Exploring Output Capacitor Impact on DC-DC Converter Performance

Selection of a suitable output capacitor plays an important role in switching-converter designs. “Some 99% of the ‘design’ problems associated with linear and switching regulators can be traced directly to the improper use of capacitors,” claims the National Semiconductor IC Power Handbook^[1]. The output capacitor is so important in switching dc-dc converters because it is (together with the main inductor) the reservoir of energy flowing to the output, and it also smoothes the output voltage.

Frequency dependence of capacitance ESR and stability with operational temperature and dc bias voltage are the important parameters of output capacitors that define performance and functionality of the complete power system. For the purpose of benchmarking, these key parameters are measured for different capacitor technologies. A notebook power-supply-converter evaluation kit based around Maxim's MAX1537 was chosen as the real application example for the evaluation of different capacitor technologies.

EVALUATION PARAMETERS

Initially, the frequency characteristics of capacitance and ESR of two capacitor groups were measured. The first group included different capacitors specified for the 3.3-V output evaluation kit, with capacitance at C = 220 µF. The second group contained capacitors for the 5-V evaluation kit where C = 150 µF. Electrical parameters were measured by an HP 4194A impedance/gain-phase analyzer^[2] in a frequency range from 120 Hz to 1 MHz (capacitance) and 120 Hz to 10 MHz (ESR).

Temperature stability of the converter is one of industry's most common concerns. Thus, the second measurement concentrated on capacitance and ESR stability with temperature and dc voltage bias. The 3.3-V output capacitor group was measured using the same impedance analyzer and a Keithley 7002 switch system across the dc bias voltage range of 0 to 4 V, conditioned in a Votsch VC 7018 laboratory oven over the temperature range of -55° to 125°C.

Maxim's MAX1537EV Kit^[3] converter was used for the benchmarking tests. The evaluation kit provides two power outputs, 3.3 V and 5 V, both with a maximum current I_out of 5 A. A photograph of the kit is shown in Fig. 1. The recommended output capacitance, C, for the 3.3-V output is 220 µF (position C6 in Fig. 2); for the 5-V output the value of C is 150 µF. Ac ripple voltage values and wave forms have been used as the main indicator of filtering quality. A Goldstar GP-505 stabilized power supply was used to supply the kit with a fixed input voltage V _in of 20V.

An output load was set up using resistors and capacitor to draw two thirds of the maximum current. (For the 3.3-V output this was a parallel combination of a 2.2-Ω resistor (R) and a 4.7-µF tantalum capacitor (C); for the 5-V output the value of R = 3.2 Ω (Fig. 3). Voltage waveforms and relevant ac Vrms (effective value) were displayed using an Agilent Infiniium 54830B digital oscilloscope ^[5].

CAPACITOR CHARACTERISTICS FOR 3.3-V OUTPUT

The graphs in Figs. 4 and 5 show the frequency characteristics of capacitors constructed with varying technologies used for the 3.3-V evaluation kit output, with a nominal capacitance of C = 220 µF. The exception is the MLCC, where two 100-µF devices were used. The capacitor technologies chosen were tantalum-polymer, tantalum (MnO ₂ with single- and multi-anode constructions), niobium oxide (MnO₂), multilayer ceramic, and aluminum-electrolytic.

In the case of tantalum-polymer and tantalum-MnO₂ multi-anode capacitors, there is a relatively small drop in capacitance in frequencies from 10 to 100 kHz (Fig. 4), whereas tantalum-MnO ₂ and aluminum-electrolytic capacitors exhibit a larger drop across the same range. The actual capacitance of the MLCC capacitor suffers due to its dependence on the dc bias voltage, which was applied during measurement. Fig. 5 shows the very low ESR performance of the MLCCs and relatively low ESR of the tantalum-polymer devices. The ESR of aluminum-electrolytic capacitors is relatively high over the complete measured frequency range.

Fig. 6 shows the frequency characteristics of the different capacitor types used with the 5-V output evaluation kit with a nominal capacitance 150 µF (except the 100-µF MLCC and aluminum electrolytic devices). Both tantalum single- and multi-anode capacitors retain a higher capacitance at higher frequencies (above 100 kHz), whereas niobium oxide and aluminum-electrolytic capacitors lose their capacitance faster at lower frequencies (Fig. 7). The MLCCs exhibit very low ESR around the 100-kHz frequency range; tantalum multi-anode and tantalum-polymer capacitors show low ESR in the same frequency range; and the aluminum-electrolytic capacitor has a high ESR over all frequency ranges.

CAPACITANCE STABILITY

Experiments showed that the best overall capacitance stability is exhibited by the tantalum-MnO₂-technology capacitor. The capacitance of niobium-oxide-MnO₂ devices is more sensitive to dc bias voltage and tantalum-polymer is more sensitive to temperature changes. The capacitance of the MLCC is very dependent on both actual temperature and dc bias, while aluminum-electrolytic capacitors are stable with dc bias but very temperature-dependent.

We can see that ESR is relatively stable vs. dc bias voltage for all capacitors. Differences can be seen when we compare ESR stability vs. temperature. Tantalum-polymer and MLCC capacitors exhibit the most stable ESR, whereas the ESR of MLCCs is very low over the whole temperature range. With tantalum-MnO₂ and niobium oxide-MnO₂ devices, ESR decreases as temperature increases. Aluminum-electrolytic capacitors behave differently — ESR grows to very high values at low temperature (below 0°C) due to the limitation of wet electrolyte conductivity at low temperatures.

Figs. 8 and 9 show the different waveform shapes that occur when various capacitor-types are used. Comparing tantalum-polymer and tantalum-MnO ₂ capacitors shows that the ripple voltage using the tantalum-MnO₂ device has a lower level of higher harmonic components for both 3.3- and 5-V outputs. The basic frequency of the ripple voltage is naturally equal to the switching frequency of the converter, f_sw = 300 kHz. When using MLCC capacitors, both 3.3- and 5-V circuits exhibited undesirable oscillations with frequency approximately f_osc = 50 kHz and high ac Vrms due to the regulator instability. Aluminum-electrolytic types did not perform well, as can be seen on the waveforms of both outputs measured by a relatively high ac Vrms.

TEMPERATURE EFFECT ON OUTPUT RIPPLE VOLTAGE

Aluminum-electrolytic and MLCC Vrms behavior across a wide Vrms range is displayed in Figs. 10 and 11. The figures also show a much smaller range in magnified scale. For both outputs — and for most of the capacitor technologies — the output-ripple Vrms decreases nearly linearly with increasing temperature. Aluminum-electrolytic and MLCC capacitors are exceptions due to their exponential change in capacitance and ESR with temperature (from Figs. 8 and 9).

Aluminum-electrolytic capacitors exhibit too high a level of ESR across the temperature range, so their smoothing ability is limited, as the output ripple voltage will be much higher than with other technologies. When MLCCs are used, the very low ESR-levels cause circuit instabilities, making output-ripple voltage also high. Among the other technologies we can observe that ripple voltage at the output will be lower when ESR is low capacitance at switching frequency is high.

COMPARING CAPACITOR SIZE

In our benchmark, tantalum-polymer and tantalum-MnO₂ low profile capacitors were the smallest suitable capacitors followed by niobium oxide-MnO₂ with the same footprint but a little taller. Aluminum-electrolytic radial-leaded capacitors require a much bigger footprint and are bigger in volume.

SUMMARIZING THE FINDINGS

Tables 1 and 2 summarize our findings. Furthermore, you can achieve low-output ripple voltage for a dc-dc converter by using output capacitors with low ESR at the switching frequency — in our case tantalum-polymer and tantalum-MnO ₂ multi-anode capacitors. The rate at which the actual capacitance decreases with frequency (relative to the resonance frequency) is also important. Tantalum-MnO₂ capacitors are recommended in applications with variable output voltages because they offer the best capacitance stability versus dc bias voltage.

It is strongly recommended that designers consider the capacitance and ESR temperature stability of output capacitors when deciding on the system's operating temperature. From this point of view, tantalum-polymer and tantalum-MnO₂ capacitors were found to be the most stable, whereas MLCC and aluminum-electrolytic capacitors offered the least stability.

The capacitance and ESR of the output capacitor can significantly influence the dc-dc converter regulator feedback loop, which defines the stability of the converter operation. These parameters have to be in a certain range to assure stability of the system.

In our experiments, MLCC output capacitors had too low an ESR (1 to 2 mΩ), which resulted in circuit oscillations and a relatively high ripple voltage. Therefore, MLCC devices cannot be recommended in our experimental study. The use of MLCC capacitors can be recommended only under careful evaluation of their low ESR versus stability of the loop.

Using generic aluminum-electrolytic capacitors resulted in high-output ripple voltage and poor filtering due to their higher ESR characteristics. This also significantly deteriorates at lower temperatures.

Based on our measurements, we can conclude that using low-ESR output capacitors such as tantalum-polymer and tantalum-MnO₂ — especially with multi-anode construction — leads to the best results measured by ac Vrms of output ripple voltage and Vrms temperature stability. MLCC and aluminum electrolytic technologies can be used as long as attention is paid to the instability (MLCC) and output ripple (aluminum). Good cost vs. performance value can be also achieved using niobium-oxide capacitors.