Advertisement

Blog

Designing for Power Density and High Reliability

The quest for the highest efficiency, reliability and lowest footprint are at the top of the power designer’s priority list. Optimising the efficiency of power converters not only saves energy and minimises the customer’s environmental footprint, but in conjunction with enhanced thermal performance, improves reliability and delivers lower cost of ownership. As a generally accepted rule, for every 10o C rise in operating temperature, life expectancy decreases by around 50%. On the other hand, reducing operating temperature by 10o C can double life expectancy.

The major contributor to excessive operating temperature is internal heat dissipation due to inefficiency in the conversion process. In effect, the operator pays twice for inefficient power conversion: every watt dissipated is another watt that must be cooled to keep the ambient temperature within specified limits. Clearly, improving energy efficiency can improve reliability and also reduce operating costs by reducing system-cooling demand.

Power/ Current density

A vitally important figure of merit (FoM) for power converters is current density. Higher current density means smaller devices for a given power rating, which ultimately allows system architects to utilise more of the valuable board real estate for revenue-generating devices such as processors, ASICs or FPGAs that add commercially attractive functionality.

As a result, designers need a converter to be simultaneously smaller, more energy efficient, with excellent heat-dissipation properties. Dealing effectively with heat is critical to maximising current density and power-handling capability. It’s just not acceptable to pack in the power components into their designated (and usually severely limited) space on the board without paying attention to these factors, otherwise the spectre of field failures will certainly ensue.

Solving the issues – digitally

Digital technology helps considerably to overcome the challenges facing power conversion designs today. Digital converters can be smaller since they require fewer components than a conventional analog converter topology which also helps to boost reliability. In a digital converter the output voltage is sensed in the same way as in an analog design, but there is no error amplifier. Instead, the sensed voltage is digitised by an analog-to-digital converter, and the digitised values are input to a control algorithm hosted on a microcontroller. A variety of algorithms are available to optimise performance as operating conditions change.

Figure 1 illustrates the main functional blocks of a typical digital converter.

Figure 1

Digital power conversion simplifies circuit design and reduces component count

Digital power conversion simplifies circuit design and reduces component count

Figure 2

Digital power converters can deliver significantly higher efficiency at light loads

Digital power converters can deliver significantly higher efficiency at light loads

Ericsson’s 3E single-phase PoL (Point-of-Load) converters feature advanced energy-optimisation algorithms to maximise efficiency across the whole range of loads. Also, with a specific input voltage, output voltage, and capacitive load, these converters permit the control loop to be optimised for robust and stable operation. This minimises the amount of output decoupling capacitance required to achieve a given load-transient response, thus delivering optimised cost and minimised board space. In effect, this simplifies hardware design, reduces overall module size, and helps to boost reliability. Figure 2 shows just how digital converter technology enables designers to maintain high efficiency at light loads, where traditional analog converters are often less efficient.

Importantly, these 3E PoLs utilise the latest-generation power MOSFETs, featuring low internal capacitance and optimal on-resistance x gate-charge FoM (RDS(ON) x Qg ) which minimises conduction and switching losses under all operating conditions.

The latest converter in the family, the BMR466, is capable of delivering up to 60A and has been demonstrated to achieve efficiency of 94.4% with 5V input and 1.8V output at half load. The MTBF of the BMR466 is calculated at 50 million hours – based on the industry standard ‘Telcordia’ method.

Up to eight of these converters can be connected together, allowing designers to rely on a single part number when powering their applications anywhere from 60A up to 480A. It is also possible to synchronise two or more of the devices with an external clock to enable phase spreading, which helps lower input ripple current and effectively reduces capacitance requirements and efficiency losses.

Advanced thermal performance

The internal design of the DC/DC is optimised to achieve the lowest package profile. Its height of 7mm (0.276 inch) minimises interference with cooling airflow across the board. Moreover, the solder pad distribution of the LGA package provides excellent thermal performance and enables the module to dissipate heat efficiently while at the same time benefiting from an extremely compact footprint of 14mm x 25mm (0.98 x 0.55 inch). The LGA contacts are positioned symmetrically and offer extremely low inductance connections, which ensures superior mechanical contact and high reliability after soldering. Internally, the package technology eliminates connecting leads and their associated inductance. Additionally, a high number of the LGA contacts are designated as ground pins. Together, these design advances ensure outstanding noise and EMI characteristics.

Optimising thermal performance is key to minimising thermal derating which allows higher output current operation without sacrificing reliability. Figure 3 shows thermal derating curves for this DC/DC with an output of 1.0V, capable of delivering the maximum 60A at ambient temperature of 70o C using only natural convection cooling. For an ambient air temperature of 85o C, the converter can deliver 50A cooled with natural convection, or 55A with an airflow of 1.0m/s.

Figure 3

Thermal derating for BMR466 PoL converter with 1.0V output

Thermal derating for BMR466 PoL converter with 1.0V output

The derated current capability of the device is comparable to that of competing PoLs that require more than twice the surface area and occupy nearly four times the volume, even though these have higher maximum current ratings. Consider the derating curves for a competing 80A PoL, shown in Figure 4, which shows that the higher-rated converter has a real-life limit of 62A for ambient air temperature of 70o C with natural convection with a 1.0V output. While this competing converter can deliver up to 60A at 85o C ambient temperature, with 1.0 m/s airflow, this is only marginally higher than the 55A available from this DC/DC operating under the same conditions even though the BMR466 is significantly smaller.

Figure 4

Thermal derating curves for alternative 80A PoL with 1.0V output

Thermal derating curves for alternative 80A PoL with 1.0V output

The physically larger 80A PoL has little more than one-third the current density of the device. Considering that a single computing board for an application such as a data centre server may need several – sometimes 10 or more – high-current PoLs like this DC/DC, the cumulative space savings that can be achieved, without derating the maximum current or degrading reliability, deliver a clear and valuable advantage to the designer and developer.

Design tools

Ericsson’s comprehensive Power Designer software enables the digital converter to be configured via a user-friendly GUI (Graphical User Interface) to ensure optimal efficiency and performance, and optimised control loop that minimizes the amount of capacitance for a given transient response. Ericsson’s Power Designer gives system architects complete control over parameters such as switching frequency and threshold settings for input and output under/over-voltages, output over/under-current limits, over/under-temperature and the implementation of phase spreading to ensure optimum efficiency over a wide range of operating conditions. The phase spreading, as well as other sophisticated features in output tracking and sequencing, can be accomplished with just a few keystrokes using Ericsson Power Designer.

Summary

It’s no secret that the industry acceptance of digital power has driven huge advances in converter efficiency, flexibility and economy. But this, in conjunction with vastly improved thermal design, is enabling power system designers to meet the often complex and conflicting market demands for the smallest footprint with the highest current delivery and greatest reliability – historically, a massively challenging design task.

1 comment on “Designing for Power Density and High Reliability

  1. sroochi
    April 28, 2016

    I fully agree with your post.  The hard part is to achieve >95% efficiencies at voltages lower than 4.7V.  Last time I checked, Vcc=4.7V for various ICs have been designed as a standard approximately ~ 20 yrs ago.  Today's Vcc= 0.6V, 0.8V, 1.2V,  and possibly 1.8V to 2.4V??  The only way is with either buck or (LDOs) with less than 90% to ~ 50% efficiencies.  

    How about fully isolated conversions?  Going from a noisy ground of +12V to +48V to a quiet  1.2V Vcc?  Do you have products/solutions for them?

    All the best to you,

    Sam Ochi

     

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.