Noob ARMA questions

[Dragoneo]

Hello,

i'd like to ask for help because i read a lot of threads and posts but still i can't seem to find a solution. I'm currently trying to forecast gold, silver and platinum prices using an ARIMA approach. I started by checking for non-stationarity, and then i did the first difference of the logs, which solved the problem. The problem then was that i noticed that there was seasonality in gold:
i tried solving it as lngold=ln gold-lngold(-3), as the seasonality was in the 3rd quarter; in alternative, i tried to use TRAMO/SEATS to correct seasonality and funny thing - it gave me an automatic ARIMA model specification (3,0,0)x(0,1,1), which i changed to (3,1,0)x(0,2,1) because i ran the test in a I(1) series, and i tried to estimate but the results are not what i expected -probably something wrong with the estimation formula i used:
dlog(gold,2,3) ar(1) ar(2) ar(3) sma(1).

this is my first time working with time series and even though i have been able to understand the major parts, i'm still an ignorant noob in all the rest. As such,i'd like to ask several things: first, can i use the model specification obtained from TRAMO in OLS or changes significantly? second, did i treat the seasonality in any alternative? Third, what is the code to write a SARIMA model (2,1,0)x(0,1,2) for example?
finally, i followed a cookbook a colleague suggested and one of the steps after the forecast was to check the ARMA structure - the correlogram and impulse response but i can't understand why, so if anyone could explain it to me or suggest some materials i'd appreciate.

Sorry for the inconvenience, but the more i read the more confused i got....
Thank you