Improving the Performance of Flyback Power Supplies

For the PDF version of this article, click here.

The “lowly” flyback switched-mode power supply (SMPS) is still a workhorse in the power supply business for requirements below 100 W, but the requirements for a good design are getting tougher all the time. Energy-savings initiatives are now putting pressure on both mid- and full-load efficiency, and the requirements for ultralow standby power have reached new lows (Table 1) and are decreasing even further.

For example, just throwing a larger primary-side MOSFET at the problem to reduce the I² × R losses due to high peak currents in a discontinuous conduction mode (DCM) flyback SMPS can be counterproductive. That's because the fixed switching loss components due to output capacitance C_OSS pumping can make for real problems in controlling standby power or ultralight load operating losses. Yet requirements for improved power density even in low-cost applications would seem to mandate cooler running SMPSs operating at higher switching frequencies.

Protection functionality is also critical to building cost-effective power supplies, because any need to oversize components to deal with overload stresses has an adverse impact on cost. However, the alternative can be to compromise reliability, a factor we seem to take for granted in modern power electronics.

Table 1. European Commission Code of Conduct Standby Power Requirements.Rated Input PowerNo-Load Power Consumption Phase 1
1.1.2001 Phase 2
1.2.2003 Phase 3
1.1.2005 0.3 W and < 15 W 1.0 W 0.75 W 0.3 W 15 W and < 50 W 1.0 W 0.75 W 0.50 W 50 W and < 75 W 1.0 W 0.75 W 0.75 W

In this article, we'll take a look at these challenges in the context of a 60-W to 80-W flyback SMPS design, and investigate how to satisfy conflicting requirements for density, efficiency, hold-up time, reliability and cost. These challenges can be addressed with some innovative component technologies in MOSFETs and flyback controllers. We'll take an in-depth look at a few key design points for the power stage and how component behavior affects our ability to hit challenging performance, size and cost targets. We'll also see how some innovations in controller design aid in meeting standby power targets as well as robust protection.

Further Flyback Transformer Reading:

Full Flyback Transformer Spotlight

Performance Challenges

The outer envelope of the operating specification requirements for the flyback SMPS will set the boundary conditions for the design and the component challenges. The combination of the maximum output power requirement, plus the worst-case low-ac-line voltage and cycle-skip hold-up time requirements will be crucial. These factors will determine the maximum operating current for the power MOSFET, as well as the requirements for the flyback transformer and the rectified high-voltage dc bus capacitor.

Pivotal to these requirements is the selected minimum dc voltage for the flyback converter to operate without losing regulation. Key losses include the conduction loss of the MOSFET, which is aggravated by the high peak current in DCM operation relative to RMS power, and the switching crossover loss of the MOSFET, which is similarly increased by the high peak current at turn-off.

On the other side of the operating envelope, no-load operation at high line will incur losses in the MOSFET due to the combination of parasitic switching losses. Those parasitic-related losses are a function of the square of the rectified line voltage and the C_OSS output capacitance of the MOSFET, the transformer capacitance and the reflected capacitance of the secondary-side rectifier diode. Combined with the normal housekeeping requirements for the controller IC, the losses in the MOSFET make hitting standby power targets very difficult without taking special measures. In principle, meeting performance targets at both ends of the spectrum requires a switching FET with low R_DS(ON), as well as very low output capacitance and gate charge.

Let's examine a specific design example, one not too different from what is encountered frequently in practice. We'll take on an older 60-W flyback design (Fig. 1) using through-hole components, and upgrade it to an 80-W design with surface-mount components with higher efficiency and very low standby power. Our target here will be to completely eliminate the heatsink for the primary-side MOSFET switch by lowering the conduction and switching losses, while delivering significant reductions in standby power.

80-W Flyback Design

The main performance requirements for this design are spelled out in the load and line specifications. In this case, the output voltage is typical of adapter requirements for notebook computers and printers.

Minimum ac input voltage: 90 Vac
Maximum ac input voltage: 265 Vac Output voltage: 16 V
Maximum output overshoot, full load to no load: 250 mV
Maximum output power: 80 W
Target efficiency η at full load: 80%+
Hold-up time at 115-Vac drop, full load: 20 ms
No-load standby power, 90 Vac to 265 Vac: < 0.5 W

The maximum input power is a function of the output power and converter efficiency:

To design for low line and cycle-skip hold-up time requirements, a minimum dc bus regulation voltage target must be selected, and the dc bus filter capacitor requirements must be calculated. The lower the minimum dc bus regulation voltage, the smaller and lower the cost of the bulk bus capacitor. However, this places higher requirements on the MOSFET and transformer, as higher peak primary current is required. For this design, the minimum dc bulk capacitor voltage regulation target V^DCmin = 90 Vdc. For 20-ms holdup from a nominal ac line voltage of 115 V, first calculate the peak bus capacitor voltage at line dropout:

Then the required hold-up capacitance C_BULKnom is calculated from:

Selecting the nearest-size standard value, the high-voltage dc bus capacitor will be 220 µF. Next, the peak and RMS primary side current can be estimated, which will guide the MOSFET switch selection. The peak primary current I_{PRI lpk} is a function of the required input power, duty-cycle limit of 0.5 for DCM operation, and minimum bus regulation voltage V_DCminPK:

And the primary RMS current can be calculated from:

This value is an aid in estimating the MOSFET and transformer primary conduction losses.

Since the introduction of superjunction technology using the charge compensation principle for the MOSFET drain region^[1,2], significant advances have been made in chip size and parasitic capacitance of high-voltage MOSFETs. Fig. 2 illustrates the intrinsic conventional epitaxial drift region limitation versus voltage (red trace), usually referred to as the “silicon limit line.” The area-specific R^ON of the original 14-µm cell pitch superjunction transistor is shown for blocking voltages of 500 V through 800 V in the yellow plot line, and is characteristic of the CoolMOS C3 and S5 processes. The newest superjunction technology is realized with a 7.5-µm cell pitch technology (blue plot line). This technology achieves substantially lower area-specific R^ON with a value of 2.4 Ω/mm² at 625 V.

Corresponding improvements have been made in device capacitance, both the C_OSS output capacitance and the gate input capacitance, which determines gate charge. This comparison is shown in Table 2, which illustrates the difference in characteristics for state-of-the-art low gate charge conventional DMOS 600-V devices (IRFPC60LC), a mainstream superjunction transistor (SPP11N60C3), and a state-of-the-art 600-V, 7.5-µm pitch transistor (IPD60N385CP). The latter extends an R_DS(ON) class to the TO-252 (DPAK) surface-mount package, which in conventional technologies required a TO-247 package, while reducing dynamic losses related to C_OSS (Fig. 3) and gate charge (Fig. 4) even in comparison to existing superjunction MOSFETs.

Table 2. Comparing ~380-mΩ, 600-V FETs manufactured in different process technologies.Max R_DS(ON) (mΩ) Package C_O(ER) (pF) Q_G (nC) IRFPC60LC 400 TO-247 ~125 120 SPP11N60C3 380 TO-220 45 45 IPD60N385CP 385 DPAK 36 17

Let's examine the predicted loss situation with the IPD60N385CP. Conduction losses for worst-case operation can be estimated using high-temperature values for R_DS(ON):

For the DCM flyback converter, turn-on is at zero current and is nearly lossless. Turn-off losses are the primary concern, and are a function of turn-off switching time, snubber networks and the device technology. The strongly nonlinear capacitance, which contributes to the higher C_OSS energy at low voltages (Fig. 3), also acts as a nonlinear snubber during turn-off, making the calculation of turn-off losses inexact. For the flyback supply, they could be estimated using:

Here I_SWpk is the peak current on the primary at turn-off, f_S is the switching frequency (110 kHz in this design), and V_DCtypPK is the dc bus voltage at low line. These calculations suggest a worst-case MOSFET power dissipation under 2 W, which is highly amenable to surface-mount operation with a small foil area (~6 cm²) on the MOSFET drain for the heatsink.

From the peak primary current and target switching frequency (set by the controller), the primary inductance requirement of the flyback transformer can be calculated:

Assuming a maximum flux density of 0.12 T to 0.3 T (0.2 T, typical), a core can be selected from vendors' data sheets, considering core area, ambient cooling conditions, material behavior at high frequencies, winding window, etc. Many cores might be suitable for this application, such as an ETD30 or ER35W, but with parts on hand in the lab, I chose a PQ2625 from Magnetics Inc. (Pittsburgh). The core set gap must be chosen for an AL product that supports a reasonable turns count for magnetic coupling primary-to-secondary, and for the target inductance to achieve the required output power while keeping the flux within reasonable limits at a duty cycle at or below 0.5. In practice, this can be an iterative process, evaluating off-the-shelf gap selections, calculating the AL product, the required turns for the target inductance and the estimated flux excursion. With a gap around 2 mm, the PQ2625's A_L=55 × 10^-9. The number of primary turns can then be calculated as:

To be able to reach the required peak primary current, the turns should be rounded down, then, considering the estimated peak diode forward voltage V_VDIODE, reflection voltage on the primary V_RMAX (100 V), and V_OUT, the secondary turns for the first cut, can be calculated:

Now, this is another point that is somewhat “iterative,” as the secondary side current hasn't been calculated or the diode selected. Due to the right triangle shape of the current waveform, the peak current is quite high, as are the I²R losses, so minimizing the conduction loss in the rectifier diode is a key factor in achieving good overall efficiency. The peak semiconductor stresses become quite high as the power level increases.

Next, the reverse voltage for the rectifier diode (V_RDIODE), and the diode's peak and RMS currents are calculated:

Considering voltage overshoots due to parasitic inductance, a rating of 100 V or more is needed:

While low-cost fast diodes like the MUR1520 are often used in this application, a 100-V Schottky diode is a choice to consider if reducing the output diode V_F by 200 mV is attractive or necessary for hitting efficiency targets. In this design application, a 100-V Shottky diode was used (Fig. 5).

Before considering the controller, one other critical element for the flyback design is the output capacitor selection. Output capacitors are highly stressed in flyback converters. Normally, the capacitor will be selected for three major parameters:

Capacitance value, related to controlling voltage overshoot in the case of switch-off at maximum load condition (typically 10 to 20 clock cycles at f_S).
Low ESR, to meet output voltage ripple requirements as a function of the actual output current ripple current delivery.
Ripple current rating, to handle the internal dissipation and heating as a result of the high ripple current at the output of DCM flyback SMPS.

All of these criteria must be met in selecting output capacitors. For this design, the targeted maximum allowable voltage overshoot was set at ΔV_OUT=0.25 V, and to be safe, the number of clock periods for loop overshoot control was set to n_CP=20, while the maximum output load current:

Then, the required C_OUT can be calculated:

The required ESR can be estimated from the peak diode current and desired ripple voltage, where:

Keep in mind that a further spike ripple filter may be employed after the primary output filter caps.

Flyback SMPS Controller

The ICE3DS01 controller selected for this application is a stand-alone PWM controller for flyback applications. This controller is part of a family developed from the controller architecture of the third-generation integrated PWM + Power transistor CoolSET products.^[3] The development for this controller was focused on reducing standby power consumption through several measures.

For instance, an integrated high-voltage MOS startup cell is used, eliminating the losses from startup resistors. The startup cell turns off once the V_CC is within the normal regulation window from the auxiliary power winding. An active burst mode is employed for regulation at low output power levels; this is done with active control within the feedback loop, and ensures maintaining an accurate output voltage while operating the PWM switch at a very low, effective duty cycle. In Fig. 6, while operating with regulated output voltage at no-load, the burst pulse repetition rate drops to a little over 100 ms. Reducing the time scale to 2 µs/div in Fig. 7 reveals details of the gate and drain switching waveforms.

For comparison, sample switching waveforms are shown for the 60-W flyback SMPS using the SPP07N60 MOSFET (Fig. 8) and the 80-W flyback SMPS using the IPD60R385 (Fig. 9), both running at full load (60 W and 80 W, respectively) at 115 Vac. Due to similarities in the resonant frequencies from transformer leakage inductance and C_OSS, other than switching times (which are lower for the IPD60R385CP), the waveforms are very similar.

The efficiency of the 80-W flyback SMPS from Fig. 5 was measured at selected line voltages, for 20-W, 40-W, 60-W and 80-W output power. The results are shown in Fig. 10. Efficiency was over 80% for any of the tested conditions, comfortably exceeding the target requirements. Standby power was also checked at 110 Vac and 265 Vac. In both cases it was difficult to measure standby power with high accuracy because it was so low, apparently below 120 mW. This is consistent with other application boards using the ICE3DS01, which for 20-W to 40-W output are usually under 100 mW.

References

Deboy, G.; März, M.; Stengl, J.; Strack, H.; Tihanyi, J.; Wever, H. “A New Generation of High Voltage MOSFETs Breaks the Limit of Silicon,” pp. 26.2.1-26.2.3, Proc. IEDM 98, San Francisco, December 1998.
Saggio, M.; Fagone, D.; and Musumeci, S. “MDmesh: Innovative Technology for High Voltage Power MOSFETs,” pp. 65-68, Proc. ISPSD 200, Toulouse, France, May 2000.
Zoellinger, H. “CoolSET ICE3DS01 Current Mode Controller for Off-Line Switch Mode Power Supply,” Application Note AN-SMPS-ICE3DS01-1, Infineon Technologies, AG.