Okay, you have a 300w in a 3x3. If you are 100% at 17" that's almost certainly too much light without co2 supplementation. You are giving it 12/12 power over 18 hours. You are probably giving 150% of local limitations. If my guess is correct, you need to turn the light down to ~60%.
--- less likely but may need attention after fixing the light....
Looks like K-def but that should happen at middle of plant not at top, so it's possible it's too much light and the plant is cannibalizing itself where the intensity is too much -- this sort of thing can take a longn time to present itself, too.. just depends on severity.
Anything that locks out K is also on the table - see Mulder's Chart for list of possibilities.
Your pH will swing because the plant does not take in all nutrients equally. So, pH can swing. Better nutes should be pH buffered to mitigate this to some extent, but still need a rez refresh at some point. Topping off and a full drain and refill can be a cyclical thing you learn through trial and error to avoid it. MAybe, a better ph-buffered fertilizer setup could help too.
--- more details about light if you care to read... please note that the quantitative things i say about ac infinity are mathematically true and not exaggerated, i do not like ac infinity and will embellish the way in which i talk about them in other less relevant ways, lol.
Fuck, ac infinity specs on that light are total nonsense lies... 1819 ppfd? first, that requires knowing what area it is focused on, because 300w and 1819 would be incredibly small area relative to 3x3. If it were focused on half that size assuming their efficacy claims are true, which they most likely are not, 1800ppfd makes sense but makes no sense when you say it is for a 4x4 vege and 3x3 flower, lol. The efficacy of this light will be a bit reduced from samsung evo specs because they are running the diodes 150% more wattage than testing parameters. They went cheap on diodes because AC infinity is dog shit company with good reputation and very average products. They are willing to lie and misinform on their light specs, becuase they are dog shit.
I am making some guesses.. the actual umol/s PAR output of this light is around ~850 based on a 2.8umol/J efficacy guesstimate. Still good efficacy. 3.14 is just not possible unlesss it's a light with ALL 6500K CCT dioes (or whatever the CCT of testing parameters on smasung.com is... 90% sure it's 6500K), which would be a shitty CCT for a grow light, lol. They clearly have other diodes, so 3.14 is just a blatant lie - on tp of running them too "hot" at .33w/diode, give or take a hundredth.
On 18h, i'd run this light in a 3x3 around 60% and go from there (~38 DLI or thereabouts). ~18" is a good starting spot. Further fine-tuning can be made based on observing growth pattern -- if too tight, needs more height or less power, and if it is too lanky, the opposite. Getting closer is not all roses. Closer with less power generally means less penetration - has to do with out light spreads out accoring to inverse square law, i.e. it spreads out very quickly. Greater distance and more power = greater depth of more intense light -- this too is not all roses... a happy medium balances cost and useful depth. If it's heavily shielded, no amount of penetration reaches those depths anyway, so more is not always better or useful.
Their spec is actually coherent in this one regard -- 4x4 vege and 3x3 flower coverage. Though i think they err a bit on the high side by about 10%. So, i'd recommend 90% for 12h 3x3 and 60% 18h autos in 3x3. The math is in balance, but depends on my guesstimate of actual power. IT's difficult when the lightmanufacturer lies like dog shit.
It is an absolute certainty that running the diodes at .33w/diode vs the .2w/diode testing parameters on samsung.com will reduce the stated "3.14umol/J" efficacy -- which is only for the 6500K CCT chips anyway... lower CCT will also come with reduced efficacy, which only amplifies the context further. These deceitful motherufckers, lol. It's a decent light. It's just not top-shelf efficacy because they went cheap on the diode count. They needed 150% more diodes but decided to maximize profits while people falsely perceieve they are high-end equipment.
2.7-2.8umol/J (mutilply that by watts and you get umol/sPAR production and not some bullshit made up PPFD ac infinity gives with a distance they don't provide and area of focus they also neglect to mention - important for a metric that is umol/S PAR per meter-squared) is a saner guesstimate and because they run hotter, they will also degrade faster, too. So the advertised 50,000 hours before they hit 90% of original intensity is also a lie. It will dim faster than that. Longevity curves are significantly impacted by overpowering the diodes. Maybe 35-40k hours is a better guess before it dims to 90%. If they show the longevity curve on samsung.com you can make a better guess. visualize the resulting difference of the area under the curve and it'll head downhill around the same total energy blasted through it, i bet... good enough for a guess but not entirely accurate.
Read up on DLI... it's not about watts..
it's about photons. watts can mean different rates of photons produced,therefore watts are a shitty way to dictate light usage. Shoot for upper 30s DLI to start, than adjust based on observations -- local variables will make "your" max DLI different from someone else's garden. You need to observe and fine tune. This math is a tool, not an exact answer. It will be 100x better than a wild stab in the dark to start. That's the usefulness... gets you at a much better starting point with minimal adjustment from there.