Sundance FC100 v.2.3 - Manual (Page 3)

FC100 - Floating Point Fast Fourier Transform
v2.3
Fast Fourier Transform product manual
October 2005 www.sundance.com - 3 -
Data format
This core is compliant to the IEEE-754 standard for Binary floating-point arithmetic.
There is however one restriction regarding the optional denormalized numbers of this
standard that are not supported by the Floating Point FFT core.
Other data formats available for this core are coded in 2’s complement for both the
mantissa and exponent.
The 8-bit exponent ranges from -128 to 127
The p-bit mantissa ranges from -2
p-1 to 2 p-1 -1
The exponent bit width is noted Ebw. The mantissa bit width is noted Mbw.
Parameters and Ports definitions
Parameter name type Value Description
addr_width integer ≥ 8 and
≤ 20
Address width. This parameter (also noted Abw)
indicates the width of the address bus for twiddle
factors and data. If N is the maximum transform length
used for computing the FFT, then Abw=log2(N).
Please note that the transform length can be changed
on the fly by assigning a new FFT length when
restarting the core. However this new transform length
cannot be larger than 2
Abw . Assigning the smallest
address width as possible is recommended for
achieving higher clock frequencies during synthesis.
Table 1 : Parameters definition