The DC Bus and Bulk Capacitor

(Last edited 10/10/2026)

Earlier chapters examined capacitors in detail, including the X and Y capacitors used in the PSU’s input filter. Now let’s focus on the large capacitor on the primary side. This cap is often called the bulk capacitor or the primary capacitor.

After the PFC stage, the DC must be both smoothed and stored for downstream circuits, and that’s exactly what bulk capacitors do.

As discussed in the ripple chapter, ripple can stem from residual line AC or switching artifacts, and bulk caps significantly reduce this unwanted variation. They also hold charge to ride through brief AC interruptions; the larger the capacitance, the longer a PSU can maintain output during a brown-out. This ride-through capability is expressed as the PSU’s “hold-up time,” measured in milliseconds at a specified load. For instance, ATX standards require a 16.6ms hold-up at 100 % load.  This is equivalent to losing one cycle on 60Hz mains (since 60 cycles occur per second).

As for the voltage used for the capacitor; I see a lot of people incorrectly assume that we’re only dealing with operating voltages within the PSU that are equal to that of the mains’ voltage. But in reality, we are working with voltages in the neighborhood of 380 to 420V. The reason for the higher voltage is to support the voltage at the highest peak, called the “Vpeak”, of the AC wave. Let’s say the AC voltage is 264V. The formula we use is “​Vpeak = VRMS x √2”. VRMS is 264 and we can round √2 to 1.414, we can calculate 264 x 1.414 as 373.296. This is where I would use at least a 400V capacitor. Obviously, we typically see bulk cap voltages much higher than 400V. This is because we always want a considerable amount of margin in our designs… unlike the design of the 12VHPWR connector.

Sometimes, the reason behind a bulk cap being considerably larger than necessary isn't as cut and dry as the engineer trying to meet a high hold up time. Some times, they may increase the capacitance in an effort to address a problem where there may be too much ripple on the AC bus. When using an LLC topology; if the engineer finds that their APFC design does not deliver perfectly constant power, there may be some low-frequency noise. If one can clean that noise up, the LLC converter can see more stable input voltage and require less frequency modulation, more consistent gain, less transformer current variation, and easier regulation.

Higher capacitance can also be good for the ATX 3.x requirements. If a GPU requires a sudden demand of power, what we call a "power excursion, the bulk cap can act as an energy reservoir and prevent the bus voltage from drooping.

Unfortunately, there is a trade off for higher capacitance, other than the obvious higher cost and taking up more real estate. Higher inrush current is the one that will always be the trade off that will make an engineer decide against increasing the capacitance of their bulk caps.

Before the PSU can be used, those capacitors need to be charged before the PSU is usable. And while there is no formal ATX requirement that says the bulk capacitor must reach a full charge within X seconds after the PSU receives power from the wall, the PSU is required to have output with 50 to 500 ms after the PS_ON# is asserted. So many engineers will make sure the bulk caps are charged within that time frame as well. This is why experts will often tell users to NOT turn their PCs off at the wall or from the switch on the back every time they leave their PC (i.e. every day or multiple times a day).

As I explained very briefly in a previous chapter, you have an NTC or inrush limiting circuit to prevent the inrush from damaging your primary side components, which would happen VERY easily without such circuitry, but repeated abuse would just as easily cause the inrush current circuitry itself to fail. The results won't be as catastrophic, but at the end of the day you're still left with a non-functioning PC.