Electric vehicle (EV) fast charging stations are gaining considerable attention and growth in the market, which means an increasing demand for specialized passive components. Given the high-power, high-energy levels involved — particularly at the power stage — these passive components must tolerate the environment without fault.
In this article, we’ll review the role of the dc-link capacitor and why choosing the optimum component is critical to the performance and lifetime of fast chargers.
The power output
EV fast chargers generate a regulated dc output from a three-phase ac input with isolation and power factor correction. At first, they may seem like any traditional high-power ac-dc converter.
The typical arrangement rectifies three-phase ac input with an active circuit, such as a “Vienna rectifier,” which generates a boosted dc output of about 650V. This is expected to hit closer to 800-1000V in the future, correcting for the ac input power factor.
This boosted voltage is the so-called dc link and the input to a “following isolated dc-dc converter.” Although fast chargers can be rated at 400kW or higher, they’re typically assembled from paralleled modules of a lower rating (say, 30kW), which helps with the apportioning of stresses and enables a degree of redundancy. The position of the dc link is shown in Figure 1.
The dc-link capacitor’s purpose
The rectification or power factor correction stage requires a capacitor on its output to absorb the discontinuous high-frequency current it generates and smooth the voltage. At the same time, the following dc-dc converter needs a capacitor on its input to ‘stiffen’ its supply, source a high-frequency current, and minimize the ripple voltage. Fortunately, one component will do all three tasks — the dc-link capacitor.
In a conventional ac-dc converter, large aluminum electrolytic capacitors are often seen on the dc link, occupying a high proportion of the product volume. Their value in capacitance is usually defined by the hold-up time needed to ride through mains dips or on mains failure to maintain operation while the load shuts down in an orderly way.
The hold-up is often specified as 18/20ms; at 30kW, it needs about 10 millifarads. This is calculated by equating the hold-up energy required (hold-up time x output power/efficiency), with the energy drop in the capacitor between the normal operating voltage and drop-out voltage (say, from 650 to 500V). That is, (0.5 x C x 6502) – (0.5 x C x 5002).
As an additional downside, AL-electrolytics have a finite lifetime — as low as 9000 hours in the example quoted, at a maximum rated temperature, ripple current, and voltage.
Fortunately, the hold-up is not a significant concern in a battery charger. So, the dc-link capacitor can be sized for its performance in absorbing and sourcing ripple current with a low ripple voltage. This is set not only by capacitance but also by the component’s equivalent series resistance and inductance.
The $64K questions are: what value of capacitance, equivalent series resistance (ESR), and equivalent series inductance (ESL) are acceptable? And, if a much lower capacitance is required, could film types be used in the available space, with the additional benefits of their longer lifetime and lower overall cost?
Determining the dc-link capacitor parameters
This is not always as easy as it sounds. Film, or in some cases ceramic capacitors, can be used on a dc link. However, evaluating the ripple current — which defines the acceptable capacitance, ESR, and ESL — is challenging. This is because it consists of three elements:
1. The high-frequency ripple from the PFC stage output
2. The residual ac mains ripple
3. The high-frequency input ripple of the dc-dc stage
The values of the HF ripple depend on the operating conditions, the chosen topology, and whether it’s interleaved. Additionally, the PFC and dc-dc stages might not be synchronized, which means the ripple would fail to add predictably.
Here’s an exercise: acknowledging that dc links are moving to higher voltages makes it possible to look at the overall scale and guess that the dc link for the 30kW and 650V has an average current of about 50A, while allowing for inefficiencies. With a dc-dc duty cycle of 80%, this is about 25A rms sourced from the capacitor, assuming a square wave. With a switching frequency of 100kHz, simplistically, only about 2µF would be needed for a 20V-rms ripple across the capacitance.
If the capacitor ESR were about five milliohms, this would only add an extra 0.125V rms of ripple — but 3W of heating. It could be assumed that the ripple from the PFC stage is of the same order, so we could ignore the line frequency ripple. Despite these gross assumptions, the capacitance needed would only be a few tens of µF. The result is practical if it was spread across several paralleled parts, each with <10 milliohms ESR.
For example, four paralleled 20µF/700V metalized polypropylene capacitors from the Vishay MKP1848H range can handle 62.5A rms ripple with an overall ESR of less than one milliohm. This is with about 1 W dissipation in each one if run to their maximum rating. Four of these parts have a total volume of 139 cm3 or 8.5 inches3.
As another example, compared to an AL-electrolytic solution, achieving sub-10 milliohm ESR would require 10 parts from the KEMET ALC80 range in a series/parallel arrangement to achieve the voltage rating. These could be 2700µF/400V parts with about 85A ripple rating (10kHz) and an ESR of 10.8 milliohms, with about 8W of dissipation in each part, run at their maximum ripple rating. The total volume would be 2060 cm3 (125 inches3) or nearly 15x larger than the film capacitor solution, which packs to a smaller relative volume with their typically rectangular case.
Further advantages of the film capacitor types are a quoted useful lifetime of more than 100,000 hours and a degree of self-healing after over-voltage stress. The ESL of the parts is also particularly low with a few tens of nH, which only adds a volt or so to the ripple voltage waveform.
Cost is also a factor, and here the film capacitors also score, with a volume price for four of the film parts more than 4x lower than 10 of the AL-electrolytics. In practice, derating will be applied to the capacitors of either type, requiring further parallel parts, with possibly more for the electrolytics (Figure 2). The difference becomes even more striking.
Multilayer ceramic capacitors as dc links
Although not widely used in dc-link applications, advances in materials and manufacturing mean that multilayer ceramic capacitors (MLCCs) are becoming available with higher CV ratings for a given component size. Also, new packaging techniques make it easier to stack these components and connect them in parallel to achieve higher capacitance values.
However, despite the benefits of high ripple current capabilities and the ability to operate at high temperatures (up to 150° C in some instances), the approximately 10X price premium for MLCCs over film capacitors of the same CV rating means they’re unlikely to displace electrolytic or film capacitors as dc links any time soon.
Conclusion
If high capacitance for the hold-up is not required — as is the case with EV fast chargers — film types are preferable as dc-link capacitors. When the cost of changing out AL-electrolytics at the end of their life is factored in, the appeal is greater still. Finally, as the cost of MLCCs falls, these components may also be worth considering in the future.
Related links
You may also like:
Filed Under: Charging, FAQs