I'm using this 3v3 regulator in my design to power an SD card:
It has a bypass capacitor which is optional, and the value they recommend for it is 10nf, ie, .01uf.
The thing is I'm designing these boards for manufacture, and this is the only part on the board which needs a .01uf cap. Since more types of parts equals more money and lower discounts on bulk part buys, I'd like to swap it for a .1uf cap. But I'm not sure how that will affect the noise performance.
In fact, I'm not even sure if I need the cap there at all. But if you look at page 11 in the datasheet you can see two graphs side by side for the load transient response, which indicate that I'll get significant voltage spikes if the load changes rapidly without any cap there. So I'd like to use a cap because better safe than sorry, and I'll have a 3v regulator design I can use for other projects without worrying about it, but I don't know if a .1uf cap will be any better than no cap at all in this circumstance. They have some examples of other cap sizes on some of the other graphs, but they're all for smaller caps.
On page 15 there's also a warning about leakage current:
"Thus, for best output accuracy, dc leakage current through CBYPASS should be minimized as much as possible and never should exceed 100 nA."
And I'm not totally sure what that is and if a larger cap will have more of it.
TLDR; Would it be a bad idea to use a 0.1uf cap on the bypass pin of this regulator?
Chances are you'll have no problem with a cap that's within an order of magnitude of 10nF, especially since it's optional anyway. That's just my opinion though, and I'm no electronics expert. I'd be more worried about filtering the actual supply, as an SD card's current consumption can spike up to 75mA during writing.
The size of the cap relates to the frequency of the noise; higher frequency = smaller cap. Having any cap is always better than none though. Multiple caps are good too for multiple frequencies of noise.
Leakage current relates to the insulation resistance of the cap (there's always some amount of current flowing through the cap because no cap is perfect). Ceramics have very low leakage compared to electrolytic or tantalum; assuming you're using a ceramic then I wouldn't worry about it.
I'd suggest sticking with the .1uF but still design the board with an additional spot for the recommended capacitor if you find that the need arises.
I found this discussion about the cap size on TI's website but nothing specific about using a larger cap, just that the 10nf cap will have the lowest noise over the bandwidth of the regulator, whatever that means...
scswift:
I found this discussion about the cap size on TI's website but nothing specific about using a larger cap, just that the 10nf cap will have the lowest noise over the bandwidth of the regulator, whatever that means...
Yes, as I mentioned, the size of the cap is related to the frequency of the noise. In this case the regulator is running at "X" frequency thus "Y" capacitor size is recommended.
The bypass cap mentioned in this thread is only to clean up the internal voltage reference on the chip - input and output capacitors are mandated for this regulator and it has quite detailed section on the requirements on these components for stability - don't omit those!
Unless you require particularly low noise supply (powering very sensitive analog circuitry for instance) the bypass cap isn't needed I think.
I have the caps on the input and output. I just don't have a clue how much noise is too much for various applications. I don't think an SD card and the logic level shifter I'm using to access it requires a particularly low noise level but I simply don't have a clue how close the noise level of this regulator would be to an unacceptable level.
I guess what I will do if you think the cap isn't needed is leave the spot unpopulated. Then I can go back and solder one in place by hand if I get back my prototype and there's an issue.