When do I need separate MLCC (input) and AL-ELEC (bulk) capacitors on my SMPS design?

Question

I've been studying the specification of input capacitance for switching converters. I have a related question here that provides more background.

In my research into the topic I came across TI Application Note SLTA055 - Feb 2006 and an EE Times contributed article that both describe a similar approach to designing input capacitance for buck converters.

The gist is this:

• You need two types of capacitors, input capacitors (large value MLCC) and bulk capacitors (AL-ELEC in most cases).
• The input capacitors provide (much of the) current during each "on/conducting" period, reducing the input voltage ripple to roughly 75mV p-p. The bulk capacitors main role is to supply current during a load current step (until the supply inductance allows the supply current to catch up).
• The reduction of input ripple voltage is important because otherwise the ripple current in the bulk capacitors is too high, and they heat up because of their higher ESR.

Now the capacitance in the MLCCs we're talking about here is not trivial, 84µF (implemented as 4x22µF) in the example they walked through.

This sounds like an awesome approach to me, and the math and everything seems to add up, but I find precious few high-value MLCCs in my scrap bin because it appears none of the designers of my choice e-Scrap follow this logic.

So my question is: "When would I need (or perhaps want) to use high-value MLCCs for separate input capacitance in an SMPS design?"

Responses to a related question I found on-site seem to indicate "Yeah it's a great idea but no one does it because it costs an extra buck". Or perhaps it was true in 2006 when these articles were written but electrolytic capacitor technology has improved since then? Another idea I had was that most of my power scrap is off-line, and maybe it's not needed or wouldn't work when the switched voltage is close to 200V? The one place I did find some high-value MLCCs is in an old Cisco router that seemed to have some point-of-load (POL) supply circuitry on it.

Answer 1

My preferred input capacitors for SMPS circuits are ceramic (X7R); generally speaking that is sufficient for most designs but it does depend on the load current to some extent (because that determines how large a 'gulp' of current is taken from the input during switching).

I always specify X7R devices with Vin * 2 as a rated voltage as that gives me some confidence they will remain within 20% of their nominal value over time and temperature.

I do not like using dry tantalums for a number of reasons; They are prone to failure simply due to reflow even when properly derated and with a low impedance source (which is what the power input to a switch mode supply should be) they can become spectacularly pyrotechnic. In addition, they have effectively zero capacitance above perhaps 400kHz (far below the transition rate of the switches in a SMPS supply).

When using multiple input capacitors, I make a small 'mini-plane' for the power and return sides to ensure low inductance paths as the ceramic devices are the ones that can respond at the rate at which the switch transitions are operating.

Aluminium devices can be used as a bulk source, but are there for slow transitions only.

I use Aluminium devices for hold up applications at present (although I am evaluating different techmologies for the future).

My answer to the question is:

When would I not use ceramic capacitors on the input to a switch mode power supply?