in c7ca9d3, a division by zero can occur with the following bitstream : [M1stereo_-dtx_16400_32.zip](/uploads/ae83c42c7432f12da1542ebb43ff893c/M1stereo_-dtx_16400_32.zip) Command line : IVAS_dec STEREO 32 M1stereo_-dtx_16400_32.bit syn the issue happens in the function stereo_dft_dec_smooth_parameters_fx() at line 895 where **STEREO_DFT_ITD_CNG_XFADE == hStereoDft->itd_xfade_counter.** The same issue seems present in the floating point decoder. @norvell, @multrus, @vaclav any idea if the latest floating encoder can still generate this kind of bitsteam? if yes, then both float and fixed decoder should be corrected. hStereoDft->itd_xfade_step_fx = BASOP_Util_Divide3232_Scale(L_sub( hStereoDft->itd_xfade_target_fx, hStereoDft->itd_xfade_prev_fx) , L_shl(**sub( STEREO_DFT_ITD_CNG_XFADE, hStereoDft->itd_xfade_counter )**, **Q15), &q_val);` Floating-point equivalent : hStereoDft->itd_xfade_step = ( hStereoDft->itd_xfade_target - hStereoDft->itd_xfade_prev ) / **( STEREO_DFT_ITD_CNG_XFADE - hStereoDft->itd_xfade_counter )**;