As mentioned above, Base64, is not an encryption but an encoding. As you can see at the RFC that specified the standard, base64 works the following way.
- You have a stream of characters s, of length n.
- You read 3 8-bit values from the stream (now you have a total of 24 bits = 3 bytes)
- You break these 24 bits to 4 groups of 6 bits each
- Using the table of the Base 64 alphabet, you encode each of the 6-bit groups to the Base64 equivalent
Now, there is a chance that you reach the end of the stream and you don't have a 24 bit group (s mod 6 != 0). If this happens, then you add zeros to the end of your input, until you have an integral number of 6 bit groups.
Given that your input stream is ASCII encoded, so it's composed of 8-bit characters, there are only two cases where you end up in the above scenario.
1) You have 8 bits in the last group
2) You have 16 bits in the last group
In the first case, 4 zeros are added (giving you 12 bits) and the output would be two characters (2 * 6 bits = 12 bits) encoded based on the alphabet, and two "=" padding characters
In the second case, 2 zeros would be added (giving you a total of 18 bits) and the output would be three characters (3 * 6 bits = 18 bits) and one "=" padding character.
That's how sometimes you end up with one, two, or no "=" at the end of the encoded text. For more info you should really read the RFC which defined that standard and the wikipedia entry related to it.