• Fractional and integer two’s complement data formats
• Input and output data latches with user-configurable
enables
• Selectable 16-bit rounded output
• Internal 1/2 LSB rounding
• Available in 120-pin CPGA, PPGA, MPGA, or MQFP
Description
The TMC2246A is a video-speed convolutional array composed of four 11 x 10 bit registered multipliers followed by a
summer and an accumulator. All eight multiplier inputs are
accessible to the user and may be updated every clock cycle
with integer or fractional two’s complement data. A pipelined architecture, fully registered input and output ports,
and asynchronous three-state output enable control simplify
the design of complex systems.
Applications
• Fast pixel interpolation
• Fast image manipulation
• Image mixing and keying
• High-performance FIR filters
• Adaptive digital filters
• One- and two-dimensional image processing
The data or coefficient inputs to the multipliers may be held
over multiple clock cycles, providing storage for mixing and
filtering coefficients. The 25-bit internal accumulator path
allows two bits of cumulative word growth and may be internally rounded to 16 bits. Output data are updated every clock
cycle, or may be held under user control. All data inputs, outputs, and controls are TTL compatible and (except for the
three-state output enable) are registered on the rising edge of
CLK.
Logic Symbol
D1
9-0
C1
10-0
D2
9-0
C2
10-0
D3
9-0
C3
10-0
D4
9-0
C4
10-0
TMC2246A
Image Filter
ENB1-4
ENSEL
ACC
FSEL
S
15-0
OCEN
OEN
CLK
The TMC2246A is uniquely suited to performing pixel interpolation in image manipulation and filtering applications. As
a companion to the Fairchild Semiconductor TMC2301 and
TMC2302 Image Manipulation Sequencers, the TMC2246A
can execute a bilinear interpolation of an image (4-pixel kernels) at real-time video rates. Larger kernels or other, more
complex, functions can be realized with no loss in performance
by utilizing multiple devices.
With unrestricted access to all data and coefficient input
ports, the TMC2246A offers considerable flexibility in applications performing digital filtering, adaptive FIR filters, mixers, and other similar systems requiring high-speed
processing.
Fabricated in a submicron CMOS process, the TMC2246A
operates at a guaranteed clock rate of 60 MHz over the full
temperature and supply voltage ranges. It is pin- and function-compatible with Fairchild’s TMC2246, while providing
higher speed operation and lower power dissipation. It is
available in a 120 pin Plastic Pin Grid Array (PPGA), 120
pin Ceramic Pin Grid Array (CPGA), 120 lead MQFP to
PPGA (MPGA), and a 120 lead Metric Quad FlatPack
(MQFP).
REV. 1.0.3 9/11/00
PRODUCT SPECIFICATIONTMC2246A
Block Diagram
ENSEL
ACC
FSEL
OCEN
CLK
OEN
9-0
C1
10-0
ENB1
D2
9-0
ENB2D1
C2
10-0
-10
2
*
D3
LSB MSB
9-0
25
ENB3ENB4
C3
10-0
D4
9-0
*Automatic rounding function
C4
10-0
S
15-0
Functional Description
The TMC2246A Image Filter is a flexible multiplier-summer
array which computes the accumulated sum of four 11x10
bit products, allowing word growth up to 25 bits.
The inputs are user-configurable, allowing latching of either
the 10- or 11-bit input data. The data format is user-selectable
between integer or fractional two’s complement arithmetic.
Total latency from input registers to output data port is 5
clocks.
The output data path is 16 bits wide, providing the lower 16
bits of the accumulator when in integer format or the upper
16 bits of the 25-bit accumulator path when fractional two’s
complement notation is selected. One-time rounding to 16
bits is performed automatically when accumulating fractional data, but is disabled when operating in integer format
to maintain the integrity of the least-significant bits.
F3, H3, L7, C812, 20, 46, 102 Supply Voltage. The TMC2246A operates from a single +5V
8, 16, 24, 42,
H11, C7
72, 106
Clock
CLKC31
Inputs
D1
D2
9-0
9-0
M1, K3, L2, N1,
L3, M2, N2, L4,
M3, N3
J12, K13, J11,
K12, L13, L12,
K11, M13, M12,
27, 28, 29, 30,
31, 32, 33, 34,
35, 36
70, 69, 68, 67,
66, 65, 64, 63,
62, 61
L11
D3
9-0
J13, H12, H13,
G12, G11, G13,
F13, F12, F11,
71, 73, 74, 75,
76, 77, 78, 79,
80, 81
E13
D4
9-0
B4, C5, A4, B5,
A5, C6, B6, A6,
A7, B7
115, 114, 113,
112, 111, 110,
109, 108, 107,
105
C1
C2
10-0
10-0
M4, L5, N4, M5,
N5, M6, N6, M7,
N7, N8, M8
N13, M11, L10,
N12, N11, M10,
L9, N10, M9,
37, 38, 39, 40,
41, 43, 44, 45,
47, 48, 49
60, 59, 58, 57,
56, 55, 54, 53,
52, 51, 50
N9, L8
C3
10-0
E12, D13, E11,
D12, C13, B13,
D11, C12, A13,
82, 83, 84, 85,
86, 87, 88, 89,
90, 91, 92
C11, B12
C4
10-0
A8, B8, A9, B9,
A10, C9, B10,
A11, B11, C10,
104, 103, 101,
100, 99, 98, 97,
96, 95, 94, 93
A12
Outputs
S
15-0
C1, D2, D1, E2,
E1, F2, F1, G2,
G1, H1, H2, J1,
J2, K1, K2, L1
6, 7, 9, 10, 11,
13, 14, 15, 17,
18, 19, 21, 22,
23, 25, 26
Pin Function Description
supply. All power and ground pins must be connected.
Ground. The TMC2246A operates from a single +5V supply. All
power and ground pins must be connected.
System Clock. The TMC2246A operates from a single master
clock input. The rising edge of clock strobes all enabled registers.
All timing specifications are referenced to the rising edge of CLK.
Data Input Ports. D1 through D4 are the 10-bit data input ports.
The LSB is Dx
.
0
Coefficient Input Ports. C1 through C4 are the 11-bit coefficient
input ports. The LSB is Cx
.
0
Sum Output. The current 16-bit result is available at the Sum
output. The LSB is S
. See the Functional Block Diagram .
0
REV. 1.0.3 9/11/00
5
PRODUCT SPECIFICATIONTMC2246A
Pin Descriptions
Pin Name
Controls
FSELB22
ENSELA1120
ENB1–
ENB4
ACCB13
CPGA/PPGA/
MPGA
C4, A2, A3, B3118, 117, 116,
(continued)
Pin Number
MQFP
119
Pin Function Description
Format Select. Coefficients input during the current clock are
assumed to be in fractional two's complement format. Rounding to
16 bits is performed as determined by the accumulator control,
ACC, and the upper 16 bits of the accumulator are output when
the registered Format Select input (FSEL) is LOW. When FSEL is
HIGH, two's complement integer format is assumed, and the
lower 16 bits of the accumulator are presented at the output. No
rounding is performed when operating in integer mode. See the
Functional Block Diagram and the Applications Discussion.
Enable Select. The registered Enable Select determines whether
the data or the coefficient input registers may be held on the next
rising edge of clock, in conjunction with the individual input
enables ENB1–ENB4. See Table 1.
Input Enables. When ENBi (i=1, 2, 3, or 4) is LOW, registers Ci
and Di are both strobed by the next rising edge of CLK. When
ENBi is HIGH and ENSEL is LOW, Di is strobed, but Ci is held.
When ENBi and ENSEL are both HIGH, Di is held and Ci is
strobed. See Table 1. Thus, either or both input registers to each
multiplier are updated on each clock cycle.
Accumulate. When the registered ACCumulator control is LOW,
no internal accumulation will be performed on the data input
during the current clock, effectively clearing the prior accumulated
sum. If operating in fractional two's complement format (FSEL =
LOW), one-half LSB rounding to 16 bits is performed on the result.
This allows the user to perform summations without propagating
roundoff errors.
When ACC is HIGH, the internal accumulator adds the emerging
products to the sum of previous products, without performing
additional rounding.
OCEN
OEN
No Connect
NCD4 (Index Pin)Not Connected. (Optional)
Note:
1. X denotes a "Don't Care" condition.
2. Any register not explicitly held is updated on the next rising edge of CLK.
6
D34
C25
Output Register Enable. The output of the accumulator is
latched into the output register on the next clock when the Output
Register Clock Enable is LOW. When OCEN is HIGH the contents
of the output register remain unchanged; however, accumulation
will continue internally if ACC remains HIGH.
Output Enable. Data currently in the output registers is available
at the output bus S
LOW. When OEN
state.
15-0
is HIGH, the outputs are in the high-impedance
when the asynchronous Output Enable is
REV. 1.0.3 9/11/00
Loading...
+ 12 hidden pages
You need points to download manuals.
1 point = 1 manual.
You can buy points or you can get point for every manual you upload.