The Journey to Low Iq in Bandgap References

This article is in the TechXchange: Why Low Iq is the Smart Thing to Do.

Members can download this article in PDF format.

What you'll learn:

What are the key specs for a bandgap reference?
Details regarding the error amplifier.
The architecture of a conventional bandgap reference.
The ingredients for a successful bandgap-reference solution.

Modern-day CMOS converters typically operate from 3.3- or 5-V power supplies. This will limit the on-chip voltage reference to a bandgap reference.

Reference technology in integrated circuits (ICs) is typically divided among three key product solutions: bandgap, buried Zener, and XFET references. This article will focus on the bandgap reference (BGR).

Bandgap Reference

Bandgap references need a high power-supply rejection ratio (PSRR), low quiescent current (Iq), low temperature coefficient (TC), and high accuracy. These references are composed of two transistors, each with different emitter areas that are used to generate a voltage proportional to absolute temperature (Fig. 1).¹

V_BE1 and V_BE2 will have opposite temperature coefficients. The V_CC is converted to currents I₁ and I₂, which are mirrored to the output branch. The output Equation 1 is:

V_O = V_BE1 + ʎ(V_BE1 - V_BE2) (1)

where λ is the scale factor, V_BE1 is the base-emitter voltage that’s the larger of the two transistors, and V_BE2 is the base-emitter voltage of the second transistor.

Bandgap references, usually employed in systems up to 12 bits, produce a fixed (constant) voltage even in the presence of power-supply variations, noise sources, temperature changes, or circuit loading from other devices in a system. They rely on a ∆V_BE in series with the temperature coefficient of a forward-biased diode junction.

Buried Zener-diode-based references employ a Zener diode and a forward-biased diode that’s connected in series. This combination leads to a near-zero temperature coefficient, long-term stability, and low-noise performance.

Let’s take a look at the error amplifier in a low-dropout (LDO) regulator, which needs a very stable reference voltage for equivalent measurements.

Error Amplifier

An error amplifier for frequency compensation may be just a simple voltage-to-voltage amplification device, such as a traditional op amp. The op amp will require local feedback between its output and inputs to ensure stability.

Error amplifiers can also be voltage-to-current amplification devices, i.e., the gm op-amp (operational transconductance amplifier, OTA). This is an open-loop amplifier stage that has no local feedback.

Researchers have offered some strategies that would improve the power efficiency and transient response performance of output capacitorless LDO regulators. Two main approaches were proposed:

Modifying the error amplifier to improve the slew rate (SR).
Using a capacitor coupling effect for fast transient response performance.

Unfortunately, these methods will necessitate additional circuitry and higher current, leading to a serious degradation in power efficiency. Let’s see how we can further modify and improve the error amplifier.

A high-performance OTA can function as an error amplifier with single-pole characteristics. An OTA has a differential input voltage that will lead to the production of an output current. It’s primarily used as a controlled-gain block. To bring stability to such an op amp, designers need to provide local feedback that’s between the input and the output of the device.

An Ultra-Low-Iq BGR

Extending battery life in handheld devices, implantables, and handheld devices is especially crucial in devices where charging or replacing is quite difficult and in many cases not possible. The silicon area must also be reduced.

On this front, a BGR architecture using a 130-nm TSMC CMOS technology offers a resistor-less topology. Powered via a 1.8-V power supply, it reduces circuit board area and consumes a minimal current consumption of 1.6 nA.

This switched resistor-less bandgap reference uses a single bipolar junction transistor (BJT) with cascaded stages of NMOS cascodes that will generate the necessary reference voltage. Ultra-low-Iq design techniques have been used in this design without sacrificing the TC and PSRR.

A Conventional BGR

The design architecture in a conventional BGR requires a proportional-to-absolute temperature (PTAT) voltage added to a complementary-to-absolute-temperature (CTAT) voltage. The PTAT voltage is generated from the difference between the base-emitter voltages of two BJTs having different emitter areas, making the reference voltage temperature-independent. The CTAT voltage is generated by a BJT base-emitter voltage. The reference voltage is now made independent of temperature:

V_REF= V_be1 + (R₃/R₂)V_T ln(n) (2)

where n is the ratio of the emitter areas of transistors Q₁and Q₂. The correct values of n, along with the ratio between resistor values, will give a zero TC at room temperature. (Caveat: matching constraints and area will deem it difficult to reduce Iq simultaneously with the low TC.)

The design will avoid using BJTs; instead, all-MOSFET BGRs will be utilized. That’s because while the BJTs have a low Iq and small area, they actually generate CTAT voltages using MOSFETs in the sub-threshold region, making them quite susceptible to variations in process.

Ultimately, other variations of bandgap-reference designs were pursued. However, all efforts proved to be lacking the goal for ultra-low Iq with low TC without sacrificing the area, PSRR, and other variations.

Proposed Successful BGR Solution

This solution targeted an ultra-low Iq with a low TC while not sacrificing board area or PSRR (Fig. 2).

2. Shown is the most commonly used bandgap reference solution, which has proven quite successful. (Image courtesy of Reference 8)

The design has a PTAT voltage that’s generated via cascading self-cascode NMOS cells. The CTAT voltage is generated from the bipolar transistor Q₁. And the M_CK1 drives the circuit ON and OFF from a low-frequency clock (clk).

CTAT Voltage Generation

In Figure 2, M₁ and M₂ have different lengths and widths, but they don’t have different threshold voltages when the source and bulk are connected to each other. This results in the Q₁ V_BE not equally divided between them, depending on the factor R. The gate-source voltage of M₁ is given by:

V_GS,M1= RV_{BE, Q1}/2 (3)

PTAT Voltage Generation

In a self-cascode NMOS cell that has transistors operating in the sub-threshold region, the drain-source voltage of the lower transistor in Figure 2 is:

V_DS,bottom = nV_Tln(I_bottomS_top/I_topS_bottom) (4)

where n as the MOSFET sub-threshold slope; V_T is the PTAT thermal voltage; I_top is the drain current of the top transistor (M₃, M₅, M_7,or M₉ shown in Fig. 2); I_bottom is the drain current of the bottom transistor (M₂, M₄, M_6,or M₈ shown in Fig. 2); and S_top and S_bottom are the sixing factors of the top and bottom transistors, respectively.

The Voltage Reference

We generate the voltage reference by adding V_GS,M1 to the drain-source voltage of the four bottom transistors in the self-cascode cells (PTAT). The voltage reference equation is:

V_REF = V_{BE, Q1}/2 + nV_Tln(S₃S₅S₇S₉/ S₂S₄S₆S₈) (5)

The TC gets set to zero at room temperature via the proper sizing of the transistors.

Switching

By adding transistors M_CK1 and M_CK2, the system will periodically turn the BGR ON and OFF. The BGR is able to use a low-frequency clock, which will typically be available in the system. When the clock is high, the BGR turns OFF and no current will be drawn from the power supply; when the clock is low, the reference voltage is generated and stored on the C₁ capacitor.

The reference voltage will still be stored on C₁, which is separate from the BGR to ensure correct operation. If the clock duty cycle is reduced, the quiescent current will be further reduced as well. However, the output ripple must be tolerable and the BGR must have enough time for startup.

Summary

Bandgap references are temperature-independent voltage reference circuits that are widely used in integrated circuits. The BGR will produce a fixed voltage even in the presence of power-supply variations, circuit loading from devices, as well as during temperature changes.

This BGR circuit concept was initially published by David Hilbiber in 1964: “A new semiconductor voltage standard,” IEEE Conference Publication. Hilbiber found that using the avalanche diode as a reference voltage source, was well known in the art. Hilbiber said, “If suitable precautions are taken, it has been found possible to obtain short-term stabilities of 10-50 ppm over moderate ranges of temperature.”

Choosing devices that are able to provide as much as 1,000+ hours of long-term operation can be quite challenging, especially when system functionality needs are complex. Today’s BGRs offer a quantum leap in performance.

Read more articles in the TechXchange: Why Low Iq is the Smart Thing to Do.