not necciserily hard on the threshold. the threshold is input dependant. meaning if you have a strong signal going into the comp then you may not need a low tresh. it also depends on the dynamics of the original signal.

the ratio is also dependant. if you have a dynamic signal you'll need a high ratio to level it out.

thats why i suggest an opto LA2A style comp. i use the UAD la2a



its dead simple. just raise the peak reduction until your signal is at the desired "flatness2 and then crank the output. uad also model the output distortion.. which is nice