LF3320
DEVICES INCORPORATED
Horizontal Digital Image Filter
The LF3320 is capable of supporting
24-bit data and 12-bit coefficients or
12-bit data and 24-bit coefficients.
When configured for this mode of
operation, the Filter B output is scaled
by 2-12 before adding it to the Filter A
output. This mode of operation is only
valid in single filter mode.
FIGURE 11. ACCUMULATOR ACCESS MODE
12
12
I/D
I/D
DIN11-0
RIN11-0
REGISTERS
REGISTERS
FILTER
A
FILTER
B
To configure the LF3320 for this mode,
bit 3 of Configuration Register 5 must
be set to 1; this will account for the
scaling function (Table 7). For 24-bit
data, DIN11-0 becomes the MSB
(Filter A)and RIN11-0becomes the
LSB(Filter B), bit 2ofConfiguration
Register 5 must be set to 0. To insure
correct results,the coefficient sets must
be aligned appropriately; that is to say,
the coefficient set used for the MSB
must be the same for the LSB. For
24-bit coefficients, the coefficient banks
for Filter A will correspond to the
coefficient MSBand the coefficient
banks for Filter B will correspond to
the coefficient LSB;therefore, bit 2must
be set to 1.
R.S.L.
CIRCUIT
R.S.L.
CIRCUIT
16
16
DOUT15-0
ROUT3-0 / COUT11-0
reversal (Table 7). For additional
considerations, refer to the correspond- requirement to satisfy the equation,
ing mode of operation section.
need to be accumulated. Such is the
y(t) = cos(vt)+jsin(vt). Thecomplex
data can be streamed and filtered
using a respective ‘I’ filter and
Accumulator Access Mode
‘Q’ filter. To convert the complex result
into a real result, as seen at the LF3320
output, two multiplies and one accu-
mulation is required. The cosine and
sine functions are realized through the
coefficient sets;consequently,multi-
plied by the corresponding ‘I’ and ‘Q’
data streams. To satisfy the remainder
ofthe equation, Filter A and Filter B
must be accumulated.
Accumulator access allows the user to
accumulate the Filter A output with the
Filter Boutput. Therefore, this mode is
only valid when the device has been
configured for dual filter operation. To
configure the device for this mode, bit 1
and bit 5 must be set to 1; bit 5 is the
corresponding accumulator access bit
(Table 7). Writing 022H to Configura-
tion Register 5 configures the device to
accumulate the Filter A output with the
Filter B output. All remaining Configu-
ration Registers, 0 through 4 inclusive,
will depend on specific application
requirements (see Tables 2through 4).
In this mode of operation, the accumu-
lated output is realized at DOUT15-0.
The output data at
Once again, to insure correct results,
the coefficient sets must be aligned
appropriately; that is to say, the MSB
coefficients must correspond to their
LSB coefficients. The output data will
appear at DOUT15-0; output appearing
at ROUT3-0/ COUT11-0 will not be of
any value. Bit 1 is set to 0 (for single
filter mode) and bit 0 is set to 0
(cascade mode must be disabled).
Therefore, to realize 24-bit data/ 12-bit
coefficients the user must write 008H
to Configuration Register 5;conversely,
for 12-bit data/ 24-bit coefficients the
user must write 00CH to Configuration
Register 5.
As previously stated, this mode of
operation is only valid with the dual
filter mode configuration. All modes of
operation, that are valid in the dual
filter mode, are valid with the accumu-
lator access mode. For additional
considerations, refer to the correspond-
ing mode of operation section.
The Double Wide Data/ Coefficient
Mode is valid in the Matrix Multiplica-
tion Mode;however, special consider-
ations must be observed when these
two modes are combined. The LF3320
must be configured for single filter
mode only, for a maximum (8x8)
matrix. In addition, the user must
disable the cascaded filter mode, the
accumulator access mode, and the data
ROUT15-0/ COUT15-0is the Filter B
output that is normally expected;
however,the accumulated output data
(DOUT15-0) will be delayed by one
clock cycle, compared to the Filter B
output data.
FUNCTIONALDESCRIPTION
ALUs
The ALUs double the number offilter
taps available, when symmetric
coefficient sets are used, by pre-adding
data values which are then multiplied
byacommon coefficient(see Figure 12).
This type of operation is useful when
two filtered data streams (i.e. I+jQ)
Video Imaging Products
08/16/2000–LDS.3320-N
2-9