V_EB|_Q1c−V_EB|_Q2c≈I_PTAT×R_ic≈ΔV_EB.Therefore,I_D|_P3c≈I_PTAT≈ΔV_EB/R_ic≈(V_EB|_Q1c−V_EB|_Q2c)/R_ic ΔV_EB≈V_EB|_Q1c−V_EB|_Q2c≈V_T×ln(a×b).Hence:I_D|_P3c≈I_PTAT≈ΔV_EB/R_ic≈[V_T×ln(a×b)]/R_ic.

Note for example that for a=5 and b=11, then 100 mV≈VT×ln(55) could be a typical value of [V_T×ln (a×b)] term in the above equation. Let us summarize what has been described thus far about embodiment of the FIG. 3 bandgap, which pertains mostly to the operation of PTAT signal generation. Amplifier A_2c substantially equalizes its inputs, which are connected to emitters of Q_1c and Q_2c, by way of A_2c's output driving P_3c's gate. Transistor P_3c, configured as such, provides enough current via its drain terminal to the top of R_ic which produces a voltage at node 4c that is equal to the difference voltage between V_EB of Q_1c and V_EB of Q_2c. In summary, the voltage across R_ic is set to ΔV_EB=V_EB|_Q1c−V_EB|_Q2c which is a P_TAT current or I_PTAT. As noted earlier, given that A_2c drives the gates P_1c through P_4c, whose sources are all connected the V_DD, then the quiescent current of P_1c through P_4c are also scaled (by ‘b’, ‘c’, ‘d’) as a function of I_PTAT, which has a positive temperature coefficient as stated earlier.

As was the case with the embodiment in FIG. 2, the embodiment in FIG. 3 has the important benefit of segregating Q_1c and Q_2c to operate at very low currents compared to the current that flows through R_ic. Since current through R_ic is I_PTAT≈I_D|_P3c, then the current through Q_1c and Q_2c c an be I_PTAT/n. By lowering quiescent of BJTs by ‘n’, then the V_EB in Q_1c and Q_2c is lowered by V_T×ln (n), which can be in the range of 150 mV to 200 mV. Having 150 mV to 200 mV less voltage across V_EB in sub-1V power supply environments is about 15%-20% of extra headroom, which is very beneficial for bandgap operations to specifications with lower power supplies. Also note that current consumption of the FIG. 3 circuit embodiment may be a predetermined value for example in the range of 500 nano amperes to 2 micro amperes, depending on considerations such as die size and power consumption objectives amongst others.

Note that In this embodiment, the error due to BJT's low β (e.g. in the 1-5 range) is attenuated by ‘n’ or I_Ric/I_Q2c ratio, where IR_Ric is the current through R_ic (the PTAT resistor), and I_Q2c is the quiescent current through Q_2c BJT.

Similar to the operational principles of FIG. 2, in the embodiment of FIG. 3, by segregating the Q_1c and Q_2c currents from R_ic, another benefit is realized which is keeping size of R_ic small despite Q_1c and Q_2c very low currents.

Now we turn to describing the CTAT and fractional voltage generation of this embodiment. Amplifier A_1c, configured as a unity gain buffer amplifier functions to segregate V_EB of Q_2c which operates in nano-ampere current levels (a CTAT signal) and A_1c functioning as a buffer to absorb the sink and source currents through R_fc which operates with micro ampere levels. As stated earlier, this is beneficial because the same V_EB of Q_1c is used to generate CTAT and PTAT signals, V_EB of Q_1c is kept low to help operations at lower V_DD, and keep resistor R_fc size small by running normal current levels through it.

The method to arrive at the fractional bandgap voltage is described as follows. The sum of currents at node 7c must equal zero. For c=d we get I_D|_P3c=I_D|_P4c. Moreover, at node 7c it can be seen that I_ROc=V_rc/R_Oc and I_Rfc=(V_EB−V_rc)/R_fc. Similar to the principle of operations in FIG. 2, for the embodiment in FIG. 3 we also have I_D|_P3c+I_ROc+I_Rfc=0, therefore the simplified mathematical solution for V_rc can be described as:V_rc≈V_BG×(1+R_fc/R_oc)⁻¹ ≈(m×V_T+V_EB)×(1+R_fc/R_oc)⁻¹ ≈[(R_fc/R_ic)×V_T×ln(a×b)+V_EB]×(1+R_fc/R_oc)⁻¹.

As previously discussed, to get optimal manufacturing fabrication yields, it is best to have performance specification depend on geometric ratios of devices available in the fabrication process. Here also, manufacturing yields are not compromised because the V_BG and V_rc terms depend on ‘a’ which is the ratio of sizes of Q_1c and Q_2c, and ‘b” which is the ratio of W/L of P_1c and P_2c, and R_fc/R_ic and R_fc/R_oc which are resistor ratios.

Similarly to FIG. 2, the embodiment of FIG. 3 has an additional benefit which is a more stable frequency response, when compared with conventional bandgaps (such as the one depicted in FIG. 5 which require a large size R_ic in series with emitter of Q_2c). Also, a similar benefit of FIG. 3 embodiment is that it also allows for decoupling of V_SS and ground (GND), where technically V_SS can run at a lower voltage than the ground potential. Note that V_DD may be lowered to substantially similar value by which the V_SS is lowered limited by the forward biasing of base-collectors of Q_1c and Q_2c. Such capability of this embodiment can accommodate proper operations of the Bandgap reference at lower V_DD while V_rb remains referenced to the ground potential.

For similar reasons discussed earlier, contribution in noise and offset due to lower currents and the extra amplifier are mitigated when compared to conventional bandgaps (such as the one depicted in FIG. 5).

As noted earlier, the risk and error due to BJT's low β (e.g. in the 1-5 range) is attenuated by ‘n’ or I_Ric/I_Q2c ratio. However, in certain CMOS fabrication processes where the base leakage current of parasitic BJTs (Q_2c which connected to R_ic) can derail high temperature accuracy. For such fabrication cases, the embodiment of FIG. 2 may be a more rugged solution.

The embodiment disclosed in FIG. 4 enables ultra low currents while preserving optimal performance, manufacturing quality, and low costs. Operating at ultra low currents generally requires large size resistors, which can occupy large die areas and thus cost more. Generally, BJT based references that generate a bandgap voltage yield more quality performance, and typically CMOS (non-BJT, non-bandgap) based references that yield ultra-low currents are resistorless. Hence, making large size bandgaps with high value resistors (as in FIG. 5) that have reasonable quality, or making resistorless references (that are not bandgap-based) that may have sub-par quality are two of the conventional approaches to realize ultra low power references. In FIG. 4 note that the circuit diagram consists of elements such as amplifiers, resistors, BJT, and CMOS transistors (i.e. PMOS transistors can be configured to operate as current sources).

The embodiment in FIG. 4 is ultra low power and it is different, among other things, in that it does not use resistors and is bandgap based. A summary of the description on FIG. 4 embodiment is first provided. Parasitic BJTs (such as Q₁ and Q_d1) are used to generate a multiple of a thermal voltage, which is V_T ln (n×m)=ΔV_EB=(such as V_EB|Q₁−V_EB|Q₂). Such multiple of V_T is then duplicated in series a proper number of times to generate a PTAT voltage and then added in with a V_EB (or CTAT voltage) to produce a 1.25 v bandgap voltage. Cascading scaled parasitic BJT V_EB followers looped-in with CMOS amplifiers (that have no built in offset), depicted in the embodiment in FIG. 4 offers some advantages. For example, it helps keep current consumption ultra-low and parasitic leakages minimized which can extend the higher temperature span of operation. Operating headroom is maximized in part because of lowered V_EB and V_GS of CMOS and BJT transistor, which operate in nano-ampere currents levels. Importantly, this embodiment eliminates use of resistors without the need for any special device or process. Such a bandgap can operate at ultra-low currents (i.e. nano-amperes) without using (large size) resistors and without compromising frequency response stability which saves significant die area and improves frequency response in ultra-low currents. This embodiment may be implemented only using standard CMOS processes and may not require any custom or special devices and hence lowers manufacturing cost. Also, importantly, manufacturing yields should improve with this embodiment because statistical (random) contributions from CMOS amplifier noise, offset, and drift terms are not multiplied but accumulated by the square root of sum of the squares, fundamentally because the proposed topology uses summation and not multiplication to generate the bandgap's PTAT term.

FIG. 4 is a circuit schematic showing a bandgap reference according to another embodiment of this invention. Let us start by describing the connections of elements in FIG. 4 with the PTAT signal generation section first. Transistor P_i, biased at current ‘i’, has its gate and drain connected to node 2d, which biases equal sized PMOS transistors P₁, P₂, P₃, P₄, P₅, and P₆ to also be biased at current ‘i’. Current source PMOS transistors P₁ through P₆ have their sources connected to node 1, which is V_DD, and have their drains connected to nodes 6d, 9d, 12d, 15d, 18d, and 21d respectively, which feeds the emitter of nX BJT PNPs in each of V_T cell. Specifically, collectors of BJTs Q_d1, Q_d2, Q_d3, Q_d4, Q_d5, and Q_d6 are connected to V_SS and their emitters are connected to nodes 6d, 9d, 12d, 15d, 18d, and 21d respectively, which bias the emitter currents of each of nX BJT PNP to about ‘i’. Similarly, transistor P_m, biased at current ‘m×i’ has its gate and drain connected to node 3d, which biases equal sized PMOS transistors P_d1, P_d2, P_d3, P_d4, P_d5, and P_d6 to also be biased at current ‘m×i’. Current source PMOS transistors P_d1 through P_d6 have their sources also connected to node 1, which is V_DD, and have their drains connected to nodes 5d, 8d,11d, 14d, 17d, and 20d respectively, which feeds the other BJT emitter of 1×PNPs in each of V_T cell. Collectors of BJTs Q₁, Q₂, Q₃, Q₄, Q₅, and Q₆ are connected to V_SS, while their emitters are connected to nodes 5d, 8d, 11d, 14d, 17d, and 20d respectively, which biases the emitter currents of each of the 1×BJT PNP to about ‘m×i’. Amplifiers A_d1, A_d2, A_d3, A_d4, A_d5, and A_d6 have their non-inverting inputs connected to nodes 5d, 8d, 11d, 14d, 17d, and 20d respectively, and have their inverting nodes connected to nodes 6d, 9d, 12d, 15d, 18d, and 21d respectively. The output nodes of amplifiers A_d1, A_d2, A_d3, A_d4, A_d5, and A_d6 are connected to nodes 7d, 10d, 13d, 16d, 19d, and 22d respectively. While the base of Q₁ is connected to ground (GND); bases of Q_d1 and Q₂ are connected to node 7d; bases of Q_d2 and Q₃ are connected to node 10d; bases of Q_d3 and Q₄ are connected to node 13d; bases of Q_d4 and Q₅ are connected to node 16d; bases of Q_d5 and Q₆ are connected to node 19d; and finally bases of Q_d6 and Q₇ are connected to node 22d.

A ‘V_T cell’ is comprised of current source transistors P_d1 and P₁ which bias emitters of Q₁ and Q_d1, respectively, while amplifier A_d1 generates a voltage at node 7d such that its inputs at nodes 5d and 6d are substantially equalized. Because Q_d1 is ‘n’ times larger in emitter size than that of Q₁ and because Q₁ is biased at ‘m’ times higher current than Q_d1, then the difference between V_BEs of Q₁ and Q_d1 is ΔV_BE=V_T ln (n×m). This ΔV_BE is the multiple of thermal voltage, V_T, that ‘V_T cell’ number 1 generates at node 7d. Therefore, V_T cell 1 is made up of transistors P_d1 and P₁ and amplifier A_d1; V_T cell 2 is made up of transistors P_d2 and P₂ and amplifier A_d2; V_T cell 3 is made up of transistors P_d3 and P₃ and amplifier A_d3; V_T cell 4 is made up of transistors P_d4 and P₄ and amplifier A_d4; V_T cell 5 is made up of transistors P_d5 and P₅ and amplifier A_d5; and V_T cell 6 is made up of transistors P_d6 and P₆ and amplifier A_d6.

V_T cells number 1 to 6 each generate a ΔV_BE=V_T ln (n×m), which are configured in a series cascade that yields 6×V_T ln (n×m), which is the PTAT voltage. Note for example that for n=5 and m=11, then 100 mV≈VT×ln(55) could be a typical VT cell output voltage. Moreover, note that number of V_T cells may be in the range of 5 to 8 depending on ‘L’, ‘m’, and ‘n’ scale factors that can be optimized for current consumption and area objectives on the bandgap.

The CTAT signal generation is simple and generated by Q₇. Transistor P_L, biased at current ‘L×i’ has its gate and drain connected to node 4d, which biases equal sized PMOS transistor P₇. This transistor's drain terminal at node 24d supplies the bias current for emitter of Q₇ that has it collector connected to V_SS. BJT Q₇ generates the V_EB which is the CTAT voltage.

Next is a summary of the description of the FIG. 4 embodiment and a discussion of the impact of non-ideal terms such as CMOS amplifier offset voltages. The proposed V_T cell approach depicted in FIG. 4 entails generating a PTAT voltage through 2 parasitic vertical PNP pairs looped by a small CMOS amplifier that operates in subthreshold which constitute ‘V_T Cells’. Cascading multiples (i.e., G) stages of such ‘V_T cells’ generates the V_PTAT (without utilizing resistors), which is added to a V_EB at the last (or first) stage to produce the bandgap voltage, V_BG.

It is commonly known that non-ideal terms of CMOS amplifiers, such as V_OFS, dominate as compared to BJT non-ideal terms such as that of V_EB follower pairs. Hence, V_BG is can be approximated as follows:V_rd≈V_BG≈V_EB+V_OFS-total+G×[V_T×ln(n×m)]V_OFS-total≈(±V_OFSd1±V_OFSd2± . . . ±V_OFSdG).The random statistical contribution of the CMOS V_OFS summed in series along within the V_T cell can be approximated as:V_OFS-total≈[(±V_OFSd1)²+(±V_OFSd2)²+ . . . +(±V_OFSdG)²]^1/2.Given that CMOS amplifiers are geometrically identical and may be situated next to each other on the same die, the statistical variations of V_OFSd1 through V_OFSdG should fall within a similar statistical band, let's say V_OFS. As such the statistical contribution of CMOS amplifier V_OFS to the final bandgap voltage may be approximated as:V_OFS-total≈√(G×V_OFS²)≈V_OFS×√G V_rd≈V_BG≈V_EB+V_OFS×√G+G×[V_T×ln(n×m)].CMOS amplifier V_OFS in subthreshold (which is where all CMOS transistors of FIG. 4 are under) approximates V_T like variations (thus tracking the VT ln [m×n] term by first order approximation), which can be corrected as part of the V_PTAT wafer level trim (trim considerations will be described shortly). For comparison sake, it is of note that in conventional bandgaps (such as the ones in prior art FIG. 5) random statistical contribution of non idealities of CMOS amplifier A_1e, including V_OFSe1, is amplified by G+1 (G=R_fe/R_ie) instead of √G:V_rd≈V_BG≈V_EB+G×(1+R_fe/R_ie)[V_OFSe1+V_T×ln(n×m)]Therefore the embodiment of FIG. 4 (for a gain of, for example, 6), from a die yield perspective, the V_OFS statistical contribution should be 2.4 times (or 6/√6) less than a conventional bandgap of FIG. 5 (for a gain of 6). Note that the benefit of the embodiment in FIG. 4 should extend in lieu of the same mathematics being applicable to other random non-ideal terms of CMOS amplifier such noise and drift. Another operational benefit of FIG. 4 embodiment should be higher temperature spans. CMOS amplifier input noise and offset voltage is generally lower with larger W/L of especially input transistors. However, because the statistical contribution of CMOS amplifier offset voltage in FIG. 4 is attenuated by √G, then CMOS transistors can be designed with smaller W/L and (as importantly) equally sized transistors inside each V_T cell to improve high-temperature performance. Hence, small and (as importantly) equally sized CMOS transistors inside each V_T cell amplifier maximize the bandgap's operating temperature span via balancing smaller (transistor) junction areas but large enough for reasonable matching which could contain the derailing impact of leakage currents particularly at high temperatures. Another benefit of the FIG. 4 embodiment is that BJTs and small CMOS amplifiers operate in subthreshold (low V_EB and V_GS) so as to optimize bandgaps operating headroom.

Here is more specific descriptions pertaining to the advantages of the resistor-less bandgap of FIG. 4. Standard CMOS process offers gate-poly resistors with resistivity typically in the tens of Ω/square (□) (Ω is the symbol for resistivity and □ denotes unit area). For example, 700 mV (of V_PTAT voltage) for every 100 nA of current flow (translate to a 7 MΩ resistor) which is area prohibitive with gate-poly resistivity in the tens of Ω/□. There are second layer poly (i.e. salicided and un-salicided) that offer resistance in the hundreds to thousands Ω/□, but they require extra manufacturing process steps that increase die cost, and also they are less accurate compared to first gate-poly. Also note that current consumption of FIG. 4 circuit embodiment may be a predetermined value for example in the range of 100 nano amperes to 500 nano amperes, depending on considerations such as die size and power consumption objectives amongst others.

Another important benefit of the embodiment of FIG. 4 being resistor-less, is better frequency response stability when compared with conventional CMOS bandgaps or variations such as FIG. 5. Frequency response of the loop containing A_1e in conjunction with P_1e and P_2e can be made stable, but it is inherently challenged because, in the A_1e, P_ie, P_2e signal loop, P_1e and P_2e small signals (from node 6e to nodes 4e and 3e) are both inverted and fed back to the inverting and non-inverting node of A_1e. When bias currents are ultra-low and R_ie extra-large (i.e. by making a low current bandgap by utilizing the extra mask with higher resistivity poly, which is in the thousands of Ω/□), the frequency response stability of a FIG. 5 type configuration becomes even more challenging to mange for stability. The embodiment of FIG. 4 is inherently more stable, even in ultra-low currents, because it uses no resistors and it has no such A_1e, P_1e, P_2e loop.

The embodiment of FIG. 4 can offer quality and performance advantages of BJT based bandgaps, it can also have an approximate die area that is in the ball park of other resistor-less references, fabricated in typical sub-micron CMOS processes, as indicated in the referred literature. As noted earlier, small W/L and ultra-low currents of amplifiers transistors have unfavorable noise impact on the bandgap. However, just one bigger compensation capacitor (C_C, which can be made by a CMOS transistor with its source and drain shorted) may be used in the last stage of the amplifier chain to further low-pass filter out the noise.

Bandgap references are fundamental budding blocks that are used almost in every electronic device. The embodiments disclosed in FIGS. 1-4 of this application may be used in personal computers, cell phones, consumer electronic, audio, automotive, test, imaging, video, defense, and medical electronics. These are some of the systems (but not limited to) that generally use bandgap references. The embodiments disclosed in FIGS. 1-4 may be used as a stand alone function (chip), part of a SOC (system-on-a chip), part of a hybrid (multi-chip), or in an electronic system such as a cell-phone. The embodiments disclosed in FIGS. 1-4 may also be used in applications such as power-on-a-chip (PoC), battery-less & wireless, and Internet-of-things (IoT) which are emerging applications that require very low voltage, very low current, high performance, and low costs (obtainable in standard CMOS fabrication processes).

As described above in the embodiment of FIG. 1 and method thereof, an output voltage (V_ra) and V_BE are used to generate and feedback a current that supplies a needed positive temperature coefficient (TC) current to the bandgap's PTAT current loop. Such positive TC feedback current facilitates generating the residual near zero TC current which is concurrently fed-forward to an output resistor (R_OA) to produce the voltage reference (V_ra).

As described above in the embodiment of FIGS. 2 and 3 and methods thereof, by segregating and scaling currents (by n-times) in the BJTs from resistors in a bandgap voltage reference, area is saved and operating current minimized. Scarce power supply head room in sub-1V applications is freed-up in the PTAT loop from V_EB by VT×ln (n).

As described above in the embodiment of FIG. 4 and method thereof, without use of resistors, bandgap voltage is generated by adding V_EB to a PTAT voltage (G×V_T), where G V_T is produced by cascading pairs of scaled V_EB differential followers (with a built-in offset voltage which is a function of V_T) looped-in with small CMOS amplifiers (that have no built in offset) for ultra low currents that keeps parasitic leakage current contained, optimizes operating supply and temperature spans.

In summary, bandgap families disclosed in this application enable the bandgap's operations with (combination levels of) ultra-low power and low supply voltage and with the capacity to deliver optimal temperature and power supply spans.

Claims

1. A method for generating a scaled bandgap reference voltage (VSR) comprising: providing a positive and negative power supply voltage to a proportional to absolute temperature (PTAT) circuit loop having a first transistor that operates at a first current, a second transistor that operates at a second current and a first resistor (RPTAT) that operates at a third current;operating the proportional to absolute temperature (PTAT) circuit loop so that the first current of the first transistor and the second current of the second transistor are different from the third current operating through the first resistor (RPTAT) by isolating RPTAT from the first and second transistors in the PTAT circuit loop;generating a scaled bandgap reference voltage (VSR) by supplying a first terminal of a second resistor with a fourth current that is scaled with the third current operating through the RPTAT, while supplying a second terminal of the second resistor with a ground signal; andsupplying a complementary to absolute temperature (CTAT) signal to a first terminal of a third resistor while coupling the scaled bandgap reference voltage (VSR) to a second terminal of the third resistor.