Anda di halaman 1dari 16

Understanding / Reading Bode Plots

Source: Wikipedia and Motion Engineering Technical Support pages

Introduction
A Bode plot is a graph of the transfer function of a linear, ti e!in"ariant s#ste "ersus fre$uenc#, plotted %ith a log!fre$uenc# a&is, to sho% a s#ste 's fre$uenc# response( )t is usuall# a co *ination of a Bode agnitude plot, e&pressing the agnitude of the fre$uenc# response gain, and a Bode phase plot, e&pressing the fre$uenc# response phase shift( A ong his se"eral i portant contri*utions to circuit theor# and control theor#, engineer +endrik Wade Bode ,-./01-.234, %hile %orking at Bell 5a*s in the United States in the -.6/s, de"ised a si ple *ut accurate ethod for graphing gain and phase!shift plots( These *ear his na e, Bode gain plot and Bode phase plot( The agnitude a&is of the Bode plot is usuall# e&pressed as deci*els of po%er, that is *# the 3/ log rule: 3/ ti es the co on ,*ase -/4 logarith of the a plitude gain( With the agnitude gain *eing logarith ic, Bode plots ake ultiplication of agnitudes a si ple atter of adding distances on the graph ,in deci*els4, since

Most Motion 7ontrol Manufacturer8s pro"ide a Bode Tool %ith their Progra ing and/or Tuning Soft%are( A Bode Tool's ain function is to ake it eas# to get fre$uenc# response data for #our s#ste ( 9or fre$uenc# response data to *e useful, #ou need to kno% %hat it eans( This White Paper is an introduction to understanding fre$uenc# response plots( We %ill tr# to keep ath to a ini u %hile still de onstrating the concepts in"ol"ed( :eep in ind, this page is intended as a guide to understanding ho% to read fre$uenc# response plots of #our s#ste , not a definiti"e dissertation on fre$uenc# anal#sis or a control loop tuning guide(

Background
This section %ill la# so e of the foundation for ho% non ! sine %a"e *ased signals can *e read in ter s of their sine co ponents( )f #ou are co forta*le %ith this concept, or %ant to skip it, go straight to the section 9re$uenc# Response(

Making a Square Wave From Sine Waves


A linear function ,i(e( a s$uare or triangle %a"e, or the o"e ent of #our ser"o s#ste 4 can *e represented *# a series of sine %a"es( This is a critical point, so let's take a look at a couple of e&a ples( The first e&a ple is as eas# as it gets( A single sine %a"e can *e represented *# # ; sin,&4( This is ore a definition than a disco"er#(

<:, that %as eas#( 5et's see %hat %e can do %ith a s$uare %a"e( A s$uare %a"e that goes fro - to !- at the sa e fre$uenc# as the sine %a"e a*o"e can *e represented *# the series of sine %a"es represented *#

=o% don't %orr# a*out *eing a*le to co e up %ith this e$uation( 5et's >ust focus on the idea that #ou can ake a s$uare %a"e %ith >ust a *unch of sine %a"es( Another i portant point is that %e need to add an infinite nu *er ,all the odd "alues of n4 to ake a perfect s$uare %a"e out of a *unch of sine %a"es( We'll get to the ra ification of this later( Right no%, let's take a look at a s$uare %a"e ade using the a*o"e e$uation using n ; -, n ; - to --, and all the odd n's(

A couple of i portant notes here: With an infinite number of n's, the square wave made of sine waves is perfect. With on ! one sine wave "n # $%, the square wave turns into a sine wave. With n # $ to $$, the square wave is c ose to, but not e&act ! a square wave. 'he more n's !ou use, the better !our square wave. This eans that #ou get ripples in #our s$uare %a"e if #ou can't reproduce high fre$uencies ,think a*out ho% high a fre$uenc# #our achine can reproduce4(

Making Square Waves from Sine Waves, (art )


The plots a*o"e %ere sho%n in the ti e do ain( That eans that %e plot a plitude "s( ti e( A fre$uenc# response plot, like a Bode Tool, sho%s results in the fre$uenc# do ain( This eans that %e plot a plitude "s( fre$uenc#( Think of it as plotting ho% loud a piano is *# plotting a plitude per ke#( Each ke# is at a different fre$uenc# and #ou can represent the usic *# %hat ke# is pressed, and ho% hard ,in a crude sense4( 5et's take a look at a sine %a"e in the ti e do ain and the fre$uenc# do ain:

<:, this isn't too *ad( The fre$uenc# do ain chart, >ust sho%s that there is a sine %a"e at a single fre$uenc# of - radian / sec and its a plitude is -( 5et's take a look at the s$uare %a"e for the cases %here %e ha"e the sine co ponents n ; -, n ; - to --, and all odd n's(

?ou can see all the sine a plitudes and fre$uencies that ake up a s$uare %a"e ,up to n ; -.4( The i portant part is not to *e a*le to figure out the fre$uenc# content of a s$uare %a"e, *ut to understand that the t%o a*o"e plots are different %a#s of reading the sa e data and ho% to read the plots(

Fr e que nc ! * es po ns e
With so e *ackground, let's o"e to a related, *ut not identical idea of frequency response.

Frequenc! *esponse definition


5et's take the idea of aking a signal ,like the s$uare %a"e a*o"e4 out of sine %a"es and e&tend it to tell #ou %hat a s#ste %ill output %ith a gi"en input( )f #ou input a sine %a"e into a s#ste and %atch the output, #ou can easure the relationship *et%een the input and output( 5et's take a look at ho% this is done(

We co and the s#ste to follo% a sine %a"e( We then response in the ti e do ain, at the sa e ti e(

easure the input and output

9irst #ou easure the ratio of the output a plitude to input a plitude( This is usuall# e&pressed in deci*els ,dB4( )n this e&a ple, %e ha"e 3/@log,- (A / 3(/4 ; !6(- dB( This is referred to as the amp itude response. Second, #ou need to kno% ho% uch the output lagged *ehind the input in ti e( This is easured in degrees, %here 6B/ degrees is the length of one c#cle( )n the a*o"e e&a ple, the output lags A0 degrees *ehind the input( Re e *er that A0 degrees at one fre$uenc# is a different a ount of ti e than it is at 3 ti es that fre$uenc#( The ti e that the output lags *ehind the input is referred to as the phase response. When a Bode Tool uses a sine s%eep ode, it is >ust co pleting the a plitude and phase response easure ent o"er and o"er at different fre$uencies so %e can plot the frequenc! response ,a plitude and phase response together4 across a fre$uenc# range(

'he Bode ( ot

<nce %e ha"e collected the phase and a plitude response for an# fre$uencies, one of the ost intuiti"e %a#s to displa# it is to plot the a plitude response "s( fre$uenc# and the phase response "s( fre$uenc# and put it on one plot( This plot is called the Bode plot(

The a*o"e plot sho%s the fre$uenc# response of an idealiCed si plified odel of an actual otor4(

otor ,data *ased of a

9irst, look at the fre$uenc# response at fre$uencies less than 0 +C( The response is close to / dB ,- /D,/ dB / 3/4 ; -(/4( This eans that the otor's output %ill look like the otor's input( )f #ou input a 3 +C position co and, this otor %ith this control s#ste can follo% it %ith close to identical a plitude( 5ook at the phase, in the sa e fre$uenc# range( The phase is near / degrees, %hich eans that the output %ill not lag *ehind the input( / dB at / degrees i plies perfect agree ent *et%een the output and input to the s#ste ( This is %hat %e are shooting for( Second, let's take a look at the a plitude and phase response around -/ +C( The a plitude response is a*out EA dB, !B/ degrees( This eans that the output %ill *e -(B ti es as large as the output and B/ degrees *ehind the input at -/ +C( This isn't %hat #ou %ant( ?ou %ant #our s#ste to o"e as far as #ou told it to, no ore, no less( 5astl#, let's look at the fre$uenc# response at 3// +C( The a plitude response is !A/ dB, !./ degrees( This eans that the output a plitude is /(/- ti es the input a plitude ,-/D,!A//3/4 ; /(/- and the output lags ./ degrees *ehind the input( ?ou don't %ant this either( )f #ou co and the s#ste to o"e one unit, #ou %ant it to o"e one unit, not /(/- units( These are the challenges of controls s#ste s(

=o% that %e ha"e the *asics of ho% to read a Bode Plot and a si ple idea of %hat the plot is telling us, let's look at so e e&a ples to get a ore in depth understanding of %hat the Bode Plot is telling us(

+ osed ,oop, -pen ,oop, +ontro ,oop, ( ant *esponses


The a*o"e e&a ple sho%s the closed loop response of a s#ste ( 5et's take a o ent to talk a*out %hat closed loop, open loop, control loop, and plant ean( The follo%ing figure sho%s a *lock diagra of a P)F control loop( )t could >ust as easil# *e a P)G or fre$uenc# shaping control loop *# replacing P)F %ith another algorith in the sa e place(

We alread# touched on closed loop response in the pre"ious section( The closed loop response is the a ount of output ,actual position4 per input ,co anded position4 in *oth a plitude and phase( What akes it the closed loop response as opposed to another t#pe of response is that %e took the easure ent %hile the s#ste %as under closed loop ser"o control( )f %e disconnected the feed*ack of the s#ste , *ut other%ise left the s#ste alone, %e %ould ha"e a "er# different s#ste ( )f %e easured that s#ste as descri*ed a*o"e, %e %ould get the open loop response.

Most of the ti e, running the s#ste open loop like is sho%n in the diagra a*o"e results in %ild otion that is unaccepta*le( 7onsidering this, %e often deter ine the open loop response of a s#ste fro the closed loop response( So e athe atical g#rations are re$uired for this, *ut luckil#, the Bode Tool can do this for us( The control loop response is si pl# the response of the controller ,P)F section in the a*o"e diagra s4( We easure the control loop response *# easuring ho% uch control loop output %e get for an a ount of control loop input in a plitude and phase( )t doesn't reall# atter if the loop is closed or open( )t >ust atters that %e ha"e a easurea*le a ount of control loop input and output to o*ser"e( 9or that atter, the control loop is eas# to si ulate accuratel#, so %e rarel# easure the control loop response directl# !! %e >ust rel# on Bode Tool si ulations of the control loop( The plant response is si pl# the response of the s#ste ,plant4( The ter plant refers to the s#ste under control and can consist of echanical / electrical / sensor / other aspects( )t does not refer to the control loop or controller( Think of the plant as #our ser"o s#ste *efore #ou hook up the controller( ?ou a# ha"e otors, encoders, echanical s#ste s,

etc( The ter HplantH co es fro classical control theor#( The plant response is easured *# easuring the a ount of plant output per plant input in a plitude and phase( :eep in ind that the closed loop, open loop, control loop, and plant responses are all separate easure ents, *ut are not co pletel# unrelated, as the# are all part of the sa e ser"o s#ste (

.nderdamped / +ritica ! damped / -verdamped S!stems


Man# s#ste designers %ill ha"e "er# fir opinions on criticall# da ped "s( underda ped s#ste s( Sidestepping the erits of each, let's look at %hat constitutes an under / o"er / criticall# da ped s#ste ( 9irst, let's look at a plot of an under, o"er, and criticall# da ped s#ste (

This plot sho%s three s#ste s that start at position /(/ and are co anded to go to position -(/( The underda ped s#ste o"ershoots the target and co es *ack to the target( The criticall# da ped s#ste approached -(/ as fast as it can %ithout o"ershooting( The o"erda ped s#ste approaches ore slo%l# than is re$uired to not o"ershoot( 5et's co pare this to si ilar data in the fre$uenc# do ain(

)n so e %a#s, the fre$uenc# and ti e do ain plots ha"e the sa e defining characteristic that sho%s %hether the s#ste is under, o"er, or criticall# da ped( The underda ped otor sho%s a peaking in the fre$uenc# response( )t doesn't oscillate a*out a target in the fre$uenc# response, though( The criticall# da ped s#ste has / dB response until so e fre$uenc# %hen it falls cleanl# a%a# fro / dB( The o"erda ped s#ste falls a%a# fro / dB earlier and ore graduall#( Understanding ho% to identif# an under, o"er, or criticall# da ped s#ste is i portant *ecause it is an i portant easure of s#ste sta*ilit# used in control loop tuning(

Bandwidth
Band%idth is a ter that is used fre$uentl# to descri*e ho% HfastH( or ho% HgoodH a s#ste is( The *and%idth of a s#ste is erel# the fre$uenc# at %hich the closed loop a plitude response falls to !6 dB( )n the e&a ple *elo%, the *and%idth of the s#ste is -I +C(

Band%idth is not a "er# precise easure of s#ste perfor ance( Use it for rough ideas of s#ste perfor ance, *ut don't ake uch of so eone telling #ou a*out 6J difference in *and%idth across different achines or the like(

0ain and (hase Margins


Kain and phase argins are co on ter s to descri*e ho% sta*le a s#ste is( Kain and phase argins are used ore *ecause the# are si ple than *ecause the# are ideal easure ents of sta*ilit#( Kain and phase argins are easured fro the open loop fre$uenc# response of the s#ste ( This is a critical thing to note( Kain and phase argins cannot *e ac$uired directl# fro a closed loop fre$uenc# response( To easure the gain argin of a s#ste , find the point that the open loop phase response crosses !-2/ degrees( At the sa e fre$uenc# as this point, find the open loop a plitude response( The distance *elo% / dB that the open loop a plitude response reads at this fre$uenc# is the gain argin( )n the e&a ple *elo% the gain argin is 63(0 dB( To easure the phase argin of a s#ste , find the point that the open loop a plitude crosses / dB( At the sa e fre$uenc# as this point, find the open loop phase response( The distance a*o"e !-2/ degrees that the open loop phase response reads at this fre$uenc# is the phase argin( )n the e&a ple *elo%, the phase argin is 66 degrees ,-2/ degrees ! -AI degrees ; 66 degrees4(

Kain and phase argins are supplied in pairs( The e&a ple a*o"e %ould *e descri*ed at ha"ing gain and phase argins of H63(0 dB, 66 degrees(H

Instabi it!
Ser"o s#ste s ha"e the pro*le of insta*ilit#( )f the control s#ste for a ser"o is applied incorrectl#, the s#ste %ill *e unsta*le( This can ean an#thing fro the s#ste aking o*>ectiona*le noises *ecause of oscillations to a co plete ser"o runa%a# %here the ser"o %ill tr# as hard as it can to o"e away fro the target position( There are an# different %a#s to easure ser"o insta*ilities, *ut let us consider >ust one case for no%: /dB, !-2/ degrees( 9irst %e %ill talk a*out %hat akes /dB, !-2/ degrees unsta*le, then %e %ill talk a*out ho% to tell if #our s#ste is close to this condition( )f #ou %ant, skip to the section, +o% Fo ) Tell )f M# S#ste )s 7lose To / dB, !-2/ FegreesL

Wh! Is 1dB, 2$31 4egrees So Bad5


This $uestion re$uires looking at ho% a feed*ack s#ste ,ser"os are feed*ack s#ste s4 %orks( 5ooking at the t%o cases %here -4 the open loop response is / dB, / degrees and 34 the open loop response is / dB, !-2/ degrees %ill illustrate %h# / dB, !-2/ degrees is so *ad( 9irst, let's look at / dB, / degrees open loop response( This eans that %hen #ou send a co and, the otor does e&actl# that( )t akes a o"e %ith the sa e a plitude and in phase( Since this is open loop response, it eans that the s#ste does this %ithout considering the feed*ack loop( 5ooking at the follo%ing figure, consider that the sine %a"e on the input %ill sho% up *ack fro the s#ste *ecause it does e&actl# %hat %e tell it( This eans that the error *lock ,orange circle %ith a cross in it4 %ill su*tract the actual s#ste

response fro the co and and get !! CeroM This eans that the ser"o s#ste ha"e to ake an# corrections( The s#ste is sta*le and accurate(

doesn't

=o% let's look at the sa e s#ste , *ut %ith a / dB, !-2/ degree open loop response, instead of a / dB, / degree open loop response( We put in a sine %a"e and *ecause the s#ste is ! -2/ degrees out of phase, %e get the sa e sine %a"e, *ut in"ertedM When the error *lock su*tracts the t%o, %e get an error that is t%ice the siCe of the inputM This is no good( This causes the s#ste to create an actual response fro the s#ste that is *igger, causing a *igger error, and this %ill spiral out of control until the s#ste a*orts or *reaks( This is an unstable s#ste (

6ow 4o I 'e If M! S!stem Is + ose 'o 1 dB, 2$31 4egrees5


There are t%o signs: -4 large a plitudes in the closed loop a plitude response and 34 the open loop fre$uenc# response approaches / dB, !-2/ degrees( )t is eas# to see a large closed loop a plitude response( 5arge %ould *e an# "alue of N B dB or so( An e&a ple is sho%n *elo%(

There is a huge 3/ dB peak( )n practice, an# closed loop a plitude peak a*o"e B dB or so should ha"e #ou seriousl# considering the sta*ilit# of #our s#ste ( The closed loop a plitude peak is reall# a s# pto of the pro*le ,a serious s# pto M4( The real pro*le is that the open loop response is aproaching / dB, !-2/ degrees( 5et's take a look at the sa e s#ste a*o"e, *ut let's look at the open loop response(

This is the sa e s#ste and it has 0(0 degrees of phase argin ,-2/!-IA(0 ; 0(04 and 60 dB of gain argin( Most controls engineers %ould shoot for 6/ degrees of phase argin as a ini u ( We can %orr# later a*out e&actl# %hat is accepta*le or not %hen %e appl# P)F function tuning( 9or no%, it is enough to kno% %hat easure ents on the Bode Plot %ill let us decide %hether a s#ste is sta*le or not(

Anda mungkin juga menyukai