Intersil Corporation HSP43168 Datasheet

HSP43168
Data Sheet November 1999
Dual FIR Filter
The HSP43168 Dual FIR Filter consists of two independent 8-tap FIR filters. Each filter supports decimation from 1 to 16 and provides on-board storage for 32 sets of coefficients. The Block Diagram shows two FIR cells each fed by a separate coefficient bank and one of two separate inputs. The outputs of the FIR cells are either summed or multiplexed by the MUX/Adder. The compute power in the FIR Cells can be configured to provide quadrature filtering, complex filtering, 2-D convolution, 1-D/2-D correlations, and interpolating/decimating filters.
The FIR cells take advantage of symmetry in FIR coefficients by pre-adding data samples prior to multiplication. This allows an 8-tap FIR to be implemented using only 4 multipliers per filter cell. These cells can be configured as either a single 16-tap FIR filter or dual 8-tap FIR filters. Asymmetric filtering is also supported.
Decimation ofupto 16 is providedto boost theeffective number of filter taps from 2 to 16 times. Further, the Decimation Registers provide the delay necessary for fractional data conversion and 2-D filtering with kernels to 16 x16.
The flexibility of the Dual is further enhanced by 32 sets of user programmable coefficients. Coefficient selection may be changed asynchronously from clock to clock. The ability to toggle between coefficient sets further simplifies applications such as polyphase or adaptive filtering.
The HSP43168 is a low power fully static design implemented in an advanced CMOS process. The configuration of the device is controlled through a standard microprocessor interface.
File Number 2808.8
Features
• Two Independent 8-Tap FIR Filters Configurable as a Single 16-Tap FIR
• 10-Bit Data and Coefficients
• On-Board Storage for 32 Programmable Coefficient Sets
• Up To: 256 FIR Taps, 16 x 16 2-D Kernels, or 10 x 19-Bit Data and Coefficients
• Programmable Decimation to 16
• Programmable Rounding on Output
• Standard Microprocessor Interface
Applications
• Quadrature, Complex Filtering
• Image Processing
• Polyphase Filtering
• Adaptive Filtering
Ordering Information
TEMP.
PART NUMBER
HSP43168VC-33 0 to 70 100 Ld MQFP Q100.14x20 HSP43168VC-40 0 to 70 100 Ld MQFP Q100.14x20 HSP43168VC-45 0 to 70 100 Ld MQFP Q100.14x20 HSP43168JC-33 0 to 70 84 Ld PLCC N84.1.15 HSP43168JC-40 0 to 70 84 Ld PLCC N84.1.15 HSP43168JC-45 0 to 70 84 Ld PLCC N84.1.15 HSP43168JI-40 -40 to 85 84 Ld PLCC N84.1.15 HSP43168GC-45 0 to 70 84 Ld CPGA G84.A
RANGE (oC) PACKAGE PKG. NO.
Block Diagram
INA0 - 9
INB0 - 9/ OUT0 - 8
OEL
OEH
10
10
CIN0 - 9
A0 - 8
WR
CSEL0 - 4
1
MUX
10
9
COEFFICIENT
BANK A
FIR CELL A
CAUTION: These devices are sensitive to electrostatic discharge; follow proper IC Handling Procedures.
MUX
MUX/
ADDER
919
1-888-INTERSIL or 407-727-9207
COEFFICIENT
BANK B
FIR CELL B
CONTROL/
CONFIGURATION
OUT9 - 27
| Copyright © Intersil Corporation 1999
Pinouts
84 LEAD CPGA
11 10 9 8 7 6 5 4 3 2 1
GND
A
B
C
D
E
F
G
WR
RVRS
SHFTEN
MUX0 MUX1
FWRD
TXFR
V
ACCEN
CC
GND CLK
OEH
OUT27 OUT22 OUT26
OUT24 OUT23 OUT25
BOTTOM VIEW
A4A1
A0
A5
HSP43168
BOTTOM VIEW
A7 CSEL1 CSEL3 CSEL4
A8
V
CC
CSEL0
CSEL2 CIN9
CIN2 CIN1
INA8 INA9
INA7 INA5
A2A3
A6
CIN8
CIN5
CIN7
CIN6 CIN4
GND
CIN3
CIN0
V
INA6
HSP43168
PIN 'A1' ID
L
A
B
C
D
E
CC
F
G
GND OUT15 OUT14 OUT12
K
OUT18
J
OUT19
H
OUT21
G
OUT24 OUT23 OUT25
F
OUT27 OUT22 OUT26
OEH
E
84 LEAD CPGA
11 10 9 8 7 6 5 4 3 2 1
V
OUT16
CC
OUT17
OUT20
TOP VIEW
OUT10 OUT11
V
INB0
CC
OUT9
OEL
HSP43168
INB5
INB7
INA7
INA8
INB6
INB8
INA0
INA3
INA5
INA9
INB4
INB1
GND
INB2OUT13
INB3 INA2
TOP VIEW
CIN2
GND
CLK
CIN1
INB9
INA1
INA4
INA6
V
CIN0
L
K
J
H
G
F
CC
E
OUT21 OUT20
H
J
OUT19 OUT17
V
OUT18
K
L
CC
GND OUT15 OUT14 OUT12 OUT10 OUT11 INB1 INB4 INB5 INB6
11 10 9 8 7 6 5 4 3 2 1
OUT9
V
INB3 INA2
OEL
CC
CSEL 3
CSEL 4
CIN 9
CIN 8
109 8 7 6 5 4 3 2 1 848382818079
11
CIN 7
12
CIN 6
13
CIN 5
14
CIN 4
15
GND
16
CIN 3
17
CIN 2
18
CIN 1
19
CIN 0
20
INA 9
21
INA 8
22
INA 7
23
INA 6
24
INA 5
25
V
26
CC
27
INA 4
28
INA 3
29
INA 2
30
INA 1
31
INA 0
32
INB 9
33 34 35 36 37 38 39 40 41
INA3 INA4
INA0
INB8
CSEL 1
CSEL 0
CSEL 2
H
D
C
J
INA1INB7GNDINB2OUT13 INB0OUT16
INB9
B
K
A
L
84 LEAD PLCC
TOP VIEW
CC
A 6
A 7
A 5
A 8
V
42 43 44 45 46 47 48 49
CIN9
CSEL3
GND
CIN6
CIN7
CSEL4
V
ACCEN
CC
A5
TXFR
FWRD
SHFTEN
MUX0 MUX1
RVRS WR
11 10 9 8 7 6 5 4 3 2 1
A 4
A 3
A 2
A 1
A 0
GND
GND
WR
A0
A1
MUX 1
MUX 0
A6
A2A3
A7 CSEL1
A4
CSEL0
V
CC
A8
CSEL2
767778 75
74
RVRS
73
FWD
72
SHFTEN
71
TXFR
70
ACCEN V
69
CC
68
CLK
67
GND
66
OEH
65
OUT 27
64
OUT 26
63
OUT 25
62
OUT 24
61
OUT 23
60
OUT 22
59
OUT 21 OUT 20
58
OUT 19
57
OUT 18
56
OUT 17
55
V
54
CC
50 51 52 53
CIN3
CIN4
CIN5
CIN8
D
C
B
A PIN
'A1' ID
INB 8
INB 7
INB 6
INB 5
GND
INB 4
INB 3
INB 2
INB 1
INB 0
OEL
OUT 9
OUT 10
OUT 12
OUT 11
OUT 13
OUT 14
OUT 15
GND
OUT 16
CC
V
2
Pinouts (Continued)
HSP43168
100 LEAD MQFP
TOP VIEW
CIN8
NC
CIN7
NC CIN6 CIN5 CIN4 GND GND CIN3 CIN2 CIN1 CIN0 INA9 INA8 INA7 INA6 INA5
V
CC
V
CC
INA4 INA3 INA2 INA1 INA0
NC
NC INB9 INB8 INB7
CCVCC
CIN9
CSEL4
99 98 97 96 95 94 93 91 89 87 85 84 83 818286889092100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
CSEL2
CSEL3
CSEL1
CSEL0
V
A8
A7
A6
A5
A4
A3
A2
A1
A0
GND
GND
80 79 78 77 76 75 74 73 72 71 70 69 68 67 66 65 64 63 62 61 60 59 58 57 56 55 54 53 52 51
WR
MUX1 MUX0 RVRS NC FWRD SHIFTEN TXFR ACCEN V
CC
V
CC
CLK GND GND OEH OUT27 OUT26 OUT25 OUT24 OUT23 OUT22 OUT21 OUT20 OUT19 OUT18 OUT17 NC V
CC
V
CC
GND GND
32 33 34 35 36 37 38 40 42 44 46 47 48 50494543413931
INB6
INB5
GND
GND
INB4
INB3
INB2
INB1
INB0
OEL
OUT9
OUT10
VCCV
CC
OUT11
OUT12
OUT13
OUT14
OUT15
OUT16
3
HSP43168
Pin Description
SYMBOL TYPE DESCRIPTION
V
CC
GND Ground.
CIN0-9 I Control/Coefficient Data Bus. Processor interface for loading control data and coefficients. CIN0 is the LSB.
A0-8 I Control/Coefficient Address Bus. Processor interface for addressing Control and Coefficient Registers. A0 is the
WR I Control/Coefficient Write Clock. Data is latched into the Control and Coefficient Registers on the rising edge of
CSEL0-4 I Coefficient Select. Thisinput determines which of the 32 coefficient sets areto be used by FIR A and B. Thisinput
INA0-9 I Input to FIR A. INA0 is the LSB. INB0-9 I/O Bidirectional Input for FIR B. INB0 is the LSB and is input only. When used as output, INB1-9 are the LSBs of the
OUT9-27 O 19MSBs of Output Bus. Data format is either unsigned or two's complement depending on configuration. OUT27
SHFTEN I ShiftEnable.This active lowinput enables clockingof data intothe part andshifting of datathrough the Decimation
FWRD I Forward ALU Input Enable. When active low, data from the forward decimation path is input to the ALUs through
RVRS I Reverse ALU Input Enable. When active low, data from the reverse decimation path is input to the ALUs through
VCC: +5V power supply pin.
LSB.
WR.
is registered and CSEL0 is the LSB.
output bus, and INB9 is the MSB of these bits.
is the MSB.
Registers.
the “a” input. When high, the “a” inputs to the ALUs are zeroed.
the “b” input. When high, the “b” inputs to the ALUs are zeroed.
TXFR I Data Transfer Control. This active low input switches the LIFO being read into the reverse decimation path with
the LIFO being written from the forward decimation path (see Figure 1).
MUX0-1 I Adder/Mux Control. This input controls data flow through the output Adder/Mux. Table 5 lists the various
configurations.
CLK I Clock. All inputs except those associated with the processor interface (CIN0-9, A0-8, WR) and the output enables
(OEL, OEH) are registered by the rising edge of CLK.
OEL I Output Enable Low. This three-state control enables the LSBs of the output bus to INB1-9 when OEL is low.
OEH I Output Enable High. This three-state control enables OUT9-27 when OEH is low.
ACCEN I Accumulate Enable. This active high input allows accumulation in the FIR Cell Accumulator. A low on this input
latches the FIR Accumulator contents into the Output Holding Registers while zeroing the feedback pass in the Accumulator.
4
TXFR
FIR A REVERSE PATH
FIR A FORWARD PATH
SHFTEN
5
INB1-9/ OUT0-8
INA0-9
INB0
FWRD
RVRS
FIR B INPUT
SOURCE
10
9
DELAY 3
DELAY 3
M
DELAY 3
U X
DELAY 3
DELAY 3
10
10
AB
ALU
11
REG
DELAY
1-16 ††
DELAY 1-16
††
AB
REG
1
DELAY 4
0
DELAY 3
DECIMATION REGISTERS
DELAY
1-16 ††
DELAY
1-16
††
ALU
DATA REVERSAL ENABLE
M U X
FIR A ODD/EVEN # TAPS
DELAY 1-16 ††
DELAY 1-16
††
AB
ALU
REG
AB
ALU
REG
DATA FEEDBACK
CIRCUITRY
LIFO A
M
LIFO B
U X
DELAY 1-16 ††
DELAY 1-16 ††
ODD/EVEN
SYMMETRY
MODE SELECT
ODD/EVEN SYMMETRYMODE SELECT
D
M
E
U
M
U
X
X
M U X
AB
ALU
REG
DECIMATION REGISTERS
DELAY
1-16
††
DELAY
††
1-16
AB
ALU
REG
DELAY 4
DELAY 3
DELAY 1-16
DELAY 1-16
1
0
††
††
M U X
AB
ALU
REG
FIR B
ODD/EVEN # TAPS
DELAY
1-16
††
DELAY
††
1-16
AB
ALU
REG
DATA FEEDBACK
CIRCUITRY
LIFO A
M
LIFO B
U X
DELAY
1-16
DATA REVERSAL ENABLE
FIR B REVERSE PATH FIR B FORWARD PATH
ODD/EVEN
SYMMETRY
D E M U X
ODD/EVEN
NUMBER OF TAPS
M U X
HSP43168
11
CSEL0-4
REG
COEF
X
BANK
COEF
X
BANK
5
DELAY 4
10
21
REG
REG
COEF
X
BANK
REG
COEF
X
BANK
REG
COEF
X
BANK
3210
REG
COEF
X
BANK
REG
COEF
X
BANK
210
REG
COEF
X
BANK
3
CLK
ACCEN
MUX0-1
CIN0-9
A0-8
WR
OEL
OEH
FIR A
ACCUMULATOR
0
M
R
U X
DELAY 5
2
DELAY 6
10
9
CONTROL
ADDER
E G
22
OUTPUT HOLDING
REG
REGISTER
MODE SELECTODD EVEN SYMMETRYFIR A ODD/EVEN # TAPSFIR B ODD/EVEN # TAPSFIR B INPUT SOURCEDATA REVERSAL ENABLEROUND ENABLEDECIMATION FACT OR
FIR CELL A FIR CELL B
MUX/
ADDER
28
DELAY 2
9
ROUND ENABLE
19
FIR B
ACCUMULATOR
0
M
R
U X
ADDER
E G
OUTPUT HOLDING
REG
REGISTER
OUT9-27
Processor control words
††Decimation factor
FIGURE 1. DUAL FIR FILTER
HSP43168
Functional Description
As shown in Figure 1, the HSP43168 consists of two 4-multiplier FIR filter cells which process 10-bit data and coefficients. The FIR cells can operate as two independent 8-tap FIR filters or two 4-tap asymmetric filters at maximum I/O rates. A single filter mode is provided which allows the FIR cells to operate as one 16-tap FIR filter or one 8-tap asymmetric filter. On board coefficient storage for up to 32 sets of 8 coefficients is provided. The coefficient sets are user selectable and are programmed through a microprocessor interface.Programmable decimation to 16 is also provided. By utilizing Decimation Registers together with the coefficient sets, polyphase filters are realizable which allow the user to trade data rate for filter taps. The MUX/Adder can be configured to either add or multiplex the outputs of the filter cells depending upon whether the cells are operating in single or dual filter mode. In addition, a shifter in the MUX/Adder is provided for implementation of filters with 10-bit data and 20-bit coefficients or vice versa.
The Dual FIR Filter has a “pipeline” delay of 8 CLK periods, once normal filtering operations begin. Five typical filtering operation examples are provided in the Applications Examples Section as a guide to configuration and control of the Dual FIR Filter.
During normal filter operations, the location and duration of the
TXFR signal assertions are determined by the filter configuration and operation mode. Once set, these signal parameters must be maintained during normal operation to ensure proper data alignment in the part. Once the part is reset, donot change again.
NOTE: The fixed or periodic relationship between the TXFR signal and CLK must be maintained for valid filter operation. This relationship can only change when CLK is halted and new configuration control words are loaded into the device.
TXFR unlessyou load the configuration
Preparing the Dual FIR for Operation
Two configuration steps are requiredto prepare the Dual FIR Filter for normal operation: 1) loading the Configuration Control Registers, and 2) loading the FIR Filter Coefficients.
Configuration Control Registers are loaded by placing the control register address on address lines A0-8, placing the configuration data on the configuration input lines CIN0-9, and asserting the assertion). This action creates a rising edge on the which clocks the address and configuration data into the part. The details of the “Load Configuration” process areoutlined in the Microprocessor Interface Section.
FIR Coefficients are loaded by placing the address of the Coefficient Data Bank on the address lines A0-8, placing the FIR 10-bit coefficient values on the configuration input lines CIN0-9 and then asserting the release of the assertion). This action creates a rising edge on the
WR line, which clocks the FIR Coefficient Band address and FIR Coefficient data into the part. The details of the “Load FIR Coefficient” process are outlined in the FIR Filter Cells Section, Coefficient Bank Subsection.
Both the Configuration Load and FIR Coefficient Load can be done as a sequence of asynchronous write commands to the Dual FIR Filter. Once these actions are complete, the part is ready for normal filter operation. The CLK, FWRD, RVRS, ACCEN, and SHFTEN signals must be asserted in a manner determined by the application. MUX0-1 must meet the setup and hold times with respect to clock for proper filter operation. Details of the MUX1-0 control can be found in the Output MUX/Adder Section. Details of the ACCEN control can be found in the Fir Cell Accumulator Section. Bit locations for the various filter control/configuration signals can be found in the Input/Output Formats Section.
WR line (followed by a release of the
WR line,
WR line (followed by a
TXFR,
Microprocessor Interface
The Dual FIRhas a 20 pin write only microprocessor interface for loading data into the Control Block and Coefficient Banks. The interface consists of a 10-bit data bus (CIN0-9), a 9-bit address bus (A0-8), and a write input ( into the on-boardregisters on a rising edge. The configuration control and coefficient data loading is asynchronous to CLK.
WR) to latch the data
Control Block
The Dual FIR is configured by writing to the registers within the Control Block. Figure 2 shows the timing diagram for writing to the ConfigurationControl Registers. These Control Registers are memory mapped to Address 000H (H = Hexadecimal) and 001H on A0-8. The Filter Coefficient Registers are mapped to 1XXH (X = value described in the “Coefficient Banks” chapter of the ALU Section).
RESET
WR
A8-0
C9-0
FIGURE 2. LATCHING C9-0 VALUESINTOADDRESS A8-0
The format of the Control Registers is shown in Table 1 and Table 2. Writing toany of the Control/Configuration Registers causes a reset which lasts for 6 CLK cycles following the assertion of the Control Block will not clear the contentsof the Coefficient
000H
001H
REGISTERS
WR. The reset caused by Writing Registers in
6
HSP43168
Bank. As shown in Figure 2, either Configuration Control Register can be written to during reset.
T ABLE1. CONFIGURATION/CONTROLWORD 0 BIT DEFINITIONS
CONTROL ADDRESS 000H
BITS FUNCTION DESCRIPTION
3-0 Decimation Factor (N) R = N + 1
0000 = No Decimation. 1111 = Decimation by 16.
4 Mode Select 0 = Single Filter Mode.
1 = Dual Filter Mode. (also 20-Bit Coefficient Filter)
5 Odd/Even Filter
Coefficient Symmetry
6 FIR A Odd/Even
Number of Taps
7 FIR B Odd/Even
Number of Taps
8 FIR B Input Source 0 = Input from INA0-9.
9 Not Used Set to 0 for Proper Operation.
NOTE: Address locations002H to 011H are reserved, and writing to these locations will have unpredictable effects on part configuration.
T ABLE2. CONFIGURATION/CONTROLWORD 1 BIT DEFINITIONS
CONTROL ADDRESS 001H
BITS FUNCTION DESCRIPTION
0 FIR A Input Format 0 = Unsigned.
1 FIR A Coefficient Format (Defined same as FIR A input). 2 FIR B Input Format (Defined same as FIR A input). 3 FIR B Coefficient (Defined same as FIRA input). 4 Data Reversal Enable 0 = Enabled.
8-5 Round Position 0000 = 2
9 Round Enable 0 = Enabled.
NOTE: Address locations 002H to 011H are reserved, and writing to these locations will have unpredictableeffects on part configuration.
0 = Even Symmetric Coefficients. 1 = Odd Symmetric Coefficients.
0 = Odd Number of Taps in Filter. 1 = Even Number of Taps in Filter.
(Defined Same as FIR A Above).
1 = Input from INB0-9.
1 = Two's Complement.
1 = Disabled.
-10.
1011 = 2 (See Figure 4)
1 = Disabled.
1.
The 4 LSBs of the control word loaded at address 000H are used to select the decimation factor. The Decimation Factor is programmed to one less than the number of delays between filter taps
DF CLK delays between taps()1=
(EQ. 1)
and B before entering the reverse paths of Filters A and B (see Figure 1). Coefficient symmetry is selected by bit 5. Bits 6 and 7 are programmed to configure the FIR cells for odd or even filter lengths (number of taps). Bit 8 selects the FIR B input source when the FIR cells are configured for independent operation. Bit 9 must be programmed to 0.
NOTE: When the filter is programmed for even-taps, the TXFR signal is delayed by only three CLKS (see Figure 1). For odd-taps, the
TXFR signal is delayed by four CLKS.
The 4 LSBs of the control word loaded at address 001H are used to configure the format of the FIR cell's data and coefficients. Bit 4 is programmed to enable or disable the reversal of data sample order prior to entering the Reverse Path Decimation Registers. Data reversal is required for symmetric filter coefficient sets of both even or odd numbers of filter taps. Asymmetric filters and some decimated symmetric filters require the data reversal to be off. Bits 5-9 are used to support programmable rounding on the output.
FIR Filter Cells
Each FIRfilter cell is based on an array of four11x10-bit two's complement multipliers. One input of the multipliers comes from the ALU’s which combine data shifting through the Forward and Re verse Decimation Registers. The second multiplier inputcomes from the user programmablecoefficient bank. The multiplier outputs are fed to an accumulator whose result is passedto the output section where it ismultiplexed or added with the result from the other FIR cell.
Decimation Registers
The Forwardand ReverseDecimation Shift Registerscan be configured for decimation factors from 1 to 16 (see Table 1, bits 0-3). NOTE: Setting the decimation factor only
affects the Delay Registers between filter taps, not the filter controlmultiplexers. Example 4 and Example 5 inthe
Applications Section discuss how to configure the part for actual decimation applications.
The Reverse Shifting Registers with the data reversal logic are used to takeadvantage of symmetry in linear phase filters by aligning data at the ALUs for pre-addition prior to multiplication by the common coefficient. When the FIR cells are configured in single filter mode, the Decimation Registers in FIR cell A and FIR cell B are cascaded. This extended filter tap delay path allows computation of a filter which is twice the size of that capable using a single cell. The Decimation Registers also provide data storage for polyphase or 2-D filtering applications (See Applications Examples Section).
For example , if the 4 LSBs are prog r ammed with a value of 0010, theForward andReverse ShiftingDecimation Registers are each configured with a delay of 3. Bit 4 is used to select whether the FIR cells operate as two independent filters or one extended length filter . Dual filter mode assumes Filter A and FilterB are separate independentfilters. In thesingle filter mode, thedata is routed through the forw ardpaths of Filters A
7
The Data FeedbackCircuitry in each FIR cell is responsible for transferring data from the Forward to the Reverse Shifting Decimation Registers.This circuitry feeds blocks of samples into the reverse shifting decimation path in either reversed or non-reversed sample order. The MUX/DEMUX structure at the input to the Feedback Circuitry routes data to the LIFOs or the delay stage depending on the selected
HSP43168
configuration. The MUX on the Feedback Circuitry Output selects which storage element feeds the Reverse Shifting Decimation Registers.
In applications requiring reversal of sample order, the FIR cells are configured with data reversal enabled (see Table 2, CW5, bit 4 = 0). In this mode, data is transferred from the forward to the backward Shifting Registers through a pingponged LIFO structure. While one LIFO is being read into the backward shifting path, the other LIFO is written with data samples. The MUX/DEMUX controls which LIFO is being written, and the MUX on the Feedback Circuitry output controls which LIFO is being read. A low on and
SHIFTEN, switches the LIFOs being read and written,
TXFR
which causes the block of data to be read from the structure in reversedin sample order (See Example4 in the Application Examples Section).
The frequency with which
TXFR is asserted determinessize of the data blocks in which sample order is reversed. For example, if
TXFR is asserted once every three CLKs, blocks of 3 data samples with order reversed, would be fed into the Backward Decimation Registers. NOTE: Altering the
frequency or phase of
TXFR assertion once a filtering
operation has begun will invalidate the filtering result.
In applications which do not require sample order reversal, the FIR cells must be configured with data reversal disabled (see Table 2, CW5, bit 4 = 1). In addition, TXFR must be asserted to ensure proper data flow. In this configuration, data to the backward shifting decimation path is routed though a delaystage instead of the pingpong LIFOs. The number of registers in the delay stage is based on the programmed decimation factor. NOTE: Data
reversal must be disabled and TXFR must be asserted for filtering applications which do not use decimation.
The shifting of data through the Forward and Reverse Decimation Registers is enabled by asserting the input. When
SHFTEN is high, data shifting is disabled, and
SHFTEN
the data sample latched into the parton the previous clockis the last input to the filter structure. The data sample at the filter input when
SHFTEN is asserted, will be the next data
sample into the forward decimation path. When operating the FIR cells as two independent filters, FIR
A receives input data via INA0-9 and FIR B receives data from either INA0-9 or INB0-9 depending on the application (see Table 1).
When the FIR cells are configured as a single extended length filter, the forward and reverse decimation paths of the two FIR cells are cascaded. In this mode,data is transferred from the forward decimation path to the reverse decimation path by the Data Feedback Circuitry in FIR B. Thus, the manner in which data is read into the reverse decimation path is determined by FIR B's configuration. When the decimation paths are cascaded, data is routed through the fourth delay stage in FIR A's forward path to FIR B.
The configuration of the FIR cells as even or odd length filters determines the point in the forward decimation path from which data is multiplexed to the Data Feedback Circuitry. For example, if the FIR cell is configured as an odd length filter, data prior to the last register in the third forward decimation stage is routed to the Feedback Circuitry. If the FIR cell is configured as an even length filter, data output from the third forward decimation stage is multiplexed to the Feedback Circuitry.This isrequired to ensure properdata alignment with symmetric filter coefficients (See Application Examples).
ALUs
Data shifting through the forward and reverse decimation paths feed the “a” and “b” inputs of the ALUs respectively. The ALUs perform an “b+a” operation if the FIR cell is configured for even symmetric coefficients or an “b-a” operation if configured for odd symmetric coefficients. Control Word 0, Bit 5 is used to set the ALU operation.
Forapplications in which a pre-add or subtract is not required, the “a” or “b” input can be zeroed by disabling RVRS respectively. This has the effect of producing an ALU output which is either “a”, “-a”, or “b” depending on the filter symmetry chosen. For example, if the FIR cell is configured for an even symmetric filter with
FWRD low andRVRS high, the data shifting through the Forward Decimation Registers would appear on the ALU output.
Table 3 details the ALU configurations, where “a” is the ALU data inputfrom the front decimation delayregisters and “b” is the ALU data from the back decimation delay registers.
TABLE 3. ALU CONFIGURATIONS
ALU
OUT SYMMETRY FWD RVS DESCRIPTION
a+b 0 (Even) 0 0 EvenNumber ofTaps, Even
Symmetry (Example 1) +b 0 (Even) 0 1 Even Symmetry +a 0 (Even) 1 0 Even Symmetry
- 0 (Even) 1 1 Even Symmetry
b-a 1 (Odd) 0 0 Even Number of Taps, Odd
Symmetry (Example 2) +b 1 (Odd) 0 1 Odd Symmetry
-a 1 (Odd) 1 0 Odd Symmetry
- 1 (Odd) 1 1 Odd Symmetry
Coefficient Bank
The output of the ALU is multiplied by a coefficient from one of 32 user programmable coefficient sets. Each set consists of 8 coefficients (4 coefficients for FIR A and 4 for FIR B). CSEL0-4 is used to select a coefficient set to be used. Coefficient sets may be switched every clock to support polyphase filtering operations.
The coefficients are loaded into On-Board Registers using the microprocessor interface, CIN0-9, A0-8, and multiplier within the FIR Cells is driven by a coefficient bank
FWRD or
WR. Each
8
Loading...
+ 18 hidden pages