# MLCC as SMPS filter caps

This question is out of curiosity. I know fairly well that mains SMPS can suffer from bad electrolytic caps. Now I wonder if replacing all electrolytics with MLCCs work? Will it make a more durable SMPS?

If MLCCs work as filter caps, why the manufacturers are still using electrolytics? Here I mean SMPS a sold as parts of industrial equipment or long-lasting equipment (like computer servers that can be designed to last for 20-50 years)

• They'd work. But the existing MLCC's have way lower voltage rating/capacitance figures. – Laszlo Valko Feb 6 '15 at 11:51
• @laszlovalko I did found 47uF/450V MLCC from local suppliers. So did 100uF/50V ones. Big packages and a bit expensive, yes, but ceramic and way more durable. – Maxthon Chan Feb 6 '15 at 12:00
• @MaxthonChan: "a bit expensive" ... the world is driven by cost. Now add 2 and -2 ... – PlasmaHH Feb 6 '15 at 12:21
• @PlasmaHH In a total cost of ownership sense, electrolytics have a smaller initial cost which is only guaranteed for, like, 5 years. If we are talking about equipment designed to last, like, 20 years, the maintainence cost start to mount up. – Maxthon Chan Feb 6 '15 at 14:03
• @MaxthonChan: you were asking "why the manufacturers are still using electrolytics?" which I interpret as "why isnt everyone using MLCCs?" – PlasmaHH Feb 6 '15 at 14:26

Yes it is possible to replace all electrolytic caps with MLCCs. There are some things you'll have to consider in your design, but it is possible. And yes, the lifetime will be greatly increased in most cases, but even with electrolytic caps it is possible to build very reliable switch mode power supplies.

While it is true, that MLCC offer a more reliable solution to filtering the output, most product designs are driven by cost and space.

MLCCs are far more expensive per capacity than electrolytic caps. I haven't found a 47µF/450V MLCC, but a search on Digikey turned up with some points:

47µF/25V MLCC - 0.28$@100k 47µF/25V aluminium - 0.05$ @100k

So the aluminium electrolytic cap is only a sixth of the cost of the MLCC. And these are quite low requirements, it gets very very worse for higher capacity and higher voltages.

Another point which should be made is, that if you actually need 47µF @25V, you can't use a MLCC which is marketed as 47µF/25V cap, as the capacitance of (some) dielectrics used for MLCCs are extremely voltage dependent. Let's take a look at one of those datasheets for MLCCs(page 54). As you can see, the capacitance at the rated voltage is only half of what is advertised. So you basically need two 47µF/25V MLCC for 47µF at 25V, which doubles the costs for MLCCs.

Yet another point which might go wrong with MLCCs is the piezoelectric effect, they can act as microphones or vibration sensors and give off additional noise into your system. Or even emit some sound if the ripple frequency is in the right spot.

Edit

If you are aiming for really long lasting power supplies, you will increase safety margins on every component, so you make your capacitors bigger and suited for a higher voltage and most importantly rated for higher temperatures. If you combine that with good thermal management, you can go a long way even with electrolytic caps.

Still, from my experience people in the industry tend not to buy a really really expensive thing which is advertised to last forever but rather like a redundant solution. So they'll buy two normal sensors, two normal power supplies and if one fails they will replace it. It's in most cases much more cost efficient to do so. Also the question always arises: How much can you trust the manufacturer to really make a "lasts forever" product.

So in the end, there is only a really low demand for super expensive, super reliable switch mode power supplies, so you never see them. I'm sure they are around somewhere.

• For consumer electronics that is designed to last 5 or 10 years then maybe. For equipment that is designed to last 20 or even 50 years that cap replacement can be very expensive, justifying MLCCs in initial design. – Maxthon Chan Feb 6 '15 at 14:07
• Electrolytic capacitors are also able to survive that long depending on the outer circumstances. You haven't specified what exactly you are targeting at, mains SMPS sounded a bit like consumer stuff. Today no company is going to replace a cap, they'll just change the whole unit. The answer is as always: it depends. – Arsenal Feb 6 '15 at 14:11
• For some industrial environment a SMPS can be buried in the middle of some piece of non-stop equipment and stopping it means lots of loss. – Maxthon Chan Feb 6 '15 at 14:14
• @MaxthonChan Sure, I guess replacing a SMPS on a satellite is also quite expensive. But your question was not very specific what applications you have in mind. I updated my answer to be more clear about the possibility to do such a design. If you need an ultra-reliable power supply, you design one, maybe make it redundant. The possibilities are all there, it's just that everything has different requirements, if you have something special in mind, rephrase your question to be more clear. – Arsenal Feb 6 '15 at 14:30

It is often possible to substitute MLCC capacitors in place of electrolytic caps but I'd rather just replace the old capacitors with high-quality parts.

I'm extremely fond of Sanyo's low-ESR capacitors: we started out using their "Os-Con" family but have since migrated to the AX and GX parts.

Sanyo publishes detailed data sheets which allowed me to choose the best devices for my applications. One very important factor for me is that many of their parts are rated for operation down to -55C - most electrolytic caps stop working at -20C or thereabouts.

If you do decide to replace some of the electrolytic caps with ceramic, be sure to check the exact part's datasheet for voltage-dependent effects. Many ceramic capacitors exhibit significant loss of capacitance as the applied voltage increases. Also note that this effect is specific to EACH part you might look at - you can NOT substitute to a different / less expensive part without ensuring that it is going to work for you under all of your operating conditions.

One final note: many ceramic capacitors (both large and small) can exhibit piezo-electric effects: they can act as both a microphone and a speaker. Ceramic capacitors in sensitive analog circuits can cause you all manner of nasty problems: sensitivity to acoustic noise or vibration is just one.

MLCCs are often preferred in Point of Load SPMS designs for several reasons e.g. i) they can be distributed around the board, under ICs; ii) they have improved reliability compared to electrolytics; ii) improved dielectric stability e.g X5R. But their Capacity per unit area counts against them and their cost.

Importantly, Switch Mode Power Supply loop stability often depends upon the phase lead introduced by the ESR zero of the Electrolytic capacitors. So swapping them out would make the loop unstable in many cases or if not unstable then at least significantly different from the designed loop dynamics.

So in summary; No, it is not usually OK just to swap out electrolytics for MLCCs in SMPS designs.