Hi, all, I met a probelm in my project, please help me.
In my project, many signed data ( each is 28 bit) are added. In order to avoid overflow, the data are extended to 32-bit. Thus the result is 32-bit. But the output is only 28 bit, how to choose the 28-bit in the final result? the most siginificant 28 bits?
That's a decision you'll have to make based on your system specifications.
You need to tradeoff between allowable distortion from saturation/overlow and dynamic range. It sounds like you have calculated the worst-case bit growth and that you cannot tolerate any saturation/overflow. In which case, you will need to use the MSBs.