lf3370 LOGIC Devices Incorporated, lf3370 Datasheet - Page 10

• Manufacturer Part Number: lf3370
• Description: High-definition Video Format Converter
• Manufacturer: LOGIC Devices Incorporated

DEVICES INCORPORATED
reset. HF
loading is discussed in the LF Interface™.
Please note, using HBLANK is the
recommended way of clearing HF
HF
RESET, normally performed during device
initialization. RESET will not reset the
counter.
Input/Output Bias Adder
The programmable Input/Output Bias
Adders can be used to subtract or add a
13-bit offset to the data. Input and output
data formats for the two sections are
shown in Figure 3. By using INBIAS
the user may select one of four pro-
grammed Input Bias Adder values (see
Figure 8). By using OUTBIAS
may select one of four programmed
Output Bias Adder values (see Figure 9).
A value of 00 on INBIAS
selects Input/Output Bias Adder Register
0. A value of 01 selects Input/Output
Bias Adder Register 1 and so on.
INBIAS
every clock cycle if desired. If a bias is not
desired, then bits 11 & 12 of Configuration
Register 1 can be set up to independently
disable the input and output bias values.
Thus, effectively zeroing the function.
The total pipeline latency from the input
to the output for each of the two sections
is one CLK cycle. Input/Output Bias
Adder Register loading is discussed in the
LF Interface™ section.
3 x 3 Matrix Multiplier
Processing almost 550 million colors, three
simultaneous 13-bit input and output
channels are utilized to implement a 3 x 3-
matrix multiplication (triple dot product).
Each truncated 20-bit output is the sum of
all three input words multiplied by the
appropriate coefficients (see Figure 11).
These outputs are then fed into the RSL
circuitry (see Figure 13). Input/Output
formats are shown in Figure 3.
For each of the nine multipliers, up to four
user-defined 13-bit coefficients can be
1
flags but they can be cleared by
1-0
0
/OUTBIAS
and HF
1
count value register
1-0
1-0
may be changed
/OUTBIAS
1-0
, the user
0
and
1-0
,
1-0
programmed and selected by CA
value of 00 on CA
0 on each of the 9 coefficient banks. A
value of 01 selects Coefficient Set 1 and so
on. CA
cycle if desired. Coefficient bank loading is
discussed in the LF Interface™.
The total pipeline latency from the input of
the Matrix Multiplier to the output of the
RSL Circuitry is 6 CLK cycles and new
output data is subsequently available
every clock cycle thereafter.
If matrix multiplication is not desired,
using the appropriate combination of
coefficient values while keeping in mind
bit weighting, an identity matrix may be
set up to bypass the Matrix Multiplier
section (see also First Operation Select in
the Bypass Options discussion).
Key Scaler
The Key channel is equiped with a 13 x 13-
bit Key Scaler (see Figure 11) producing a
truncated 20-bit output which is then fed
into the RSL circuitry (see Figure 13). Up to
four user-defined 13-bit coefficients can be
programmed and selected by CA
Input/Output formats are shown in
Figure 3.
F
C
B
IGURE
1-0
13
13
12. 1:2 I
may be changed every clock
1-0
INTERPOLATION
INTERPOLATION
10
CIRCUIT
CIRCUIT
selects Coefficient Set
High-Definition Video Format Converter
NTERPOLATION
VARIABLE LENGTH BYPASS DELAY
VARIABLE LENGTH BYPASS DELAY
1-0
1-0
CONFIGURATION / CONTROL
. A
.
127 x 13-Bit
127 x 13-Bit
REGISTERS
55-TAP
FILTER
55-TAP
FILTER
/ 2:1 D
FIR
FIR
The total pipeline latency from the input of
the Key Scaler to the output of the RSL
Circuitry is 6 CLK cycles and new output
data is subsequently available every clock
cycle thereafter. If scaling is not desired,
load and select a Key Scaler Coefficient
value of 1
the Bypass Options duscussion).
Half-Band Filter
There are two internal Half-Band filters in
the LF3370. These Half-Band filters can
either interpolate, decimate, or pass
through data found on channel B and
channel C. Data on channel A and
channel D in this section pass through a
programmable 127 x 13-bit delay (see
Bypass Section). The filter section (as
show in Figure 12) is a fixed-coefficient,
linear-phase half-band (low-pass)
interpolating/decimating digital filter.
The filter in this section is a 55-tap
transversal FIR with 13-bit coefficients as
shown in Table 3. The frequency re-
sponse (Figure 14) is in full compliance
with SMPTE 260M. This section can be
configured for 2:1 interpolation, 1:2
decimation, or pass-through mode by
setting bits 5-8 in Configuration Register 0
(see Table 5). This section can also be
placed before or after the matrix multi-
Video Imaging Products
ECIMATION
(see also First Operation Select in
DECIMATION
DECIMATION
CIRCUIT
CIRCUIT
H
ALF
R S
R S
-B
AND
L
L
03/13/2001–LDS.3370-F
LF3370
F
ILTERS
13
13
B'
C'