Forum Discussion
Altera_Forum
Honored Contributor
16 years agoparrado
Thanks for the reply. Yes, I implemented the scaling logic to shift the BFP FFT outputs by the exponents out so that all results are in correct relationship to each other. I am a little confused by your integer divider example. I understand the representation of fractional bits by the multiplication by 2^7. However, when you say 32/73 would return Q=2 Rem=14, I would think the divider would return Q=0 Rem = 32. I'll play around with my simulations. I also had a major conceptual error in that algorithm. The output of the final divider (normalizing the correlation value) produces numbers that range 1.0 to -1.0. The Quotient out of the divider is not the value I'm really interested in. The remainder contains the fractional info that is supplying the phase shift between the two images. I'll have to use the quotient to know when the output is exactly 1.0 or -1.0 but all other values are in the remainder. mmeyers