There are 3 main versions so far:

The first version just used intervals and didn't allow customization of stream properties beyond font size and theme

digitalrain - version1.html

The second version was updated to use requestAnimationFrame() instead of intervals, and allowed a lot of customization of the stream properties through a UI. The interval version is actually still implemented and you can switch between versions

digitalrain - version2.html

The third and current version the UI was separated from the main logic better, and is 3 main classes I made this so it would be easier to implement by just calling the class passing in a target container element, and possible to have multiple instances of the animation, each assigned to and managing their own canvas and properties

Example 1
Should function exactly as version 2, except implements v3 class and attaches UI differently
digitalrain - version3.html

Example 2
Multiple canvas, each with their own instance of the animation, with intervals setup to randomize each one
digitalrain - version3 - multiple canvas

classes for version3

class DigitalRain
The main class that can be called passing a container element as a parameter It will create its own canvas elements inside it, and setup a resize observer on the parent container It contains all methods for generating streams, intervals, stream-controllers, updating the instances properties

class Stream
Represents an individual stream, its X & Y locations, length, and first 2 characters (since they are reprinted different colors)

class StreamController
For the requestAnimationFrame() version I needed a way to call requestAnimationFrame on a group of streams, so this class was actually created for that version

One drawback is StreamController & Stream both rely on methods from the parent DigitalRain class, so they are tightly coupled

I think there are still some opportunities to improve the performance. I read that drawing font with fillText() might not be great, and I actually created an experimental version where I print each character ONCE using fillText() on a different canvas, save each character as an image, and then print using those images with drawImage(). I think this made a performance impact when I was testing on a lower end PC, but only with a very high number of streams and 5px font size, but I have to test more.