Yes, Victor, that is it. lpanal does not like long silences. I too cut the silence in the beginning of the sample and now the analyzing process runs.  Then, to test it further, i transformed a few words in the middle of this sample into silence, to see if in the lp analysis would continue once it had started... but lpanal could not be tricked; running the example i heard the first few words and then the buzz opcode sort of hangs in a  low amplitude state. So it ran for a while but in the Csound output: found only 0 poles...sorry.