How SMPS evolution has sparked a PSU design revolution
Back in the tumultuous days of 1976 when “power to the people” advocated revolution over evolution in human affairs, Swedish engineer Magnus Lindmark kindled a PSU design revolution with a resonant circuit that increased the speed of switch-mode power supplies (SMPS) by an order of magnitude. Power bipolar-junction-transistors (BJTs) hampered switching speeds at a time that predates commercial use of power MOSFETs by several years, choking SMPS transient-response performance in particular. Yet six years after the US authorities published Lindmark’s patent in June of 1978, linear regulators continued to dominate the power-supply landscape with switchers normally serving only applications that demanded more than 250W. As an era of cheap energy supply burned itself out, heat dissipation remained linear’s major issue with supplies routinely contributing 60% of their input energy toward global warming. As for low-power SMPS, cost and complexity restricted uptake to military and aerospace projects needing minimal bulk and power consumption. Exceptions include design classics such as 1972’s HP-35 pocket scientific calculator and the Apple II personal computer of 1977, both using switchers to efficiently generate multiple rails within minimal area.
Pre-Lindmark, a representative switcher ran at around 20 kHz leaving it with, perhaps, a mere 2 kHz of open-loop bandwidth to maintain regulation. But line-frequency power transformers and heat management measures are heavy consumers of aluminium, copper, and iron whose costs were continually rising. A shock price increase as demand outstripped supply contributed toward silicon vendors developing SMPS control ICs to rationalise design. In turn, the growth in SMPS uptake encouraged investments in silicon and design methods that saw linear power supplies relegated to local low-current provisioning, serviced by a people’s army of three-terminal regulators; or to niche low-noise applications in data-acquisition and audio/RF power amplification. Power MOSFETs arrived to solve multiple BJT issues – such as device matching – while furnishing greater efficiency and reliability, setting the course for a stream of evolutionary improvements to SMPSs that we now see on a plateau. Accordingly, another revolution is taking hold as analogue control methods finally cede to programmable mixed-signal technology, with “smarts” appearing – such as on-chip firmware that minimises power consumption by continually adapting a converter to its environment as it runs. Naturally, the output powertrain remains analogue and may even be identical between analogue and digital converters of similar output power.
Resonance boosts switching speed by an order-of-magnitude
While today’s SMPS designers expend massive efforts to extract an honest 1% improvement in virtually any parameter, early designs were lauded for efficiency metrics of as much as 80%, which was double that of many linear regulators. The path to maturity is way-marked by patent applications such as Donald A. Paynter’s "bridge-type transistor converter", a self-biasing multivibrator (as shown in the figure) from US patent 3,080,054 of March 5th, 1963:
Figure 1. Paynter’s multivibrator bridge converter (US patent #3,080,054 of March 5th, 1963)
More than a decade passed before Lindmark’s design circumvented fundamental issues with hard-switching at relatively high voltages by creating a resonant converter that soft-switches the power transistors at close to zero volts. At a stroke, this refinement mitigated electromagnetic compatibility (EMC) issues and relieved component stresses, to allow a maximum switching speed of around 200 kHz using BJTs of the period, together with a proportionate times-ten improvement in transient-response performance. Unlike today’s nonisolated low-voltage converters that may run at several MHz, a few hundred kHz remains good for many AC/DC front-ends and intermediate-bus converters (IBCs) regardless of switching method. This band hits a sweet-spot between the characteristics of everyday capacitors, magnetics, and power MOSFETs that with proven design delivers efficient and affordable power conversion that’s relatively easy to filter for EMC artifacts.
Unlike hard-switched topologies such as buck regulators and flyback converters that operate fixed-frequency pulse-width-modulation (PWM) control schemes, resonant converters depend upon switching-frequency changes to maintain regulation. Benefits should include making it easier to meet EMC regulations by distributing lower peak emissions across a wider bandwidth. Crucially, Lindmark’s switching cycle passes through zero, administering a sharp turn-off kick to cut power-switch losses that impact switching speed and efficiency. His revelations appear in US patent number 4,097,773 of June 27th 1978, where the lower sketch in figure 2 reflects subtle adjustments to the ‘prior art’ sketch above.
Figure 2. Lindmark’s modifications to the push-pull converter (top sketch) exploit resonant techniques (lower sketch) that enable a 10x increase in switching speed (US patent #4,097,773 of June 27th, 1978).
Fortune favours the brave
It was at a similar time to Lindmark’s work being published that telecom infrastructure provider Ericsson decided to design a high frequency on-board DC/DC converter for its new distributed power system architecture. By 1983, Ericsson Power Modules was ready to launch its first products, the miniaturized 25-40W DC/DC power modules for 24 and 48 VDC systems. [photograph at the head of this article; the original PKA “blue module” from 1983] Running a push-pull respective half-bridge converter with power MOSFETs at a rapid 300 kHz, reducing the size of capacitive and inductive components, the PKA series captured attention with its small size and use of aerospace-inspired thick-film technology on ceramic substrates. Requiring no external support components, a metal case provided a thermal path to route heat away from the device in all six planes and dispense with expensive, heavy, and potentially toxic thermal-encapsulation materials. A five-year warranty exuded confidence in a 2-million-hour mean-time-between-failure (MTBF) calculation for 40W full-power operation at 85°C maximum ambient temperature. Typical 82% efficiency vaulted the PKA series over the 80% hurdle to become the first entry in Figure 3’s landmarks that reflect the state-of-the-art for modular power over the 30 years to 2013.
Figure 3. Efficiency improvements reflect 30 years of development in modular power.
The PKC series followed in 1987, adapting PKA’s design for half the power output. The series raised the efficiency bar 3% to 85%, overcoming the tendency for lower output power levels to degrade efficiency as uncompensated conversion losses assume a greater proportion of the energy budget. As figure 3 plots, a 1% efficiency improvement is now next-to-impossible to achieve in top-flight analogue converters that are intended for mass production, signalling the practical limit for this technology. Radical improvements to circuit techniques and/or component technology are the only way forward, and is most often fulfilled by the state of continuous revolution that characterises the semiconductor industry.
The future is digital?
Fast-forward to 2006, when Ericsson questioned the potential for digital inner-loop control to pave over efficiency roadblocks and provide customers with a solid development path for the future. Following heavy investment primarily in software/firmware development, 2008’s release of the BMR453 was definitively revolutionary: at a stroke the digital IBC hit 96% efficiency over a far wider envelope than its predecessors; at 396W, it doubled the power density of ±2% tightly-regulated quarter-brick IBCs; and while capable of analogue-style standalone operation, it furnished a comprehensive PMBus control and monitoring subsystem to become a true systems-capable component. Test boards and graphical-user-interface development software quickly followed to provide unprecedented flexibility and development ease within a digital workflow that can service the logistics of end-product manufacture through to end-of-life decommissioning.
Completing analogue’s overthaul, the BMR453 parents a family of systems-compatible IBCs and point-of-load regulators (POLs) that is currently in its second generation. Evolutionary improvements such as onboard firmware that fine-tunes a converter’s core in realtime to maximise efficiency within its environment extend the adaptive dead-time capability that largely accounts for digital’s greater efficiency over a larger portion of the loadline. Unlike the earlier comparatively static load model, these points are significant for demand-driven systems that employ power-saving measures such as slowing system clocks or disabling circuit blocks during periods of low demand. This is where customer-authored software will differentiate products, in the near-term exploiting themes such as dynamic-bus-voltage control to optimise energy use. Ultimately, all but the lowest cost modular power converters will be systems-capable and standards-compliant as a matter of course, and software will assume similar precedence as in other branches of industry.
Hardware trends include new research into 98%+ efficient converters that source kW+ levels from smallest-possible footprints. Several approaches vie for dominance that – like today’s intermediate-bus-architecture (IBA) – will ultimately win through global field trials. And we can of course expect to see materials technology contribute heavily, with the commercialisation of gallium-nitride (GaN) and silicon carbide (SiC) under way; graphene and its structures including carbon nanotubes are also on the near-term agenda.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
