Designing a Band-Gap Voltage Reference

EE 420L Engineering Electronics II 


Author: Matthew Meza

Email: mezam11@unlv.nevada.edu

May 8th, 2015  


Project Guidelines and Requirements

Using as many diodes, resistors, and capacitors as needed, along with two CD4007 chips from the same production

lot (see date code on the top of chip) to ensure current mirrors are possible, design and build a bandgap voltage

reference (BGR). Your report, in html, should detail your design considerations, simulation results (using the models

you generated in lab 8), and measured results showing the BGR's performance (how the reference voltage changes

with VDD). It would be good, but it's not required, if you could also characterize the BGR performance with

temperature. Your report is due at the end of lab on Friday, May 8. Access to your CMOSedu.com accounts

will be removed at this time.

Introduction to Voltage References
A voltage refernce is a circuit used to generate a fixed voltage which is often referred to as 'Vref'. The fixed
voltage 'Vref' is independent of the power supply voltage (VDD), temperature, and process variations. More
specifically, a bandgap reference is a type of voltage reference which uses both a proportional to absolute
temperature (PTAT) quality and a complementary to ablsole termparture (CTAT) quality. When a voltage
reference has a PTAT quality it means that the reference voltage increases with increasing temperature; while
a voltage reference that has a CTAT quality means that the refernece voltage decreases with increasing
temperature. Since the bandgap voltage reference incorporates both qualities, the net change in the voltage
reference with increasing temperature is ideally zero! Shown below is figure 23.1 from "CMOS Circuit Design,
Layout, and Simulation", by R. Jacob Baker, which illustrates PTAT and CTAT voltage references.



Design Considerations
For the Band-Gap Reference (BGR) design, a diode-referenced self-biasing circuit  will be used for the CTAT
quality and thermal voltage-referenced self biasing circuit will be used for the PTAT quality. Ideally, a cascode
structure should be used (higher output resistance) to force the same current through each branch of of the
reference as seen below!



Unfortunately, only two CD4007 chips can be used which significantly limits the design since the BGR with the
cascode structure, shown above, cannot be implemented! For this project a regular current mirror must be used to
design the BGR which will have reduced performance as compared to the previously shown design. Shown below is the
BGR using a regular current mirror where K = 8 which is the number of diodes used in parallel. The two left most
branches (current mirror) is referred to as a Thermal Voltage-Referenced Self Biasing circuit which provides the
PTAT quality. Using K diodes in parallel in the middle branch increases the current flowing in the branch such that
the current flowing through the reference is non-zero. The same effect could be realized by increasing the
MOSFETS size on the leftmost branch. The PTAT current generated from the current mirror is then driven into
a resistive load R*L (PTAT quality) and a diode (CTAT quality) to create a voltage reference which ideally does
not change with temperature, VDD variations, or process variations!



The hand calculations to find the appropriate resistance value (R) and length (L) can be seen below:


In our build and simuluations, we used PNP BJT's which were grounded at the base and collector instead of a diode.

Simulations
Below is our circuit in LT-Spice along with simulations!
Click on this picture for the full image!


VREF VS VDD
with variations in temperature

(click on pictures for full size)
VREF VS VDD
ZOOM


IREF VS VDD
with variations in temperature


IREF VS VDD
ZOOM



Note: The simulation models must have a lambda value to be accurate. A lambda value of 0.5 was used in the model.

From our simulation results we see the voltage reference changes approximately 282mV per 1V increase in VDD!
In addition, from simulations we see that the current changes approximately 442nA per 1V increas in VDD!

Experimental

Shown below is a picture of the built BGR circuit. Notice the big blue resistor on the top right; it is used as a
start up circuit by leaking a small amount of current into the NMOS gates so that the BGR circuit is in operation.


After measuring the voltage and current for several VDD, we graphed the data in Excel.
The graph below plots the changes in Vref with varying VDD. From our experimental data, Vref increases
approximately 220mV per 1V in VDD. This is better than our simulation value!


The graph below plots the changes in Iref with varying VDD. From our experimental data, Iref increases
approximately 350nA per 1V in VDD. This is also better than our simulation value!


Conclusion
Although the changes in VRef and Iref are not as small as we would like, we can see first hand that the BGR does
work. Being limited to only 2 CD4007 chips definitly inhibits the possible performance of the BGR. As stated
above, it would be ideal to have a cascoded current mirror for a higher output resistance. Nonetheless, the project
proved to be helpful in seeing how process variations can also change the performance of a BGR. Overall I score
this Lab-Project a 10/10.

Return to EE 420 Labs