Bulk Capacitor

In Figure 2-10, we Anally break up the input capacitance into a high-frequency capacitor and a relaAvely low-frequency bulk capacitor. The current distribuAons are shown, as well as how they all add up eventually. The mystery is clear now, and in the process we also understand how the decoupling capacitors are supposed to behave. Now we can also start to understand how this delicate balance can be easily shattered by lack of proper decoupling ... 

Figure 2-11 Input Ripple of a Buck Converter Showing the Effect of a Non-ideal Bulk Capacitor...

You may need to parallel several aluminum electrolytics to lower the ESR sufficiently, and you may also have to substantially increase the capacitance just to stay within the total 1% limit somehow. Also remember that the ESR of aluminum electrolytic capacitors gets significantly worse over time. So if you have a customer return after several months in the field, it may well be because of the aging of the electrolytic bulk capacitor Try replacing the capacitor and then recheck. 

Note that in the Buck and the Buck-Boost, the input capacitor is included in the critical path. That implies we need very good input decoupling in these topologies (for the power section). So, besides the necessary bulk capacitor for the power stage (typically a tantalum or aluminum electrolytic of large capacitance), we should also place a small ceramic capacitor (about 0.1 to IpF) directly between the quiet end of the switch (i.e., at the supply side) and the ground—and also as close as possible to the switch. 

Recommendation 5 (Figure 6-12) The bulk capacitor has been placed close to where the input supply leads came in, but the ceramic decoupling capacitor must be very close to the IC as indicated here. 

One of the problems we faced was a mounting standoff on the primary side that we just couldn t get around satisfactorily. We had EMI filters and other items taking up all remaining space. The only way out was to position the bulk capacitor and the switch on opposite sides of the standoff, as shown in Figure 7-2. Unfortunately, these long traces were... 

We did it somehow, almost strangulating ourselves in the process. Now when I look back at this incident, I wonder why we didn t place a ceramic decoupling capacitor close to the switch, as shown in the lower half of the figure. The bulk capacitor could have successfully managed to provide the low-frequency current components, whereas the high-frequency capacitor could have really decreased the effective loop area in which the high-frequency components were circulating. 

Once, while working as an Apps engineer at this maker of high-voltage Flyback monolithic switcher ICs, I happened to express concern about the way they were selecting the input bulk capacitors for their evaluation boards. To this day, in my opinion, they are making several misleading/erroneous recommendations to their customers via their eval boards. 

Many switcher ICs are in fact designed with a certain minimum on-time (especially the current mode control types). They also keep to the minimum pulse width until about 0.2 to 0.3V on the feedback pin. In such cases, with a reasonably large output bulk capacitor, you will see a huge inrush of current into the output capacitor, even before the latter starts to rise appreciably. You should also be aware that inrush current into the input capacitor of any topology is very high, and no switch action can even hope to prevent that. 

For intermediate time resolutions (of the order of r ) the bulk capacitor has become impermeable, and the boundary circuit is relevant for the time dependence (second term in Eq. (64)) term 1 is constant, while term 3 is still zero. Finally in the long-time regime, at t rs, the stoichiometric polarization occurs while both bulk and boundary responses constitute the initial voltage jump from U= 0 to 11 = If (/ , + R ) note that both corresponding capacitors are completely impermeable, i.e., terms 1 and 2 are constant. In the steady state (f rs) all the capacitors block, and R + Kon 316 obtained as the stationary resistance value. Obviously time-resolved dc experiments allow the partial conductivities and the capacitances to be measured together with the chemical diffusion coefficient (ts ccl/Cf). The switching-off behavior is analogous. 

---