7 Capacitors You Should NEVER Use in a Product
Capacitors are the most common part in your product, and the part nobody thinks about.
They rarely get a second look, so the same seven mistakes show up in schematic after schematic that I review.
One of them will fail safety certification outright, and three of them make a product flaky in ways you can never reproduce on the bench.
And one of them can actually catch fire when it fails, which is why some engineering teams have banned it from every board they build.
And for every single one, I’ll tell you exactly what to use instead.
So I’m counting down seven capacitors you should never use in a product.
Capacitor #7 – Y5V and Z5U Ceramics
If you’ve ever bought one of those cheap capacitor assortment kits, there’s a good chance most of it was Y5V or Z5U.
Those letter codes tell you how the capacitor behaves over temperature, and for these two, the answer is badly.
A Y5V capacitor is allowed to lose up to 82% of its capacitance across its rated temperature range, and Z5U isn’t much better at 56%.
So that 10uF cap you put on a regulator output might be acting like a 2uF cap inside a hot enclosure.
These dielectrics also lose a big chunk of capacitance from DC bias, which we’ll get into in a minute, so the two effects stack on top of each other.
That means your bulk filtering and decoupling capacitors end up specified around a number the circuit never actually sees.
In a product, stick with X7R or X5R for general decoupling and bulk capacitance, and use C0G for anything small and precise.
They cost a little more, but X7R stays within plus or minus 15% across its whole temperature range, which you can actually design around.
Capacitor #6 – An Ultra-Small Package When You’re Not Space Constrained
Product creators love to reach for the smallest capacitor package they can find, because smaller feels more professional and it leaves more room on the board.
But an 0402 is only 1mm long, and an 0201 is 0.6 by 0.3mm, which is basically a speck of dust.
Drop one and it’s gone, hand soldering an 0201 is nearly impossible, and reworking one on a finished board takes a microscope and a very steady hand.
Even in mass production, the smaller the part, the more you invite defects like tombstoning, where one end of the part lifts off its pad during reflow, and solder bridging between pads that are almost touching.
Some lower-cost assembly houses either won’t place 0201 parts at all or charge extra for them, and any rework on parts that small costs more per board.
For most general-purpose capacitors, 0603 is a good default, and you only go smaller when you truly need the space or the better high-frequency performance that a smaller package gives you.
Squeezing a large capacitance value into an ultra-small package also hurts the circuit electrically, and the reason is something called DC bias.
A 10uF ceramic capacitor in an 0402 package looks like a great deal, because it’s tiny, it’s cheap, and the datasheet says 10uF.
But that 10uF number was measured with almost no voltage across the part.
Put a normal 3.3V rail on it and it can lose more than half of its value, so your 10uF cap is now doing the job of a 3 or 4uF cap.
The reason is that these high-value ceramics use a dielectric that gets less effective as the voltage across it goes up, and the thinner the layers, the worse the effect.
Squeezing 10uF into a package that small means very thin layers, so the capacitance drops fast even at low voltages.
Where this bites you is in a switching regulator or a bulk filter, where the design counted on a certain amount of capacitance for ripple, transient response, or stability.
Murata, TDK, and Samsung all publish DC bias data and have free online tools that will show you this curve for most of their capacitors.
So check the curve at your actual operating voltage, and if you’re losing more than you can live with, move up to an 0603 or 0805 package, or try a different part number with a better curve.
Capacitor #5 – A Ceramic Cap on an LDO That Needs ESR
Every capacitor has a little bit of resistance inside it called ESR, which stands for equivalent series resistance.
Most of the time you want that as low as possible, but some older linear regulators actually depend on that resistance to stay stable.
Their control loop was designed back when the output capacitor was always a tantalum or an electrolytic, and the ESR in those parts acts like a built-in damper for the loop.
Put a ceramic capacitor on one of those regulators, with almost no ESR, and the loop can start oscillating, which shows up as a noisy output, a regulator that runs hot, or a microcontroller that resets for no obvious reason.
The classic example is the LM1117, which asks for an output capacitor with an ESR between 0.3 and 22 ohms, and a typical ceramic capacitor sits far below that.
And any time the load current jumps, that current flows through the ESR and creates a momentary dip in the output voltage, so a regulator that depends on ESR gives up some transient response to get its stability.
When I was designing LDOs at TI, almost all of them were built to be stable with a low-ESR ceramic output capacitor, and most newer, higher quality LDOs are the same, but plenty of older designs are still out there.
Check the datasheet for the exact part you’re using, and if the regulator needs ESR, give it a traditional tantalum or an aluminum electrolytic with an ESR inside the range the datasheet calls for.
And if that dip matters for your circuit, pick an LDO that’s stable with a low-ESR ceramic, and check its load transient response in the datasheet before you commit.
Capacitor #4 – X7R or X5R Where a C0G Belongs
X7R and X5R are what’s called class 2 ceramics, which means they’re built to pack a lot of capacitance into a small size, and they give up some accuracy to get there.
For decoupling and bulk filtering, that trade is exactly what you want.
But there’s a whole category of circuits where they’ll wreck your design, and that’s anything where the exact capacitance value matters.
That means RC timing circuits, oscillators, crystal load capacitors, active filters, and anything in a precision analog signal path.
The problem is that X7R doesn’t hold a value, it holds a range.
You’re starting with a tolerance of 10 or 20%, then adding about 15% of drift over temperature, then the DC bias loss we already talked about, and then aging, because these dielectrics slowly lose capacitance just sitting on the shelf.
Stack all of that up and your filter corner or your timing interval can move by 30% or more.
Class 2 ceramics are also piezoelectric, meaning they physically flex when a voltage is applied, and the reverse is true too, so vibration or even a tap on the board shows up as a voltage spike in your signal.
The right part here is C0G, also called NP0, which is a class 1 ceramic, meaning it’s built for accuracy instead of size.
It’s stable to within about 30ppm per degree C, it has no DC bias effect, it has no piezoelectric effect, and it barely ages.
You can get it up to around 100nF in small packages and higher in larger ones.
And in the pF to low nF range where most timing and filter circuits live, it only costs a few cents more than X7R.
So use X7R where you need bulk, and C0G anywhere the exact value matters.
Capacitor #3 – General-Purpose Electrolytics Next to Anything Hot
Aluminum electrolytic capacitors have a lifespan, and it’s shorter than most people think.
The cheap general-purpose ones might have only a 2,000-hour endurance rating at 85C.
An electrolytic has liquid inside it, and that liquid slowly evaporates through the seal, faster when it’s hot.
The rule of thumb is that every 10C increase in temperature roughly cuts the life in half.
At 85C, that 2,000-hour rating is less than three months of continuous operation.
Drop the capacitor temperature to 45C and that same rule of thumb gives you an estimated life of about 32,000 hours, a little under four years of always-on use.
Now think about where these capacitors end up on a lot of boards, right next to a hot linear regulator, a power resistor, or a switching transistor, inside an enclosure with no airflow.
When one finally dries out, the ESR climbs, the ripple gets worse, and you get a product that works for a year and then starts rebooting or humming.
Electrolytic capacitors are one of the few parts on a board that wear out over time, so plan for that from the start.
Use 105C parts with a 5,000-hour rating or better, which usually adds a few cents at most, and keep them away from heat sources in the layout.
Or in a lot of low-voltage circuits that don’t depend on ESR, replace it with a ceramic, or a solid polymer aluminum, which still ages but has no liquid to dry out.
Capacitor #2 – A Regular Cap Where an X or Y Cap Belongs
If your product plugs into the wall, there are two common spots in the design where a regular capacitor becomes a shock and fire hazard.
The first is across the incoming AC line, between line and neutral, where a capacitor keeps noise from getting into or out of your product.
The second spot is a capacitor that crosses from the AC side over to the low-voltage side of your product, which is done to reduce conducted emissions.
That capacitor sits right on the isolation barrier, which is the only thing keeping mains away from everything the customer can touch.
A capacitor across the line sees every surge and spike the power grid throws at it, and a regular film or ceramic capacitor was never tested for that.
If it fails as a short circuit across the line, you get a fault current that can start a fire.
And if the one crossing the barrier fails as a short circuit, mains voltage ends up on the low-voltage side of your product, which is the side your customer is touching.
An X-rated capacitor goes across the line and is tested to survive those surges, and because it sits between line and neutral, the danger if it ever does fail is fire rather than shock, so the standard tests for exactly that.
A Y-rated capacitor is the one that crosses the barrier, and it’s built and tested so that a short circuit is extremely unlikely, because a short there is what puts mains voltage on the outside of the product.
Their safety approvals are documented by the manufacturer, and a certification lab will want to see them.
Put a standard capacitor in either of those spots and the product fails safety certification, no matter how well the rest of the design was done.
The X and Y classes are split into subclasses based on how big a surge they can take and how much insulation they provide, and for most consumer products that means an X2 across the line and a Y1 across the barrier, since Y1 is the grade rated to bridge that barrier on its own, and together they cost well under a dollar.
And if this sounds like a risk you’d rather avoid, it’s a good reason to use a pre-certified external adapter and keep mains out of your enclosure entirely.
Capacitor #1 – Solid Tantalum Capacitors Without Heavy Derating
I saved this one for last on purpose, because it’s the one capacitor on this list with a fire risk baked into its chemistry, even in a low-voltage circuit with no mains anywhere near it.
The traditional solid tantalum capacitor uses manganese dioxide for one of its internal layers, and that’s an oxidizer.
When one of these capacitors fails, it usually fails as a short circuit, and if enough fault current is available, that spot heats up until the oxidizer and the tantalum react and the part ignites.
These parts often fail right at power-up, when the inrush surge hits a tiny weak spot in the dielectric.
That’s why some engineering teams have banned tantalums entirely.
My advice, if you do use them, is to never run them above half of the rated voltage, and on any rail that may see a voltage surge, closer to a third.
A 16V tantalum on a 12V rail looks fine on paper, but under that rule it’s way over the line.
Polymer tantalum capacitors fit in the exact same footprint, but they replace the manganese dioxide with a conductive polymer, so there’s no oxidizer inside.
They can still fail as a short circuit, but without the oxidizer that short is far less likely to turn into a fire, and depending on the part, you can typically run them at 80 to 90% of the rated voltage instead of 50%.
They do cost more, usually somewhere in the tens of cents per part in volume, which is cheap insurance against a fire.
Polymer does have higher leakage current than the manganese dioxide type, so in an ultra-low-power standby circuit, a heavily derated traditional tantalum may still be the right call.
But in most circuits, if you need a tantalum, go with a polymer, unless it’s the output capacitor on an old regulator that needs ESR, which is one of the few places a traditional tantalum still has a job.
And if you truly have to use a manganese dioxide part, be sure to derate it, put some resistance in series to limit inrush, and treat it as the most dangerous component on your board, because it probably is.
