View FIRFILTERS_1071473.PDF datasheet online --- IC-ON-LINE

Datasheet File OCR Text:

altera corporation 1 fir filters january 1996, ver. 1 functional speci?ation 1 a-fs-01-01 features n high-speed operation: up to 105 million samples per second (msps) n 8-, 16-, 24-, 32-, and 64-tap finite impulse response (fir) filters n parameterized data widths, pipelining, and symmetry n optimized for flex 10k and flex 8000 devices n coefficients implemented as look-up table (lut) vectors n parallel and serial design versions n supported by schematic and text design entry methods, including vhdl, verilog hdl, and the altera hardware description language (ahdl) n useful for a variety of applications, including high performance, real- time video and image filtering; radio frequency (rf) filtering; radar; magnetic resonance imaging (mri) applications; multi-rate digital signal processing (dsp); and spread spectrum filtering general description table 1 shows the performance and size of specific fir filters in flex 8000a devices. the input, coefficient, internal, and output precision values are fully parameterized. the filter name specifies the number of taps in the filter, and the architecture and suffixes follow the naming convention: p = parallel, s = serial. the video convolution filter, fir_3x3 , is discussed on page 16 of this document. table 1. fir filter performance & size in flex 8000a devices fir filter input precision taps precision/bits size (logic elements) performance (msps) coef. internal output pipelined non- pipelined a-2 a-4 a-2 a-4 fir_08tp 8 8 8 17 17 296 101 66 28 15 fir_16tp 8 16 8 10 10 468 101 75 20 13 fir_24tp 8 24 8 10 10 653 100 74 18 12 fir_32tp 8 32 8 10 10 862 101 75 18 12 fir_16ts 8 16 8 18 18 272 7.0 4.9 3.4 2.4 fir_64ts 8 64 8 24 24 920 6.5 4.1 2.4 1.7 fir_3x3 8 9 8 18 18 327 102 67 24 14
2 altera corporation fs 1: fir filters adding more taps to the filter does not significantly change the speed for any given precision width. see figure 1 . figure 1. speed for 8-bit input filters in flex 8000a devices altera?s fir filters provide the designer with a great deal of flexibility because they can be customized for specialized applications by modifying the source code. in addition, the filters? flexibility allows designers to implement decimating or interpolating filters, and to use the vector multiplier from these reference designs to implement any operation that has multiply and accumulate (macs) functions, where one set of multiplicands is constant. all linear-phase fir filters have symmetric coefficients; therefore, the designer can take advantage of this symmetry by adding taps prior to multiplication. for example, an 8-tap fir filter can be implemented with a 4-input vector multiplier. the vector multiplier performs the following multiplication: y = ( h (1) s (1)) + ( h (2) s (2)) + ( h (3) s (3)) + ( h (4) s (4)), where h ( n ) are fir coefficients, and s ( n ) are pre-added data samples. the coefficients must be fixed prior to compilation, although you can change the coefficients by reconfiguring the device. thus, it is possible to change the filter function in response to user controls. the device can be reconfigured in less than 100 ms. to be implemented in flex devices, all fir filters require max+plus ii version 6.0 or higher. f see application note 73 ( implementing fir filters in flex devices ) for a complete description of vector multipliers and the fir filter architecture. 0 0 speed (msps) 8 16 24 32 20 40 60 80 100 120 64 taps serial filters parallel filters
altera corporation 3 fs 1: fir filters parameters the following fir filter parameters can be defined by the user: n data width ( width )?he width (in bits) of incoming data. n coefficient precision ( rom_p )?he width of the coefficients used for the fir filter, independent of the data width setting. n internal precision ( internal_p )?he precision (in bits) of internal calculations in the fir filter. n output precision ( output_p )?he precision (in bits) of the output data. this value can be different from the internal precision. n pipelined/non-pipelined design ( pipelined )?hen designers choose between a fully pipelined or non-pipelined design, the main filter design program, firgen, creates the coef.inc file accordingly. a fully pipelined design results in the maximum operating frequency with no extra device area usage. n symmetrical/antisymmetrical filters (" sym " or " anti ") designers can specify whether the filter is symmetric or antisymmetric. symmetrical fir filters implement even functions with symmetric tap adders. designers can create an antisymmetric fir filter that implements an odd function by using symmetric tap subtractors. firgen allows the designer to specify symmetrical or antisymmetrical coefficients. n coefficients ?irgen automatically generates an ahdl include file ( coef.inc ) that specifies vector table representations of filter coefficients. designers specify the original file containing coefficients to firgen, which then computes the vector tables and adds them to coef.inc . although the coefficients are specified at compilation, they are not parameters because they are defined by constants in the coef.inc file. if more than one filter exists in the design, modify a copy of the top-level text design file ( .tdf ) file (e.g., fir_08tp.tdf ) and change the name of the include file ( coef.inc ). parameters vs. ports parameters, which are variables established at design time, determine the configuration of the filter. parameters cannot be changed once the filter has been designed and is operating in-circuit. in contrast, ports refer to the in-circuit i/o ports, including the i/o data, clock, and reset.
4 altera corporation fs 1: fir filters design flow overview the 8-tap parallel fir filter design flow in figure 2 is common to all fir filter reference designs supplied with the altera dsp design kit. the designer can specify the filter requirements using any commercially available fir filter design software that can store the coefficients in an ascii file. this file should contain the filter coefficients one-per-line and should have the extension .dat . the listing below shows the coefficients for an 8-tap lowpass filter in the lp8.dat file, available from the altera dsp design kit: -5.144663e-03 -2.279553e-02 9.638777e-02 .4296515 .4296515 9.638777e-02 -2.279553e-02 -5.144663e-03 next, the designer runs firgen to read the coefficients from the data file. then, firgen calculates the vector table for the lut and creates coef.inc . using max+plus ii, the designer can compile the ahdl include file and set the parameters for the particular filter. to insure the smallest size and maximum performance, set the global project synthesis style to fast in the global project logic synthesis dialog box. this option uses carry chains for high-speed operation. firgen also generates a swept sine wave (or chirp signal) vector file, called sweep.vec . this file can be used as a simulation source file for the max+plus ii simulator to generate the filter? frequency response. the response is then viewed with any plotting program, such as microsoft excel or gnuplot, a freely distributed program included with the altera dsp design kit. 1 for a sample design, see the fir filter design walkthrough in the readme file on the altera dsp design kit cd-rom.
altera corporation 5 fs 1: fir filters figure 2. fir filter design flow firgen firgen plotting program fir filter design software max+plus ii compiler max+plus ii simulator sweep.vec fir_08tp.tdf sweep.tbl sweep.out coef.inc ascii file (< file name >.dat) flex 10k or flex 8000 device
6 altera corporation fs 1: fir filters instantiation the designer can compile a fir filter as a stand-alone design or can insert an instance of a fir filter in a design. the filter function is placed and connected in the max+plus ii graphic editor, an ahdl tdf, a third-party verilog hdl file, or a vhdl file. figure 3 shows an instance of an 8-tap fir filter in a graphic design file ( .gdf ). figure 3. 8-tap filter in a gdf figure 4 shows an instance of an 8-tap fir filter implemented in an ahdl tdf. fir_08tp y[output_p..1] rom_p = 8 width = 8 clk aclr xin[width..1] output_p = 17 pipelined = "yes" internal_p = 17 sym = "sym" xin[8..1] clk aclr y[17..1]
altera corporation 7 fs 1: fir filters figure 4. 8-tap filter in an ahdl tdf include "fir_08tp" constant input_width = 8; constant coefficient_p = 8; constant output_p = 17; constant internal_p = 15; constant sym = "sym" subdesign test_tdf ( xin[width..1] :input; clk :input; aclr :input; y[output_p..1] :output ) variable fir : fir_08tp with ( width = input_width, rom_p = coefficient_p, output_p = output_p, pipelined = "yes" internal_p = internal_p sym = "sym" ); begin fir.xin[] = xin[]; fir.aclr = aclr; fir.clk = clk; y[] = fir.y[]; end;
altera corporation 8 tr_08tp, tr_16tp, tr_24tp, tr_32tp & tr_64tp parallel fir filters features n parallel fir filter reference designs n high-speed operation: > 100 msps n parameterized data widths, pipelining, and symmetry n symmetrical or antisymmetrical coefficients n parallel design versions for maximum performance general description altera parallel fir filters are linear-phase designs with operation speeds of over 100 msps. by efficiently using the altera flex architecture, these filters provide extremely fast, flexible, and cost-effective design solutions. the coefficients are stored in luts and their values are determined at compilation; however, the coefficients can be changed by reconfiguring the device. figure 5 shows the symbol for the 8-tap parallel fir filter. all other parallel fir filters in the altera dsp design kit have the same symbol. coefficients are defined in an include file ( .inc ) called coef.inc . figure 5. 8-tap parallel fir filter symbol function prototype the ahdl function prototype for the fir_08tp function is shown below and is the same for all other parallel fir filters in the altera dsp design kit: function fir_08tp (xin[width..1] clk, aclr) with (width , rom_p , internal_p , output_p , pipelined) returns (y[output_p..1]) ; fir_08tp y[output_p..1] rom_p = width = clk aclr xin[width..1] output_p = pipelined = internal_p =
altera corporation 9 fs 1 : fir filte r s parameters parameters for the parallel fir filter functions are provided in table 2 . in the default value of the output_p parameter, the constant blocks refers to the number of 8-tap blocks in the filter. ports input and output ports for the parallel fir filter functions are provided in table 3 . block diagrams figures 6 through 9 show block diagrams for the fir_08tp , fir_16tp , fir_24tp , and fir_32tp filters. f see application note 73 ( implementing fir filters in flex devices ) for more information on how to implement vector multipliers. table 2. parallel fir filter parameters name default value description width 8 integer input word width (in bits) rom_p 8 integer coefficient precision pipelined "yes" " yes " or " no " pipelined/non-pipelined sym "sym" "sym" or "anti" symmetrical/antisymmetrical output_p width+blocks+rom_p integer output precision (in bits) internal_p output_p integer internal precision (in bits) table 3. input & output ports port type name description input xin[width..1] data input input clk clock input input aclr asynchronous clear output y[output_p..1] data output
10 altera corporation fs 1: fir filters figure 6. fir_08tp filter block diagram c(1) c(2) c(3) c(4) parallel vector multiplier w xin[width..1] s (1) s (2) s (3) s (4) w + 1 w + 1 w + 1 w + 1 y[output_p..1] w = width
altera corporation 11 fs 1: fir filters figure 7. fir_16tp filter block diagram figure 8. fir_24tp filter block diagram figure 9. fir_32tp filter block diagram 8-tap filter block w w w w 8-tap filter block xin[width..1] y[output_p..1] w w xin[width..1] 8-tap filter block w w w 8-tap filter block 8-tap filter block w w y[output_p..1] w 8-tap filter block w w w w 8-tap filter block 8-tap filter block w w 8-tap filter block w w xin[width..1] y[output_p..1] w
altera corporation 12 tr_16ts & tr_64ts serial fir filters features n serial fir filter reference designs n serial implementation for minimum resource usage n high performance: > 6 msps n parameterized data widths, pipelining, and symmetry n symmetrical or antisymmetrical coefficients general description the altera serial fir filters are small, linear-phase designs with operation speeds of over 6 msps. by efficiently using the altera flex architecture, these filters provide extremely fast, flexible, and cost-effective design solutions. in addition, the filters use bit-serial computation, making them significantly smaller than parallel implementations. the filters are implemented using the vector-multiplier approach described in application note 73 ( implementing fir filters in flex devices ) . the coefficients are stored in luts and their values are determined at compilation; however, the coefficients can be changed by reconfiguring the device. other tap lengths are available by modifying a copy of the ahdl tdf. these designs convert parallel data input into serial internal data for processing. thus, each input sample must be applied to the input ( xin) for width + 1 clock cycles. figure 10 shows the symbol for the 16-tap serial fir filter. all other serial fir filters in the altera dsp design kit have the same symbol. coefficients are defined in an include file ( .inc ) called coef.inc . figure 10. serial fir filter symbol fir_16ts y[output_p..1] rom_p = output_p = clk aclr xin[width..1] pipelined = sym = width =
altera corporation 13 fs 1: fir filters function prototype the ahdl function prototype for the fir_16ts function is shown below and is the same for all other serial fir filters in the altera dsp design kit: function fir_16ts (xin[width..1] clk, aclr) with (width, pipelined, sym, rom_p, output_p) returns (y[output_p..1]); parameters parameters for the serial fir filter functions are provided in table 4 . in the default value for the output_p parameter, the constant blocks refers to the number of 8-tap blocks in the filter. ports input and output ports for the serial fir filter functions are provided in table 5 . table 4. serial fir filter parameters name default value description width 8 integers input word width (in bits) rom_p 8 integers coefficient precision pipelined " yes " " yes " or " no " pipelined/non-pipelined sym " sym "" sym " or " anti " symmetrical/antisymmetrical output_p width+blocks+rom_p integers output precision (in bits) table 5. input & output ports port type name description input xin[width..1] data input input clk clock input input aclr asynchronous clear output y[output_p..1] data output
14 altera corporation fs 1 : fir filte r s block diagrams figures 11 and 12 show block diagrams for the fir_16ts and fir_64ts fir filters. figure 11. fir_16ts filter block diagram n + 1 shift register serial adder n + 1 shift register n + 1 shift register n + 1 shift register n + 1 shift register n + 1 shift register n + 1 shift register n + 1 shift register parallel- to-serial shift register cin cout dq clr serial adder cl r sum serial adder cl r sum serial adder cl r sum 8 16 lut scaling accumulator carry_clr plsr_load accum_clr add_sub latch_result result register rom_p control block 8- t ap serial fir filter block y[output_p..1] plsr_load xin[width..1] 1 8- t ap serial fir filter block
altera corporation 15 fs 1: fir filters figure 12. fir_64ts filter block diagram 8 parallel-to-serial shift register scaling accumulator accum_load result register control block 8 8 8 8 8 8 8 8 8 8 8 8 plsr_load carry_clear 8-tap serial fir filter block 8-tap serial fir filter block 8-tap serial fir filter block 8-tap serial fir filter block 8-tap serial fir filter block 8-tap serial fir filter block 8-tap serial fir filter block y[output_p..1] xin[width..1] latch_output
altera corporation 16 tr_3x3 symmetrical video fir filter features n symmetrical video fir filter reference design n high performance: > 100 msps n parallel implementation for maximum performance n parameterized data widths and symmetry general description the altera fir_3x3 symmetrical video fir filter is a linear-phase design with an operation speed of over 100 msps. by efficiently using the altera flex architecture, this filter provides an extremely fast, flexible, and cost- effective convolution operation. these filters are implemented with the vector-multiplier approach described in application note 73 ( implementing fir filters in flex devices ) . the coefficients are stored in luts and their values are determined at compilation; however, the coefficients can be changed by reconfiguring the device. figure 13 shows the symbol for the fir_3x3 symmetrical video fir filter. figure 13. fir_3x3 video filter symbol function prototype the ahdl function prototype for the fir_3x3 function is shown below: function fir_3x3 (xin[3..1][width..1], clk, aclr) with (width, rom_p, output_p, internal_p, pipelined) returns (y[output_p..1]); fir_3x3 y[output_p..1] internal_p = pipelined = clk aclr xin[3..1][width..1] rom_p = output_p = width =
altera corporation 17 fs 1 : fir filte r s parameters parameters for the fir_3x3 function design are provided in table 6 . ports input and output ports for the fir_3x3 function are provided in t able 7 . block diagram figure 14 shows the block diagram of the fir_3x3 video filter. one input to the vector multiplier is zero and the other three inputs are the sums of the symmetric taps. the n -bit value from pixel 1 is stuffed with two leading zeroes so that all inputs to the parallel vector multiplier are n + 2 bits wide. table 6. fir_3x3 parameters name default value description width 8 integers input word width (in bits) rom_p 8 integers coefficient precision pipelined " yes " "yes" or "no" symmetrical/antisymmetrical output_p width+2+rom_p integers output precision (in bits) internal_p = output_p = output_p pipelined/non-pipelined table 7. input & output ports port type name description input xin[3..1][8..1] data input input clk clock input input aclr asynchronous clear output y[output_p..1] data output
18 altera corporation fs 1: fir filters figure 14. fir_3x3 video filter block diagram parallel vector multiplier gnd 3 2 3 2 1 2 3 2 3 w w w +2 w +2 w +2 w +2 xin[3][width..1] y[output..1] xin[2][width..1] xin[1][width..1]
altera, max, max+plus, and flex are registered trademarks of altera corporation. the following are trademarks of altera corporation: max+plus ii, ahdl, and flex 10k. altera acknowledges the trademarks of other organizations for their respective products or services mentioned in this document, specifically: verilog and verilog-xl are registered trademarks of cadence design systems, inc. mentor graphics is a registered trademark of mentor graphics corporation. synopsys is a registered trademark of synopsys, inc. viewlogic is a registered trademark of viewlogic systems, inc. altera products are protected under numerous u.s. and foreign patents and pending applications, maskwork rights, and copyrights. altera warrants performance of its semiconductor products to current specifications in accordance with altera?s standard warranty, but reserves the right to make changes to any products and services at any time without notice. altera assumes no responsibility or liability arising out of the application or use of any information, product, or service described herein except as expressly agreed to in writing by altera corporation. altera customers are advised to obtain the latest version of device specifications before relying on any published information and before placing orders for products or services. copyright 1996 altera corporation. all rights reserved. 2610 orchard parkway san jose, ca 95134-2020 (408) 894-7000 a pplications hotline: (800) 800-epld customer marketing: (408) 894-7104 literature services: (408) 894-7144 fs 1: fir filters 20 altera corporation printed on recycled paper.

▲Up To Search▲

Price & Availability of FIRFILTERS

	To Download FIRFILTERS Datasheet File
If you can't view the Datasheet, Please click here to try to view without PDF Reader .