Help with custom analyzer/tokenizer

Hi - I have some strings in ES (I cant control the string format) , in the format A_AValue_B_BValue_N_NValue

I'm trying to define an analyser to split them into A_AValue, B_BValue .. N_NValue tokens .
I have very little experience with custom analysers and have only used the pattern analyser before ... but as the pattern analyser defines the "separators" instead of the token patterns , I cant figure out how to accomplish this .

Any help?
Thank you

OK - actually not a very difficult regex to put together blush

     "type": "pattern"