04.06.2015 Views

Optimization and implementation of a multi-level buck converter for ...

Optimization and implementation of a multi-level buck converter for ...

Optimization and implementation of a multi-level buck converter for ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

International Workshop on Power Supply On Chip<br />

September 22nd - 24th, 2008, Cork, Irel<strong>and</strong><br />

<strong>Optimization</strong> <strong>and</strong> <strong>implementation</strong> <strong>of</strong> a <strong>multi</strong>-<strong>level</strong> <strong>buck</strong> <strong>converter</strong><br />

<strong>for</strong> st<strong>and</strong>ard CMOS on-chip integration<br />

Vahid Yousefzadeh,<br />

Toru Takayama<br />

Dragan Maksimović<br />

Colorado Power Electronics Center<br />

ECE Department, 425 UCB<br />

University <strong>of</strong> Colorado<br />

Boulder, CO 80309-0425<br />

maksimov@colorado.edu<br />

Gerard Villar<br />

Eduard Alarcón<br />

Dept. <strong>of</strong> Electronic Engineering<br />

Technical University <strong>of</strong> Catalunya<br />

Campus Nord UPC – Building C4<br />

08034 Barcelona, Spain<br />

ealarcon@eel.upc.edu


Outline<br />

• Introduction <strong>and</strong> motivation<br />

• Series-connected <strong>multi</strong>phase <strong>multi</strong><strong>level</strong> <strong>buck</strong> <strong>converter</strong><br />

– Ideal topology. Amplifier <strong>and</strong> regulator operation<br />

– Self-driven low-floating-capacitor PFM-operated 3-<strong>level</strong> <strong>converter</strong><br />

• Design-space optimization<br />

• Mixed-signal <strong>implementation</strong> in 0.25µm TSMC CMOS<br />

– Air-core bondingwire-based inductor, tapered buffer <strong>and</strong> transistor design<br />

– Inductor current zero-crossing detection circuit<br />

• Conclusions


Outline<br />

• Introduction <strong>and</strong> motivation<br />

• Series-connected <strong>multi</strong>phase <strong>multi</strong><strong>level</strong> <strong>buck</strong> <strong>converter</strong><br />

– Ideal topology. Amplifier <strong>and</strong> regulator operation<br />

– Self-driven low-floating-capacitor PFM-operated 3-<strong>level</strong> <strong>converter</strong><br />

• Design-space optimization<br />

• Mixed-signal <strong>implementation</strong> in 0.25µm TSMC CMOS<br />

– Air-core bondingwire-based inductor, tapered buffer <strong>and</strong> transistor design<br />

– Inductor current zero-crossing detection circuit<br />

• Conclusions


3-Level (2-cell) Buck Converter<br />

• 3-<strong>level</strong> (2-cell) <strong>converter</strong><br />

has been proposed <strong>for</strong> high<br />

voltage inverters [Meynard<br />

et al., 1992]<br />

•“g 1 -g 2 ” & “g 3 -g 4 ” are<br />

complementary switches<br />

• g 1 <strong>and</strong> g 3 have the same<br />

duty cycle<br />

• V C = ½ V in<br />

g 1<br />

g 3<br />

V SW<br />

V C V in V in -V CV in V C<br />

(D > 0.5)<br />

T s<br />

V in<br />

V in /2<br />

0 T s /2<br />

t<br />

• Objective:<br />

Investigate potential <strong>for</strong><br />

lower ripple/ higher<br />

efficiency / lower reactive<br />

component size / higher<br />

b<strong>and</strong>width realization <strong>of</strong><br />

DC-DC <strong>converter</strong>


Ripple comparison <strong>of</strong> 3-<strong>level</strong> <strong>and</strong> <strong>buck</strong> <strong>converter</strong> with same f s , L, C<br />

Inductor current ripple i<br />

1<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

Switching Ripple in the 3-Level Buck Converter<br />

0<br />

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />

Duty cycle D<br />

4 times smaller peak inductor<br />

current ripple<br />

1<br />

∆imax_<br />

n = ∆i<br />

2<br />

n −1<br />

i 2<br />

i 3<br />

( )<br />

max_ 2<br />

Output voltage ripple v (mV)<br />

120<br />

100<br />

80<br />

60<br />

40<br />

20<br />

0<br />

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />

8 times smaller peak capacitor<br />

voltage ripple<br />

1<br />

∆vmax_<br />

n = ∆v<br />

3 max_ 2<br />

n −1<br />

( )<br />

Ripple results similar to two-phase <strong>converter</strong>, but with a single, smaller inductor


3-Level (2-cell) Buck Converter<br />

Comparison <strong>of</strong> 3-<strong>level</strong> <strong>and</strong> <strong>buck</strong> <strong>converter</strong><br />

• For ∆v max<br />

= 12 mV, f s_3<br />

= 200 kHz, f s_2<br />

= 560 kHz ⇒ η 2<br />

= 0.83 <strong>and</strong> η 3<br />

= 0.92<br />

⇒ improved efficiency<br />

• Same f s<br />

, ∆v o<br />

∆i L<br />

⇒ L 3<br />

= 0.25L 2<br />

, C 3<br />

=<br />

0.5C 2<br />

⇒ reduced area<br />

Example application to<br />

switching power amplifier


Two-Tone Signal Generation Using a<br />

3-Level <strong>and</strong> a Buck Converter<br />

A two-tone signal generated with a three-<strong>level</strong> <strong>and</strong> a <strong>buck</strong><br />

<strong>converter</strong>.<br />

Switching frequency, f s = 1MHz, f sig = 100kHz, f o = 550kHz, Q = 1


Experimental Envelope Tracking Wave<strong>for</strong>ms<br />

V out<br />

Frequency spectrum<br />

V out<br />

<strong>of</strong> V out<br />

Frequency spectrum<br />

<strong>of</strong> V out<br />

Harmonic components<br />

<strong>of</strong> rectified sinusoid<br />

Switching harmonics<br />

Harmonic components<br />

<strong>of</strong> rectified sinusoid<br />

Switching harmonics<br />

St<strong>and</strong>ard <strong>buck</strong> <strong>converter</strong><br />

3-<strong>level</strong> <strong>buck</strong> <strong>converter</strong><br />

• St<strong>and</strong>ard <strong>buck</strong> <strong>and</strong> 3-<strong>level</strong> <strong>buck</strong> compared <strong>for</strong> the same open-loop<br />

b<strong>and</strong>width <strong>and</strong> the same switching frequency<br />

• Modulation: rectified sine-wave at f m = 20 kHz<br />

• 30dB lower switching-frequency harmonic in the 3-<strong>level</strong> <strong>converter</strong>


Flying capacitor voltage control<br />

g 1<br />

sample<br />

sample<br />

V SW<br />

+<br />

_<br />

g 3<br />

V SW<br />

V in -V C<br />

V C<br />

V in -V C<br />

V C<br />

0 T s /2<br />

t<br />

• Digital (verilog) controller<br />

<strong>implementation</strong> using FPGA<br />

T s<br />

• x <strong>and</strong> y are sampled at<br />

V sw = V C or<br />

V sw = V in - V C<br />

V in<br />

V in /2<br />

C x<br />

g 3<br />

g 4<br />

g 1<br />

+<br />

V C<br />

V in -<br />

g 2<br />

g 1 =d(t)<br />

g 3 =d(t)+ d<br />

y<br />

x<br />

V in /2 - V q /2<br />

V SW<br />

+<br />

_<br />

V in /2 + V q /2


Experimental wave<strong>for</strong>ms <strong>for</strong><br />

flying capacitor voltage control<br />

V sw<br />

g 1<br />

V in<br />

-V c<br />

V c<br />

Uncontrolled<br />

capacitor C x<br />

voltage<br />

V in<br />

-V c<br />

V c<br />

D < 0.5<br />

g 3<br />

D > 0.5<br />

V sw<br />

g 1<br />

V in<br />

-V c<br />

V c<br />

Controlled<br />

capacitor C x<br />

voltage<br />

V in<br />

-V c<br />

V c<br />

g 3<br />

D > 0.5<br />

D < 0.5


Outline<br />

• Introduction <strong>and</strong> motivation<br />

• Series-connected <strong>multi</strong>phase <strong>multi</strong><strong>level</strong> <strong>buck</strong> <strong>converter</strong><br />

– Ideal topology. Amplifier <strong>and</strong> regulator operation<br />

– Self-driven low-floating-capacitor PFM-operated 3-<strong>level</strong> <strong>converter</strong><br />

• Design-space optimization<br />

• Mixed-signal <strong>implementation</strong> in 0.25µm TSMC CMOS<br />

– Air-core bondingwire-based inductor, tapered buffer <strong>and</strong> transistor design<br />

– Inductor current zero-crossing detection circuit<br />

• Conclusions


3-<strong>level</strong> self-driving PFM low-Cx <strong>buck</strong> <strong>converter</strong><br />

Low-C x resonant 3-<strong>level</strong> <strong>buck</strong> <strong>converter</strong> in DCM. Self-driving transistor-<strong>level</strong> topology<br />

Self-driving scheme to interconnect power transistors <strong>and</strong> drivers, which<br />

reduces the voltage across the power MOSFETs gate dielectric.


3-<strong>level</strong> self-driving PFM low-Cx <strong>buck</strong> <strong>converter</strong><br />

• Use <strong>of</strong> core transistors to implement the power MOSFETs<br />

• Use <strong>of</strong> core transistors to implement power drivers <strong>of</strong> P 2 & N 2<br />

• Reduces the power consumption <strong>of</strong> the power drivers<br />

37 MHz switching frequency<br />

26 nH inductance<br />

Driver supply voltage <strong>and</strong> vgs <strong>for</strong> all power MOSFETs,<br />

Cadence transistor-<strong>level</strong> simulations.


3-<strong>level</strong> self-driving PFM low-Cx <strong>buck</strong> <strong>converter</strong><br />

Output voltage ripple as a function <strong>of</strong> V o , <strong>for</strong><br />

the 3-<strong>level</strong> (C x sweep) <strong>and</strong> the classical<br />

(dotted line) Buck <strong>converter</strong>s.<br />

L=35nH<br />

C o<br />

=30nF<br />

f s<br />

=25MHz<br />

V bat<br />

=3.6V<br />

V o<br />

=1V<br />

I o<br />

=100mA<br />

Control signal-to-output voltage transfer<br />

function comparison between the 3-<strong>level</strong> (C x<br />

sweep) <strong>and</strong> the classical (dotted line) Buck<br />

<strong>converter</strong>s.<br />

Representative wave<strong>for</strong>ms corresponding to a DCM<br />

operated 3-<strong>level</strong> Buck <strong>converter</strong>, duty cycle below 50%


Outline<br />

• Introduction <strong>and</strong> motivation<br />

• Series-connected <strong>multi</strong>phase <strong>multi</strong><strong>level</strong> <strong>buck</strong> <strong>converter</strong><br />

– Ideal topology. Amplifier <strong>and</strong> regulator operation<br />

– Self-driven low-floating-capacitor PFM-operated 3-<strong>level</strong> <strong>converter</strong><br />

• Design-space optimization<br />

• Mixed-signal <strong>implementation</strong> in 0.25µm TSMC CMOS<br />

– Air-core bondingwire-based inductor, tapered buffer <strong>and</strong> transistor design<br />

– Inductor current zero-crossing detection circuit<br />

• Conclusions


Circuit topology considerations<br />

Optimized design space exploration (II)<br />

1) Model each per<strong>for</strong>mance index (ripple,<br />

efficiency <strong>and</strong> area) as a function <strong>of</strong> design<br />

parameters: inductor, capacitor <strong>and</strong><br />

switching frequency<br />

+<br />

i in<br />

+<br />

V in<br />

-<br />

iL<br />

R W +R C L<br />

Q 2<br />

Q 1<br />

T s =1/f s<br />

d(t)<br />

C<br />

i out<br />

+<br />

V out<br />

-<br />

R<br />

D=T on /T s<br />

∆v<br />

o<br />

=<br />

VoDT<br />

RC<br />

s<br />

=<br />

VoD<br />

RCf<br />

s<br />

=<br />

T on<br />

)<br />

A priori parameters <strong>and</strong> assumptions are application-oriented:<br />

topology, V in , V out , <strong>and</strong> the target IC technology parameters.<br />

Output voltage ripple<br />

f<br />

∆v<br />

o<br />

( L,<br />

C,<br />

fs<br />

)<br />

Efficiency<br />

Occupied area<br />

Vin<br />

I<br />

L<br />

− PL<br />

−DC<br />

− PL<br />

−core<br />

− PQ<br />

−all<br />

η =<br />

=<br />

V I<br />

in<br />

L<br />

f<br />

η<br />

( L,<br />

C,<br />

f<br />

s<br />

Area = AC<br />

+ AL<br />

+ 2AQ<br />

= farea<br />

( L,<br />

C,<br />

fs<br />

)


Circuit topology considerations<br />

Optimized design space exploration (III)<br />

2) Define a merit figure encompassing the<br />

per<strong>for</strong>mance indexes to be maximized or<br />

minimized<br />

Output<br />

voltage<br />

ripple<br />

f<br />

L<br />

C<br />

f<br />

C<br />

L<br />

generic merit figure<br />

Γ(<br />

x<br />

x ) =<br />

∏<br />

i<br />

∏<br />

β f<br />

γ i<br />

i<br />

1,...,<br />

n<br />

γ j<br />

β<br />

j<br />

f<br />

j<br />

j<br />

i<br />

( x<br />

( x<br />

1,...,<br />

1,...,<br />

Boost <strong>converter</strong> case example merit figure<br />

x<br />

x<br />

n<br />

n<br />

)<br />

)<br />

Efficiency<br />

Occupied<br />

area<br />

∆v o<br />

η<br />

A<br />

f<br />

L<br />

L<br />

f<br />

C<br />

C<br />

f<br />

f<br />

C<br />

C<br />

L<br />

L<br />

fη<br />

( L,<br />

C,<br />

fs<br />

)<br />

Γ( L,<br />

C,<br />

fs)<br />

=<br />

2<br />

f ( L,<br />

C,<br />

f ) ⋅ f ( L,<br />

C,<br />

f<br />

A<br />

s<br />

∆v<br />

Note that the area dependence is square-weighted<br />

so as to solve the ill-conditioned solution <strong>of</strong><br />

∆vo→∞ when A→0.<br />

o<br />

s<br />

)<br />

Merit<br />

figure<br />

Γ<br />

f<br />

L<br />

f s<br />

={1MHz,100MHz}, L={10nH,1uH}, C={10pF,10nF}<br />

C<br />

f<br />

C<br />

L


Circuit topology considerations<br />

Optimized design space exploration (IV)<br />

Design example <strong>for</strong> a st<strong>and</strong>ard 0.35 µm CMOS technology<br />

3) Obtain optimum point within<br />

design space (L,C, f s<br />

) as regards<br />

efficiency, occupied area,<br />

functionality<br />

V out<br />

(V)<br />

I L<br />

(A)<br />

Specs: ∆ vo<br />

=0.1 V i out<br />

=0.4 A<br />

<strong>Optimization</strong> result: L= 30 nH, C= , f s = 50 MHz<br />

+<br />

i L<br />

i in 0.82 Ω 30 nH<br />

2.5 V<br />

W=3352 µm<br />

L=0.35 µm<br />

k core=221⋅10 -11<br />

δ L=0.04 H/m 2 W=3352 µm<br />

L=0.35 µm<br />

W=2500 µm<br />

L=2500 µm<br />

i out<br />

+<br />

3.3 V<br />

- 16.5Ω


Optimized design space exploration<br />

3-<strong>level</strong> <strong>converter</strong> design example <strong>for</strong> a st<strong>and</strong>ard 0.25 µm CMOS technology<br />

Design space exploration <strong>of</strong> a CMOS-compatible 3-<strong>level</strong><br />

<strong>converter</strong>: 70% efficiency, 5mm 2 silicon <strong>and</strong> f s =37 MHz


Outline<br />

• Introduction <strong>and</strong> motivation<br />

• Series-connected <strong>multi</strong>phase <strong>multi</strong><strong>level</strong> <strong>buck</strong> <strong>converter</strong><br />

– Ideal topology. Amplifier <strong>and</strong> regulator operation<br />

– Self-driven low-floating-capacitor PFM-operated 3-<strong>level</strong> <strong>converter</strong><br />

• Design-space optimization<br />

• Mixed-signal <strong>implementation</strong> in 0.25µm TSMC CMOS<br />

– Air-core bondingwire-based inductor, tapered buffer <strong>and</strong> transistor design<br />

– Inductor current zero-crossing detection circuit<br />

• Conclusions


Bondwire triangular spiral inductors in st<strong>and</strong>ard CMOS<br />

•Area underneath inductor is usable <strong>for</strong> capacitors <strong>and</strong> power MOSFETs


Power MOSFETs<br />

Complete loss optimization <strong>of</strong> on-chip CMOS synchronous rectifier<br />

W P = 3092µm<br />

W N = 2913µm power<br />

drivers with 7.59 <strong>and</strong> 7.48 tapering factors<br />

Overall losses 37.1mW<br />

Breakdown <strong>of</strong> loss distribution, corresponding<br />

the optimized design <strong>of</strong> power MOSFETs <strong>and</strong><br />

their associated drivers.


Power MOSFET gate drive design<br />

Additional degree <strong>of</strong> freedom: impact <strong>of</strong> W p<br />

upon efficiency <strong>and</strong> delay<br />

Q i<br />

<strong>and</strong> Q e1<br />

variation as a function <strong>of</strong> the PMOS channel<br />

width <strong>of</strong> the minimum inverter (W n<br />

= 0.3µm)<br />

t fri<br />

<strong>and</strong> t fre1<br />

parameter variation<br />

Total energy losses as a function <strong>of</strong> the number <strong>of</strong> inverters n <strong>and</strong> the<br />

minimum inverter PMOS channel width W p<br />

. The area includes all the<br />

designs constrainted to a propagation delay lower than 1.15 ns.


iL=0 detection circuit. Event detection<br />

i L >0<br />

i L


iL=0 detection circuit. Circuit <strong>for</strong> time adjustment. Inductor current observer


iL=0 detection circuit Mixed-signal <strong>implementation</strong> in 0.25µm CMOS<br />

i L >0<br />

N 1<br />

N 2<br />

7.9<br />

i L


Time-domain per<strong>for</strong>mance <strong>of</strong> iL=0 detection circuit<br />

Be<strong>for</strong>e adjustment<br />

After iL=0 adjustment<br />

7.10


Complete integrated 3-<strong>level</strong> CMOS switching power <strong>converter</strong><br />

P 1<br />

2630 µm<br />

2370<br />

µm<br />

P 2<br />

N 1<br />

N2 7.17


Full-transistor-<strong>level</strong> circuit results (I)<br />

7.19


Full-transistor-<strong>level</strong> circuit results (II)<br />

7.20


Full-transistor-<strong>level</strong> circuit results (III)<br />

7.21


Experimental results


Outline<br />

• Introduction <strong>and</strong> motivation<br />

• Series-connected <strong>multi</strong>phase <strong>multi</strong><strong>level</strong> <strong>buck</strong> <strong>converter</strong><br />

– Ideal topology. Amplifier <strong>and</strong> regulator operation<br />

– Self-driven low-floating-capacitor PFM-operated 3-<strong>level</strong> <strong>converter</strong><br />

• Design-space optimization<br />

• Mixed-signal <strong>implementation</strong> in 0.25µm TSMC CMOS<br />

– Air-core bondingwire-based inductor, tapered buffer <strong>and</strong> transistor design<br />

– Inductor current zero-crossing detection circuit<br />

• Conclusions


Conclusions<br />

• Three-<strong>level</strong> <strong>converter</strong> results in favorable trade-<strong>of</strong>fs in terms <strong>of</strong> decreasing the switching ripples,<br />

decreasing the switching frequency, reducing the size <strong>of</strong> the filter elements, increasing the<br />

<strong>converter</strong> open-loop b<strong>and</strong>width, or increasing the <strong>converter</strong> efficiency.<br />

• The 3-<strong>level</strong> <strong>converter</strong> with low-Cx, self-biased drivers <strong>and</strong> operating in DCM/PFM has been<br />

presented as a c<strong>and</strong>idate <strong>for</strong> DC-DC <strong>converter</strong> integration<br />

• The use <strong>of</strong> the self-driving scheme to supply the drivers allows the use <strong>of</strong> thin-oxide transistors<br />

which increases the per<strong>for</strong>mance <strong>of</strong> the switches.<br />

• Design optimization results in the 3-<strong>level</strong> <strong>converter</strong> outper<strong>for</strong>ming the Buck <strong>converter</strong>.<br />

Future research lines<br />

• Linear-assisted scheme <strong>for</strong> <strong>multi</strong><strong>level</strong> <strong>converter</strong>s<br />

• Explore extending the approach to more intermediate <strong>level</strong>s<br />

• Use different modulations (e.g. asynchronous sigma delta)<br />

• Applying time optimal control

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!