The intrinsic performance of the ATLAS barrel and extended barrel calorimeters for the measurement of charged pions is presented. Pion energy scans (E = 20, 50, 200, 400 and 1000 GeV) at two pseudo-rapidity points ($eta$ = 0.3 and 1.3) and pseudorapidity scans ($-0.2 < eta < 1.8$) with pions of constant transverse energy ($E_T = 20$ and 50 GeV) are analysed. A simple approach, that accounts in first order for non-compensation and dead material effects, is used for the pion energy reconstruction. The intrinsic performances of the calorimeter are studied: resolution, linearity, effect of dead material, tails in the energy distribution. The effect of electronic noise, cell energy cuts and restricted cone size are investigated.