Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
15 years ago

FFT Megafunction Fixed Point

I'm trying to use Altera's FFT Megafunction in Fixed Point mode but I didn't find any documentation defining how many bits it uses for decimal part. I'm using the Variable Streaming because its the o...
Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
15 years ago

the FFT core only has a sense of decimals when using floating point mode. fixed notation is largely an abstraction

since the FFT is a linear operator, you should be able to imagine multiplying your 26:6 input word by 1,000,000 to force it to 32:0, a full integer. then you can divide the FFT output by 1,000,000 to rescale

the FFT output is somewhat relative due to processing gain and any windowing. if you convert to decibels you could just subtract 20*log10(1,000,000)=60dB from each value