BGR Project
BGR Project
BGR Project
The goal of the project has been to understand the functionality of Bandgap reference
circuit(s) and to verify the performance using HSPICE.
Motivation:
Bandgap reference circuits operate on a very simple principle, yet has a lot of
applications.
Widely used in ADC, DAC, and voltage regulators.
It is an essential component in data acquisitions. The conversion accuracy is
dependent on the stability of the reference voltage.
Personal interest to learn about Bandgap reference circuits
2
Table-1 shows the resistor and p-channel transistor values. Table-2 compares the
deviations in the topologies with different parameters. Table-3 compares the performance
of the topologies when resistors were halved, while keeping ratio constant.
Table 3 compares the performance of the topologies when resistors were halved, while
keeping ratio constant
* The upper limit value was calculated at 130C, but the values are reasonably constat
from 125C to 130C.
** The opamp has an offset voltage of 269uV.
When the resistance values were reduced to half, there was no change in the power of the
topologies (with the resistances I used) except for topology 2. There was a power
(average) difference of about 5.325W (figure 30, page 32).
3
Results are shown in more detail after each plot, pages 22-33.
The Key Performance parameters of the opamp used for the project were the following:
offset voltage: 269.2V
The phase margin: 52.
Unity Gain bandwidth: 11.8MHz
Gain: 10K
Poles are located at 1630Hz and 25.1MHz
The open loop output impedance is 461K
4
Section 1: Introduction to voltage reference
Reference voltage generators are used in DRAMs, flash memories, and analog devices.
The generators are required to be stabilized over process, voltage, temperature variations,
and also to be implemented without modification of the fabrication process.
An important part in the design of analog integrated circuits is to create reference voltages
and currents with well defined values. To accomplish this on-chip, Bandgap reference
circuits are commonly used. These circuits allow the design of temperature independent
reference voltages. A typical application for this reference voltage is in analog to digital
conversion, where the input voltage is compared to several reference levels in order to
determine the corresponding digital value.
(a) Zener diode: Making use of a Zener diode that breaks down at a specific reverse
junction voltage. The following figure shows the configuration using a Zener diode.
5
(b)Making use of the threshold voltage of an MOS device. The implementation is as
shown in the following figure:
h
Where k Boltzmann constant 1.38065x10-23 J/K
q- Electric charge 1.6*10-19 As
Vt threshold voltage
VFB-Flatband voltage
Fn-Fermi potential
Cox ox/tox where tox -thickness of the oxide layer
ox- permittivity of the oxide layer
s- Relative permittivity
NA- p-substrate doping concentration
mn* mp* - effective mass of electron and hole
h planks constant 6.63x10-34 Js
ni intrinsic carrier concentration of silicon (1.5*1010 cm-3)
6
(c) Making use of the threshold voltage between enhancement transistor and a depletion
transistor: This technique cannot be used in CMOS technology because depletion
transistors are not available. Instead, threshold voltage of a p and n-channel devices can
be used. The following figure shows the concept of this technique [9].
(d)The Bandgap reference (BGR) is one of the most popular reference voltage generators
that generates a temperature independent voltage. This method involves the generation of
a voltage with a positive temperature coefficient. The base-emitter voltage, VBE, has a
negative temperature coefficient. Therefore, when the two voltages are added together,
the sum has a zero temperature coefficient. For silicon, this is achieved when the total
voltage equals roughly 1.22V. This value is the Bandgap voltage of silicon. Hence, this
method is called Bandgap reference.
7
Section 2: Theory Bandgap Reference Circuit
In a conventional Band Gap Circuit, the output voltage is the sum of the built-in voltage
of the forward biased (Base-emitter) diode which has a negative temperature coefficient
and a voltage proportional to the absolute temperature (PTAT), which is the thermal
voltage multiplied by a constant.
8
I c = GT
VGo
4 n VTH
I s = BT e
Where VGO - Bandgap voltage extrapolated to 0oK 1.2V
n - Depends on the doping level in Base 3/2
is typically in-between 0 and 1
VTH Thermal voltage = kT/q 26mV at room temperature
G and B are proportionality constants
GT
Vbe VTH ln V
VTH
Go
BT e
4 n
G ( 4n ) VVGo
Vbe VTH ln T e TH
B
G
Vbe = VGo VTH (4 n ) ln T ln (eqn 2)
B
Output voltage after summer, as shown in figure (4), is given by the following equation:
Vref = Vbe + VTH K (eqn 3)
To determine the temperature dependence of the output voltage Vref, [from notes (page VI-3)]
Consider equation (3): Vref = Vbe + VTH K . Since the base emitter junction is forward
biased, it is the same as a forward biased p-n junction, hence Vbe and Vd (diode voltage)
can be used interchangeably.
dVref dVbe V dV V
= + K TH = d + K TH (eqn 6 )
dT dT dT dT dT
9
In order to obtain the temperature coefficient of the output voltage Vref, the temperature
coefficients of Vd and VTH should be determined.
(i)To obtain dVd/dT, consider the following. From the diode equation (1b) (page 8), we
can express Vd as:
kT I d I
Vd = ln = VTH ln d (eqn 7 )
q Is Is
By taking the partial derivative of Vd, (equation 7) with respect to temperature, we get:
Vd VTH I d I I I I
= ln + VTH s d d2 s (eqn 8a )
T T Is I d I s T I s T
Vd Vd VTH T I d 7 EG
= + (eqn 11)
T T T I d T 2 kT
T
If I d = I d 0 and T = 300 K , using Vd 0.7V (diode voltage)
T0
Vd Vd VTH 5 E G
= + 1.8mV / o k (eqn 12 )
T T T 2 kT
(ii)Similarly, the temperature coefficient of thermal voltage at room temperature is given
by:
VTH k
= 0.085mV / o k (eqn 13)
T q
Substituting the results equation (12) and (13) in equation (6), we get:
dVref dVd V
= + K TH = 1.8mV / o k + K * 0.085mV / o k (eqn14)
dT dT dT
10
From the equation (14) it can be seen that in order to obtain a temperature independent
voltage K has to be chosen such that the dVref/dT = 0.
By equating equation (14) to zero and by solving, we get K 21.17
The Bandgap reference circuit originally proposed uses bipolar transistors. A basic
CMOS compatible bipolar implementation of a Bandgap reference circuit is shown in the
following figure. Here the bipolar devices are just being used a p-n junctions.
Most of the state of the art technology uses CMOS technology; hence it is essential to
implement Bandgap reference circuits compatible to CMOS technology. Figure (6) shows
the diode implementation of the bipolar transistors in CMOS technology.
11
The resulting CMOS equivalent block diagram of figure (5) is shown in the following
figure (7).
The topologies I implemented for this project were Voltage-mode Bandgap reference
circuits. A topology depicting the current-mode is shown in the following figure:
12
Section 3: The topologies implemented and verified in HSPICE
The following three voltage-mode Bandgap reference circuit topologies were
implemented and verified in HSPICE:
Bandgap topology 1a: This topology is the CMOS equivalent of the figure (5).
Figure 9Topology 1a
The output voltage of this topology is given by: Vref = Vd + I1 R1 (eqn 15 )
If R1 is chosen to be equal to R2 (R1 = R2), the currents I1 and I2 are equal (I1 = I2). The
current can be expressed as the following:
(V V3 ) (V2 Vdn )
R1 = R2 I 1 = I 2 = 2 = (eqn 16a )
R3 R3
Considering the opamp to be ideal, the voltage at the inputs of the opamp should be
equal, therefore equation (16a) can be written as: I 1 = I 2 = (V1 Vdn ) = (Vd Vdn ) (eqn 16b )
R3 R3
V1 = V2 Vd = Vdn + I 2 R3 (eqn 17 )
Using the diode equations (1a) on page 8, the currents I1 and I2 can be written as:
qVd
kT I 1
I1 = I s e KT
Vd = ln
q Is
qVdn
kT I
I 2 = nI s e KT
Vdn =
ln 2
q nI s
Subtracting Vdn from Vd to compute current I1/ I2, we get
kT I 1 nI s kT
Vd Vdn = ln = ln (n ) (eqn 18)
q I s I 2 q
By substituting equation (18) in equations (15) and (16b), we get:
(V Vdn ) kT R
Vref = Vd + I 1 R1 = Vd + d R1 = Vd + ln (n ) 1 (eqn19 )
R3 q R3
To check the temperature dependence of the output voltage Vref (equation 19), let us
consider the derivative of Vref with respect to temperature.
13
By taking the derivative of equation (19) with respect to temperature, we get:
dVref dV k R
= d + ln (n ) 1 (eqn 20 )
dT dT q R3
To obtain a temperature independent output voltage Vref, we need to choose n, R1,R2,R3
such that the above equation (20) is zero.
Bandgap topology 1b
To obtain similar results as Topology 1a, I selected the following values for this topology:
R0 = 150K R1 = 100K, R2, = 100K (=> R1eff = R2eff =2*150 + 100 K =
400 K), and R3 = 54K, n=16 (consider this as configuration i)
14
The same performance is obtained by choosing: R0 = 100K R1 = 200K, R2, = 200K
(=>R1eff = R2eff =2*100+200 K = 400 K) (consider this as configuration ii)
By comparing the total resistance of the different configurations (i and ii listed above) and
topology 1a, it can be seen (from the following table) configuration i is the most area
effective solution
Resistance as sum Total resistance
configuration i (150+2*100+54) K 404K
configuration ii (100+2*200+54) K 554K
Topology 1a (100+2*200+54) K 854 K.
Apart from this difference, the principle is still the same, and all equations derived for the
previous topology (1a) still apply to this topology.
Therefore, the output voltage for this topology is given by:
kT R1eff
Vref = Vd + ln (n ) (eqn 21) When (R1 = R2)
q R3
Bandgap topology 2
Figure 11Topology 2
This topology is similar to the previous topologies 1(a) and 1(b), except the output of the
opamp is not feedback.
Advantage: The opamp has to drive only the p-channel device. In the previous topologies
1(a) and 1(b), the opamp should drive the output load which could be a large. Also, this
topology has better driving capability as the output resistance is higher than previous
topologies 1(a) and 1(b).
Apart from this difference, the principle is still the same, and all equations derived for the
previous topology (1a) still apply to this topology.
Therefore the output voltage for this topology is given by:
kT R kT R
Vref = Vd + ln(n ) 2 = Vd + ln(n ) 1 (eqn 22) When (R1 = R2)
q R3 q R3
15
Section 4: Opamp section:
I chose a simple two stage opamp with a p-channel input differential stage. The following
steps were followed:
Step1: Select a topology
Step2: Select n-channel or p-channel input differential stage
Step3: Select W/L(s) and then obtain W/L(s) of the devices
Step4: Implementation of the bias circuitry
Step1 Selecting a topology: I used a two stage opamp because of the robustness.
Consider the p-channel input differential stage shown in figure (13). The input voltage at
m1 is about 0.7V (diode voltage), when m5 is in saturation the Vds for m5 0.2V*.
Therefore, the voltage at the node Vd is about 0.2 to 0.3V below Vdd (3V). Therefore, the
Vgs of m1 is about -2.7/-2.8V, hence there is no problem to turn m1 ON, unlike in the n-
channel input differential stage, as Vtp for a p-channel is about -0.9V
16
Figure 13 opamp with p-channel differential input stage
17
Step4 (a): Initially I used an ideal current source of 20A and a p-channel device mb1
(W/L = 100/5) as shown in figure (14) as the bias circuitry. And the configuration was
simulated in HSPICE. The phase margin of this circuit was not good, and the poles were
located at 4.3 kHz and 3 MHz. Hence, a compensation circuitry was necessary.
18
The following equations were used to obtain the value of the Resistor Rb.
1 W
I mb1 = cox (Vb1 Vb 2 | Vtp |) (eqn 23a )
2
2 L
Where Vb 2 = Vdd I bias Rb (eqn 23b )
1 W
I mb 2 = cox (Vb1 Vdd | Vtp |) (eqn 23c )
2
2 L
By equating Imb2 (equation 23c) and Imb1 (equation 23a), Rb can be calculated. The
following equation (24) shows the relation.
1 W 1 W
cox (Vb1 Vdd | Vtp |)2 = cox (Vb1 (Vdd I bias Rb ) | Vtp |)2 (eqn 24)
2 L 2 L
By substituting Ibias = 20A in the above equation, Rb can be calculated. This value was
then used in the HSPICE simulation to verify. It was found that when Rb was equal to
680, the bias current was close to 20A, and the biasing voltage was close to the
simulation value obtained when an ideal current source was used.
19
The most important parameter in this application is the offset voltage. Since this is used
to create a constant DC voltage, Slew Rate, Gain, and other parameters are not very
crucial.
Opamp simulation results:
The following plot shows the result of the opamp without phase compensation. It can be
seen that the poles are located at 4831Hz and 3.059MHz.
Unity Gain bandwidth: 25.12MHz
Gain: 10.6K
The phase margin was very low.
20
The following plot shows the result of the opamp with phase compensation.
Now the poles are located at 1630Hz and 25.1MHz (pole splitting phenomenon)
The phase margin: 52.
Unity Gain bandwidth: 11.8MHz
Gain: 10.6K
The pole splitting phenomenon can be clearly seen from the two plots. In the previous
case the poles were located at 4831Hz and 3.059MHz. In the compensated case they are
located at 1630Hz and 25.1MHz.
At the same time, the UGB is less, as this decreased from 25.12MHz to 11.8MHz.
21
Section 5: HSPICE simulation results of the three topologies
Power (with real opamp), simulated with different offset voltage remains the same.
22
Figure 21HSPICE simulation of Bandgap reference topology 1a, variation with supply
voltage
Topology1a At At Deviation Deviation due to
2.7V 3.4V due to Vdd offset
At 2.7V At 3.4V
Ideal opamp no offset 1.1882 1.1882 0
Ideal opamp with offset = 269uV 1.1907 1.1907 0 2.5mV 2.5mV
Real opamp with offset = 269uV 1.1901 1.1903 0.2mV
Real opamp with offset = 538uV 1.1926 1.1928 0.2mV 2.5mV 2.5mV
Power, when simulated with different offset voltage (with real opamp) remains the same.
23
Figure 22 HSPICE simulation of Bandgap reference topology 1a, variation with change
in resistors, but keeping the ratio constant
Power, when simulated with different resistors and at different offset voltage remains the
same.
24
Figure 23HSPICE simulation of Bandgap reference topology 1a, variation with change
in diode parameters Is 10f - 20f
The deviation with respect to change in diode parameter, at room temperature: 1.1929-
1.1719 V = 21mV.
25
Figure 24HSPICE simulation of Bandgap reference topology 1b, variation with
temperature
Topology1b At At Deviation Deviation due to
-25C 125C due to offset
temperature At -25C At 125C
Ideal opamp no offset 1.1818 1.1959 14.1mV
Ideal opamp with offset = 269uV 1.185 1.1991 14.1mV 3.2mV 3.2mV
Real opamp with offset = 269uV 1.1849 1.1973 12.4mV
Real opamp with offset = 538uV 1.1881 1.2005 12.4mV 3.2mV 3.2mV
Average power found to be 199.21W
Power (with real opamp), simulated with different offset voltage remains the same.
26
Figure 25HSPICE simulation of Bandgap reference topology 1b, variation with supply
voltage
Topology1b At At Deviation Deviation due to
2.7V 3.4V due to Vdd offset
At 2.7V At 3.4V
Ideal opamp no offset 1.1879 1.1879 0
Ideal opamp with offset = 269uV 1.1911 1.1911 0 3.2mV 3.2mV
Real opamp with offset = 269uV 1.1904 1.1906 0.2mV
Real opamp with offset = 538uV 1.1936 1.1939 0.3mV 3.2mV 3.3mV
Power, when simulated with different offset voltage (with real opamp) remains the same.
27
Figure 26HSPICE simulation of Bandgap reference topology 1b, variation with change
in resistors, but keeping the ratio constant
Power, when simulated with different resistors and at different offset voltage remains the
same, 199.21W.
28
Figure 27HSPICE simulation of Bandgap reference topology 1b, variation with change
in diode parameters Is 10f - 20f
The deviation with respect to change in diode parameter, at room temperature: 1.1907-
1.1697 V = 21mV.
29
Figure 28HSPICE simulation of Bandgap reference topology2, variation with
temperature
Topology2 At At Deviation Deviation due to
-25C 125C due to offset
temperature At -25C At 125C
Ideal opamp no offset 1.1836 1.1977 14.1mV
Ideal opamp with offset = 269uV 1.186 1.2001 14.1mV 2.4mV 2.4mV
Real opamp with offset = 269uV 1.1793 1.2 20.7mV
Real opamp with offset = 538uV 1.1817 1.2025 20.8mV 2.4mV 2.5mV
Average power found to be 16.821W
Power (with real opamp), simulated with different offset voltage remains the same.
30
Figure 29HSPICE simulation of Bandgap reference topology 2, variation with supply
voltage
Topology2 At At Deviation Deviation due to
2.7V 3.4V due to Vdd offset
At 2.7V At 3.4V
Ideal opamp no offset 1.1908 1.1914 0.6mV
Ideal opamp with offset = 269uV 1.1933 1.1939 0.6mV 2.5mV 2.5mV
Real opamp with offset = 269uV 1.1925 1.1928 0.3mV
Real opamp with offset = 538uV 1.195 1.1953 0.3mV 2.5mV 2.5mV
Power, when simulated with different offset voltage (with real opamp) remains the same.
31
Figure 30HSPICE simulation of Bandgap reference topology 2, variation with change in
resistors, but keeping the ratio constant
When the resistance values were reduced to half, the change in the average power =
22.146-16.821 W = 5.325W.
32
Figure 31HSPICE simulation of Bandgap reference topology 2, variation with change in
diode parameters Is 10f - 20f
The deviation with respect to change in diode parameter, at room temperature: 1.1899-
1.1695 V = 20.4mV.
33
Summary: The summary of the results (for a real opamp case) are shown in the following
three tables. Table1 shows the resistor and p-channel transistor values.
Table 1
Values (resistor/p-channel) used
Topology M1 R0 R1 R2 R3
(W/L)
1(a) 400K 400K 54K
1(b) 150K 100K 100K 54K
2 60/6 400K 400K 54K
Table2 compares the deviations of the topologies with different parameters.
Table 2:
Deviation with respect to
Topology Temperature Power supply Offset Diode parameter
-25oC to 2.7V to +269.2V Is from 10f-20f
o
125 C 3.4V (at room temp)
1(a) 12.8mV 0.2mV 2.4mV 21mV
1(b) 12.4mV 0.25mV 3.2mV 21mV
2 20.8mV 0.3mV 2.5mV 20.4mV
Table3 compares the performance of the topologies when resistors were halved keeping
ratio constant.
Table 3
Topology 1a Topology 1b Topology 2
Deviation due to Deviation due to Deviation due to
Offset Resisto tempe resistors at - Average tempe resistors at - Average tempe resistors Average
voltage rs used rature 25 and Power rature 25 and Power rature- at -25 and Power
** -25 to 125C (mV) over -25 to 130C*(mV) over 25 to 130C*(mV) over
125C temp 125C temp 130C temp
( mV) (W) (mV) (W) *(mV) (W)
269uV Actual 12.8 199.21 12.4 199.21 20.8 16.8
version
Halved 23.6 16.7 27.5 199.21 21.9 16.5 25.9 199.21 32.8 16.5 28.5 22.1
version
538uV Actual 12.8 12.4 20.9
version
Halved 23.6 16.7 27.5 21.9 16.5 25.9 32.9 16.5 28.5
version
* The upper limit value was calculated at 130C, but the values are reasonably constat
from 125C to 130C.
** The opamp has an offset voltage of 269uV.
Results are shown in more detail after each plot on pages 22-33.
34
The Key Performance parameters of the opamp used for the project were the following:
offset voltage: 269.2V
The phase margin: 52.
Unity Gain bandwidth: 11.8MHz
Gain: 10K
Poles are located at 1630Hz and 25.1MHz
The open loop output impedance is 461K
Conclusions: The three topologies function as predicted by theory. The results match with
the predictions: when resistances were reduced to half and also when the diode model
parameter current Is was changed from 10f to 20f.
Observations:
1. The simulations show that the results are process dependent; these results are
valid for 0.5um process.
2. To achieve more accurate results, we need precise diode models.
3. The power of the opamp dominates the power in all the topologies. If a less power
Bandgap circuit is needed, the current in the opamp has to be reduced.
4. When the resistance values were reduced to half, there was no change in the
power of the topologies (with the resistances I used), except for topology 2. There
was a power difference of about 5.325W.
5. HSPICE simulations reveal that topology 2 has less the output impedance than
topologies 1a and 1b. Therefore it can be concluded that topology 2 has a better
driving capability.
6. The Average power of topology 2 was found to be 16W, while topologies 1a and
1b were 199.21W.
Acknowledgements:
I would like to take this opportunity to thank Dr Fischer and others who have assisted me
in completing this project.
References
1. Discussions with Dr Fischer
2. ELE539 notes
3. ELE432 notes (Spring 2008- Dr Fischer)
4. Analog Integrated circuit design David Johns, Ken Martin
5. A CMOS Bandgap Reference Circuit with Sub-1-V Operation Hironori Banba,
Hitoshi Shiga, Akira Umezawa, Takeshi Miyaba, Toru Tanzawa, Shigeru Atsumi,
and Koji Sakui,
6. Band-Gap References for near 1-V operation in standard CMOS technology
Andrea Pierazzi, Andrea Boni, Carlo Morandi, Universtiy of Parma, italy.
7. Low-Voltage Bandgap Reference Design Utilizing Schottky Diodes David L.
Butler and R. Jacob Baker
8. DESIGN AND IMPLEMENTATION OF BANDGAP REFERENCE CIRCUITS
Master thesis performed in Electronics Systems by Ramanarayana Reddy
Sanikommu
35
9. A Robust Smart Power Bandgap Reference Circuit for Use in an Automotive
Environment Wolfgang Horn, Heinz Zitta. Infineon Technologies
10. A novel supply-independent biasing scheme for use in CMOS voltage reference
Luis Toledo, Carlos Dualibe, Pablo Petrashin, and Walter LancioniLaboratorio de
Microelectrnica, Facultad de Ingeniera, Universidad Catlica de Crdoba,
Argentina
Future work:
1. Implementing a current-mode Bandgap reference circuit. Following figure (32)
shows a current-mode Bandgap reference circuit. The concept of a Current-mode
Bandgap reference is to generate two currents, proportional to Vf and VTH. The
reference voltage is the drop across the resistor R4.
V f dV f
Vref = R4 +
R
2 R3
36
Figure 33 plots showing the performance parameters of the opamp
37