I studied it from youtube videos posted by Alegorithmic . They are all there. Sometimes too long and annoying but still ok .
There are few things that covered there not well enough imo:
1. scale relations/primary input
each node of several inputs have a primary one marked with a dot . it's where texture scale goes from one input to another . If you leave it unconnected like "diffuse color" in material nodes your scale would be back to default . So primary should be always connected.
2. 2d transform node with it's default "Automatic" mip map level does much blurrier bitmap scaling than automatic in Photoshop. in material transform node too. I suggest to set it to manual Zero level
3. Don't waste you time studying FX-map constructions . Imo it;s from soft past. Now "shape splatter" is everything you may need and could do all the same.
4. All textures start from creating depth and then go ahead with gradient node remapping like gradient filter in Photoshop to produce all other channels including color one.
5. All the height generation is based on few basic manipulations:
a. Shifting a pixel along a vector(direction) defined by "slope" , fading gray-scale values in a secondary input ( usually a noise texture)
or by normal map itself where each red and green value also represent a vector (direction) considering zero as a starting point of such vector.
it's Warp , vector warp, vector morph, etc
Pixel shifting occurs only in shades of gray of secondary input . 0r rather where gray values of pixels differ in neighboring pixels. Once it's flat black, white or any other flat value like flat normal color in vector inputs the shifting stops immediately
And Directional warp where shifting direction is constant and the secondary input introduces not the direction but rather distance of a pixel shifting .
For some uncertain reason there is no node that dose both kind of pixel shifting at the same time. But you can use pixel processor and a bit of basic trigonometry math to make it yourself
b. Distance node working like glow effect in Photoshop. Making fading glow around white pixels toward blackness of a background
Bevel node makes the same basically with a bit more conveniences
c. edge detect Speak for itself . Something to feed into distance node for example
d flood fill node Making isolated areas to fill
e. pixel processor to apply simple arithmetic on pixels , just allowing more complex arithmetic constructions than Blend mode
float - numbers like 1.5
float2 - 2 such numbers like x and y coordinate of a pixel for example
integer - a number like 0, 1, 2 or -1
get float or get integer - something to get those numbers from other inputs or parameters . getFloat2 with $pos from drop down list is a way to get XY position of each pixel to be than fed into "sampler" node is most common starting point in pixel processor that make an image input appear in the node for further manipulations.
f. blur - speak for itself and "Slope blur" - most used node in SD probably . If set to min or max blending modes it makes gradual shift of pixels in direction defined by "slope" , a secondary image grayscale fading same as warp node but doing this in multiple (up to 32) iterations of such warp each time in more transparent way fading eventually too. Used to fake erosion like features and weathering drips/ smears
g. MAx and Min blending nodes - same as lighter or darker blending modes in Photoshop. Basic things to combine depth/height . And 'Subtract' one to subtract one of initial height from two already mixed by MAx to make a mask of such height based blending for color info.
Once you figure out this basic concepts everything other is just a few extra conveniences
Replies
And Directional warp where shifting direction is constant and the secondary input introduces not the direction but rather distance of a pixel shifting .