Farayola Praise Ololade Pallavi-Sugantha Ebenezer

24
Farayola Praise Ololade Pallavi-Sugantha Ebenezer Buck Converter with PWM Voltage – Mode – Control May, 2019

Transcript of Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Page 1: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Farayola Praise Ololade

Pallavi-Sugantha Ebenezer

Buck Converter with PWM Voltage – Mode – Control

May, 2019

Page 2: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Table of Content

Table of content for the project is listed below to make transition between the design blocks

and simulation results easier.

Content Page Number

Introduction -03

Basic Calculations -03

Buck Converter -05

Buffer Design -06

Power Transistor Design -07

Power Inductor and Output Capacitors -10

Comparator Design -10

Ramp and Clock Generator Design -12

Operational Amplifier Design -13

Compensation Network Design -14

Simulations and Measurements -17

Calculations and Efficiency Measurements -23

Conclusion -24

Page 3: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

INTRODUCTION:

Having spent most of the semester learning about Buck Converters, it is imperative that we

build one ourselves with real components; hence this final project.

Over the semester, chunks and part of the Buck Converter has already been built. This final

project adds a few blocks to it and brings all the individual part all working together.

In Homework 4, we designed a closed loop Type-III compensated buck converter operating in

CCM and verified the design using the AC model and transient model. In this final project, we

would be replacing all the ideal components used in the design with real transistor level

circuits.

Specification for the Buck Converter to be designed are set and the final design is adjusted to

ensure that the specifications are met.

The specifications for the project are listed below;

Specifications

Technology UMC 0.18πœ‡π‘š (Preferred) 𝑉𝑔 3.3V ( I/O _33_MM devices for UMC 0.19πœ‡m )

Inductor ESR 20mΞ©

Capacitor ESR 10mΞ©

Vo ( Steady – State ) ~1.8V Β± 5% ( Load Regulation )

π‘‰π‘œπ‘Ÿπ‘–π‘π‘π‘™π‘’ (Steady – State ) <10mV

𝑓𝑠𝑀 1MHz

πΌπ‘œ 300mA - 800mA

Vo under / over- shoot <100mV during transient with 10ns step between 300mA and 800mA

Operation CCM only

Efficiency ( Include Everything ) >85%, Higher is better

The project is designed in the UMC 0.18πœ‡m process and ALL the specifications were met.

All blocks designed are real but we were allowed to use current bias and voltage references.

(Idc and Vdc) Also the inductors, capacitors and resistors used are ideal. However for the

inductors and capacitors in the power stage to be properly modelled, we put a reasonable ESR

(20mΞ© for the inductor and 10mΞ© for the capacitor)

BASIC CALCULATIONS:

Various calculations are done to pick the value of the components used. They are shown below;

1) Picking the Value of L and C

Page 4: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

βˆ†π‘‰π‘œ

π‘‰π‘œ=

1 βˆ’ 𝐷

8πΏπΆβˆ—

1

𝑓2

Where D is the Duty Cycle

𝐷 =π‘‰π‘œ

𝑉𝑔=

1.8

3.3= 𝟎. πŸ“πŸ’πŸ“πŸ’

𝐷′ = 1 βˆ’ 0.5454 = 𝟎. πŸ’πŸ“πŸ’πŸ“

To pick the value of inductor L so that the inductor ripple current βˆ†IL β‰ˆ 300mA

𝐷 = ( π‘‰π‘œ

𝑉𝑔 )

𝐷 = ( 1.8

3.3 ) = 0.5454

βˆ† 𝐼𝑙 =π‘‰π‘œ

𝐿𝐷′𝑇 =

π‘‰π‘œ

𝐿𝑓𝑠𝑀

(1 βˆ’ 𝐷)

𝐿 =π‘‰π‘œ(1 βˆ’ 𝐷)

βˆ†πΌπ‘™π‘“π‘ π‘€

𝐿 =1.8(1 βˆ’ 0.5414)

300π‘š βˆ— 1𝑀

𝑳 = 𝟐. πŸ•πŸ“ 𝝁𝑯

To pick the output capacitance C so that the output voltage ripple is about 5mV

βˆ†π‘‰π‘œ =(1 βˆ’ 𝐷)π‘‰π‘œ

8𝐿𝐢𝑓2

𝐢 =(1 βˆ’ 𝐷)π‘‰π‘œ

βˆ†π‘‰π‘œ8𝐿𝐢𝑓2

𝐢 =( 1 βˆ’ 0.5454 )1.8

5π‘š βˆ— 8 βˆ— 2.75πœ‡ βˆ— 1𝑀2

π‘ͺ = πŸ•. πŸ’πŸ’ 𝝁𝑭 These values could be adjusted later to reduce voltage or current ripple.

After the design was completed and optimized to meet the specification, the value of C and L

was adjusted. The final value used for L and C are listed below

Inductor (L) 1.365𝝁𝑯

Capacitor (C) 7.44𝝁𝑭

Page 5: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

BUCK CONVERTER:

The Buck converter is one of the major switching converters. It employs switching action on one

or more inductors to regulate the output voltage to a predefined value even when the input

voltage ( 𝑉𝑔 π‘œπ‘Ÿ 𝑉𝐷𝐷) or the load current changes.

They are better than LDO’s in the sense that they have better efficiencies which we would show

in this project.

The Buck Converter is made up of the following design blocks.

(i) Driver Stage ( Buffer Design )

(ii) Power Transistor Design

(iii) Power Inductor and Output Capacitors

(iv) Comparator Design

(v) Ramp and Clock Generator

(vi) Operational Amplifier

(vii) Compensation Network

(viii) Digital Logics

Figure 1 Block Diagram of the Buck Converter

Page 6: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

DRIVER STAGE AND POWER TRANSISTOR DESIGN

The driver stage has the buffer and the non-overlapping clock generation circuit design.

BUFFER DESIGN

The design of the buffer used is from project 1. Since the power transistors are of large

dimensions, the PWM control signal cannot drive them effectively. Hence the buffer stage is

designed with gradually increasing sizes by each stage to drive the power transistors, starting

with the minimum sized inverters and the largest sized buffer drives the power stage.

Figure 2 Buffer circuit

NONOVERLAPPING CLOCK DESIGN

A feed-back break before make structure is used to design the nonoverlapping clock

generation circuit. This circuit helps turn off the NMOS before PMOS turns on and vice versa.

The resistor helps the NMOS turns on slower and PMOS turns off faster during low to high

transition. Similarly, for high to low transition, the PMOS turns on slower and NMOS turns off

faster. Hence this logic turns off the PMOS completely before the NMOS is turned on. Similarly,

the NMOS is turned off completely before the PMOS is on. The design of these transistors are

also gradually increased as to drive the power stage. This significantly reduces shoot-through

loss.

Page 7: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

POWER TRANSISTOR

The sizing of the power transistor is very important. Here they are represented with β€œMF” and

sweep is done to give us the best sizing that will guarantee the maximum efficiency. This is

because there is a trade-off between Conduction Loss and Switching Loss.

Figure 5 Power Transistors

To NMOS

To PMOS

Figure 4 Break before Make Circuit

Page 8: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

COMBINATION

I/O_33_MM transistors were used as speculated in the instructions and the variable (MF) was

used in the sizing of the power transistors to make it easy to change and sweep.

Other variables were also used like Ratio1, Ratio2, Ratio3, Ratio4, Ratio5 and Ratio6 to break

down the large power transistor sizing for other transistors used in the buffers.

The buffers were designed using a combination of the two methods to minimize shoot through

loss.

Figure 6 Buffer and Power Transistor Design

OPTIMIZATION OF THE POWER TRANSISTOR SIZING

After the entire design, we swept MF (the Sizing of the Power Transistor and ratioed sizing of the

buffers) from 10k to 100K in 10 steps to pick the best sizing for the buck converter.

Buffer Design Power Transistors

Page 9: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Figure 7 Parametric Sweep for MF

Here is the results that we got after sweeping MF

It is observed that at around 10K to 30k, the PVDD power is decreasing and then later starts increasing

back again after 55K. Hence our choice of picking 50k. The PVDD power is observed to be lowest at that

point and the output power is about the same for all the ranges.

Figure 8 Optimization of the Power Transistor Sizing

Hence this is the final result used.

Figure 9 Final values used for the Buffer and Power Transistor used

Page 10: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

POWER INDUCTOR AND OUTPUT CAPACITORS

To ensure that our simulation could be as close to the real life situation as possible, we

modelled the inductor having an ESR of 20mΞ© and the capacitor having an ESR of 10mΞ©. This

can be seen clearly from the diagram below as the resistors connected in series to the capacitor

and the inductor.

COMPARATOR DESIGN

The comparator is used to generate the required PWM signal by adjusting D. It compares the

Vramp and Va to generate a signal Vcmp. Vcmp is high if Va is greater in amplitude than Vramp and vice

versa.

Figure 11 Comparator Block

The comparator is designed as shown below; Care was taken in the sizing of the comparator. We made

sure that the ratio of (M25/M24) / (M22/M23) matched with M35/M37 as corrected in Homework3

Figure 10 Power Stage

Page 11: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Figure 12 Comparator Design

The comparator used in this design is a cross coupled NMOS input pair comparator. The

transistors are minimum sized to provide the least propagation delay. The length of the input

pair is made larger to have better matching and thus have reduced offset. The cross coupled

transistors have size smaller than the transistors in the main branch to have hysteresis.

Here are the results of the comparator after test

The propagation delay (td): Td_LH =8.7ns

Td_HL =1.1ns

Hence, Td = Td_LH + Td_HL = 9.8ns (< 10ns)

Figure 13 Transient Simulation of Comparator

Page 12: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

RAMP AND CLOCK GENERATOR DESIGN

The double boundary ramp generator is used in this design as shown in Figure 8. It is based on

the Vm which was used in the AC model simulation. The upper boundary is 1.5 V and the lower

boundary is 0.5V. A current of 12uA is generated from the Vref (=1.2 V) using the voltage to

current converter with a resistor of 100kΞ©. This is mirrored to the ramp and clock generation

circuit.

Since the clock required for the design is 1MHz (dt =1us)

The range of the ramp is between the upper and lower boundary (dv =1V)

The capacitor value chosen:

𝐼 = 𝐢𝑑𝑣

𝑑𝑑

𝐢 = 𝐼 𝑑𝑑

𝑑𝑣 = 12u * 1u / 1

C= 12 p F

The capacitor gets charge until the upper boundary until which the NMOS is off. At this point,

the output of both the upper and lower comparators are high. Since the set and reset signals of

the SR latch are high, the output of the latch holds the previous output and hence the NMOS

remains turned off. When the voltage of the comparator hits the upper boundary limit the

output of the upper Comparator is low. Now the reset signal is low and the output of the latch

is high, which turns on the NMOS. The capacitor discharges through the NMOS and ramp signal

gets pulled to ground. When the signal hits the lower boundary, the output of the lower

comparator becomes low and the latch outputs a low signal and turns off the NMOS. The

capacitor is charged up again, thus generating the ramp signal and clock signals at a frequency

of 1MHz.

Figure 14 Ramp and clock generation circuit

Page 13: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

The transistor level ramp and clock generation circuit was designed and its performance was

verified as shown in Figure 9.

OPERATIONAL AMPLIFIER DESIGN

A single stage transistor level implementation is shown in the figure below. The gain from this

op-amp is measured to be around 52 dB and with a phase margin of 87 deg. This is good

enough.

Figure 16 Single stage Opamp

Figure 15 Ramp and clock signals

Page 14: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

We then went ahead to plot the gain and phase plot of the op amp to show that it is good enough and

can be used in out design.

Figure 17 Stability plot of the op-amp

COMPENSATION NETWORK DESIGN

Type III Compensator is required to have a high system bandwidth and good stability. It helps

provide a fast response with no ringing as the system bandwidth is greater than the LC resonant

frequency

Figure 18 Type III Compensator

Page 15: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

How the resistors and capacitors were picked are referenced in Homework 4 as the exact

capacitors and resistors are used.

C1 500p

C2 42p

C3 2p

R2 100k

R3 1k

We then went ahead to plot the gain and phase margin.

It can be observed that the gain margin of the system is 58db and the phase margin of the

system is 63.8 degrees.

DIGITAL LOGIC

The Digital Logic used is a simple SR latch. A simple transistor level minimum sized SR latch has

been designed to provide the PWM signal for the Power transistors.

The schematic of the SR latch is shown below;

Figure 19 Stability plot of the system

Page 16: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Figure 20 SR Latch Circuit

BUCK CONVERTER CIRCUIT

The Entire Buck Converter circuit is shown again with all the block arranged together.

The complete transistor level buck converter has been designed with real amplifiers, ramp and

clock circuit and other logic circuits as shown in Figures above.

Buffer

Nonoverlapping clock

SR Latch

Compensator

Ramp and clock

Comparator

Power stage

Figure 21 Transistor level Implementation of the Buck Converter Circuit

Page 17: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

SIMULATION AND MEASUREMENT RESULTS

AC AND TRANSIENT MODEL SIMULATION

We are required to show the AC and Transient Simulation result and proof that the

specifications in the tables are met. The transient and AC model simulations are shown when

the load is 300mA and 800mA; since these are the two extremes, it is observed that they both

meet specifications. For the transient simulation has all things real while the AC simulation uses

the transformer model for the power stage to make simulations faster

At 300mA

Figure 22 Transient Simulations at 300mA to 800mA transition

Figure 23 AC simulation with 300mA to 800mA transition

For better comparison, we go forward to compare the two together especially the output

voltage and current.

Page 18: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Figure 24 AC and Transient Simulation of the Output Voltage

1) It can be observed that at steady state the voltage ripple is (1.802-1.791) which is 9mV.

This meets specification of less than 10mV and β‰ˆ 1.8𝑉 Β± 5%

2) For the Transient Simulation, the voltage undershoot is (1.789 – 1.7408 ) which is

48.2mV. The overshoot is (1.8511-1.791) which is 60.1mV. This meets the specification

of the undershoot/overshoot being less than 100mV. It can be seen that the AC

simulation has better results for overshoot and undershoot because of the transformer

model

The Output current of the AC and Transient Model is also plotted together

Figure 25 AC and Transient of Output current

This is amplified and shown below

Page 19: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Hence it proofs that the output current can range from 300mA to 800mA.

The output current ripple is (788.353 – 782.8152) = 5.5mA which is according to specification.

TOP LEVEL / SYSTEM SIMULATIONS

AC FREQUENCY RESPONSE

With the transformer model, we went ahead to plot the AC frequency response to show that

with the compensation network, the design was stable.

Figure 26 AC Frequency Response

Page 20: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

From the AC response, it is observed that the Phase margin is 61.8 degree and the gain margin is 34 db.

This also is to specification of the phase margin above 60 degree and the gain margin above 12dB.

Load transient

With real transistors, the transient results with the inductor current is plotted.

Figure 27 Load Transient

AT STEADY STAGE

At Steady State, the inductor current, output voltage and the switching node voltage is shown

more clearly.

Figure 28 At steady State Showing Inductor Current, Output Voltage and Switching Node

This can be seen more clearly

Page 21: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Figure 29 Zoomed in version of the steady state voltages and current

CALCULATIONS AND EFFICIENCY MEASUREMENT

For each of the simulation, the load was changed to match the required output

current.

LOAD REGULATIONS

We are required to calculate the Load Regulations in V/mA of the Buck Converter at load of

300mA, 500mA and 800mA.

πΏπ‘œπ‘Žπ‘‘ π‘…π‘’π‘”π‘’π‘™π‘Žπ‘–π‘œπ‘› =πΆβ„Žπ‘Žπ‘›π‘”π‘’ 𝑖𝑛 𝑂𝑒𝑑𝑝𝑒𝑑 π‘‰π‘œπ‘™π‘‘π‘Žπ‘”π‘’

πΆβ„Žπ‘Žπ‘›π‘”π‘’ 𝑖𝑛 𝑂𝑒𝑑𝑝𝑒𝑑 πΆπ‘’π‘Ÿπ‘Ÿπ‘’π‘›π‘‘=

βˆ†π‘‰π‘‚

βˆ†πΌπ‘‚

πΏπ‘œπ‘Žπ‘‘ π‘…π‘’π‘”π‘’π‘™π‘Žπ‘‘π‘–π‘œπ‘› =1.800914 βˆ’ 1.800032

800 βˆ’ 300= 𝟏. πŸ•πŸ”πŸ’ βˆ— πŸπŸŽβˆ’πŸ” 𝑽/π’Žπ‘¨

EFFICIENCY

We are required to calculate the Efficiency of the Buck Converter at 300mA, 500mA and

800mA.

𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 =𝑂𝑒𝑑𝑝𝑒𝑑 π‘ƒπ‘œπ‘€π‘’π‘Ÿ

𝐼𝑛𝑝𝑒𝑑 π‘ƒπ‘œπ‘€π‘’π‘Ÿβˆ— 100%

Since we have input power going to the Power Transistor from PVDD and input power going to

the Logic from AVDD, hence the total input power will be an addition of the two.

Page 22: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 =𝑂𝑒𝑑𝑝𝑒𝑑 π‘ƒπ‘œπ‘€π‘’π‘Ÿ

𝑃𝑉𝐷𝐷 π‘ƒπ‘œπ‘€π‘’π‘Ÿ + 𝐴𝑉𝐷𝐷 π‘ƒπ‘œπ‘€π‘’π‘Ÿβˆ— 100%

To get the Power Measurements from the Cadence simulation, we integrated and averaged the

Power signal over 80cycles (From 19us to 99us). The input power signal was gotten from the

multiplication of the transient voltage and current signals from the respective voltage source

while the output power signal was gotten from the multiplication of the voltage and current

signals across the load resistor.

A) At 300mA

Figure 30 Power Calculations at 300mA

Figure 31 Waves at 300mA

𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 =538.9

567.3 + 1.072βˆ— 100% = 94.8% β‰ˆ πŸ—πŸ“%

B) At 500mA

Figure 32 Power Calculations at 500mA

Page 23: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

Figure 33 Waves at 500mA

𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 =895.8

944.7 + 1.07βˆ— 100% = 94.72% β‰ˆ πŸ—πŸ“%

C) At 800mA

Figure 34 Power Calculations at 800mA

Figure 35 Waves at 800mA

Page 24: Farayola Praise Ololade Pallavi-Sugantha Ebenezer

𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 =1.429

1.529 + 1.07π‘šβˆ— 100% = 93.39% β‰ˆ πŸ—πŸ‘%

Contribution of the Project Mates

Since the basic design was common to both of us we went ahead and used Praise’s design. The

new blocks were designed by Praise and verified by Pallavi. The report and presentation were

made by Pallavi and missed details were provided by Praise.

CONCLUSION

We were able to get the efficiency of the Buck Converter at around 93% for 800mA Load and

around 95% for 500mA and 300mA load. This is well beyond the required specification.

We also observed that the Load regulation was highest with a load of 300mA ( πŸ“. πŸ—πŸ—πŸ•πŸ‘ βˆ—

πŸπŸŽβˆ’πŸ‘ 𝑽/π’Žπ‘¨) This reduced down to 𝟐. πŸπŸ”πŸ• βˆ— πŸπŸŽβˆ’πŸ‘ 𝑽/π’Žπ‘¨ at a load of 800mA.

It is also seen that the power that is consumed by the logic remains pretty the same (1.07mW)

but as the current increases then the power consumption of the power transistors increases in

proportion.

We also observed that the efficiency dropped as the Load Current increased. This agrees with

what we have learnt in class.

It was a nice project and we learnt a lot of things. The course in general was an interesting

course and we look forward to applying the lots of things that we have learnt in areas of

research and other aspects of Life.