For the PDF version of this article, click here.
On Dec. 15, 2004, the California Energy Commission (CEC) amended the state's Appliance Efficiency Regulation. The revised regulation defines mandatory efficiency requirements for 19 groups of equipment, including items as diverse as water heaters, plumbing fixtures, lighting, exit signs, traffic signals, external power supplies (EPSs), and consumer audio and video equipment. The EPSs covered by the standard are single-output, self-enclosed units that reside outside of the product they power, with output voltages ranging from 1.5 V to 24 V. 
The new CEC standards are intended to reduce the amount of energy waste that is currently tolerated within the state. Ecos Consulting, conducting research for the CEC, found that more than 3.1 billion power supplies were currently in use nationwide with an estimated 1.3 billion of them being external chargers and adapters. [2, 3] Proportionally, about 145 million are in use in California.  About half of the adapters and chargers are highly inefficient linear-transformer-based models, and so the energy-savings opportunity from raising the efficiency standards on all EPSs sold in California is substantial.  It is estimated that implementation of the efficiency standards will save 96 million kWh per year — the equivalent output of a small power plant. 
Voluntary energy-efficiency regulations are in effect in other regions of the world. Many marketers consider it an advantage for their product to have the logos that denote compliance, such as the well-known U.S. ENERGY STAR label. However, the new CEC standards are generally stricter than some of the voluntary standards (Fig. 1). Therefore, to ensure that sufficient energy-waste reduction occurs, compliance with the new standards was made mandatory.
Additionally, many believe that California's mandatory compliance requirement will set a new precedence for the rest of the world. For example, the China Energy Conservation Project recently passed regulations similar to the ENERGY STAR program. If initial compliance with its standards are low, it also may make meeting them mandatory, to ensure that the desired level of energy conservation is achieved. 
The manufacturers of linear-transformer-based battery chargers and adapters, and the OEMs that power their products with them, will feel the greatest impact of California's new standards. Most chargers and adapters get plugged into an ac wall socket and are left there indefinitely, even when not in use. Because they are built around low-tech copper and iron line-frequency transformers, they typically waste more than a watt, even when they are disconnected from the device they power.
This no-load consumption is due to the transformers magnetizing hysteresis and the eddy currents that are induced in its core. Linear-transformer cores are made of a silicon steel alloy, because it has relatively high values of permeability and saturation flux density, moderately high resistivity, is fairly immune to asymmetrical magnetization and suffers only a small percentage change in its effective permeability from the usual air gaps in the magnetic path.  To minimize the flow of unwanted eddy currents in the core, it is made of thin sheets that are insulated from each other. Even so, the amount of power dissipated in most linear-transformer cores prevents them from meeting the CEC's 500-mW no-load consumption specification.
The Death of Linears?
California's new standards have fairly demanding requirements for active mode operating efficiency and no-load power consumption (Table 1). Starting in July 2006, compliance will require a maximum no-load consumption of a half-watt from all EPSs that deliver less than 10 W. In January 2008, the half-watt limit will be extended to all EPSs up to 250 W.
However, even if a linear supply could meet the no-load limit, the active-mode efficiency of the series pass transistor in a linear power supply cannot exceed the ratio of its output voltage to its input voltage (VOUT/VIN). Therefore, few linears, if any, will meet the active-mode efficiency specification. Attempting to make linears compliant would result in a cost penalty that would make equivalent switched-mode power supplies (SMPS) more economical. Therefore, California's new standards have effectively made linear-transformer-based chargers and adapters obsolete.
With the July 1, 2006, compliance date less than 18 months away, manufacturers must redesign noncom-pliant EPS models, obtain certification from the required regulatory agencies and start mass production, all within a short time frame. This could be challenging for those who do not have expertise in SMPS design. However, recent developments in integrated power conversion IC technology can provide assistance in meeting the compliance deadline.
Many factors affect power-supply efficiency (η), such as output voltage and current, the converter topology, the operating input-voltage range and its bill-of-materials (BOM) cost constraints. However, when considering how to make a given power supply more efficient, two different modes of operation must be considered: no-load to light load, and the active or “ON” mode.
It is valuable to identify the sources of loss within a typical SMPS, and its load and/or line dependence. This is important because reducing load-independent losses can help increase both no-load and active-mode efficiencies, while reducing load-dependent losses rarely helps reduce no-load losses and can actually increase them. Additionally, some line-dependent sources of loss may not be obvious.
There can be a dozen sources of loss within a typical SMPS, depending on its power level, topology and circuit complexity (see Fig. 2). The approximate loss contribution range of each is shown, as a percentage of the input power, for a typical supply that delivers 10 W of load power.
In a well-designed supply, the losses usually split about 50-50 between the primary and the secondary, with the two biggest offenders being the MOSFET switch and the output rectifier. The dissipation in the output rectifier is primarily load dependent. The MOSFET's conduction losses are load dependent, but its switching losses are independent of the load. The power transformer and the output filter circuit account for the next largest losses, with the filter tending to be more load dependent than the transformer.
Although the other losses are much smaller, their contributions detract from overall supply efficiency and can significantly impact light-load and no-load operation in sub 10- W supplies. Because BOM cost reduction is the goal of most power-supply designers, the elimination of X capacitors can reduce their cost and eliminate their energy consumption. Although their loss contribution is small, if a supply can meet EMI with fewer, smaller or preferably no X-caps, both their cost and their loss contribution is reduced or eliminated.
The PWM control IC is indirectly responsible for two sources of load-independent loss. The first is its required start-up current. Typically, a string of resistors feeds current to the IC from the filtered dc input. If that current is not eliminated after the IC is receiving sufficient current from its operating bias source, the power dissipated in that string is simply wasted. The IC's MOSFET gate-drive current will be discussed later, when the MOSFET switching losses are addressed.
The dissipation in the primary-side clamp depends on the transformer leakage inductance and its turns ratio, both of which need to be optimized for each power supply. Power transformer optimization requires years of experience or the use of one of the newest of CAE tools for designing transformers, such as PI's Transformer Designer. The dissipation in the output snubber depends on the type of rectifier chosen, the duration of its reverse-recovery time and the value of the transformer's leakage inductance. Typically, the higher the leakage inductance, the more the rectifier will ring as it turns off, making the snubber circuit dissipate more energy. Therefore, transformer optimization also is critical for keeping snubber dissipation low.
Input filter capacitor loss is due to the product of its ripple current and its ESR. The remaining contributors, such as EMI filter chokes, input rectifier diodes and the output feedback circuit, all make relatively small contributions to the inefficiency of the supply and little can be done to dramatically reduce their losses.
Tips for Meeting CEC Standards
As previously mentioned, eliminating or reducing the number or size of X capacitors is an easy way to lower cost without lowering efficiency. In supplies that deliver 10 W or less, this can be important since the other small contributors do not scale down proportionally as the output power level drops. However, meeting EMI without X capacitors can be difficult.
On the secondary side, output rectifier losses can be reduced several percentage points by using synchronous rectification. However, synchronous rectifiers can substantially increase the cost of the supply and its no-load consumption, due to their MOSFET gate-drive losses. Because many of the supplies that will need to be redesigned or replaced are in the under 10-W power range, the increases in cost and no-load consumption that come with synchronous rectification make it an impractical solution for improving efficiency at low power.
However, designing the power supply around a highly integrated power-conversion IC is something that can be easily and cost effectively done to reduce a number of losses. ICs, such as Power Integrations' TOPSwitch-GX, have a modulated-switching frequency that spreads the switching noise over an 8-kHz range, which can effectively lower both the quasi-peak and the average EMI readings significantly. This reduction in EMI often allows one or more X capacitors to be eliminated. Even if all of the X-caps cannot be removed, it usually will allow a lower value of capacitance to be used, which lowers the cost of the capacitor and the amount of power it consumes when ac is applied to it.
As mentioned earlier, the MOSFET switching losses are not load dependent. The power lost when the MOSFET switches is mathematically described as:
PSWLOSS = ½CP · V2 · FSW
where CP is the parasitic capacitance on the drain node, V is the average dc value of the rectified input voltage and FSW is the switching frequency. From this equation, it can be seen that the one variable that can be most readily and easily modified to compensate for load changes is the switching frequency. Because half of the MOSFET losses and all of the transformer and output diode losses are load dependent, they are reduced significantly as the switching frequency is lowered in light to no-load operation.
Although the switching frequency of a PWM IC driving an external MOSFET can be made lower as the load lightens, this technique requires additional components, design effort and verification to ensure proper operation. The newer power-conversion ICs typically have a switching frequency-reduction feature built in.
Another factor mentioned earlier was the considerable dissipation that can come from the resistor string that provides start-up current to the control IC. Again, additional circuitry can be put in place so that this current can be interrupted once the IC receives sufficient current from its bias supply. However, some of the new chips have a built-in bias current source that is disabled when it is not needed, reducing circuit design and verification time.
One more reason to consider a highly integrated power-conversion IC to improve power-supply efficiency is the consumption of the chip itself. For most PWM controllers, much of the current it draws is used to drive the MOSFET switch. Some IC makers integrate a lateral MOSFET with their controller, which requires far less gate charge than does a vertical MOSFET, due to its lower capacitances (Fig. 3a, 3b and 3c). This makes the IC consume less power, especially at no-load where the load independent losses have the highest impact.
Is SMPS Cost Effective?
Even if redesigning a power supply around a highly integrated chip can help make it meet the new CEC standards, can a SMPS cost effectively replace a small, low-power, inexpensive linear-transformer-based charger or adapter? A substantial amount of a regulated linear supply's inefficiency occurs in the energy that is intentionally “burnt off” in the pass transistor. For the purposes of this comparison, an unregulated linear supply will be compared to a SMPS that was designed to replace it.
Table 2 compares some electrical performance, energy-efficiency and logistical data on two units with equal output voltage and current ratings. The difference in annual energy costs at no-load demonstrates why stringent energy-efficiency standards are relevant today. Additionally, there are some hidden, shipping-related costs for the linear. Lastly, the SMPS has integrated, automatically resetting safety features that protect the supply, the load equipment and the user better than the linear's one-time thermal fuse. Fig. 4 depicts the simplicity of the SMPS replacement circuit, as it only uses 18 components.
Tighter energy efficiency regulations such as California's will likely become the norm. Small, cost-competitive, low-component count-switching EPSs can be designed to replace inexpensive, low-power linear chargers and adapters. Highly integrated, high-voltage power-conversion ICs can help achieve the no-load and active-mode efficiency an EPS needs to meet the new CEC energy-efficiency standards easily and cost effectively.
Ecos Consulting, Davis Energy Group, Energy Solutions, comps. Codes and Standards Enhancement Initiative for PY2004: Title 20 Standards Development Analysis of Standards Options for Single-Voltage External AC to DC Power Supplies, May 3, 2004, p. 3.
Ecos Consulting comp. Power Supply Efficiency: What Have We Learned? February 2004, p. 1.
Ecos Consulting, Davis Energy Group, Energy Solutions, comps. Codes and Standards, Market estimates made by Travis Reeder and Chris Calwell of Ecos Consulting, and Carrie Weber, LBNL, in August 2003. May 3, 2004, p. 6.
Ibid, p. 7.
Staff Report, comp. Update of Appliance Efficiency Regulations, CEC November 2004, p. 42.
Flanagan, William M. Handbook of Transformer Design & Applications, Second edition, McGraw Hill, 1992, pp. 6.11 and 6.13.
|Standards for Power Supplies, Effective July 1, 2006|
|Nameplat Output (W)||Minimum Efficiency in Active Mode of Operation|
|< 1 W||0.40 * Nameplate output (Value in watts)|
|≥ 1 (W) ≤ 49 W||0.09 * Ln (Nameplate output) + 0.49|
|> 49 W||0.84|
|Maximum Energy Consumption in No-load Mode|
|0 to < 10 W||0.50 W|
|≥ 10 to < 250 W||0.75 W|
|Where Ln (Nameplate output) = Natural logarithm of the nameplate output expressed in watts.|
|Output specification||2.7 W, 9 V||2.7 W, 9 V|
|BOM Cost||1 ×||1 ×|
|Input voltage||98 to 132 Vac||85 to 264 Vac|
|Full-load efficiency (115 Vac)||53%||75%|
|No-load input power (115 Vac)||1.6 W||200 mW|
|Annual energy cost (2.7 W load)||$5.34||$3.80|
|Annual energy cost (no-load)||$1.68||$0.22|
|Short-circuit current||2.3 A||50 mA|
|Short-circuit protection||One-time thermal fuse||Self-resetting auto-restart|
|Weight||9.4 oz/267 g||2 oz/56 g|
|Volume||11 in3/176 cm3||2.45 in3/40 cm3|
|Shipping cost by sea (per unit)||1 × (reference)||0.4 ×|
|Shipping cost by air (per unit)||10 ×||4 ×|