Hi guys,

I have a question about filter capacitor size in power supplies. So if you look at the datasheet on a regulator like the 7805 or something, they always say put a 0.1uF cap on the input if it is more that 6 inches away from the regulator. So basically you see caps like 0.1uF, 10uF, 22uF, etc... used to filter out transients in the voltage at either the input or output stages of the regulators.

I'm wondering how the size of the capacitor affects the transient rejection performance. If I put a very small cap, say 1pF that means the cap will not have enough energy to stabilize the voltage if it drops for a few milliseconds right? So a bigger cap is better for this purpose? Lets say I replaced it with a 1mF capacitor, will that be much better at rejecting transients or will that cause other problems?

I tried looking around but couldn't find anything definitive.

BTW, I'm not talking about the smoothing capacitor. I don't know if these transient filter caps have a better name than simply filter capacitor.

Thanks.

 

Hello,

The small capacitors near the stabelizer are used to prevent the stabelizer from oscillating.

Greetings,
Bertus

 

Bigger caps have more inductance. They might start to act like RF chokes in series with the capacitance. In other words, the cap will not be able to charge and discharge fast enough to bypass very high frequencies. It would be like putting a large resistor in series with the cap. Makes it hard for the cap to do it's job.
So you use two caps in Parallel - one for high frequencies, one for low.

 

Bigger caps have more inductance. They might start to act like RF chokes in series with the capacitance. In other words, the cap will not be able to charge and discharge fast enough to bypass very high frequencies. It would be like putting a large resistor in series with the cap. Makes it hard for the cap to do it's job.
So you use two caps in Parallel - one for high frequencies, one for low.

Whilst all this is true, it is not applicable in this case.

Bertus has the right of it. The IC contains a high gain feedback amp. Any high gain feedback amp will oscillate if conditions are right - not a desired situation for a regulator.

 

Hi, thanks for the replies.

Can you explain why the IC will oscillate? Does that mean the output voltage will vary periodically? Why does this happen?

Also how did you choose values of Ci and Co? Thats what i'm really wondering.

Bigger caps have more inductance. They might start to act like RF chokes in series with the capacitance. In other words, the cap will not be able to charge and discharge fast enough to bypass very high frequencies. It would be like putting a large resistor in series with the cap. Makes it hard for the cap to do it's job.
So you use two caps in Parallel - one for high frequencies, one for low.

Whilst all this is true, it is not applicable in this case.

Why is this not the case? I'd really like to know the details about making power supplies.

Thanks.

 

I'd really like to know the details about making power supplies.

A worthy aim.

Why is this not the case?

I didn't notice any large value ripple smoothing caps in your circuit. You would still need the anti oscillation ones if you fed the IC from a DC battery.

Can you explain why the IC will oscillate? Does that mean the output voltage will vary periodically? Why does this happen?

I've already observed that the IC contains a feedback amplifier. The circuit connections, leads, pcb tracks etc will introduce unpredictable stray capacitance and perhaps inductance into the regulator circuit. This stay reactance will produce a phase shift between input and output of the feedback amp. At some frequency the phase shift will make the feedback positive. This is the condition for oscillation, if the gain is high enough. Because the stray reactance is not large the frequency of oscillation where this happens will be relatively high many Khz or even tens of Khz.

Also how did you choose values of Ci and Co?

The capacitors are not feed back capacitors, they provide a short to earth at the high frequencies involved, without offering a significant load to the IC, at lower frequencies the power supply might encounter as loads vary.

Most people just use the manufacturers recommended values, gained from experience. Special values are not normally recommended in circuit design for repeatability reasons.
Of course higher value electrolytics are unsuitable for the reasons given by flat5.

 

Check the datasheet for the specific regulator you will be using, they show the recommended values for different voltages and loads. Some assume ripple has already been removed, other datasheets have information to help with that as well.

The only "rule" is that the capacitors should be present, and not the same value. Typically input capacitor >> output.

 

You are looking for some design formula so you can plug in values. It does not exists with experience you will find out that a small .1ufd ceramis cap will do you more good then an axial 10ufd especialy if located close to the chip pins.

 

Thanks for the info guys, that was really useful.

I'm using an LM338 which indicates on the datasheet that the input should be 0.1uF and the output should be 1uF.

I'm wondering if it is beneficial to put capacitors in parallel, for example on the input put a 1uF and a 0.1uF and on the output put a 10uF and a 1uF. Assuming I want the best possible precision and lowest possible noise, ignoring cost (more of a thought experiment i guess...)

Also, I read that tantalum capacitors have better ripple rejection performance than ceramic or aluminum capacitors. Why is this? The Wikipedia page didn't go into details about ripple rejection.


Thanks a lot.

 

Have a look at these:

http://www.analog.com/library/analogDialogue/archives/41-05/ldo.pdf

www.aeng.com/Articles/LDOArticle.pdf

www.national.com/appinfo/power/files/f4.pdf

 

The following links tell u how n why to select the capacitors in switching and linear regulators
http://www.national.com/nationaledge/jul02/article.html
http://www.national.com/nationaledge/jul02/article2.html

 

Thanks for the links, that info was great.