your experiences with caps switching to SMD?

[EATyourGuitar]

making an SMD version of a distortion pedal that has had every single poly film chicklet cap tweaked to what sounds best. my plan is to use the prescribed values in XR7 ceramic %5 0805 up to 10n but maybe up to 100n. the 470n will be through hole poly. my question relates to miller effect. have you ever noticed SMD ceramic caps to sound consistently smaller or larger in value than the through hole equivalent parts? should I be buying 0805 %5 caps with the lower voltage rating 16v? or should I simply scale the values so that 2n2 through hole film = 4n7 XR7 0805?

mostly concerned with keeping the SMD version sounding the same as the through hole version. thanks!

[PRR]

I do not see Mr. Miller in this.

X7R has significant Voltage Coefficient. The in-use value will normally be -less- than marked. I would use way-high voltage rating, 50V or 100V, in 9V work. Then I would just use same uFd as your film-version.

[Transmogrifox]

Murata and Kemet both have online "simulation" utilities so you can quickly view the DC bias vs. capacitance the PRR is talking about if you want to really understand this relationship for yourself and pick specific SMT caps for specific parts in the circuit based on applied signal and bias levels.

Paul gave the closest you can get to a generalized answer, that is to pick caps with voltage ratings several times greater than the max DC voltage available in your circuit.

Otherwise you have to look at the datasheet or online sim and look at capacitor value vs applied voltage.  This is also important when considering how much distortion will be added when, for example if it's a high pass filter but the incoming signal has low frequency components (blocked by the cap) modulating the capacitance value (which will result in a form of intermodulation distortion on the pass-band frequencies). 

As an aside there have been some topics on the board about possibly capitalizing on this characteristic to make sweeping filter effects by application of LFO to one side of an MLCC cap.  Don't know of any threads where the effect has been reported, but seems plausible that something useful could be done with this.

[Rob Strand]

SMD caps have evil behaviors you never see with through holes.

Check out this old thread:
http://www.diystompboxes.com/smfforum/index.php?topic=111219.0

The solution to the "decrease in capacitance with voltage issue"  is to larger packages.
For small values use COG rather than X7R.

I've also seen X7R's that have significantly less capacitance (say 30%) at 0V than they do at 1V.  You will see this for AC levels and DC bias - there is like a hill in the capacitance vs voltage curve.   There is no work around. You have to check the datasheet or measure it.   There is often no data on this type of behaviour.

[bool]

COG is ok
X7R; try 50 or 100V versions.

[dschwartz]

I am one of those "heretic" builders that use mlcc X7R SMD caps for almost everything..and honestly, i cant hear a difference (maybe there is a difference if you use measuring tools, but the ear is by far, the one i trust more).. The voltage/capacitance variance becomes important when you need absolute precision .. If you're using a 470nF cap for stage coupling, 20% variation is undetectable by ear
The only drawback i found with mlcc is microphonics.. I avoid them at the first stages of high gain distortion and replace them with film.. No mojo there..microphonic caps are quite real and you can hear it..

[bool]

... The voltage/capacitance variance becomes important when you need absolute precision ..

... The only drawback i found with mlcc is microphonics..

Yes & yes

a simple "rule-of-tumb" cure (that dosent really cure all); see my post above

[DDD]

I use smd more than 6 years in the 9-Volt stompboxes but didn't hear any difference ever.

[miech]

Using a higher voltage rating in an attempt to decrease bias drift is absolutely the wrong way. You usually need to decrease the voltage rating to have less bias drift. It's not what you would expect, but higher voltage ratings have much higher voltage drift, even at relatively low voltages.
A 100n 6V3 MLCC will have about 60% capacitance left at maximum rated voltage, while a 50V MLCC in the same package size may already be down to less than 40% at the same voltage.

"X7R" tells you exactly nothing about bias drift. It is only an indication for operating temperature and temperature drift. You MUST read the datasheet!

Larger packages usually have less bias drift, but you still need to read the datasheet to know. MLCC's also suffer from capacitance loss over time, and are indeed microphonic.

For decoupling an MLCC is overkill for most audio applications. The ESR is extremely low, so excellent decoupling becomes just a matter of PCB design.
For anything that needs a stable capacitance, an MLCC is a no-go. I do not ever use them in filters, audio paths and the like.

I do use very small MLCC's in compensation networks for buck/boost converters, which means I spend more time reading and calculating than actually doing the electrical design.

[Rob Strand]

Using a higher voltage rating in an attempt to decrease bias drift is absolutely the wrong way. You usually need to decrease the voltage rating to have less bias drift. It's not what you would expect, but higher voltage ratings have much higher voltage drift, even at relatively low voltages.
A 100n 6V3 MLCC will have about 60% capacitance left at maximum rated voltage, while a 50V MLCC in the same package size may already be down to less than 40% at the same voltage.

I've found very little correlation between voltage rating  (higher or lower).  It all comes down to specifics.    There is a high correlation between the package size and the capacitance voltage dependency.  There is also a correlation to the thickness of the package (capacitances come in different thicknesses).  The bigger the package the higher the voltage before the capacitance drops.   This at least makes sense from physics in that the E-field is lower when the part is larger.

Nonetheless the same size part across the different models of a given manufacturer, or across different manufacturers, has a wide variation - +/-50% or more.   I see two causes: the dielectric properties and the amount of "dead space" in the part  as the cap layers don't fill the entire space of the outer dimension.   Then even after that, when you measure the capacitance voltage profile you find wide deviations from the datasheet - as can be seen in articles on this issue. So, for the cap/voltage dependency I don't trust the datasheets outright.   I've spend a lot of time on this fitting curves and looking at many datasheets.  In this process I realized one of the manufacturers, I can't remember if it was the website or one of the supplied programs, used a standard voltage profile for all data in their datasheet, they just scale the voltage and capacitance on the graphs!