With an auto, 24" and 100% on an 18/6 light cycle might be about right... it's a good place to start. Simply observe the plant. The distnace between growth nodes (internode) will tell you if you need to get the light closer or further away -- you judge how short or long the internode is and adjust accordingly... once you find a happy zone it'll be nearly the same every time with that light. I'd still suggest always letting the resulting internode length determine and final adjustments for each plant grown.
With photoperiods, you want to cover an area in vege that is 66% of what you povide in bloom -- if you want to maximize DLI vs maximize growing space. In the end you give the same amont of light to a mature vege plant and a plant in flower phase... you can see that 12 hours is 66% of 18. You need 66% of umol/s (this is a rate) over 18hours... and 150% of that when you go to 12/12, or 66% * 1.5 = 100%.... This is best if working backward from a general "max" DLI, which will be 35-40DLI for atmospheric co2 and loosely controlled environment, otherwise.
Read up on daily light integral.. think of light as indivudual photons of energy... umol/s gives the rate at which they are produced from lights...
if you give "50 umol/s" over 12 hours or 18 hours it's the same amount of energy per 24 hour cycle. One just recieves it a bit more slowly, but the plant can handle the same amount per 24 hours regardless of 12 hours or 18 hours per day of light.
Daily Light Integral - read up on it... umol/s // PPFD are good to understand.. you don't need to memorize the math.. but understanding the gist of it will allow you to better understand how best to apply light. Maybe you want to try to cover more area with less DLI... might work out to a larger yield, but won't be as dense of buds, for example. maybe it's an irrelevant difference? trial and error.. i usually work backward from providing 35-40 DLI to my area of garden. doesn't mean you have to.. and i don't know if it is mathematically the best either.. the cost effectivness of it would need some data points and then some sort of calculus with multiple indendent variables could easily show a curve that would definite "optimum" of it all... to my knowledge i've not seen such research.