# Basic info - Fixed point: - Encoder (fixed): f6daa6390a963ba3a8110b18cf08057832c56f0b - Decoder (fixed): f6daa6390a963ba3a8110b18cf08057832c56f0b # Issue description Static code analysis shows there is potentially a problem with the fucntion `stereo_memory_dec_fx `. in the part: ``` div_m = BASOP_Util_Divide3216_Scale( hCPE->output_mem_fx[n][i - delay_comp_DFT], hCPE->hStereoDft->win32ms_fx[STEREO_DFT32MS_STEP * ( hCPE->hStereoDft->dft32ms_ovl - 1 + delay_comp_DFT - i )], &div_e ); hCPE->output_mem_fx[n][i - delay_comp_DFT] = L_shl( div_m, add( div_e, 16 ) ); /* div_e + Q16 */ move32(); // hCPE->output_mem_fx[n][sub(i, delay_comp_DFT)] = hCPE->hStereoDft->win32ms_fx[i_mult(STEREO_DFT32MS_STEP, ( sub(add(sub(hCPE->hStereoDft->dft32ms_ovl, 1), delay_comp_DFT), i) ))]; ``` Values in buffer `hCPE->output_mem_fx ` is altered but the respective exponent in the struct is not. It is not exactly clear if this is causing problems or not but after inspecting a single testcase, it became clear that `hCPE->q_output_mem_fx `is carrying data but the shifting operation does not always yield the same exponent or Q value. It would be great if we can shed ligth in this problem. Please make a comment if this is not a concern.