Are you really fine with IC voltage regulators ?

[MarcelvdG]

Proof?

I just looked at datasheets of some manufacturers (Samsung, Tayo Yuden, Kemet). No other difference than temperature specified between X5R and X7R.

Murata 22 uF 16 V 1206 X7R

Murata 22 uF 16 V 1206 X5R

Both are pretty awful to me, even though the X5R capacitor is indeed worse than the X7R in this example. The working voltage and manufacturer also make a big difference, most of Murata's class-II capacitors are even worse than those of their competitors. You typically have to search the websites for characteristic curves to find data like these, the datasheets barely contain any useful information.

 

[jean-paul]

With ceramic class II caps it is indeed not the nicest to see but we need that low ESL. In that aspect Cornell Dubilier's FCA1210C105M-G2 is OK. Not liked by some, I don't know why. Maximum value is 1 µF 16V however. What you conclude about Murata is right, same counts for tantalum caps of various manufacturers. I recall Murata having the highest values in certain sizes but I may be wrong. I use ceramic caps because I have to and reading their datasheets is not my hobby. If everything could be done with film caps I would choose for those.

The waiting is for a 4.7 µF 16V C0G in size 0805 or 1210

 

[bohrok2610]

View attachment 1026657
Murata 22 uF 16 V 1206 X7R


View attachment 1026658
Murata 22 uF 16 V 1206 X5R

Both are pretty awful to me, even though the X5R capacitor is indeed worse than the X7R in this example. The working voltage and manufacturer also make a big difference, most of Murata's class-II capacitors are even worse than those of their competitors. You typically have to search the websites for characteristic curves to find data like these, the datasheets barely contain any useful information.

But when these are used as output caps for regulators how much variation do you expect in the DC bias?

 

[jean-paul]

Just derate their capacitance values when DC bias is around (honestly, when isn't it?), also choose larger casings and higher voltage ratings than strictly necessary. Using caps with a better dielectric is advisable. I think my remark why I use X7R as a minimum standard holds some truth or not?!

From our friends at Murata:

https://www.murata.com/en-eu/support/faqs/capacitor/ceramiccapacitor/char/0005

 

[bohrok2610]

I hope nobody is using either X5R or X7R capacitors close to their rated voltage. I typically use at least 15V rated caps for 5V output.

 

[EUVL]

4.7 µF 16V C0G in size 0805 or 1210

Rubycon 16MU335MB33225, 3.3µ, close enough.

The working voltage and manufacturer also make a big difference,
most of Murata's class-II capacitors are even worse than those of their competitors.

See attached.


Patrick

 

[bohrok2610]

My point was that once you have done the derating for the voltage the only variance for regulator output cap comes from the temperature. In this respect X5R and X7R are about the same (X5R slightly worse).

 

[bohrok2610]

We do know that some humans can hear certain types of effects in dacs at very low levels, yet not hear other things at higher levels.

I also know some people who keep hearing things. My advice to them is to be silent about it.

 

[Evenharmonics]

but we still need to check audibility experimentally.

How is such experiment done?

 

[MarcelvdG]

As no-one guarantees what the capacitance-voltage characteristic looks like and as class-II capacitors also have aging effects, you can't predict very well what the capacitance is going to be, not even if you know the type number, voltage and temperature.

 

[gerhard]

I simply put them on the bridge and feed in a DC bias. Works for varicaps, too.

 

[Bonsai]

Here is the OPA1642 ; 50 dB PSRR at 20 kHz in the worst case

_

That’s still superb at 20 kHz

Here’s the 4562

 

[Bonsai]

As no-one guarantees what the capacitance-voltage characteristic looks like and as class-II capacitors also have aging effects, you can't predict very well what the capacitance is going to be, not even if you know the type number, voltage and temperature.

You don’t have to for DC decoupling though do you? If you want 0.1uF, just look at the curves for V vs C and select as appropriate taking into account th eg worst case scenario.

 

[Bonsai]

AVCC; 0dB PSRR in the general case.

That’s why a local reg makes perfect sense- I had the same issue with X-Altra since both NC and MM stages had essentially zero PSRR.

 

[Bonsai]

What about including an estimate of whether or not humans could possibly hear such a thing? Without that what good is an estimate for the magnitude of electrical effect in a dac?

We do know that some humans can hear certain types of effects in dacs at very low levels, yet not hear other things at higher levels. ESS is very clear that there is good experimental evidence for that, and they continue to try to refine their chips to satisfy what they think some audiophiles can hear, not just what is measured at ASR.

The message to me is that we can calculate and measure electronics all we want, but we still need to check audibility experimentally.

Agree you should listen but I have a problem when we try to explain differences in sound with hypothesis that don’t stand to to scrutiny.

 

[MarcelvdG]

You don’t have to for DC decoupling though do you? If you want 0.1uF, just look at the curves for V vs C and select as appropriate taking into account th eg worst case scenario.

True, except that you don't really know what the worst case is when no-one guarantees anything. You can do sandbagging and hope for the best.

 

[jean-paul]

Rubycon 16MU335MB33225, 3.3µ, close enough.



See attached.


Patrick

Not C0G but a film cap nonetheless. That is a good tip, Patrick. Thank you. Did you use these yourself?

 

[bohrok2610]

You can do sandbagging and hope for the best.

As a (former) chess player I know what sandbagging in chess means but what does it mean in electronics?

 

[Bonsai]

True, except that you don't really know what the worst case is when no-one guarantees anything. You can do sandbagging and hope for the best.

Come on Marcel. Curves are provided. As a designer you are supposed to use your experience and size the caps accordingly.

 

[jean-paul]

As if Marcel does not know