The alternating current (AC) line voltage around the world varies in terms of voltage and frequency. The U.S. and Japan supply voltages are around 100V, while Europe and China use ~220V. The frequency across the world also varies from ~50 to 60Hz.

Having a device that works all over the world is pretty convenient. Many devices take advantage of this concept, such as mobile phone chargers and laptop supplies. However, when was the last time you took a big-screen TV on a flight with you from the U.S. to Europe? In that case, it does not make much sense. Larger items that require high power and do not need to travel from zone to zone can take advantage of having a power supply that is specifically designed for the input voltage range.

There are a number of factors to consider which involve the AC input range when designing a power supply:

• Size.
• Cost.
• Performance.
• Regional regulatory markings.
• Electromagnetic interference (EMI).
• Power factor correction (PFC).

A supply designed to operate over the universal AC input range (85-265V) will be more expensive, larger and less efficient than one designed to operate over a certain range (high line or low line). It is also costly for the product to pass the regulatory checks in each region where it will be sold. If the device is only going to be used in Europe, why spend the extra money to qualify it in China? There are also some regions which require special qualifications, such as power factor correction, that other regions don’t. Adding a PFC circuit to a product that only gets used in the U.S. can be a large, unnecessary expense.

As an example of power supply configurations, let’s consider a 250W audio amplifier for a home theater. Table 1 lists some of the trade-offs of the power supply configurations.

Table 1: Power supply parameters for a 250W design

Table 1 takes into account the differences between the configurations.  The most important result is in the relative cost.  There could be up to a 3x price difference for a universal line configuration.  In addition to the cost, there will also be differences in performance, size and complexity.

In some cases it really makes sense to use a universal input, but in many cases you can realize great benefits by using separate supplies. For more information on this topic, read the EE Times Power Tip post “insert title and link”.

Additional resources

• Explore more power-supply topics.
• Find a Power Tips video to help with your next design challenge.

This post was originally published on EPC's GaN Talk blog. Discover more about the LMG5200 and TI GaN solutions here. 

Alex Lidow, Ph.D., CEO and Co-founder, EPC
David Reusch, Ph.D., Executive Director of Applications Engineering
John Glaser, Ph.D., Director of Applications Engineering

The demand by our society for information is growing at an unprecedented rate. With emerging technologies, such as cloud computing and the internet of things (IoT), this trend for more and faster access to information is showing no signs of slowing. What makes the transfer of information at high rates of speed possible are racks and racks of servers, mostly located in centralized data centers.

In 2014, data centers (in the U.S.) consumed about 100 billion kilowatts hours (kWh) of energy and it is projected by the Natural Resources Defense Council (NRDC) that data center electricity consumption will increase to roughly 140 billion kilowatt-hours annually by 2020, the equivalent annual output of 50 power plants.

The power needed to support this rapidly growing demand comes from our electrical grid, and goes through multiple conversion stages before it actually feeds the remaining energy into a digital semiconductor chip. In figure 1 is an illustration of this “journey.” Also shown in this figure are the losses due to the transmission and conversion of electricity – from the power plant to the computer chip that does all the work.

Multiplying these efficiency numbers shows that the power grid needs to supply 150 watts of power to meet the demands of a digital chip that may need only 100 watts of power. Therefore the combined waste across the U.S. due to power conversion for servers is 33 billion kWh, equivalent to almost a dozen power plants. But the overall wasted energy within the server farm is even more, because every watt of power loss through power conversion is actually energy that is converted into heat, and removing this heat demands even more power.

Figure 1: Typical multi-stage power conversion architecture used in modern server farms, which takes 150 watts of power from the electrical grid to supply 100 watts to a digital processor used in servers.

Whereas the electrical grid has been around for more than a century, the various stages of conversion have been built with technologies based on the semiconductors developed post World War II. These semiconductors are based on silicon crystals, and it is the properties and limitations of silicon that shaped the architecture of figure 1.

In this article, we will demonstrate the benefits of enhancement-mode gallium nitride (eGaN^® technology) based power converters in solutions for existing data center and telecommunications architectures centering around an input voltage of 48 V_DC with load voltages as low as 1 V_DC. We will explore the capability of high performance GaN power transistors to enable new approaches to power data center and telecommunications systems with higher efficiency and higher power density than possible with previous Si MOSFET based architectures.

Getting from 48 V_IN– 1 V_OUT Directly

Since the adoption of the 12 V intermediate bus architecture (IBA), bus converters are currently approaching about an order of magnitude increase in output power, from around 100 W to current designs of around 1 kW in a quarter brick footprint. This means that the amount of current on the 12 V bus to the POL converters has also increased by a factor of 10 and, without reductions in busing resistance, a two order of magnitude increase in busing conduction losses follows. GaN technology-based solutions have already demonstrated significant efficiency improvements compared to silicon based solutions in traditional IBA systems.

However, with the increasing conversion losses in the 48 V_IN bus converter, the mounting 12 V busing losses on the motherboards, and the higher performance of GaN technology, different architectures may now be considered, such as going directly from 48 V_IN to load using non-isolated POL converters, as shown on the bottom of figure 2.

Figure 2: Intermediate bus architecture (IBA) and a direct conversion DC bus architecture.

The first approach evaluated to convert 48 V_IN directly to 1 V_OUT is a traditional buck converter. The buck converter is one the simplest topologies and the approach taken in the vast majority of current 12 V_IN systems. For the 48 V input, an 80 V eGaN monolithic half-bridge IC (EPC2105), embedded in an EPC9041 demonstration board, was selected for the much higher step-down ratio. A second approach to convert 48 V_IN direct to 1 V_OUT, from Texas Instruments, uses a transformer-based design to improve converter efficiency. An LMG5200 GaN-based half bridge is used on the 48 V_IN input side, and four 30 V EPC2023 eGaN FETs are used on the low voltage output side.

The efficiencies and power densities of the two 48 V_IN to 1 V_OUT designs are shown in figure 3. The buck converter efficiency is that of the entire power train, including the inductor (Würth Elektronik 744 301 033), capacitors, and PCB losses. At a switching frequency of 300 kHz, a peak efficiency of 84% is achieved, and at 20 A the efficiency is around 83.5%. The power density of the buck building block (excluding controller) is approximately 300 W/in3. For the transformer-based approach operating at 600 kHz, efficiencies over 90 % are achieved, with the efficiency almost 88% at 50 A of output current. For the transformer-based building block (excluding controller), the power density is approximately 80 W/in³.

A comparison of estimated efficiencies and power densities for the single stage 48 V_IN to 1 V_OUT POL converters and the traditional two-stage IBA approach using the best GaN technology-based design is shown in figure 4, and summarized in table 1 (Silicon-based solutions are significantly less efficient than these GaN technology-based solutions). The GaN IC-based IBA’s power converters have an estimated 1.5% efficiency improvement over the direct 48 V_IN to 1 V_OUT conversion for the buck-based approach. However, when adding in the losses from the 12 V bus, estimated to be 2%, the total system efficiencies are very similar. The traditional IBA approach and 48 V_IN direct conversion buck-based approach have similar power densities. For the 48 V_IN transformer-based approach, the system efficiency is over 7 % higher than the traditional IBA approach and the system has a lower power density, about a third of the conventional IBA GaN based approach.

The DC bus architecture also has a potential cost advantage, since the cost of the IBC can be eliminated and the cost increase of the 48 V_IN POL converter over the 12 V_IN POL converter will be minimal as they use a similar number of power devices, controllers, and drivers.

Figure 4: Performance comparisons of GaN technology-based 48 V_IN intermediate bus architecture and 48 V_IN DC bus architecture

In figure 5 we revisit figure 1 while applying the single-stage efficiencies demonstrated with eGaN FET and IC based designs. The direct savings by eliminating just the last stage in the server farm power architecture is not only a cost reduction, but also a reduction of power consumed between 7-15% depending on the GaN-based approach. This correlates to direct savings of up to 21 billion kWh per year when compared with silicon-based solutions. This savings is increased further when air conditioning energy costs inside the server farm are added, bringing the total to almost 25% of the 140 billion kWh consumed by servers in the U.S. alone.

Conclusion

The impact of eGaN technology in our post-silicon world is even greater than the savings possible in U.S. server farms today and is but one example of the impact that this new, emerging technology makes to the efficient use of electrical power. eGaN technology provides a path to higher performance semiconductors, re-opening the possibilities of Moore’s Law for driving innovation – just as Moore’s Law falls off the tracks. For example, eGaN technology is enabling new applications such as wireless power transmission, 5G cellular, autonomous vehicles, and colonoscopy pills. And, as the electronics industry gains experience and knowledge in the inherent attributes high performance capabilities, the resulting high performance eGaN semiconductor devices will enable many unforeseen applications, just as silicon, its predecessor, brought about in the post-WWII era.

eGaN^® FET is a registered trademark of Efficient Power Conversion Corporation.

General Reference

[1] D. Reusch and J. Glaser, DC-DC Converter Handbook, Power Conversion Publications, 2015. ISBN 978-0-9966492-0-9

Example of PMIC powering an SoC

As more and more designs are surpassing the capabilities of simpler microcontrollers, today I’m going to walk you through the basic considerations for powering a system on chip (SoC). The first thing you’ll need to do is pull up the data sheet or technical reference manual for the SoC you have selected. Within these documents are five conditions that can help you define the power scheme for your particular processor or SoC:

1. Absolutemaximum ratings.These ratings define the conditions that, if exceeded, will damage the device. If you’re worried about breaking your new processor, you definitely need to know what the device can tolerate
   
   
2. Typical power domains.SoCs integrate features that often operate at different voltage levels, requiring separate supplies. For instance, ARM devices often use a dedicated power supply for the ARM core separate from the SoC domain, which contains the rest of the bells and whistles integrated into the device. Other common domains include the input/output (I/O) domains for logic and communication peripherals, and the memory domain, which will depend on the type of memory selected. You’ll find specific voltages listed in the data sheet under a title similar to “Recommend Operating Conditions,” often containing some combination of minimum, typical, and maximum values.
   
   
3. Power-sequence requirements.Not only do you need to provide adequate voltages to the required domains, but you also need to ensure that you enable and disable the internal domains in the correct order to avoid non-ideal behavior such as current leakage or latch-up conditions. This is often accomplished by delaying the state change of specific domains until the prerequisite domains have stabilized or surpassed a specific voltage. Sometimes power-sequencing requirements will list the order in which to supply the domains, while other data sheets may detail the maximum allowable voltage differences or the minimum required delay times between specific domains. You can create delays by enabling regulators with a previous supply or corresponding power-good signal, with specific delay times accomplished through resistor/capacitor RC networks and additional logic gates. Alternatively, reference designs can assist you if your project has an aggressive schedule, and many power-management integrated circuits (PMICs) (like the TPS65910 PMIC) even have sequences pre-configured for various devices.
   
   
4. Current requirements.Defining the expected currents for the application will greatly simplify the selection process when choosing acceptable voltage regulators. If available, the absolute maximum supply currents will detail the most demanding conditions that a regulator should support. If these currents are not clearly defined within the datasheet, or the values presented seem excessive for your application, you might be able to find an additional document online that contains power-consumption measurements for various operating conditions, which can help you set more realistic expectations for typical values. Often, you can group domains with equivalent voltages together as long as they do not violate the sequencing requirements. This flexibility for combining or separating domains enables you to redistribute load currents across available regulators.
   
   
5. Additional features.Features like dynamic voltage scaling (DVS) and different modes of operation expand SoC functionality and can reduce power consumption when appropriate. You may not need these features for simpler designs, which could operate with discrete voltage regulators permanently configured for one voltage value. Other designs may be more consumption-conscious, however, and would benefit from a flexible power scheme that enables different operating modes such as sleep states or high-performance conditions. PMICs generally integrate multiple regulators into a single package that you can control on the fly, making them flexible for platforms that require separate voltage domains in a compact space. The added versatility of a PMIC often comes at a higher cost than discrete devices, but adaptable devices such as the TPS65023 are affordable for even cost-conscious designs, expanding the capabilities of the SoC compared to simpler power architectures.

Of course, you may need more details depending on the processor, but these generic guidelines can quickly help you understand the basic power requirements for a new platform. The number of unique voltage domains will correlate to the number of voltage regulators you need, and summing the individual current limits for grouped modules will define the power capability you need for shared supplies.

You can revisit specific parameters such as tolerances and transient conditions to help narrow down component selections once you have selected a few possibilities. Additionally, searching the datasheet or technical reference documents for keywords such as “must” and “require” can help spotlight other inconspicuous necessities.

Additional resources

• Learn more about PCB design and layout strategies
• Learn more about the portfolio at TI.com/PMIC

USB Type-C changed the USB ecosystem in a major way by making both ends of the USB cable interchangeable (not just flippable). This enables USB devices such as laptops or smartphones to have different behavior depending upon what other USB device they are connected to, since the data role and power role can be exchanged independently. The USB Implementers Forum has now released version 1.2 of the USB Type-C specification. There are several key changes over version 1.1 that I’ll summarize in this post, but the biggest difference is a shift in the terminology used to describe this new USB ecosystem. You might say the USB Type-C terminology got a “do-over” with this version.

The new terminology does a better job explaining this new USB world, and is meant to clarify and emphasize that data roles and power roles are orthogonal to each other (except that the initial power role determines the initial data role). In other words, USB is fully embracing this new two-dimensional ecosystem. The following table has some key terms to know.

Figure 1 categorizes many of the possible applications and shows where they fall on this two-dimensional grid. 

Figure 1: USB Type-C version 1.2 example applications

 USB Power Delivery enables swapping or changing data roles. For example, there may be a dual-role data (DRD) system that is always a source of power but can be either DFP (host) or UFP (device). Alternatively, there may be a dual-role power (DRP) system that is a DFP (host) while sourcing power, but does not support data while sinking power. Discussing all of the possible applications is a topic for another post, but here I’m just trying to highlight the separation of data roles and power roles.

The initial power role and the initial data role are still associated, as they always have been in Type-C. The device that initializes as the source is either the DFP or not data capable; it may not initialize as a UFP. Likewise, the device that initializes as the sink is either the UFP or not data capable. In order to swap data roles after the initial connection, you must use the USB Power Delivery message DR_Swap. 

There are also two features in USB Type-C called default source and default sink. (Version 1.2 adds some clarifications about these features, which were previously referred to as try source and try sink, respectively.) The default source feature is intended for systems that primarily deliver power, but that also sink power at times such as a power bank. The power bank should be providing power unless it is connected to source-only or its battery runs out.  The default sink feature is intended for systems such as smartphones that primarily sink power, but can source power if connected to a sink-only accessory. These two features can be leveraged dynamically, meaning that depending upon the charge level of its battery or some other criterion, the system may change to a sink-only or default source. The following table summarizes the different kinds of roles for USB Type-C devices.

What else is new in USB Type-C version 1.2? Table 1 lists several other changes.

The max source VBUS capacitance change is worth discussing. It demonstrates how seriously the USB-IF takes compatibility with legacy USB. Legacy USB Type-A ports always have 5V on VBUS even while not attached to anything. As a result, when attaching a USB Type-B port, there is an in-rush current into the VBUS capacitance on the Type-B port. USB-IF has long required that USB Type-B ports not have more than 10µF of capacitance in order to limit that in-rush current.

 Since a USB Type-C system with a receptacle can connect to a USB legacy Type-A receptacle, all USB Type-C receptacles must also limit the capacitance on VBUS to less than 10µF. Without this requirement, legacy USB Type-A systems may not be able to supply the in-rush current and their internal voltage rails could droop, which could cause a blue screen if the droop is severe enough. Large in-rush currents have also been known to cause connector damage if repeated many times.

 TI’s latest USB Type-C Power Delivery devices, the TPS25740 and TPS25740A, are both compliant with the latest USB Type-C version 1.2. Explore TI’s full portfolio of USB Type-C solutions.

 Additional resources

• Read the blog post, “Why does Iq matter for USB Type-C?”
• Jumpstart your design with the USB Type C 5V/12V/20V Control Card Reference Designand USB-C DFP, 20V/3A Out, Universal AC Input Reference Design

My cousin introduced me to the Texas two-step a couple years ago. I enjoyed swing dancing in college but had never tried two-step before. My first few attempts were quite abysmal. Fortunately, I was with friends and we could laugh about it. After several more tries, I was able to get the hang of it.

It can also be a little daunting to learn a new converter topology.You might be familiar with the conventional buck converter. The simplicity and beauty of this converter has made it popular for decades. TI recently introduced the TPS54A20 based on the series capacitor buck converter. It is a new topology that enables efficient, high-frequency operation of small point-of-load voltage regulators.

Figure 1: Thetwo-phase series capacitor buck converter

Today we are going to learn the “steps” of the series capacitor buck converter shown in Figure 1. Like any new dance, it may be challenging at first. After walking through the steps of steady-state operation a few times, I think you will find that it is not that difficult. You might even like it! This will be a brief beginner’s class; if you want more details, check out this application note. So let’s begin by considering a converter with a 12V input switching at 5MHz per phase.

The first step, or time interval, occurs when the high side switch of phase A (Q_1a) is on as shown in Fig. 2. The series capacitor (C_t) connects to the input by switch Q_1a. Because the nominal voltage across the series capacitor is half the input voltage (approximately 6V in this case), the phase A switch-node voltage (V_SWa) is roughly half the input voltage, shown in blue in Figure 2. The phase A inductor current (I_La) rises in a triangular fashion just like a normal buck converter (no resonant behavior) and simultaneously charges C_t. In fact, the series capacitor current (I_Ct) is equal to I_La during this step. The differential series capacitor voltage (V_Ct) increases by a few hundred millivolts due to the added charge. During this step, the phase B low-side switch (Q_2b) is on, connecting the phase B switch node (V_SWb) to ground. The phase B inductor current (I_Lb) decreases linearly as a result.

Figure 2: High side switch of phase A (Q_1a) on (step 1) 

Both low-side switches (Q_2a and Q_2b) are on during step two, as shown in Fig 3. This connects both V_SWa and V_SWb to ground just like a conventional two-phase buck converter. Both I_La and I_Lb have negative slopes. Because the series capacitor has no current flowing through it (because I_Ct is zero), V_Ct remains constant.

Figure 3: Both low side switch on (step 2)

Step three is where things get interesting, so pay attention to Fig. 4. Switch Q_2a is still on, connecting V_SWa to ground. Switch Q_2a is also connecting the negative side of C_t to ground. When the phase B high-side switch (Q_1b) turns on, the positive side of the series capacitor connects to V_SWb. Now the series capacitor is acting like an input capacitor for phase B! I_Lb ramps up and simultaneously discharges the series capacitor. This is evident from the negative I_Ct and the small decrease in V_Ct. I_La continues to ramp down.

Figure 4: High side switch of phase B (step 3)

Step four is identical to step two as shown in Fig. 5. Q_2a and Q_2b are on and V_SWa and V_SWb are grounded. Both I_La and I_Lb ramp down. V_Ct remains fixed because I_Ct is zero. After step four, the whole cycle repeats.

Figure 5: Both low side switches on (step 4)

How was that? Not too hard, right? Check out the additional resources to learn more about this exciting new topology. Now it’s time to hit the design floor and take the series capacitor buck converter for a spin.

Additional resources

• Explore the “Tiny, Low Profile 10 A Point-of-load Voltage Regulator Reference Design.”  
• Read the white paper “Breakthrough power delivery for space constrained applications.”
• Start a design now with WEBENCH® power designer.
• Order the TPS54A20evaluation module.

Need to replace your fan?  Simply trade it out.  Want to add more storage capacity?  No problem – just exchange that 500GB SSD for 4TB. 

But have you ever worried during one of these activities that your server might spontaneously combust?  Probably not.  Unknown to some end users, many modern electronics provide protection against current and voltage spikes during what is known as a “hot swap” event. 

What is hot swapping and why do you care?

Hot swapping is when the user connects an external device or module to expand system capabilities or provide regular maintenance without powering down the host system.  While you see hot swap activity in a variety of applications, protection against these kinds of events is critical in complex systems such as servers.  As you can see in Figure 1, a current or voltage spike from a hot swap event can result in hardware damage, expensive repairs, server downtime, or physical injury to yourself or others. 

Figure 1. Damage to IC due to over voltage and over current events.

Where can this happen in your server?

Many servers are designed to be highly configurable – comprising modules that you can swap in and out as needed including fans, storage devices (HDD and SSDs), and power supply units (PSUs) as shown in Figure 2.  You must carefully consider protection near these modules against hot swap events.

Figure 2. Server components including storage, fans, and PSUs are commonly hot swapped.

Typically, you can place protection against hot swapping events on either the module or the host system as shown in Figure 3.  Due to the highly configurable nature of servers, the host system or backplane vendor is often different from the module vendors.  This makes it difficult to know where protection already exists, but if you are designing a module or backplane, it never hurts to have redundant protection against surges in your server.

Figure 3. You can place hot swap protection on both the host system and module

How can you protect against hot swap events in your server?

There are many options to protect against hot swap events in a server.  Let’s take a look at a few common solutions.

Fuses and polyfuses can serve as low-cost solutions – but a large footprint (shown in Figure 4), degraded performance, and increased maintenance costs can outweigh this benefit over time. 

Figure 4. Comparison of fuse to TI TPS25942 eFuse

Hot swap controllers are another common solution.  These devices provide control logic to an external FET and sense resistor that enable design flexibility specifically when setting R_DSON and upper current limits.  However, for many space-constrained server applications, an eFuse can provide the necessary protection and save precious board space by integrating the external components. 

In addition to integration, eFuses add protection features critical for servers.   For example, TI’s TPS25942 eFuse offers common protection features required for a hot swap event including adjustable current limit, over voltage protection, and thermal shutdown, as well as general system protection such as programmable soft start, under-voltage protection, and reverse current blocking.  Should a hot swap event occur, the TPS25942 recovery options include latched and auto retry versions. 

Figure 5. TPS25942 simplified schematic.

Along with its protection features, the TPS25942 also offers system status monitoring by providing outputs to the system for power good, fault, and current monitoring.  With all these features packed in a 3mm x 4mm QFN, it’s easy to see why an eFuse is the right choice for hot swap protection in your server!

So the next time you swap out your fan or install the latest in memory technology – and your super sweet server doesn’t erupt into flames – just remember that it’s all thanks to hot swap protection.  Don’t forget to include an eFuse for hot swap protection in your next server design!

Learn more about designing with eFuses:

• See how traditional fuses and polyfuses compare to modern TI eFuses, check out: “Get out of the dark: Upgrade your fuse!”
• For more information on eFuse integration, read “How to integrate your hot swap with an eFuse”
• To learn more about eFuse fault recovery, read this three-part blog series: “eFuses: clamping and cutoff and auto retry, oh my!”

Multilayer ceramic capacitors (MLCCs) are popular in power electronics designs compared to traditional polymer capacitors for many reasons:

An MLCC provides:

• A small profile with relatively higher capacitance.
• Very low equivalent series resistance (ESR).
• Very low equivalent series inductance (ESL).
• Lower impedance at higher frequencies.
• Non-polarization for easier mounting and manufacturing.
• Higher reliability over time compared to tantalum and aluminum electrolyte capacitors.
• Lower unit costs.

However, MLCCs do not always sit quietly on the board and do their job. Sometimes they get bored and start to “sing.” This is due to the piezoelectric effect of the ceramic material, which has the same characteristics as other ferroelectric dielectrics. When an electric potential or field applied on the surface of an MLCC causes a deformation at a frequency range from 20Hz-20kHz, it could be audible to humans. This is called MLCC acoustic noise, or singing noise.

Major contributions to acoustic noise include:

• Electric potential operating at a frequency within an audible range.
• Smaller case size tends to be lower sound levels than lager case size.
• Ceramic dielectric constants (K); a higher K has higher ferroelectric properties.
• Less ceramic layers producing lower sound levels due to less deformation.

Image courtesy of Murata.com

Figure 1: MLCC and board deformation when applying an electric field

The MLCC itself should be quiet to human ears. However, it can be loud when mounted on a printed circuit board (PCB). Let’s say you had a ceramic capacitor at the input of a switching power converter. The switching behavior creates a high-frequency voltage change on the ceramic capacitor; as the voltage increases and decreases, the MLCC will expand and contract. The deformation of the MLCC creates vibration of PCB, which causes to buzz amplifying. The higher the electric potential change, the larger the deformation (piezoelectric effect), which will result in a louder sound when the frequency occurs in the audible range.

Some applications can use electrolyte or tantalum-type capacitors, preferably thru-hole types when acoustic noise is problematic. But for applications that are more cost-sensitive or size-constrained (such as personal electronic devices), you cannot avoid thin, small ceramic capacitors, and the need to reduce noise immediately becomes critical.

Here are a few available solutions that can minimize or reduce noise to acceptable levels:

• Use acoustically quieter capacitors. Capacitor manufactures have already developed ceramic capacitors with low distortion dielectric material, which exhibit lower ferroelectric properties and smaller deformation in regards to a voltage change. And there is a series manufactured by Murata that the capacitor is on interposer substrate to reduce the acoustic noise (Figure 2). Murata also has a series with a special mechanical configuration; it uses metal terminals to mount the capacitor on the PCB board to achieve noise reduction by absorbing mechanical impact (Figure 3). Unfortunately, this kind of capacitor tends to be more expensive, which prevents its wide use by end-equipment manufacturers. The effect of noise suppression depends on the type of capacitors. (Figure 4) 

Figure 2: Mechanical configuration of “interposer” ceramic capacitors

Image courtesy of Murata.com

Figure 3: Mechanical configuration of “metal terminal” ceramic capacitors 

Figure 4: Acoustic noise reduction effect by each capacitor (Typical value)

• Reducing noise by optimizing PCB layout. The origin of the noise is the interaction of MLCCs with the PCB. Optimizing component placement on the PCB can be effective. Using a thicker PCB allows the sound frequency to shift due to weight change. Some articles also suggest placing the components at the edge of the PCB to lower the sound pressure level. Similarly, placing components symmetrically on top and bottom of the PCB can also help to reduce noise level, since the two vibrations will cancel each other out, due to the cancellation-of-vibration effect (Figure 5) when the voltage applies to both capacitors simultaneously.

Figure 5: Capacitors on each side of a PCB to create vibration cancellation

• Reducing voltage amplitude variation on capacitors. In most cases, end device manufacturing is limited by cost or size, which makes the first two methods described to reduce acoustic noise not practical. However, the other main factor that determines noise is how high or fast the voltage variation is across the capacitors. This is something that can be optimized through proper system design, by either improving the load-transient response or line-transient response.

Considering line-transient response as an example, an experiment was conducted using the TPS51622, one of TI’s DCAP+™ Vcore controllers, by measuring the noise level with a sound meter when varying the output voltage change with fast (48mV/µs) and slow (12mV/µs) slew rates using the Intel voltage regulator (VR) tool. Sending an I²C command to the TPS51622 changes the output voltage from 0.5V to 1.5V, and the input-voltage ripple was measured shown in Figure 5.

Figure 6: Input-voltage ripple on input ceramic capacitors with fast/slow slew rates

The voltage amplitude with a fast slew rate is much higher than with a slow slew rate; this voltage difference across the capacitor directly translates into an increase of noise in decibels. Measured data shows that the noise dropped to a lower, quieter level, from ~40dB to ~50dB. See table 1 for other sound levels and effects.

Table 1: Noise sources and their effects

The wide usage of conventional ceramic capacitors brings acoustic noise issues to power system designs. However, there are solutions that approach the problem from different angles: changing the electronic characteristics of the MLCC itself, or minimizing its interaction with the PCB. These methods either reduce the noise to an acceptable level or remove the noise from the source by using more expensive “anti-noise” capacitors. 

My cousin introduced me to the Texas two-step a couple years ago. I enjoyed swing dancing in college but had never tried two-step before. My first few attempts were quite abysmal. Fortunately, I was with friends and we could laugh about it. After several more tries, I was able to get the hang of it.

It can also be a little daunting to learn a new converter topology.You might be familiar with the conventional buck converter. The simplicity and beauty of this converter has made it popular for decades. TI recently introduced the TPS54A20 based on the series capacitor buck converter. It is a new topology that enables efficient, high-frequency operation of small point-of-load voltage regulators.

Figure 1: Thetwo-phase series capacitor buck converter

Today we are going to learn the “steps” of the series capacitor buck converter shown in Figure 1. Like any new dance, it may be challenging at first. After walking through the steps of steady-state operation a few times, I think you will find that it is not that difficult. You might even like it! This will be a brief beginner’s class; if you want more details, check out this application note. So let’s begin by considering a converter with a 12V input switching at 5MHz per phase.

The first step, or time interval, occurs when the high side switch of phase A (Q_1a) is on as shown in Fig. 2. The series capacitor (C_t) connects to the input by switch Q_1a. Because the nominal voltage across the series capacitor is half the input voltage (approximately 6V in this case), the phase A switch-node voltage (V_SWa) is roughly half the input voltage, shown in blue in Figure 2. The phase A inductor current (I_La) rises in a triangular fashion just like a normal buck converter (no resonant behavior) and simultaneously charges C_t. In fact, the series capacitor current (I_Ct) is equal to I_La during this step. The differential series capacitor voltage (V_Ct) increases by a few hundred millivolts due to the added charge. During this step, the phase B low-side switch (Q_2b) is on, connecting the phase B switch node (V_SWb) to ground. The phase B inductor current (I_Lb) decreases linearly as a result.

Figure 2: High side switch of phase A (Q_1a) on (step 1) 

Both low-side switches (Q_2a and Q_2b) are on during step two, as shown in Fig 3. This connects both V_SWa and V_SWb to ground just like a conventional two-phase buck converter. Both I_La and I_Lb have negative slopes. Because the series capacitor has no current flowing through it (because I_Ct is zero), V_Ct remains constant. 

Figure 3: Both low side switch on (step 2)

Step three is where things get interesting, so pay attention to Fig. 4. Switch Q_2a is still on, connecting V_SWa to ground. Switch Q_2a is also connecting the negative side of C_t to ground. When the phase B high-side switch (Q_1b) turns on, the positive side of the series capacitor connects to V_SWb. Now the series capacitor is acting like an input capacitor for phase B! I_Lb ramps up and simultaneously discharges the series capacitor. This is evident from the negative I_Ct and the small decrease in V_Ct. I_La continues to ramp down.

Figure 4: High side switch of phase B (step 3)

Step four is identical to step two as shown in Fig. 5. Q_2a and Q_2b are on and V_SWa and V_SWb are grounded. Both I_La and I_Lb ramp down. V_Ct remains fixed because I_Ct is zero. After step four, the whole cycle repeats.

Figure 5: Both low side switches on (step 4)

How was that? Not too hard, right? Check out the additional resources to learn more about this exciting new topology. Now it’s time to hit the design floor and take the series capacitor buck converter for a spin.

Additional resources

• Explore the “Tiny, Low Profile 10 A Point-of-load Voltage Regulator Reference Design.”  
• Read the white paper “Breakthrough power delivery for space constrained applications.”
• Start a design now with WEBENCH® power designer.
• Order the TPS54A20evaluation module. 

The smart home is no longer just for movie and millionaires. New products are hitting the market every day to connect you to every part of your house. There are products to turn on your lights from the other room and you can even find a device to monitor your dog from afar. With advancing technology it’s now easier than ever to create your own automated system using a microcontroller. You can even add light emitting diodes, or LEDs, for visual feedback and status indication! In this blog, we’ll take a further look at designing LEDs for home automation designs.

When you’re first taught to use a microcontroller with higher current LEDs, you’re taught to use a transistor for the input control, and to place a resistor in series with the LED to set the forward current. This is fairly simple to do with a single LED, but adding additional LEDs takes up a lot of space. Soon, you will be looking for a bigger microcontroller with more general-purpose input/output (GPIO) pins. This becomes costly and requires more processing effort to control the multiple LEDs.

Figure 1: A discrete implementation to drive an LED using a microcontroller unit (MCU)

LED drivers help solve this problem by simplifying the control of multiple LEDs. They also add to the overall solution by adding features such as blinking, pulse width modulation (PWM) dimming, and error detection. A home automation system can communicate its status to the user using the blinking and PWM signaling using the functionality of the LED driver instead of relying on the microcontroller.

One of the biggest advantages of using an LED driver over traditional current setting resistors is the ability to control a large number of LEDs with minimal GPIO pins. TI’s TLC59116 I²C LED driver can control 16 different channels on a single device using only three microcontroller pins. The four hardware address pins allow the user to go from 16 channels to 224 channels using the same three pins. This means that as the home automation project becomes more complex, the number of LEDs needed to communicate the device status can be scaled as needed.

Figure 2: Multiple TLC59116 can be added using the same SCL, SDA, and RESET pins.

Not only can the number of GPIO pins for the microcontroller be minimized, the processing effort required by the microcontroller to create the LED effects can also be reduced. The TLC59116 supports both PWM dimming and blinking of LEDs using internal frequency control. Combined with the functionality to set every LED at the same time, the TLC59116 can control a variety of LEDs using basic I²C commands.

So how will you enhance your home automation project? Maybe you want to warn that a room is getting too cold, or notify that your favorite pet hasn’t been fed. Feel free to add some RGB LEDs for status indication. Multiple pets? Add more LEDs to make sure that no puppy goes hungry. With LED drivers like the TLC59116, you can scale your project with ease!

Additional resources:

• Learn more about TI’s LED driver portfolio.

Looking back at my childhood in one of coastal India’s cities, I remember dense humidity paired with tolerable heat in the low 90s – a classic equatorial climate. But when I visit now, I notice that the temperatures are much higher, often exceeding 100°F. I also notice split air conditioners installed almost everywhere; these were rare during my childhood. Split air conditioners consist of an indoor and outdoor unit. The outdoor unit is installed outside the wall of the room and houses the condenser coils and compressor. Finally, I noticed many signs and advertisements from major air-conditioning suppliers.

Air-conditioner units draw a lot of power, so their increasing popularity makes it imperative to keep energy costs lower. The only way to do this is to have energy-efficient air-conditioners and a well-insulated home. When selecting the right air conditioner, its capacity and seasonal energy efficiency ratio (SEER) rating are key. A higher SEER rating means greater energy efficiency. Fortunately, a majority of air-conditioner brands have high energy-efficiency ratings.

Energy-efficient air conditioners use variable frequency drives (VFDs) in compressors and condenser fans. The other usage is the inclusion of active power factor correction (PFC) in air-conditioning power supplies. Both trends represent the adoption of switched-mode power conversion in systems that previously relied on less-efficient line-frequency electronics to power compressors and fans. VFDs have a motor in the compressor unit, whose speed can vary by modulating the voltage, current and frequency of the power delivered to the compressor. As a result, the system does not have to run at full speed. This dramatically cuts down on energy costs and is by far the largest benefit of VFDs. And since air conditioners operate at relatively high power levels, typically on the order of a few kilowatts, implementing active PFC improves power quality and reduces harmonic distortion. This reduces the amount of reactive power that ends up going back to the grid as wasted or unutilized power.

Both VFDs and PFC in air-conditioning units require high-voltage systems that are trending towards implementing integrated circuit (IC)-based solutions. Using the right power supply ICs, such as a high-voltage isolated driver  like the UCC21520, can provide an ecosystem for air conditioners that enables high system-level efficiency and lower bill-of-materials cost, allowing you to stay cool, efficiently.

Additional resources:

• Get to the know the UCC21520. 
• Explore TI’s PFC controllers, including UCC28180, UCC28070 and UCC28950.
• Learn more about TI system solutions for indoor and outdoor air conditioner designs.

Wide-bandgap power devices such as gallium nitride (GaN) and silicon carbide (SiC) field effect transistors (FETs) have become commercially available in recent years. Compared with high-voltage (≥600V) silicon FETs, GaN and SiC FETs generally have lower on-resistances (R_ds(on)), lower output capacitances (C_oss) and fewer/no reverse recovery charges (Q_rr). Due to their lower switching losses, you can greatly increase the efficiency of a hard-switching converter with wide-bandgap power devices.

Applying GaN FETs to resonant converters improves efficiency by reducing magnetic losses. Let’s take an inductor-inductor-capacitor series resonant converter (LLC-SRC), shown in Figure 1, as an example. An LLC-SRC uses the energy stored in the resonant inductor (L_r) to discharge MOSFET output capacitors in the input switch network. If the output capacitor voltage discharges to zero before the MOSFET gate signal goes high, you can achieve zero turn-on loss.

Figure 1: An LLC SRC

Figure 2 shows key waveforms of an LLC-SRC. During the MOSFET’s switching transient, i_Lr equals the maximum current flow through L_m, expressed as Equation 1:

The current I_Lm– assumed constant during dead time – discharges the C_oss of one MOSFET and charges the C_oss of another MOSFET. Assuming the C_oss of the two MOSFETs of the half bridge are the same and that you can ignore the interwinding capacitance of the transformer, Equation 2 expresses the maximum inductance with which you can achieve zero turn-on losses:

Figure 2: LLC-SRC switching waveforms

Now let’s presume you’re choosing between a GaN FET and silicon FET on the same 400V_IN to 12V_OUT conversion specification using an LLC-SRC. TI’s LMG3410 GaN device has a 70mΩ on-resistance and a 95pF output capacitance (energy related). One 70mΩ silicon FET I found has a 140pF output capacitance. If your selected turns ratio is n = 16 and the target maximum switching frequency of the LLC-SRC is 750kHz, L_m,max will be 134µH for TI’s LMG3410 and 91µH for the silicon FET with the 140pF output capacitor. As the input switches, the air gap of the LLC-SRC transformer with the silicon FET will be wider than the transformer with the LMG3410 if you apply the same core. Because of the wider air gap, there will be more eddy current losses on the transformer wires.

Figure 3 shows the thermal performance of the same LLC-SRC with different transformer air gaps under the same test conditions. As you can see, the wire losses on the transformer with the wider air gap are much higher than the one with the narrower air gap. Therefore, using GaN devices with lower C_oss can help reduce magnetic losses in resonant converters.

Figure 3: LLC-SRC transformer thermal performance at 400V_IN, 12V/42A output with L_m = 100µH (narrower air gap) (a); and L_m = 70µH (wider air gap) (b)

While in this post I discussed a benefit of using GaN devices on resonant converters – lower output capacitance enabling fewer transformer losses – TI GaN devices such as the LMG3410 offer not only low R_ds(on) and C_oss but also incorporate several protections such as overcurrent and overtemperature protection. With all of these protections, converter reliability improves greatly.

In my next post, I will discuss power-supply reliability with TI GaN devices in more detail.

Additional resources

• More detail information of TI GaN devices can be found in the following articles:
  • Advancing power supply solutions through the promise of GaN
  • Optimizing GaN performance with an integrated driver
  • A comprehensive methodology to qualify the reliability of GaN products
• If you are interested in using TI GaN devices in your power supply designs, take a look at the following reference Design
  • 400V – 12V/500W High-Frequency LLC Series Resonant Converter reference design (PMP20289).
  • GaN FET-Based CCM Totem-Pole Bridgeless PFC

Figure 1: Automotive exterior lighting with LEDs

In one of my previous blog posts, I introduced different dimming methods for light-emitting diode (LED) drivers: analog and pulse-width modulation (PWM) dimming. In this post, I will elaborate more on PWM dimming and how to get its frequency fast enough in order to reduce its effect on humans or other living creatures.

Let’s review why PWM dimming is preferable to analog dimming in some applications, such as rear lamp. The brightness and color of an LED depends on the current flowing through. Once the current changes, the brightness and color change accordingly. Some older red LEDs might even change from red to orange because of flowing current-level changes. This is not desirable in rear lamp that requires precise color control per governmental regulations. The concept of PWM dimming is to leverage the slow response of human eyes by quickly turning the LED on and off (at a frequency above 100Hz) without changing the LED current flowing through during turn-on time. Since human eyes are not responsive to any frequency above 30 Hz, brightness seems linearly changing according to the duty cycle of the PWM signal as seen by human eyes. LED color is preserved this way.

While PWM dimming provides advantages in LED brightness control, are there any drawbacks? Yes:  it comes down to the quality of light with PWM dimming and its effect on different objects. PWM dimming involves frequency. Low-frequency dimming is in the 100Hz to 2kH range, making it perceptible by humans (subtly) and inducing eye strain or fatigue. A banding effect might occur when taking photos or recording videos if the dimming frequency is in such a range. PWM dimming can also create the stroboscopic effect, which is when moving/rotational objects look stationary. In short, to use PWM dimming and avoid its drawbacks, you should set the PWM dimming frequency higher than 2kHz.

To achieve high-frequency dimming, most LED drivers have a PWM dimming input. However, the bandwidth of the LED driver limits the dimming frequency and contrast ratio. For a fixed-frequency switched-mode power supply-type LED driver using a DC-to-DC conversion architecture, the loop bandwidth is typically designed at or below 50kHz. That imposes a limit on the contrast ratio to about 25-to-1 with a 2kHz PWM dimming frequency. To achieve a better contrast ratio, you must either use a lower PWM dimming frequency or try to further increase the loop bandwidth.

How about connecting a switch parallel to the LEDs? When the switch is off, current flows through the LEDs and shines light; when the switch is on, current flows through the switch and there is no light. This method is called shunt-FET dimming. You will need a fast-responding LED driver that enables abrupt changes in load, yet maintains good current regulation and prevents current overshoot from destroying the LEDs. With the nature of hysteretic converters or constant on-time converters that don’t form feedback loops, the response on the variation of output voltage is fast and the load change is instant. By using shunt-FET dimming with these type of drivers, current is continually flowing and regulated accordingly, regardless of whether the dimming switch is on or off. From the light-output perspective, the brightness depends on the duty cycle of the switch off-time. Thus, you can achieve linear brightness changes according to the duty cycle of a high-frequency (>2kHz) PWM signal.

To enable high-frequency PWM dimming with a shunt-FET dimming approach, TI provides fast-responding LED drivers such as the LM3409, a controlled off-time hysteretic PFET buck LED controller; the TPS92641, a constant on-time synchronous buck controller; and the TPS92515, a monolithic 2A controlled off-time hysteretic buck LED converter. All of these devices work well with shunt-FET dimming and exhibit superb performance given their small board size and high power conversion efficiency. Shunt-FET dimming with a hysteretic LED driver is the way to improve the quality of LED lighting by enabling high-frequency PWM dimming.

Additional resources

• Read other blogs on LED lighting.
• Discover what’s new with LED lighting illumination.
• Check out the TPS92515 product folder.
• Read this white paper to learn more about stroboscopic effects.

The current-limit specification for a DC/DC converter can sometimes cause confusion for designers not familiar with this type of regulator. In this two-part series, I hope that I can help clear away some of the confusion.

First, the current-limit specification in a DC/DC converter data sheet does not mean the same thing as it does for a low-dropout regulator (LDO). For an LDO, the current-limit value is the maximum current that the device will provide to the output when the regulator is in an overload or short-circuit condition. For a step-down converter, the data sheet will specify the limit on either the peak or valley of the inductor current. However, it is the average inductor current that represents the output current of a buck converter. Equations 1 and 2 translate the inductor current limits into the maximum output current:

As an example, let's take the LMR16030 converting a 24V input to a 5V output. As shown in Figure 1, the minimum peak-current limit given in the data sheet is 3.8A.

Figure 1: Current Limit Specification for LMR16030

Using Equation 1, with the inductor chosen in the data sheet example, you have:

Note that this is a “3A” regulator, yet you can get somewhat more than 3A of load current under these conditions.

Take the LM43601 as another example. As Figure 2 shows, the data sheet specifies a typical peak-current limit of 2.45A and a typical valley-current limit of 1.1A.

Figure 2: Current Limit Specification for LM43601

Using Equations 1 and 2, you have:

You would take the lower value of 1.58A as your maximum output current for this device under these conditions.

The previous calculation represents the maximum output current before the converter becomes overloaded and the output falls out of regulation. For the more severe condition of a short circuit on the output, most regulators will reduce the maximum output current even more. This "fold-back" mode helps prevent overheating of the converter and damage to the power-stage components.

In part 2, I will consider the step-up or boost converter. In the meantime, consider one of TI’s DC/DC converters for your next design.

Whenever you have a job to do, it’s important you have the right stuff, whether it’s knowledge, tools, supplies or equipment are at the basic list. In addition to working in power-supply design, I do a lot of hands-on projects, including automotive repair. Not because I’m cheap (well, maybe a little), but usually because I’m not satisfied with the job when someone else does it.

Recently, I got my motor home back from the transmission shop and noticed an oil leak. After a while the oil started pouring out from the cooler adapter seal. I knew the shop must have removed it to get other work done. After disassembling the filter and adapter, I found the petrified remains of the original gasket underneath. So much for a quality job. So I headed to the parts store for a gasket set. The replacement O-ring seemed a bit skimpy, and it leaked even worse. I went back to the parts store and the clerk said I needed to get the right stuff – some expensive form-a-gasket material that comes in a can. He handed me the can and there it was, right on the label: “The Right Stuff.” Ah ha. It worked great.

Doing a good job requires knowledge and expertise, which is why I have been a fan of today’s worldwide Power Supply Design Seminars since I first attended in the 1980s. Before Texas Instruments, Unitrode’s applications experts hosted the seminars. I remember attending as a young engineer and soaking up as much as I could. The best part was a chance to have lunch with one of the authors or presenters. I remember thinking, oh, if I could just be like those guys. So now I guess I am one of those guys.

While working for National Semiconductor prior to the TI acquisition, I developed a “Loop Compensation Made Easy” training course for our field applications engineers. Later, my colleague Louis Diana incorporated the presentation for TI training around North America, and for the 2016 seminar we have expanded the material to include isolated topologies in a topic titled, “Switch-Mode Power Converter Compensation Made Easy.”

There was one thing that always bothered me about frequency compensation for current-mode control. The simplified approach uses a single-pole approximation for the power stage and output filter, with the assumption that the inductor pole moves out to some higher frequency near the switching frequency. But where exactly does it go? After attending R.D. Middlebrook’s course, “New Structured Analog Design,” I was able to quantify the inductor pole simply using impedance analysis. If I lost you here, don’t worry. You don’t need to know how to derive the equations in order to understand and use the results you will learn in this year’s seminar.

By the way, our presentation has a list with links to compensator design topics within the entire Power Supply Design Seminar library by some of the greats – like Lloyd Dixon and Bob Mammano – along with more recent resources.  The library is open to all and can be accessed at www.ti.com/psds.

If you have a chance to attend the kick-off session in Boston, the original authors of the material typically make the presentations in that city. In addition to local and regional power experts, many of the authors will present at other locations; for example, I will be in Boston, Irvine, Seattle and San Jose. The schedule of topics and registration for the 2016 Power Supply Design Seminar, also known as SEM2200, is here: www.ti.com/powerseminars. This year, we have a new format with two rooms: one for seminar topics and the other for demonstrations and related topics. I look forward to seeing you and am sure you will find the right stuff to help with your power designs.

So you know you need a voltage reference, but you are not sure how to pick the best one for your application. You have come to the right place! In this post, I will discuss some of the key voltage-reference parameters and help you weigh them based on your application needs in your search for a “Goldilocks” voltage reference that’s “just right.”

First, think about your application and whether you are going to need a shunt reference or a series reference. You don’t have to settle on a topology right away, but it will be helpful to know the best use cases for each. If you are not familiar with the difference between a shunt reference and a series reference, I have written a blog post “Understanding Voltage References: Shunt vs series. Which topology is right for you?” and a white paper “Shunt versus series: How to select a voltage-reference topology” on this topic, so check those out.

Second, define the system boundaries of your application. Understanding the physical environment, ambient temperature changes and whether any system calibration will take place are important considerations when looking for a voltage reference. I have outlined a few key parameters below, with some corresponding calculations and questions.   At the end of each section are example devices and applications where those devices are often used.

Initial accuracy

• Relevant questions:
  • “How accurate do I need my reference to be?”
  • “Do I need to perform system calibration?”
• Description:
  • “Initial accuracy” is a specification of how close the actual reference voltage will be to the listed voltage at room temperature. Initial accuracy is typically specified as a percentage of the target voltage. For example, a 2.5V reference with an initial accuracy of +/-1% will have an initial room-temperature voltage between 2.475V and 2.525V, or +/-2.5mV. Keep in mind that this initial accuracy specification may not a high priority if you are performing system calibration.
• Examples:
  • LM4132 series reference: 0.05% accuracy.
  • LM4030 shunt reference: 0.05% accuracy.
• Applications:
  • The LM4030 is popular in factory automation and test/measurement applications because of its high initial accuracy.

Temperature coefficient

• Relevant question:
  • “Will the system experience high or low temperatures?”
• Description:
  • “Temperature coefficient,” also known as TempCo or temperature drift, is a measure of the change in the reference voltage due to a change in ambient temperature. This delta is typically listed in parts per million/degrees Celsius, which may be a little confusing at first. To calculate the change in the reference voltage, multiply the temperature coefficient by the tested temperature range. For example, the LM4132 has a temperature coefficient of 20ppm/°C over a temperature range of -40°C to 125°C, so the maximum deviation due to temperature is 3,300ppm, or 8.25mV for a 2.5V reference. Even if the application temperature range is less than the tested temperature range, you must use the full tested temperature range in your calculations because the deviation is not necessarily linear with temperature.
• Examples:
  • LM4132 series reference: 20ppm/°C.
  • REF50xx series reference: 3ppm/°C.
  • LM4050 shunt reference: 50ppm/°C.
• Applications:
  • The LM4132’s low temperature coefficient makes it popular in factory automation and control applications, where the environment can stray from a comfortable room-temperature setting.

Thermal hysteresis

• Relevant question:
  • “Will the system cycle hot and cold temperatures?”
• Description:
  • “Thermal hysteresis” is the difference in reference voltage at room temperature after a full cycle of ambient temperature. A full ambient temperature cycle starts at room temperature and ramps the ambient temperature to the device’s minimum and maximum before returning to room temperature. The reference voltage is measured at a temperature of 25°C before and after the temperature cycle, and the difference is recorded as thermal hysteresis. Thermal hysteresis is caused by stresses imposed on the package and the die from thermal effects such as thermal expansion. If your application will be in a fixed temperature environment, thermal hysteresis is less of a concern.
• Example:
  • LM4140 series reference: 20ppm.
• Applications:
  • The LM4140 has very low thermal hysteresis, making it great for factory automation applications such as sensor transmitters.

Long-term stability

• Relevant question:
  • “How stable will the reference voltage be in the long run?”
• Description:
  • “Long-term stability” is an estimate of the reference voltage deviation over time. This parameter is characterized during product development, where several parts are tested under normal operating conditions for 1,000 hours. The reference voltage is measured before and after the 1,000-hour test, and the difference is recorded in parts per million as the long-term stability. Placing the part on an area of the board where mechanical stresses are minimal can reduce the factors affecting long-term stability. The edges of a typical printed circuit board will see the least deformation for a given stress, while the center will see the most deformation.
• Examples:
  • LM4030 shunt reference: 40ppm.
  • REF50xx series reference: 45ppm.
• Applications:
  • The LM4030 is popular in factory automation and test/measurement applications because of its low long-term drift, where voltage references need to last a long time.

Low-frequency noise

• Relevant question:
  • “How much noise should I expect to see on the reference voltage?”
• Description:
  • “Low-frequency noise,” characterized as a peak-to-peak value between 0.1Hz and 10Hz, is typically not filtered out due to its low-frequency nature. On the other hand, wideband noise is often characterized between 10Hz and 10kHz, and is easier to filter out with a properly sized output capacitor. Low-frequency noise is therefore typically the dominant source of noise for a voltage reference.
• Examples:
  • LM4140 series reference: 2.2µVpp.
  • REF50xx series reference: 3µVpp/V (3µVpp per volt on V_OUT; for the REF5025, 3µVpp per volt on V_OUT = 7.5µVpp).
• Applications:
  • The low noise of the LM4140 makes it a popular series reference in medical, server and factory-automation applications.

Now that you have a basic understanding of some of the most important voltage-reference parameters, you can more confidently browse the library of voltage references available from Texas Instruments. If you are interested in more detail about topics discussed in this blog, read the white paper, “Voltage Reference Selection Basics.”

In part 1 of this two-part series, I talked about the maximum output current of a step-down or buck DC/DC converter. In this installment, I will look at the step-up or boost converter. Calculating the maximum output current for a boost regulator is somewhat more involved, but still straightforward.

The first thing to understand about a boost converter is that the average inductor current is not equal to the output current, as it is in a buck converter. A boost regulator will still control the inductor current, but this represents the input current of the converter, not the output current. Thus, boost converters are usually specified with a maximum MOSFET current rather than a maximum output current.

As an example, the LMR62421 is called a “2.1A step-up voltage regulator.” This refers to the MOSFET switch current and not the output current. You can use Equation 1 to estimate the maximum output current for a boost converter:

First, you need to estimate the efficiency of the converter, η, by looking at the efficiency curves in the data sheet and finding one close to the conditions required in your application. Let’s take an example from the LMR62421 data sheet and find the maximum output current when converting 5V to 12V, using info from the data sheet (Figure 1).

Figure 1: Datasheet excerpts for LMR62421

Filling in equation 1 with the data in figure 1 and our input and output conditions, we can calculate the maximum output current:

Does this seem like the wrong result? Does it really say that you can only get 0.73A from a “2.1A” boost converter? Yes – and it makes no difference what boost converter you use.

It is not hard to understand this result. For any DC/DC converter, the input and output power will be nearly equal. If your output voltage is higher than the input, then the input current must be higher than the output current, by about the ratio of V_OUT/V_IN, to give equal power on both sides. Since the MOSFET sees the input current, its rating must be much larger than the required output current. Equation 1 says the same thing mathematically, while taking into account the efficiency of the converter.

Because it is a little tricky to find the maximum output current for a given boost regulator, or to find a boost with a given output current, using the TI WEBENCH^® Power Designer tool is the best approach to the problem. The WEBENCH tool will make all of the calculations for you and find a range of suitable converters that match your requirements.

Additional resources

• For more information about boost regulators, see the application note, “Working with Boost Converters.”
• Start a design now with WEBENCH Power Designer.
• Consider one of TI’s DC/DC converters for your next design.

Smart meters are next-generation meters that replace existing meters, which still use technology that was developed decades ago. Smart meters use a secure connection network to automatically and wirelessly send the energy usage to the utility companies. This means that customers will no longer receive estimated energy bills or have meter readers come into their homes to read the meter.

Smart meters use more advanced communication interfaces as compared to conventional IR (Infra-Red) and IrDA (Infra-red Data association) interfaces; they also require more memory and a more powerful microcontroller. These features increase the power consumption which necessitates the use of   switch-mode power supply(SMPS) rather than a capacitive-drop power supply .The single-phase energy meter has to  operate from minimum 100V_AC up to 500V_AC.The three-phase energy meter has to operate with minimum of single-phase present (100V_AC) up to 300V_AC in each of the phases . Compliance with efficiency standard - and more importantly compliance with lower power-consumption requirements – poses more challenges for SMPS designers as you cannot bill customers for the energy that the meter use. On the other hand the smart meter energy usage should not place unacceptable power demands on utilities either.

Utility companies have been experiencing revenue loss due to tampering of energy meters worldwide. Since the first electronic energy meter was deployed in the field, unethical people have attempted to alter meters to steal electricity without paying for it. They will devise some ways to fail the power-supply .So this tampering may be the most important problem that a smart meter power-supply designer has to confront. Most power supplies use ferrite core based transformers which are cheap and efficient but susceptible to strong magnetic fields generated by permanent rare-earth magnets placed a few centimeter away. As soon as the magnet comes near the vicinity of the transformer, it saturates the transformer and creates an over-load condition, damaging the switch (MOSFET/Bipolar junction transistor) and hence destroying the power supply. Most of the power supply controllers available today have an integrated function of overcurrent protection built in. If transformer becomes saturated, a fast current comparator turns off the switch thus protecting the power supply. But the drawback is that there will be no power available to the metrology block and thus no metering. However this is what energy meter’s foremost function is. To maintain continuous operation during any attempt at tampering, one option is to shield the transformer with a magnetic shield material. But this is an expensive alternative and adds to the assembly cost –as each ferrite core based transformer would need shielding.

Another option is to use a high reluctance powdered iron core instead of ferrite core in the transformer. A powdered iron core has a much higher flux density of 1.2 – 1.4 Tesla as compared to 0.4 – 0.5 Tesla for ferrite core. It costs less than putting a magnetic shield around the transformer. But suffers from higher core loss which can significantly reduce power-supply efficiency. This drawback may not be particularly significant in smart-meter power supplies, however.

 Peak power consumption in smart meters jumps to 1W-10W levels depending upon the wireless communication protocol used.

For sub 1 GHz peak power consumption is around 0.5W; for ZigBee it is around 1W and for Global System for Mobile communication (GSM) it is around 10W.

But the meter typically consumes less than 1W for most of its operating life. The efficiency of the power-supply at low/light loads is critical in determining the overall power consumption of the meter. This imposes high light-load efficiency as a target for the power-supply designer. Several approaches can increase efficiency at lower/ light loads such as reducing the quiescent current of the switching controller, using a primary-side controller with a high voltage start-up integrated circuit, or using a switching controller that contains a mixture of frequency and amplitude modulation. The reduction in switching frequency and thus current with the reduction in load reduces core loss, thus justifying the use of powdered iron core based transformers.

A high operating voltage is essential in smart meters to protect against the accidental connection of a single-phase meter to two phases (500V_AC) or the presence of a 300V_AC in each of the three phases of a three-phase meter. But this high voltage increases power complexity, component count and cost.

A flyback topology is the simplest and the cheapest topology for a smart meter power-supply design. In a flyback topology, the switching component has to withstand up to 1000V: 730V_DC  (300VAC *√3 *√2) plus 220V_DC clamp voltage plus 50V_DC overshoot due to the snubber diode conduction delay. The switching component should have a voltage rating of 1200V assuming 15% derating. The switching converter (with a built-in 700V/800V MOSFET) uses another 500V MOSFET in cascode configuration to meet the 1200V requirement. Another low-cost technology uses a using a single 1200V BJT with switching controller to take care of high-voltage protection. The cost of BJT with the same voltage and current rating price is one-third the price of a MOSFET.

Thus the optimal SMPS design would feature a powdered iron core based transformer, control law in switching converter to take care of light-load efficiency and a cascode configuration to handle high input voltages.

Additional resources:

• Discover the UCC28722 and start designing with the Ultra-wide Input Range, Dual Output, Offline AC/DC Bias Supply with PSR Control and BJT Switch Reference Design and Dual-Output 5V at 300mA and 15V at 100mA Isolated Flyback From 85V to 440V AC Mains Reference Design.
• Read the white paper, “Smart power opens door to more efficient electrical use”

If you think power-management integrated circuits (PMICs) only power a system’s processor, allow me to introduce the application-specific PMIC. Application-specific PMICs have the same great system benefits of a general-purpose PMIC – including system-cost reduction, space savings, power sequencing and platform scalability – but they are generally smaller devices designed for end-equipment systems. Additionally, application-specific PMICs have ultra-low leakage current to help preserve battery life in portable applications. In this post, I’ll describe two example applications for an application-specific PMIC.

Compact Camera Modules (CCM) for dual-camera applications

Current versions of portable electronics such as smartphones, tablets and notebooks now use two cameras: a “world-facing” camera and a “user-facing” camera. Integrating both cameras into end equipment such as smart phones, tablets, and detachable notebooks has created the need for an integrated and highly efficient power solution. An application-specific PMIC like the TPS68470 can power a compact camera module (CCM) in a dual-camera application: generating the clock for the image sensor, driving light-emitting diodes (LEDs) for camera flashes and various indicators, and incorporating LED drivers for privacy indicators.

Because camera-sensor modules are sensitive to local electrical noise, system designers must consider ways to reduce the noise. Camera PMICs integrate clean power rails to mitigate this noise. While a discrete power implementation would require that you design additional logic components onto the board, the PMIC has integrated power-sequencing components, resulting in a reduced solution size and less sequencing design effort.

Figure 1 shows a high-level block diagram for a PMIC that can power a dual-camera module.

Figure 1: TPS68470 block diagram

Electronic Paper Display (EPD applications

Electronic paper displays (EPDs) can display an image even without a power connection. EPDs are also incredibly thin (60µ), giving them an advantage in space-constrained applications. With these benefits, you can add displays to products with challenging power and space limitations.

E-ink works by moving positive and negatively charged microcapsules suspended in a clear solution when electric charge is applied. EPD applications require several output power rails such as a low-input supply for their display including ±15V.

Figure 2 is an application schematic for the TPS65185.

Figure 2: TPS65185 typical application schematic

An application-specific PMIC like the TPS65185 integrates necessary power rails into a single device to provide a highly efficient and space-saving solution for an EPD. The TPS65185 handles sequencing and is I²C-controlled to accommodate specific power requirements.

These are two examples of application-specific PMICs. PMICs are not just power solutions for powering your entire system. Application-specific PMICs help integrate a small number of power rails into a single IC to power a dedicated system block, while still giving the same great benefits of a general PMIC.

Additional resources

• Ask questions about application-specific or general-purpose PMICs on the TI E2E™ Community PMIC forum.
• Find the right PMIC for your application

Have you ever had to do the same calculations more than once? As an electrical/electronics engineer, I am pretty sure that you have. If you are doing calculations manually, it can be quite tedious and very time-consuming. When designing power-management circuits, you can change a lot of parameters and choose from a whole variety of topologies, which can increase the amount of repetitive calculations.

The brand-new Power Stage Designer Tool 3.0 calculates the necessary voltage and current information for 17 power-supply topologies instantly. In Figure 1 you can see the topologies, which are supported. There is no need to wait for results after entering your inputs – as you would have to with a simulation – because the tool is based on equations, assuming perfect coupling and considering only diode losses. This means that you should pick your components with a reasonable margin so that effects like ringing or spiking caused by parasitics won’t cause any damage in the final design.

The tool also displays waveforms for all of the important components of the power stage. If you are not yet familiar with a topology, Power Stage Designer can also help you understand its mode of operation. The tool can calculate results and waveforms both in continuous conduction mode (CCM) and discontinuous conduction mode (DCM) for 14 topologies. The synchronous buck, Weinberg and phase-shifted full-bridge topologies are only available in CCM.

Figure 1: Topology overview of Power Stage Designer 3.0

Figure 2 shows the information available in the main window exemplified by the Zeta converter main window.

Figure 2: Zeta converter main window

You can view the peak values for currents and voltages, root-mean-square (RMS) current, alternating current (AC) and waveforms for each component individually by clicking on its symbol in the schematics. This information, like displayed in Figure 3 for L2 of a Zeta converter, enables you to choose the appropriate device rating and number of components for your design. After all, the components you choose need to be able to withstand current and voltage stress under all specified conditions. A slider in the graph window enables you to alter the input voltage within the entire input-voltage range. Calculations are independent of a specific controller, which gives you the flexibility to evaluate different topologies first. Thus, you will get a good overview without already having to dig too deep into data sheets during the beginning of the design phase. Of course, after picking a topology and a controller/converter, you have to make sure that your inputs do not violate certain integrated circuit (IC) limits, such as minimum on time, off time or maximum duty cycle.

Figure 3: Graph window for L2 of the Zeta converter

When you are happy with your design and want to keep the parameters, you can save your setup in a file and load it again whenever you need it. Another option is to search TI’s website for already built and tested TIDesigns, which might match your specification. Or just transfer the design parameters to WEBENCH® Design Center, where you can look for a controller/converter and optimize your design.

Additional resources

• Check out Power Stage Designer 3.0 and enjoy designing faster and much more conveniently.
• Watch our latest training video: Introduction to Power Stage designer

Power factor correction (PFC) forces the input current to follow the input voltage (V_IN)so that any electrical load appears like a resistor. This action requires sensing the input voltage and modulating a current reference based on that sensing. The current loop will force the input current to follow the reference. This is called average current-mode control, as shown in Figure 1.

Figure 1: PFC average current-mode control

You can find lots of commercial PFC controllers with low total harmonic distortion (THD) using this average current control algorithm in the market. However, these PFC controllers need a dedicated pin to sense V_IN and a precision analog multiplier for current-reference modulation.

Another PFC control algorithm that does not need V_IN sensing but can still provide average current-mode control has become very popular recently. TI’s UCC28180 belongs to this family. Because it lacks a V_IN sense pin and the precision analog multiplier, it comes in a smaller package, enabling lower system cost, and is very easy to use.

But when we introduce the UCC28180 to designers, many times their first response is, “What? Without V_IN sensing? How does that work?” In this post, I will try to answer this question.

Figure 2 shows the control algorithm used in the UCC28180. A low-bandwidth voltage loop regulates the output voltage. The input current is measured as V_Iin and compared with a saw-wave V_ramp. The amplitude of V_ramp is proportional to the voltage-loop output. Because PFC uses the boost topology, the input-voltage information is already there, but hidden. The control algorithm shown in Figure 2 employs the hidden information.

Figure 2: PFC without V_IN sensing

The PWM output signal always starts low at the beginning of the switching cycle, triggered by the internal clock, as shown in Figure 3. The PWM output stays low until V_ramp rises linearly to intersect the V_Iin voltage. The V_ramp/V_Iin intersection determines switch turning off time t_OFF. 

Figure 3: PWM generation

From Figure 3:

Here T is the switching period. For boost converter operating in continuous conduction mode (CCM):

Combining Equations 1 and 2 gives you Equation 3:

The voltage output, V_OUT, is a constant in steady state. Since the PFC voltage loop is very slow, V_ramp is also a constant in steady state. Thus, the input current is solely proportional to V_IN. If V_IN is sinusoidal, the input current must be sinusoidal, achieving good PFC. It looks like magic, doesn’t it?

Learn more about TI’s PFC solutions.

Additional resources:

• Get started designing with the UCC28180:
  • 230-V, 400-W, 92% Efficiency Battery Charger w/PFC and LLC for 36-V Power Tools Reference Design
  • 230-V, 3.5-kW PFC with >98% Efficiency, Optimized for BOM and Size Reference Design
  • 230V, 900W, PFC with 98% Efficiency for Inverter Fed Drives Reference Design. 
• Get to know the UCC28051, UCC28063, UCC28019A and UCC28070
• Read more of Bosheng's blogs on PFC:
  • It is not just a PFC controller, it is also a power meter
  • How to reduce PFC harmonics and improve THD using harmonic injection: Part 1, Part 2
  • How To: Improve power factor and THD using DFF control