欢迎访问ic37.com |
会员登录 免费注册
发布采购

LF2250QC20 参数 Datasheet PDF下载

LF2250QC20图片预览
型号: LF2250QC20
PDF下载: 下载PDF文件 查看货源
内容描述: 12 ×10位乘法器矩阵 [12 x 10-bit Matrix Multiplier]
分类和应用:
文件页数/大小: 15 页 / 370 K
品牌: LOGIC [ LOGIC DEVICES INCORPORATED ]
 浏览型号LF2250QC20的Datasheet PDF文件第1页浏览型号LF2250QC20的Datasheet PDF文件第3页浏览型号LF2250QC20的Datasheet PDF文件第4页浏览型号LF2250QC20的Datasheet PDF文件第5页浏览型号LF2250QC20的Datasheet PDF文件第6页浏览型号LF2250QC20的Datasheet PDF文件第7页浏览型号LF2250QC20的Datasheet PDF文件第8页浏览型号LF2250QC20的Datasheet PDF文件第9页  
LF2250  
DEVICES INCORPORATED  
12 x 10-bit Matrix Multiplier  
BIT WEIGHTING  
SIGNAL DEFINITIONS  
Power  
TABLE 1. MODE SELECTION  
The internal sum of products of the  
LF2250 can grow to 23 bits. However,  
in order to keep the output format of  
the matrix multiply mode (Mode 00)  
identical to the input format, the X, Y,  
and Z outputs are truncated to 12-bit  
integer words. In the filter modes  
(Modes 01, 10, 11), the cascade output  
is always half-LSB rounded to 16 bits  
(12 integer bits and 4 fractional bits).  
The user may half-LSB round the  
output to any size less than 16 bits by  
simply forcing a “1” into the bit  
position of the cascade input immedi-  
ately below the desired LSB. For  
example, if half-LSB rounding to 12  
bits is desired, then a “1” must be  
forced into the CASIN3 bit position  
(CASOUT4 would then be the LSB).  
MODE1-0  
OPERATING MODE  
3 x 3 Matrix Multiplier  
9-Tap FIR Filter  
VCC and GND  
00  
01  
10  
11  
+5 V power supply. All pins must be  
connected.  
3 x 3 Convolver  
4 x 2 Convolver  
Clock  
CLK — Master Clock  
OPERATING MODES  
The rising edge of CLK strobes all  
enabled registers. All timing specifi-  
cations are referenced to the rising  
edge of CLK.  
The LF2250 can realize four different  
user-selectable digital filtering architec-  
tures as determined by the state of the  
mode (MODE1-0) inputs. Upon  
selection of the desired function, the  
LF2250 automatically chooses the  
appropriate internal data paths and  
input/output bus structure. Table 1  
details the modes of operation.  
Inputs  
A11-0, B11-0, C11-0 — Data Inputs  
A, B, and C are the 12-bit registered  
data input ports. Data presented to  
these ports is latched into the multi-  
plier input registers for the current  
operating mode (Table 1). In the filter  
modes (Modes 01, 10, 11), the rising  
edge of CLK internally right-shifts  
new data to the next filter tap.  
In all four modes, the user may adjust  
the bit weighting, by applying an  
identical scaling correction factor to  
both the input and output data  
streams. If the coefficients are re-  
scaled, then the relative weightings of  
the cascade-in and cascade-out ports  
will differ accordingly. Figure 1  
illustrates the input and output bit  
weightings for all four modes.  
DATA FORMATTING  
The coefficient input ports (KA, KB,  
KC) are 10-bit fractional two’s comple-  
ment format regardless of the operating  
mode. The data input ports (A, B, C)  
are 12-bit integer two’s complement  
format regardless of the operating  
mode.  
KA9-0, KB9-0, KC9-0 — Coefficient Inputs  
KA, KB, and KC are the 10-bit regis-  
tered coefficient input ports. Data  
presented to these ports is latched  
into the corresponding internal  
coefficient register set defined by  
CWE1-0 (Table 4) on the next rising  
edge of CLK. Table 3 shows which  
coefficient registers are available for  
each coefficient input port.  
In the matrix multiplier mode (Mode  
00), the data output ports (X, Y, Z) are  
12-bit integer two’s complement  
format. In the FIR filter and convolver  
modes (Modes 01, 10, 11), the X, Y, and  
Z ports are configured as the cascade-in  
(CASIN15-0) and cascade-out  
(CASOUT15-0) ports. These ports  
assume 16-bit (12-bit integer, 4-bit  
fractional) two’s complement data on  
both the inputs and outputs. Table 2  
shows the data port formatting for  
each of the four operating modes.  
DATA OVERFLOW  
Because the LF2250’s matched input  
and output data formats accommodate  
unity gain (0 dB), input conditions that  
could lead to numeric overflow may  
exist. To ensure that no overflow  
conditions occur, the user must be  
aware of the maximum input data and  
coefficient word sizes allowable for  
each specific algorithm being per-  
formed.  
TABLE 2. DATA PORT FORMATTING  
PIN NAMES  
MODE1-0  
A11-0  
A11-0  
A11-0  
A11-0  
A11-0  
B11-0  
B11-0  
A11-0  
B11-0  
B11-0  
C11-0  
C11-0  
NC  
KA9-0  
KA9-0  
KA9-0  
KA9-0  
KA9-0  
KB9-0  
KB9-0  
KB9-0  
KB9-0  
KB9-0  
KC9-0  
KC9-0  
KC9-0  
KC9-0  
KC9-0  
XC11-0  
YC11-8  
Y7-4  
Y7-4  
NC  
NC  
NC  
YC3-0  
ZC11-0  
00  
01  
10  
11  
X11-0  
Y11-8  
Y3-0  
Z11-0  
CASIN15-4  
CASIN15-4  
CASIN15-4  
CASIN3-0  
CASIN3-0  
CASIN3-0  
CASOUT3-0  
CASOUT3-0  
CASOUT3-0  
CASOUT15-4  
CASOUT15-4  
CASOUT15-4  
C11-0  
NC  
Video Imaging Products  
08/16/2000–LDS.2250-L  
2