MLCC Microphonics

[Bo Deadly]

Just how microphonic are MLCC capacitors?

I just bought a ReVox B77 reel-to-reel that has old Tantalum caps all over the place in the speed control, drive control, monitor amp, input amp, ... pretty much all of the boards. I'm thinking anything in the audio path should just be replaced with modern Tantalums or selectively film if there's room. But for everything else, why not MLCCs? The only downside of MLCCs that I know of is that they're microphonic. But for most things, does it really matter?

Is it possible that say an MLCC used to debounce a button or used as a timer could trigger unexpected logic from vibration like a solenoid closing or just knocking the machine? We're only talking about a couple of millivolts right?

The alternative is to use electrolytics which can be a tight fit or film which in many instances is just totally overkill or to just replace with modern Tantalums.

What do you think? Is there even a remote possibility that the microphonics of  MLCCs could break circuits like those in the B77?

 

[gyraf]

Good question.

I have tried measuring several different manufacturers 10uF's - only found very few pieces that were microphonic, too few to tell if it follows manufacturer or is statistically distributed.

Among these, at least, I saw none that were more than a few mV, even when directly hit. So I wouldn't worry about malfunctioning by this.

Another related question is: What failure mode do MLCC have? Especially at overvoltage - do I need to experiment myself?

Jakob E.

 

[Bo Deadly]

Actually it looks like capacitance varies greatly with DC voltage. That pretty much rules out anything used for timing or coupling audio in the B77 without complicated re-computation of capacitance values and voltage ratings. So microphonics are not the problem. The problem is that an MLCC can loose 80% of it's capacitance with 20V across it.

But regardless of my particular project, I did find some extra info about MLCC microphonics:

1) The diaelectric matters. A 'Y' diaelectric like a Y5V is way worse. The MLCCs I was looking at are X5R which exhibits less microphonics and the temp curve is pretty flat. No such thing as C0G MLCCs. X is as good as it gets AFAICT.

2) The microphonics works both ways. Meaning they can emit noise as well as pickup. For example, if you put a high current 5kHz square wave through a one (like in a charge pump circuit of an LCD display) you will hear it clear as day.

3) Suface mount vs through-hole makes a big difference because the piezo effect is amplified by the ridged connection of the SMD part to the PCB. So through-hole should exhibit lower microphonic behavior. Especially if it's not seated low. Meaning solder it high with some bus wire between the MLCC and board.

So if it's through-hole, low current and X5R, microphonics would probably be non-existant.

But good point, I don't know what the failure modes are. Just from some quick googling it sounds like they do fail short like tantalums but that failures are very uncommon (especially if you derate voltages). The most common failures being cracking do to physical stress (such as cooling too quickly after reflow).

 

[volker]

Typed the post while you were editing...


The voltage dependence of the capacitance varies greatly with the dielectric used.

More info here.

For the signal path NP0 works great, it is actually superior to any film capacitor. The other dielectrics find its use in supply decoupling (and guitar pedals ). I don't use anything other than X7R/X5R, just too cumbersome. The worse the dielectric's voltage dependence, the better the nominal capacitance per volume. But I've never done the math comparing different dielectrics for derated capacitance and use of space to achieve a certain amount of actual capacitance at working voltage.

With regard to your original question, microphony is only a problem with class 2 capacitors, NP0 is not affected. I also remember an EEVblog video where the microphony of one of the oscilloscope's inputs was demonstrated. It's really a question of application and the surrounding circuitry, I've never experienced it.

 

[JohnRoberts]

Just how microphonic are MLCC capacitors?

I just bought a ReVox B77 reel-to-reel that has old Tantalum caps all over the place in the speed control, drive control, monitor amp, input amp, ... pretty much all of the boards. I'm thinking anything in the audio path should just be replaced with modern Tantalums or selectively film if there's room. But for everything else, why not MLCCs? The only downside of MLCCs that I know of is that they're microphonic. But for most things, does it really matter?

Is it possible that say an MLCC used to debounce a button or used as a timer could trigger unexpected logic from vibration like a solenoid closing or just knocking the machine? We're only talking about a couple of millivolts right?

Generally no... Logic is all about having high thresholds to ignore noise.

JR

The alternative is to use electrolytics which can be a tight fit or film which in many instances is just totally overkill or to just replace with modern Tantalums.

What do you think? Is there even a remote possibility that the microphonics of  MLCCs could break circuits like those in the B77?

 

[Bo Deadly]

For the signal path NP0 works great, it is actually superior to any film capacitor.

Unfortunately, as you allude to, C0G / NP0 larger than ~1u are not common or incredibly expensive. My project is a lot of 3.3u, 10u and similar. I might use X5R if DC is not an issue but it looks like I'll just go with new tantalums.

A lot of people put electrolytics in place of the tantalums in their B77s but I wonder if they fail to consider that electrolytics are usually derated quite a bit to account for low frequency distortion and decay over time. So instead of a 10u tantalum, one might have to go to 47u electrolytic which could easily be too large for a spot that had a little tant.

I'm going to go through each cap board-by-board now and eval. Maybe I'll post a list when I'm done. B77 is pretty popular and it looks like there is a strong following to keep them going so maybe someone else can benefit.

 

[JohnRoberts]

Unfortunately, as you allude to, C0G / NP0 larger than ~1u are not common or incredibly expensive. My project is a lot of 3.3u, 10u and similar. I might use X5R if DC is not an issue but it looks like I'll just go with new tantalums.

A lot of people put electrolytics in place of the tantalums in their B77s but I wonder if they fail to consider that electrolytics are usually derated quite a bit to account for low frequency distortion and decay over time. So instead of a 10u tantalum, one might have to go to 47u electrolytic which could easily be too large for a spot that had a little tant.

I'm going to go through each cap board-by-board now and eval. Maybe I'll post a list when I'm done. B77 is pretty popular and it looks like there is a strong following to keep them going so maybe someone else can benefit.

Tantalum were liked (way back when) for generally low ESR and modest tolerance to reverse voltage.

Modern aluminum electrolytic capacitors have much lower esr than they did decades ago, so might be good substitutes for replacing old tantalum.

JR