RM0400
x(n)
Register
FIR Section
37.4.6.1
Rounding
The Decimation Filter performs rounding operations in two different locations, as shown in
Figure
374:
•
to obtain the filter output result with 16 bits
•
to obtain the IIR feedback result to be stored in tap4 registers with 24 bits
The rounding mechanism implements the Convergent Rounding methodology (also known
as round-to-nearest even number), which makes the decision on rounding up or down
based on the value of the lower portion of data to be rounded (LS_WORD). The rounding
up/down condition is equal to the traditional rounding except when the LS_WORD has the
format {1000...00}. In this particular case, the rounding procedure is like the example of
Figure
375. If the MS_WORD is odd, the value is rounded up. Otherwise the value is
rounded down.
Figure 374. Filter Configuration Paths (FIR or 1x4Poles IIR)
Coefficient 0
-1
Z
Tap0
Coefficient 1
-1
Z
Tap1
Coefficient 2
-1
Z
Tap2
Coefficient 3
-1
Z
Tap3
Coefficient 4
DocID027809 Rev 4
+
+
Scale Factor S
Coefficient 5
Coefficient 6
Coefficient 7
Coefficient 8
Decimation Filter
By-
00
Round/Sat
IIR
Round/Sat
FI
01
10
B
FTYPE[1:0]
-1
Z
Tap4
-1
Z
IIR Section
Tap5
-1
Z
Tap6
-1
Z
Tap7
A
y(n)
831/2058
841
Need help?
Do you have a question about the SPC572L series and is the answer not in the manual?
Questions and answers