可編程邏輯系統(tǒng)通常部署在可能存在噪聲的應(yīng)用中。這種噪聲會(huì)影響可編程邏輯設(shè)計(jì)接收的信號(hào)。例如，它可能會(huì)導(dǎo)致信號(hào)故障或跳動(dòng)，如果處理不當(dāng)，可能會(huì)導(dǎo)致設(shè)計(jì)和操作出現(xiàn)問題。

毛刺的持續(xù)時(shí)間是隨機(jī)的，并且與時(shí)鐘沿不同步。因此，它們可能會(huì)導(dǎo)致下游信息損壞。

處理此問題的最常見方法是使用毛刺濾波器來濾除毛刺和反彈。

毛刺濾波器核心是使用長(zhǎng)度可變的移位寄存器，噪聲信號(hào)被放到寄存器中，直到移位寄存器的所有值都一致。此時(shí)，信號(hào)可以視為穩(wěn)定。當(dāng)然，我們必須確定潛在毛刺和反彈可能持續(xù)多長(zhǎng)時(shí)間，以確保時(shí)鐘周期的寄存器大小正確。這就是為什么我們的毛刺濾波器需要非常靈活，并且需要確保其大小能夠適合每個(gè)應(yīng)用程序的要求。

濾波器應(yīng)該能夠接收噪聲輸入并濾除持續(xù)時(shí)間為多個(gè)時(shí)鐘脈沖的毛刺。

libraryieee;
useieee.std_logic_1164.all;
useieee.numeric_std.all;

entityglitch_filteris
generic(
G_FILER_LEN:integer:=8
);
port(
i_clk:instd_ulogic;
i_noisy:instd_ulogic;
o_clean:outstd_ulogic
);
endglitch_filter;

architecturebehaviourofglitch_filteris

signals_delay_line:std_ulogic_vector(G_FILER_LEN-1downto0);
signals_delay_and:std_ulogic;
signals_delay_nor:std_ulogic;
signals_output_clean:std_ulogic;

begin

o_clean<=?s_output_clean;

????--Delay?disctete?using?delay?line
????synchroniser_process?:?process?(i_clk)?begin
????????if?rising_edge(i_clk)?then
????????????s_delay_line?<=?s_delay_line(G_FILER_LEN?-?2?downto?0)?&?
????????????????????????????i_noisy;
????????end?if;
????end?process;

????--Generate?AND?and?NOR?of?delay?line?bits
????s_delay_and?<=?'1'?when?to_01(s_delay_line)?=?
????????????????????????????(s_delay_line'range?=>'1')else'0';
s_delay_nor<=?'1'?when?to_01(s_delay_line)?=?
????????????????????????????(s_delay_line'range?=>'0')else'0';

--Setdiscretebasedondelayline
output_process:process(i_clk)begin
ifrising_edge(i_clk)then
ifs_delay_nor='1'then
s_output_clean<=?'0';
????????????elsif?s_delay_and?=?'1'?then
????????????????s_output_clean?<=?'1';
????????????end?if;
????????end?if;
????end?process;

end?behaviour;

為了測(cè)試這個(gè)模塊，創(chuàng)建一個(gè)簡(jiǎn)單的測(cè)試文件，它將隨機(jī)數(shù)量的毛刺注入信號(hào)中。在信號(hào)改變狀態(tài)后，許多隨機(jī)毛刺被輸入到信號(hào)中。如果濾波器運(yùn)行正常，則這些毛刺將在毛刺濾波器輸出干凈的信號(hào)。

libraryieee;
useieee.std_logic_1164.all;
useieee.numeric_std.all;
useieee.math_real.all;

entityglitch_filter_tbis
end;

architecturebenchofglitch_filter_tbis

componentglitch_filter
generic(
G_FILER_LEN:integer
);
port(
i_clk:instd_ulogic;
i_noisy:instd_ulogic;
o_clean:outstd_ulogic
);
endcomponent;

--Clockperiod
constantclk_period:time:=10ns;
--Generics
constantG_FILER_LEN:integer:=8;

--Ports
signali_clk:std_ulogic:='0';
signali_noisy:std_ulogic;
signalo_clean:std_ulogic;

begin

i_clk<=?not?i_clk?after?(clk_period/2);

??glitch_filter_inst?:?glitch_filter
????generic?map?(
??????G_FILER_LEN?=>G_FILER_LEN
)
portmap(
i_clk=>i_clk,
i_noisy=>i_noisy,
o_clean=>o_clean
);

uut:process

variableglitch_duration:integer;
variableseed1:positive:=1;
variableseed2:positive:=283647823;

impurefunctioninteger_random(min,max:integer)returnintegeris
variablerandom:real;
begin
uniform(seed1,seed2,random);
returninteger(round(random*real(max-min)+real(min)));
endfunction;

begin
i_noisy<=?'0';
????wait?until?rising_edge(i_clk);
????wait?for?G_FILER_LEN?*?clk_period;
????test:?for?i?in?0?to?1?loop
????????i_noisy?<=?'1';
????????wait?until?rising_edge(i_clk);
????????glitch_duration?:=?integer_random(1,5);
????????for?x?in?0?to?glitch_duration?loop
????????????i_noisy?<=?not?i_noisy;
????????????wait?until?rising_edge(i_clk);
????????end?loop;
????????i_noisy?<=?'1';
????????wait?for?20?*?clk_period;
????????report?"loop?high?completed"?severity?note;
????????i_noisy?<=?'0';
????????wait?until?rising_edge(i_clk);
????????glitch_duration?:=?integer_random(1,5);
????????for?x?in?0?to?glitch_duration?loop
????????????i_noisy?<=?not?i_noisy;
????????????wait?until?rising_edge(i_clk);
????????end?loop;
????????i_noisy?<=?'0';
????????wait?for?20?*?clk_period;
????????report?"loop?low?completed"?severity?note;
????end?loop;
????report?"Simulation?complete"?severity?failure;
????
end?process;

end;

運(yùn)行仿真后顯示在信號(hào)狀態(tài)改變后隨機(jī)數(shù)量的脈沖便增加。檢查輸出信號(hào)表明濾波器已正確濾除輸入信號(hào)中可能存在的毛刺。

正如在一開始所說的，這樣的濾波器對(duì)于部署在可能存在電噪聲的環(huán)境中非常有用。與 BRAM 上的 EDAC 等其他緩解策略相結(jié)合，這是可用于實(shí)現(xiàn)設(shè)計(jì)彈性的關(guān)鍵方法之一。

審核編輯：劉清