ST STM32G4 Series Reference Manual page 454

Filter Math Accelerator (FMAC)
Load X2 buffer
This function pre-loads the X2 buffer with N + M values, starting from the address in
X2_BASE. Successive writes to the FMAC_WDATA register will load the write data into the
X2 buffer and increment the write address.
The function can be used to pre-load the buffer with the elements of a vector, or the
coefficients of a filter. In the case of an IIR, the N feed-forward and M feed-back coefficients
are concatenated and loaded together into the X2 buffer. The total number of coefficients is
equal to N + M. For an FIR, there are no feedback coefficients, so M = 0.
Parameters
•
The parameter P contains the number of values, N, to be loaded into the X2 buffer
starting from address X2_BASE.
•
The parameter Q contains the number of values, M, to be loaded into the X2 buffer
starting from address X2_BASE + N.
• The parameter R is not used.
The function completes when N + M writes have been performed to the FMAC_WDATA
register.
Load Y buffer
This function pre-loads the Y buffer with N values, starting from the address in Y_BASE.
Successive writes to the FMAC_WDATA register will load the write data into the Y buffer
and increment the write address. The read pointer will point to the address Y_BASE + N
when the function completes.
The function can be used to pre-load the feedback storage elements of an IIR filter.
Parameters
•
The parameter P contains the number of values to be loaded into the Y buffer.
•
The parameters Q and R are not used.
The function completes when N writes have been performed to the FMAC_WDATA register.
17.3.6 Filter Functions
The following filter functions are supported by the FMAC unit. These functions are triggered
by writing the corresponding value in the FUNC field of the FMAC_PARAM register with the
START bit set. The P, Q and R fields must also contain the appropriate parameter values for
each function as detailed below. The filter functions continue to run until the START bit is
reset by software.
Convolution (FIR filter)
Y = B*X
R
y = 2
n
This function performs a convolution of a vector B of length N+1 and a vector X of indefinite
length. The elements of Y for incrementing values of n are calculated as the dot product,
454/2083
N
∑
⋅
b x
k
n k –
k = 0
RM0440 Rev 1
RM0440