# Basic info - Float reference: - Encoder (float): 9a10acdb8466f56d9f53e21963bd96e32638d6ac - Decoder (float): 9a10acdb8466f56d9f53e21963bd96e32638d6ac - Fixed point: - Encoder (fixed): 69eb54668b0ee7424b6210d0c6113c727606ae76 - Decoder (fixed): 69eb54668b0ee7424b6210d0c6113c727606ae76 # Issue description In function `unclr_calc_corr_features-fx()` variable `hStereoClassif->ave_ener_L_fx` is computed: `hStereoClassif->ave_ener_L_fx = Mpy_32_32( hStereoTCA->E1_mem_fx, 13421772 /* 1/length in Q31*/ ); /* Q31-hStereoClassif->ave_ener_L_fx_e */` Whereas in the floating point counterpart uin it computed using a dynamic denominator that is also present in fixed point implementation. `hStereoClassif->ave_ener_L = hStereoTCA->E1_mem / length;` by tracing length to higher stacks it is not exactly clear how the two implementation can yield the same result. Please fix it as it can create bugs.