Design and Analysis of Bandgap Voltage Reference

Reference sources are widely used in various analog integrated circuits, mixed-signal integrated circuits and system integrated chips. Their accuracy and stability directly determine the accuracy of the entire system. In the design of integrated circuits such as analog/digital converters (ADCs), digital/analog converters (DACs), and dynamic memory (DRAMs), the design of reference sources with low temperature coefficients and high power supply rejection ratios (PSRR) is very critical. In
the early days of integrated circuit technology development, reference sources were mainly implemented using Zener reference sources, as shown in Figure 1(a). It uses the voltage across the Zener diode when it is reversely broken down. Due to packaging reasons such as contamination on the semiconductor surface, the Zener diode has severe noise and is unstable. Later, people moved the Zener junction below the surface to support the buried Zener reference source, which greatly improved the noise and stability, as shown in Figure 1(b). Its disadvantages: First, the normal operating voltage of the Zener diode is 6 to 8 V, which cannot be used in low-voltage circuits; and high-precision Zener diodes have strict process requirements and are relatively expensive.

In 1971, Widlar first proposed the bandgap reference structure. It uses the positive temperature coefficient of VBE and the negative temperature coefficient of △VBE, and the sum of the two can get a zero temperature coefficient. Compared with the Zener reference source, the Widlar type bandgap reference source has a lower output voltage, smaller noise, and better stability. In 1973 and 1974, Kujik and Brokaw respectively proposed improved bandgap reference structures. In the new structure, operational amplifiers are used for voltage clamping, which improves the accuracy of the reference output voltage.
The above classic structures lay the foundation for the bandgap reference theory. This paper introduces the basic principles and basic structures of the bandgap reference source, and designs a bandgap reference source based on the Banba structure. Compared with the Banba structure, a self-starting circuit module and an amplification circuit module are added, so that it can automatically enter the normal working state and increase its stability.

1 Working principle of bandgap reference source
Because the bandgap voltage reference source can achieve high power supply rejection ratio and low temperature coefficient, it is the best reference source circuit among various reference voltage source circuits.
In order to obtain a voltage source that is independent of temperature, the basic idea is to add the difference △VBE between the base-emitter voltage VBE of a bipolar transistor with a negative temperature coefficient and the VBE of a bipolar transistor with a positive temperature coefficient with different weights, so that the temperature coefficient of △VBE just offsets the temperature coefficient of VBE, and obtains a reference voltage that is independent of temperature. Figure 2 shows a basic CMOS bandgap reference source structure circuit.

Among them, Vref is the reference voltage of the output; VBE is the base-emitter voltage of Q1 in Figure 2; the positions of R1 and R2 in the circuit are shown in Figure 2. The working principle of the circuit in Figure 2 is: the operational amplifier, PMOS tubes M1 and M2 form a negative feedback, so that the voltages at the positive and negative input terminals of the operational amplifier are equal. The VBE difference △VBE of the two transistors Q1 and Q2 with an emitter area ratio of n is added to the resistor R1. The input current of the operational amplifier is zero, so the voltage on the resistors R1 and R2 is also proportional to the absolute temperature, which can be used to compensate for the part of the VBE of the Q1 tube that decreases linearly with the absolute temperature. Reasonable selection of the values ​​of R1, R2 and n can obtain an input voltage that is independent of temperature.

The relationship between the output voltage and temperature obtained by the above circuit is generally a parabola that opens upward or downward. It is easy to think that if a certain curve is superimposed, the temperature effect of the output voltage can be further eliminated, making the voltage more stable.
This idea was proposed by B. S. Song and P. R. as early as 1983. Grav proposed it, and then many new reference source circuits were born based on the combination of different curves or the application of different processes. It is also a method with great development potential. Among them, in 2003, Leung used the temperature-related resistance ratio, one using a high-resistance polycrystalline resistor and the other using a diffused resistor. In this way, by adding the voltage drop on these two resistors to VBE, the nonlinearity of the VBE temperature coefficient can be eliminated.

2 A reference source based on Banba structure
2.1 Basic structure
A bandgap reference source circuit designed in this paper is based on the reference source structure published in JSSC in 1999, and a self-starting circuit and an amplifier circuit are added, as shown in Figure 3.

Composition: The first part is the startup circuit, which is mainly determined by the performance of the three tubes MSA, MSB, and MSC to determine the self-start of the circuit; the second part is the amplifier, which uses a two-stage Miller circuit and obtains bias current from the bandgap part; the third part is basically the same as the Banba structure.
The advantages of this structure are reflected in the following aspects:
(1) In the traditional bandgap reference circuit, the output voltage VBE is about 1.25 V, which limits the application of the power supply voltage below 1 V. The Vref of this structure is realized by the voltage drop of the sum of two currents on the resistor: one current is proportional to the VBE of the transistor, and the other is proportional to VT. The generated reference current is mirrored to the output current through the MOS tube M3, and then the output reference voltage is determined by the output load resistor R4, which is convenient for changing the required voltage value.
(2) The use of Miller compensation in the amplifier can increase stability. Hironori Banba et al. used a single-stage operational amplifier with NMOS as the differential output tube. In order to achieve a lower power supply voltage, non-standard depletion-type devices are required, and the conversion to the process is poor. Therefore, PMOS tubes are used as differential inputs in this article. Since the role of the amplifier in the circuit is to ensure that the voltages of 1 and 2 are equal and have no effect on the core part, this structure is an improvement on the Banba structure.
(3) The startup circuit can automatically enter the normal working state when the circuit node is in a degenerate state. In the Banba structure, its self-starting method is to use an additional pulse (Power On -Reset Signal) to achieve this. This is rarely used in analog and mixed circuits, so the startup circuit is added in this article. Although the number of components is increased, it can make the manufacturing and startup process simple and practical.
2.2 Analysis of the self-starting module and the amplifier circuit module
In the bias circuit of the amplifier, if the voltage of node 2 is 0 in the initial state, degeneration occurs and it will not work without external stimulation. This is unacceptable in practical applications, so the degenerate point must be removed. The method is shown in Figure 4. The start-up circuit is formed by 3 MOS tubes. Since the gate of PMOS tube MSA is grounded, MSA is always turned on, which makes the level of point S rise. S is also the gate of MSB tube, so MSB tube is turned on and its drain level is reduced. In this way, if the starting point is the gate of PMOS tube, the PMOS tube is turned on and the circuit can start working. Finally, MSB must be disconnected. When the circuit starts to work normally, MSC tube is turned on, which makes the level of node 5 drop again, and MSB tube is turned off and disconnected from the startup part.

The main function of the amplifier in the bandgap circuit is to make the levels of the two input points equal, so as long as the gain is sufficient, the phase margin must also be sufficient to prevent oscillation. Other indicators are not particularly important. Figure 5 shows the core part of the amplifier. The functions of each part are: MA1 and MA2 are the first-stage differential amplification, MA6 is the second-stage amplification, and MA5 and MA7 distribute the bias current from the bandgap part to the MOS tube of the amplification part. Cc is the Miller capacitor, which separates the primary and secondary poles and can also increase the phase margin.

2.3 Spectre simulation results and analysis
Figure 6 is a graph showing the relationship between the output reference voltage and temperature of a Bandgap based on the Banba structure shown in Figure 3 using Cadence's simulation software Spectre in TSMC's 0.18μm process. It can be seen that the results are: within -50 to 100℃, the corresponding two points of the reference voltage with the largest difference change to 96.71℃ and 901.176μV, and the corresponding temperature coefficient is From

a practical point of view, that is to say, within the temperature range of 70℃, this circuit has an accuracy of 2-11. However, this is the result of simulation in TT mode without considering the layout, parasitic resistance and capacitance, etc., and the actual situation may be somewhat biased.

3 Conclusion
Design and application of reference source The reference voltage source is the basic module of analog integrated circuits. It provides high-precision voltage reference for other functional modules in the circuit system, or converts it into high-precision current reference. A qualified reference voltage source is insensitive to power supply voltage, operating temperature, output load changes, and manufacturing process. It can provide accurate reference points for other circuit modules and is an extremely important component of contemporary analog integrated circuits. It provides reference voltage for series voltage regulator circuits, A/D and D/A converters, and is also the voltage regulator power supply or excitation source for most sensors.