It very much depends on the voice and the track, but I tend to go back to certain things.
Doubling is very voice and performance dependent. Our Singer's voice is quite harmonically rich with a lot of gravel so it takes up lots of space, and doubling it tends to sound messy even if the performances are very close together. My voice is more mellow and thin and doubling works better, even if I don't perform that tightly between the takes.
There's also a rule that I learned from my girlfriend/partner/overlord who does a lot of choral singing with orchestras etc. It's basically that if you're going to have more than one instrument be it voice or viola doing exactly the same part, it should be three or more. The reason is that if there's only 2 doing the same thing the differences between them sound like mistakes. Once there's three or more, there's enough going on that the aggregate sound becomes one entity that just has extra depth and thickness. I find that to be very applicable when doubling vocals.
If I do double, I do one main take and two background takes that may end up anywhere from -6 to -12db quieter than the main take. They're just there to add depth and thickness and this way they don't end up confusing out brains which are
very very very good at identifying and processing human voices, to the point that an unnatural effect such as a multi-tracked voice is immediately flagged as strange sounding.
Generally I'll put a lot of compression on the main vocal - and I do mean a lot. This is where mic technique and selection becomes very important as plosives, breath noises, handling noise, sibilance etc will all be brought forward by the compression. I find pop shields can increase sibilance, so often I'll avoid using them all together if I can get away with it.
Usually I'll compress in two stages. The first stage is about reducing the dynamic range of the whole performance in a relatively transparent way. So if you imagine you sing;
"
I am singing an example of a song"
You might find that the highlighted syllables are stressed and on record might peak 6db or more higher than softer syllables. It might have sounded fine in a room but ultimately in the final mix a 6db difference might result in the quiet parts being totally lost in the mix. So my first move with the compressor will be to even out the volume difference. I'll use a relatively mild ratio, maybe 2:1 or a little more, and set the threshold so that the quiet syllables barely brush the db reduction meter - anything louder than that gets a 2:1 gain reduction, which means the 6db difference turns into a 3 db difference. Attack and release will be very performance dependent but usually I use the quickest attack and release I can get without artefacts being audible - pumping, unnatural transients etc.
Then I'll use a second compressor in series with the first that is much more song dependent. For a heavier rock song it might be set to a 4:1 ratio, and the threshold set so that the entire already compressed performance lives on the gain reduction meter. Or I might just set it to come on when the vocalist really lets loose - it just depends on what I think serves the song best. I might not use it at all.
In the case of the heavier songs though, it's almost a certainty. If the compressor is doing a lot the attack and release become very important - you can really control the aggression, how much the vocal leaps out and grabs you, with the attack. You don't want to smother the emphases in each word so release time is important for maintaining control without suppressing for example the second syllable in every word.
A note about sibilance - all that compression makes it much more likely you'll need to do something. The right mic choice and performance technique go a
long way to dealing with it, but if it's a problem you've got various options ranging from a VST de-esser to EQ (though this often makes it worse) or manually moving the S sounds in the performance onto a separate track to process independently. This last option is totally tedious clownfuckery but sometimes you've just got to bite the bullet.
After all this dynamics processing automation becomes basically essential, this is a performance aspect of any mix where you as mix engineer construct a dynamic that works for the song.
As for other effects, the world's your oyster. I really like a relatively bright reverb that catches the top half of the voice's tone - you get space without muddying up the midrange. I also as a rule prefer delay to reverb for creating ambience. Quite often a stereo analog or tape sim delay, band-passed to be mainly say 500hz-3k and designed to soften up/degrade with each repeat, will make me cream myself and I assume listeners will cream themselves too. This is a vital litmus test: if you don't cream yourself, then you should include a note of apology with every copy sold.
There are various other thickening options like detuning doubles that are panned left and right, or even good old fashioned harmonising. It's all good fun!
The Coraxo EP has almost everything double tracked. Granted this approach wouldn't prolly work for a sparse mix with clean vocals, but for screaming it works pretty well. I also edited some flubs by zooming in on the wave form and moved things around a bit.
It's also good to go through a few mics to find one that sounds the best. I was originally planning on doing the vocs with an SM7B for the EP, but went with the Beyer M201 because I felt it sounded better for my voice and the mix.
If I was going to do it, I'd probably use an LFO - either on a chorus or a flanger VST - to add some movement to the duplicated track.