Capacitor

Sorry but your initial question is so vague and with very little pertinent information, I cannot directly help you. But here are a few pertinent equations that should permit you to understand some of what you seek:

1. First a few definitions
1. q≡charge in Coulombs
2. C≡capacitor value in Farads
3. t≡time in seconds
4. V≡voltage in volts
5. I≡current in amperes
6. Δ≡change of a value
2. Now a few of the equations
1. V_cap=q/C
2. I=(Δq)/(Δt)
3. (ΔV_cap)/(Δt)=I/C

This is as simple as I can get and still be reasonably accurate. If you prefer there's Wikipedia capacitor link. With some careful thought, you may realize that filtering alone will never fully remove any oscillation, just reduce the magnitude.

It depends on the load being drawn.
The simplest way is to put on 1000uF, if that works fine (at full load), try something smaller say 470uF until the ripple voltage is more than you want.
Then put on a value at leas twice as big as you need to just do the job (These Caps are often +/- 20% tollerance.
Del

Hola

Here Is a Link Where Explain How To Calculate The Power Supply Capacitor.

http://www.kpsec.freeuk.com/powersup.htm

the capacitor should be connected to the bridge rectifier according to the "+" and "-" the bridge and the capacitor

Can be calculated arbitrarily saying they are 1000uF. per ampere. But See The Link.

Greetings

I agree that the OP's post was vague, probably because the OP is not experienced in PS design. Power supply design is complex by its very nature but, on occasion, a simple filter will suffice depending on the load's requirements. In power supply design, you must make estimates of your requirements before you can go anywhere. It requires more than just a output voltage to begin with. For instance, push-pull output circuits are relatively immune to ripple, a few volts peak-to-peak are of no consequence but to a class A amplifier, such as a preamp, which has a very poor power supply rejection ration (PSRR), ripple reduction is very important (even more so is high frequency rejection).

The OP asked about how much capacitance to use, unfortunately without detail, RedFred gave basic quantities involved in power supply design, I gave a rule-of-thumb formula which generally works very well in most instances. That is about as far as one could go given the lack of data. The other replies, basically gave varying degrees of design by brute force. These, of course, work because they simply overwhelm the requirements. Del the Cat's method is the least heavy-handed of the group and is quite acceptable for the trial-and-error route. There is nothing particularly wrong with this approach but it can put an unnecessarily high demand on the rectifiers and transformer due to exceedingly high peak currents, not to mention causing lousy power factors. This is not good practice nor good engineering.

For those of you who are not well versed in power supply design, there are rules-of-thumb which are applicable in many cases, such as the one I gave. In cases where the design is more complex, rules-of-thumb are only a beginning point at best. Ripple is only one component of power supply design and many other points are involved in designing a stable, clean power supply with good high frequency rejection and transient response.

For those beginning, trial and error can be a way of learning about power supplies but it cannot substitute for good engineering knowledge either. I do recommend careful trial and error but buy yourself a good basics of electronics book to use as a guide. There are many basic tutorial books available to help and a lot of material on line as well to guide you. Always ask questions when in doubt!

I, with qualifiers, disagree with using only one cap. I (from experience) think you should have one small cap(.1Uf) in parallel with the larger value cap(i.e. 1000Uf). The small one is not a necessity but it will help. This applies for powersupplies that will be used on circuits containing chips that will be switching states in normal usage.