If you were to model an analog equalizer…say something like the Manley Massive Passive, or that Bendeth EQ, or maybe a Pultech, what are the general stages you go through in developing it? Do you start with samples of the sound then reverse engineer an algorithm that matches? Do you start with a generic algorithm then tweak it to match the components? (I think Yamaha claims they model every circuit component in compressor, then put them all together).
So excluding plugs that are convolution based, what does a modeling process look like?
I can tell you how I do it. A lot depends on what resources are available. If there is a schematic available, I use that as a guide and get the general setup done.
Then I’ll generate all sorts of test tones to see how it reacts to the test tones, and tweak the algorithm so that the plugin matches hardware on the test tones.
Then I’ll take actual music and make sure it sounds the same, and tweak it until it does. I tweak until I can’t pick out the hardware from the plugin in a blind test, then pass it on to others to make sure they aren’t hearing a difference either.
I bounce back and forth between those 3 steps until you end up with a plugin that sounds and measures the way the hardware does.
Why does the general setup consist of? Signal flow? Basically, what good is a schematic?
Just which components are in series, parallel, etc and what the order of operation is. You can also see the nature of each processing block.