Parasitic capacitance and resistance of metal routing

Get Complete Project Material File(s) Now! »

DC-DC converter requirements

Requirements for these DC-DC converters are application-dependent. How-ever an universal power management specification would be that a DC-DC converter takes up no space, cost nothing, last forever and has zero power loss. These universal specifications lead to the major trade-oﬀs seen in DC-DC conversion: power density, cost (area and bill of materials), reliability and power eﬃciency. The weights put on these key metrics are application-dependent. For instance, on the consumer electronic side, reliability is not a major concern, whereas design is cost-driven. In portable devices (hand-held consumer electronics) eﬃciency is the number one concern as it impacts both the battery lifetime and the heat generation inside the device. The key here is to limit the losses as much as possible. Besides the cost concern can be trans-lated in a footprint issue. In the case of portable devices, reducing the size of the PMU might allow to embed more battery volume, making the device last longer.
The functional integrated circuits are growing in terms of blocks and het-erogeneity. They are called Systems on Chip (SoCs). They necessitate more internal power domains linked to more DC voltage buses. Energy distribution can be seen as distributed through a power supply grid: that can be assimi-lated as a nano-grid on-chip. The interest is to serve independent DC voltage buses on the one hand. On the other hand, it comes the concerns of specific transient performances. Having a DC-DC converter that can react fast to a load, line and/or reference change is a key enabler for high eﬃciency process-ing. A fast power supply can adapt its output voltage closely to the load needs and supply just the power required for the processing to occur, reducing the wasted energy when only small computing resources are needed.
In other applications fields, reliability becomes number one concern. This is the case in automotive and aeronautics. However even in reliability-driven fields, the need for integration is present and eﬃciency and footprint remain major issues.

PowerSWIPE project

Today’s cars contain up to 70 electronic control units, using sophisticated silicon chips known as micro-controllers. Currently, these micro-controllers are supplied with a range of diﬀerent voltages from power supplies which take up a lot of space and waste energy. The vision for automotive control units in 2020 is that these power supplies will be miniaturized and integrated directly with the micro-controller chip, thereby dramatically saving space and weight while at the same time reducing energy use, CO2 emissions and manufacturing costs of next generation automotive electronics systems.
Nowadays micro-controllers in Engine Control Units (ECUs) are supplied using mostly linear regulators, that have low power processing eﬃciency. The road-map is to go from linear regulators to Switched-Mode Power Supplies (SMPSs) without impacting the footprint and the cost of the final solution. The PowerSWIPE project3 – EU FP7 318529 – addresses this challenge by proposing the combination of several institutes and companies4 that can sup-ply cutting edge technologies to address power supply integration.

Operating frequency

Once the choices for an architecture and a structure have been made, one needs to decide at which frequency the converter will operate. This is a crucial decision as it will have a major impact on the design, as well as on many metrics of the converter [Nev14]. A converter switching at high frequency will require smaller passive component values (compared to one switching at lower frequency). This implies that these passive components can be smaller, thus the total footprint of the converter is reduced, increasing its power density. Figure 1.3 depicts the expected gain in area and volume when going to higher frequency. This shows the eﬀect of using smaller components first, and then being able to integrate them inside the chip package itself.
However increasing the switching frequency is known to increase the losses [Gil02], leading to an ineﬃcient conversion and generating unwanted heat that must be addressed with an induced cost. Switching frequency also has an impact on the EMI performance of the system [Aul13] and on the size of required EMI filters for the system to comply to norms. As the operating frequency is higher, the components required to filter the perturbations can be reduced, increasing the overall compactness.
Increasing the frequency makes the converter faster to react to a load or a reference change, improving transient performance without having to use complex control strategies. The current in the inductor can change faster, as its value decreases with frequency, allowing for faster di=dt. The main trade-oﬀ is then on how much eﬃciency points a designer is willing to pay in order to improve some of the previously mentioned metrics.

Strategies for eﬃciency enhancement

In order to improve the eﬃciency of the DC-DC converters, many strategies have been proposed in literature. For high frequency conversion, these strate-gies are aiming to mainly reduce the switching losses of the converter. Some techniques are presented below. These techniques are not exclusive. They can be combined together to increase eﬃciency, usually at the expense of complexity. For instance the converter proposed in [Abe07] combines Zero Voltage Switching (ZVS) strategy in a multiphase structure.

Switching operation optimization

One way of reducing the losses is to analyze and optimize the switching cycle of the power devices. The interest of ZVS has been demonstrated in [Abe07], presenting eﬃciency figures of two similar converters, one with ZVS operation and one without. The ZVS converter presents a much better eﬃciency than its non-ZVS counterpart. Basic ZVS behavior is to turn on the power switch when the voltage across its terminals is equal to zero. This reduces losses as the V I product is null during turn on.
Another similar well known approach is to go to Zero Current Switching (ZCS). The principle of this method is to turn oﬀ the power switch when the current through it is equal to zero. Again the goal is to have a V I product null during turn-oﬀ. This behavior can be forced using a resonant or quasi-resonant operation of the converter in order to have the current naturally decreasing before the switching operation.
Additional passive components are required to introduce the necessary resonance. At high operating frequency, parasitic components are considered.
Unfortunately CMOS technology is aﬀected by variability that makes the con-trol of capacitor and inductor value diﬃcult, resulting in a quite impractical approach.

READ Validation of Modeling Technique Using Discrete Components

Power switch drivers’ optimization

The drivers in a DC-DC converter are generating a significant amount of losses. Some techniques can be used to reduce their losses.
A resonant gate driver has been proposed in [Bat12a]. It allows to re-duce the gate drive losses by reusing some of the electrical charges needed for turning on and oﬀ the power switches. A significant eﬃciency enhancement is demonstrated at the expense of power density.
Another technique is to reduce the voltage swing of the drivers in order get the optimum trade-oﬀ between switching and conduction losses of the main switch. This has been presented in [Kur04], demonstrating a +3:9 % eﬃciency gain compared to a classical design.

Further optimization strategies

Many other strategies have been proposed. The use of multiphase converter helps keeping a high eﬃciency over a wide load range by using a phase shedding scheme. The goal is to use each phase near its optimum load current. This approach has been demonstrated in [Kim15; Bur14; Son14b; Hua13].
The segmentation of the converter can also be considered at the compo-nent level. The goal is to get the best achievable eﬃciency over the maximum number of operation points, i.e. various load current and output voltage con-figurations. This technique has been used in [Mar14; Aro13].
In multiphase converters, it has been proposed to use coupled inductors in order to reduce the current ripple, thus reducing the conduction losses of the converter. Reducing the current ripple can also allow to reduce the switching frequency, thus reducing the switching losses. Various coupled structures have been tested in [Pen13; Stu13; Wib08a; Wib08b].

DC-DC principle of operation

This section present the principle of operation of some DC-DC converter struc-ture. The presented structures are limited to few relevant non-isolated DC-DC converters. The focus is on step-down capable converter. The goal is to provide a quick insight into the converter structures and the circuit behavior. Three major DC-DC converter families are presented: linear voltage converter, switched-capacitors converters and inductive converters.

Linear voltage converter

This family of DC-DC converters acts as a resistive divider. Excess power is dissipated through a variable resistor in order to ensure a regulated output voltage. Figure 1.4 depicts a streamlined linear voltage regulator circuit. The output voltage is controlled by varying the RVAR resistor in order to have an output voltage equal to the reference voltage.
There are no switching components in such a regulator, allowing for low noise regulation. Regulation can also be very fast. It is mainly dependent on the bandwidth of the feedback loop [Cou13]. However since it acts as a resistive divider, eﬃciency is strongly related to conversion ratio. This implies that the conversion eﬃciency is very low when the voltage drop across the converter is close to the input voltage. Therefore these converters are mainly used to regulate an output voltage close to the input voltage, giving them the name of Low Drop-Out (LDO) regulator.

Switched-capacitor DC-DC converters

A switched-capacitor DC-DC converter (also referred to as charge pump) uti-lizes a set of switches and capacitors to divide a voltage. The division factor is architecture-dependent but it is possible to have reconfigurable structures, allowing for multi-ratio conversion.

Table of contents :

1 State-of-the-art
1.1 Primary concerns
1.1.1 System level considerations
1.1.2 Design tools and CAD
1.1.3 Commercial trends
1.1.4 Converter structure
1.1.5 Operating frequency
1.1.6 Strategies for efficiency enhancement
1.2 DC-DC principle of operation
1.2.1 Linear voltage converter
1.2.2 Switched-capacitor DC-DC converters
1.2.3 Inductive DC-DC converters
1.3 Performance analysis
1.3.1 System view
1.3.2 Components and technologies
1.4 Conclusion and design specifications
1.4.1 Specifications of the design target
1.4.2 Technology portfolio
1.4.3 Technology suppliers
2 Design methodology
2.1 Architecture definition
2.1.1 Waveforms and equations
2.1.2 Losses breakdown and comparison
2.2 Control strategies
2.2.1 Linear control strategies
2.2.2 Sliding-mode and hysteretic regulators
2.2.3 V2, quasi-V2 and V2IC control
2.2.4 Selection of a control method
2.3 Modeling of power switches
2.3.1 MOSFET model
2.3.2 Technology evaluation
2.3.3 Expected efficiency
2.4 Output filter and decoupling capacitors
2.4.1 Circuit models
2.4.2 Dimensioning
2.5 Conclusion and design objectives
2.5.1 Active parts
2.5.2 Passive components
3 IC Design
3.1 Global system view
3.2 Design details
3.2.1 Power stage
3.2.2 Drivers
3.2.3 Level-shifter
3.2.4 Current reference
3.3 Cell design layout considerations
3.3.1 On-chip switching noise
3.3.2 Parasitic capacitance and resistance of metal routing
3.3.3 Matching issues
3.4 Layout of converters
3.4.1 Sub-circuit cells layouts
3.4.2 Full converter layouts
3.5 Chip-level considerations
3.5.1 Converter placement
3.5.2 On-chip capacitors
3.5.3 Top-level routing
3.6 Conclusion
4 Tests and measurements results
4.1 Testing particularities
4.1.1 Challenges
4.1.2 Testing equipment
4.2 Decoupling capacitors and interposer
4.2.1 Manufactured interposer
4.3 Integrated magnetic inductors
4.3.1 Designed structures
4.3.2 Inductor testing
4.4 Converter
4.4.1 Single die
4.4.2 Interposer assembly
4.4.3 State-of-the-art status
4.5 Conclusion
5 General conclusion
5.1 Summary of contributions
5.2 Perspectives
5.2.1 Short-term issues
5.2.2 Mid-term issues
5.2.3 Long-term issues
Bibliography