Singing ceramic capacitor

[shadewind]

I have designed a LED diver board using the TPS40211 boost converter. Everything is working fine except that when I'm doing PWM, there seems to be a singing noise coming from the input capacitor. This is obviously because of the on/off switching sine the noise seems to be of the same frequency as the PWM signal.

The input capacitance consists of a 330 uF Panasonic FK series rated at 35V and two ceramic 10 uF 50V caps. This is what TI have in their reference design.

Does the noise mean that there's something wrong i.e. too much RMS current or something like that? It seems to be a common problem with electrolytic caps. Is there something which can be done about it? I'm thnking I can reduce the frequency to a level where the noise isn't as audible. If I do that, will I be fine or is this a sign of an underlying problem?

[Conrad Hoffman]

When voltage is applied to two capacitor plates, force is generated between them. Can't be helped unless you repeal certain laws of physics. It's normal for capacitors to sing if an audio frequency is applied. The amount depends on the rigidity of the dielectric and various other things. Though I haven't tried it, my guess is a solid dielectric like used in an OSCON cap probably makes less noise. Ditto tantalums and, if you can get one big enough, the better films. Hi dielectric constant ceramics are absolutely the worst!

[shadewind]

Okay, so at least it doesn't mean that there's anything wrong with the design?

[Conrad Hoffman]

That I can't say without seeing it. OK, maybe can't say even if I do see it! If you excite it at an audible frequency, you'll probably hear it. There are also issues with switching supplies where you can get audible frequencies way below the actual switching frequency. It's more an annoyance than serious, but indicates some more work should be done.

[shadewind]

The switching itself does not make any audible noise. Without PWM, it's dead silent. I cannot raise the PWM frequency beyond the audible spectrum since there is a minimum on time requirement.

Any ideas that do not require a change in the design and if not, what can I change?

[Simon]

maybe you can change capacitor manufacturer / type ? or what about a few caps in parallel ?

[shadewind]

Hmm... it might actually be the ceramics. When pushing against them witha a screwdriver, the sound changes a bit.

[shadewind]

I fiddled a bit more and when pressing a screwdriver against the output ceramic caps, the noise increases a lot so my guess is that those are probably the culprit. These are the ones I'm using: http://se.farnell.com/jsp/search/productdetail.jsp?SKU=1793832

But what should I do? From what I can understand, this is an inherent problem in ceramic capacitors so I suppose changing to another brand won't help? I looked at tantalum caps but at the voltage rating I need (> 75V), they are prohibitively expensive.

[IanB]

Ear plugs?

But seriously, maybe some hot glue/silicone sealant/conformal coating may help? You just want to damp out the vibrations somehow.

[Conrad Hoffman]

IMHO, you won't be able to stop it with anything short of a thick coating of tar! Going to a lower dielectric constant ceramic will probably help- the highest constant (physically smallest) will usually have the greatest piezoelectric effect. You could also change type. A small film cap should be good to very high frequencies, though it may sing slightly.

[SiBurning]

10uF ceramic?!? The audiophile in me is horrified! Use mylar or back-to-back (or non polar) electrolytics.

[amspire]

It is possible for multilayer ceramic caps to delaminate through a manufacturing fault, or too much heat during soldering.

If that is what has happened, it would be a good idea to replace it.

I have never seen a good ceramic cap "singing", but then again I don't use 10uF ceramic caps much (probably never).

Richard.

[SiBurning]

Are you using an SMD cap? It might be that they tested this with SMD caps and your circuit is ringing because it's a discrete part that has some freedom of movement. It might also be that the circuit you chose wasn't intended or tested for audio frequencies.

[Simon]

it is occurring to me that maybe your caps are doing a lot of work, what PWM duty are you using ? are you sure that your power supply can source the current required with ease ?

[mikeselectricstuff]

This is quite common - may be worth trying different manufacturers but doubt it will make a huge difference. If you need high voltage ( as mentioned above) you don't really have a lot of options for alternative parts. Adding a low-ESR electrolytic across the cap may reduce noise by taking some of the ripple load away form the ceramic.
Potting isn't as effective as you might imagine, as you have a rigid bond between the cap and the PCB, so the sound gets conducted out via the PCB.
You may be able to soften the sound slightly by dithering the PWM by a cycle or two on each cycle.

[Gall]

High capacitance ceramic capacitors (ferroelectricity based) usually have HUGE piezo effect. Avoid use them on AC.

[shadewind]

IMHO, you won't be able to stop it with anything short of a thick coating of tar! Going to a lower dielectric constant ceramic will probably help- the highest constant (physically smallest) will usually have the greatest piezoelectric effect. You could also change type. A small film cap should be good to very high frequencies, though it may sing slightly.

Yeah, I actually did try using hot glue to silence it but with no luck really. It seems that there's a tiny bit of noise coming from the inductor as well but silencing the caps shuld be enough. I cannot use another dielectric for ceramic caps, though, since X7R is the best I can get with that capacitance.

I know nothing about film caps though. Would a film cap have low enough ESR end ESL to work? At these switching frequencies (800 kHz) I doubt the capacitance itself has much impact.

10uF ceramic?!? The audiophile in me is horrified! Use mylar or back-to-back (or non polar) electrolytics.

Horrified? Why? Ceramics have the best characteristics for power if you can get one with enough capacitance. The pizeoelectric effect also seems to be no problem as long as your switching frequency is inaudible... which it most certainly is if you're using ceramics.

It is possible for multilayer ceramic caps to delaminate through a manufacturing fault, or too much heat during soldering.

If that is what has happened, it would be a good idea to replace it.

I have never seen a good ceramic cap "singing", but then again I don't use 10uF ceramic caps much (probably never).

Richard.

No they're like this all of them, identically. So unless I've overheated all of them when soldering, this is an inherent problem. I also saw this exact same problem with another board where the minimum load wasn't met for a switcher which caused it to lower its switching frequency into the audio range. This made the ceramic buzz in that frequency. That ceramic had a different manufacturer and a different capacitance (10 µF instead of 2.2 µF which are the culprit here).

Are you using an SMD cap? It might be that they tested this with SMD caps and your circuit is ringing because it's a discrete part that has some freedom of movement. It might also be that the circuit you chose wasn't intended or tested for audio frequencies.

I'm using an SMD cap yes, 1210 package. And since the output capacitance is identical to the one in the reference design and the reference design manual speaks of a minimum on time which is not achievable without audio frequencies, I doubt this is the problem.

it is occurring to me that maybe your caps are doing a lot of work, what PWM duty are you using ? are you sure that your power supply can source the current required with ease ?

This a problem no matter the duty cycle. The sound only changes character with the duty cycle. These are the output caps in a boost converter, they are supposed to do a lot of work.

This is quite common - may be worth trying different manufacturers but doubt it will make a huge difference. If you need high voltage ( as mentioned above) you don't really have a lot of options for alternative parts. Adding a low-ESR electrolytic across the cap may reduce noise by taking some of the ripple load away form the ceramic.
Potting isn't as effective as you might imagine, as you have a rigid bond between the cap and the PCB, so the sound gets conducted out via the PCB.
You may be able to soften the sound slightly by dithering the PWM by a cycle or two on each cycle.

As I stated above, I've seen this with a totally different capacitor as well so as you say, it will probably not make much difference since this is inherent to ceramics other than NP0 as far as I can see. Potting with hot glue didn't make any difference of course.

Finding a low ESR electrolytic rated for at least 85V (as a safety margin, the max voltage is 75V) doesn't seem to be the easiest thing to do. The ones rated 100V that I found had horrible ESRs.

[FreeThinker]

Didn't Dave do a blog on this? Think it was the one about tapping the scope leads and getting ringing on your signal (or something on those lines) due to the piezo effect of the caps. Will try and find it (search bar is S**t) 

Found it! http://www.eevblog.com/2011/04/12/eevblog-162-oscilloscope-probe-shock/

[Conrad Hoffman]

This is a tough one because once you get up near 100V and several uF, the choices in low esr caps become quite limited. There are some low esr aluminum electrolytics from the usual suspects, but my guess is that because aluminum electrolytics aren't very tightly wound, they'll all sing. I really like OSCONs, but they don't make them in high voltages. Film is extremely low esr and functional to far higher frequencies than people are usually aware of, but the size would be huge compared to a 1210 pack. I have some 1 and 2.2 uF stacked film blocks that are quite compact, but still way bigger than a 1210. In the through hole world I prefer to inhabit, there are at least 5-10 different types of ceramic, but less so in the multilayer type you need. The only thought that comes to mind, and I don't like them, is tantalum. A solid tantalum cap shouldn't sing. I've seen 82V tantalums, usually mil spec parts, but they may be something lost to the dustbin of history. It's a low voltage world now! Maybe one could use a few tantalums in series, with balancing resistors, but this doesn't seem like a low esr solution. I'm thinking the singing might be a small price to pay for using available parts in a compact design.

[Zero999]

10uF ceramic?!? The audiophile in me is horrified! Use mylar or back-to-back (or non polar) electrolytics.

Horrified? Why? Ceramics have the best characteristics for power if you can get one with enough capacitance. The pizeoelectric effect also seems to be no problem as long as your switching frequency is inaudible... which it most certainly is if you're using ceramics.

High capacitance ceramics have a high dielectric absorption which causes distortion. The piezoelectric effect demonstrates their lossyness as energy is converted to sound which can lead to oscillation in the audio band and beyond. If a piezoelectric capacitor is used both on the input and the output stage of an amplifier acoustic feedback can occur, possibly even at an ultrasonic frequency.

[SiBurning]

10uF ceramic?!? The audiophile in me is horrified! Use mylar or back-to-back (or non polar) electrolytics.

Horrified? Why? Ceramics have the best characteristics for power if you can get one with enough capacitance. The pizeoelectric effect also seems to be no problem as long as your switching frequency is inaudible... which it most certainly is if you're using ceramics.

High capacitance ceramics have a high dielectric absorption which causes distortion. The piezoelectric effect demonstrates their lossyness as energy is converted to sound which can lead to oscillation in the audio band and beyond. If a piezoelectric capacitor is used both on the input and the output stage of an amplifier acoustic feedback can occur, possibly even at an ultrasonic frequency.

Yeah. What he said.

Or in simpler terms, because they distort more than any other cap, and they ring audibly.

"as long as your switching frequency is inaudible... which it most certainly is if you're using ceramics" At 10uF we're in audible territory, and your modulation is probably providing the necessary energy in the audio frequencies. Power applications aside, I've never heard of anyone using a ceramic above 105 (or possibly 106 in cheap stuff) for audio, and I wouldn't use anything above 104, and reluctantly at that. Simply put, it's the size of the thing that shocks me. I wouldn't even have tried using a ceramic of that size, and would have just used a mylar cap from the beginning. Whether or not that's the problem... Well, it's best to narrow it down by eliminating or reducing this as a possible source.

[shadewind]

I'm sure that large ceramics are horrible for audio applications but I wouldn't know since I'm mainly a digital guy myself (and a novice one at that). But if we disregard the PWM frequency which is in the audible territory, the switching frequency is 800 kHz which is way beyond audio. If there's no PWM, it's dead silent.

For power, I see a clear trend in replacing electrolytics with ceramics as much as possible because of their superior ESR, ESL and reliability. For this purpose, I don't see why one shouldn't be using them... that is if noise is not an issue

[SiBurning]

Agreed. That explains why neither you nor TI saw a problem. I'm merely suggesting that your introduction of frequencies in the audio range (say below 200kHz) through the modulation brings you into the realm of audio design. It's simple enough to reduce the ringing--the best caps in this range for audio are mylar, metallized, & polyester (though mylar IS polyester). No guarantee, but a simple enough thing to try.

Since I'm just an audio hobbyist, without serious depth of knowledge, I rely on a lot of rules of thumb, and not using ceramics is one of them.

My gut instinct is that this won't fix the problem, and you'll need to adjust values. That's because the inductor is also ringing, suggesting that the pair needs to be adjusted. But best to try the simple things first. Personally, I'd try Mike's bypass idea first, though not necessarily for the reason he suggests. If you can get the audio frequencies to rely on the other cap... Then again, it seems very backward to bypass a ceramic with an electrolytic, and the numbers (running them quickly & loosely in my head) seem to support that giving you an even harder to find cap for the bypass.

[shadewind]

But it seems film caps are my only alternative. I understand that the ESR is very good but whatabkit size and ESL? Is it at all heard of to use film caps for power?