Fundamentals of Vacuum Technology (1248463), страница 56
Текст из файла (страница 56)
The measuring advantages of the instruments, such as speed,precision and reliability, would not be completely exploited if this informationwere not inputted into an improved process monitoring system. For acoating process this means the coating rate should be kept as close andstable as possible to a setpoint. The purpose of the closed loop is to makeuse of the information flow of the measuring system in order to regulate thecapacity for a special evaporation source in an appropriately adapted way.When the system functions correctly, the controller translates smalldeviations of the controlled parameter (the rate) from the setpoint intocorrection values of the re-adjusted evaporation capacity parameter.
Theability of the controller to measure quickly and precisely keeps the processfrom deviating significantly from the setpoint.The most widespread type of controller is the PID controller. Here P standsfor proportional, I for integral and D for differential control function. In thefollowing some of the properties of this controller are described in detail.Information on the system behavior is gained through a step response to acontrol fault in certain controller settings.This response is recorded, and then improved control parameters for a newtest are estimated. This procedure is continued until a satisfactory result isachieved.
At the end the controller is optimized so that its parametersexactly match the characteristics of the evaporator source.It is a long and frustrating process to adjust a controller to an evaporationsource, requiring several minutes for stabilization and hours to obtainsatisfactory results. Often the parameters selected for a certain rate are notsuitable for an altered rate. Thus, a controller should ideally adjust itself, asthe new controllers in INFICON coating measuring units do. At thebeginning of installation and connection the user has the unit measure thecharacteristics of the evaporation source.
Either a PID controller is used asthe basis for slow sources or another type of controller for fast sourceswithout significant dead time.In relevant literature a distinction is made between three different ways ofsetting controllers. Depending on which data are used for the setting, adistinction is made between the closed loop, open loop and resonanceresponse method.Due to the simplicity with which the experimental data can be obtained, wepreferred the open loop method. Moreover, application of this techniquepermits extensive elimination of the trial and error method.The Auto Control Tune function developed by INFICON characterizes aprocess on the basis of its step responses.
After a step-by-step change inthe power the resulting changes in the rate as a function of time aresmoothed and stored. The important step responses are determined, seeFig. 6.8.In general, it is not possible to characterize all processes exactly, so severalapproximations have to be made. Normally one assumes that the dynamiccharacteristic can be reproduced by a process of the first order plus deadtime. The Laplace transformation for this assumption (transfer to the splane) is approximated:−LK p ⋅ 10 sOutput=τ ⋅ s +1Inputwith(6.8)Kp = amplification in stationary stateL = dead timeτ = time constantThese three parameters are determined through the response curve of theprocess.
An attempt has been made by means of several methods tocalculate the required parameters of the system response from curves, asshown in Fig. 6.8. This results in a 1-point accordance at 63.2 % of thetransition (a time constant), an exponential accordance at two points and anexponential accordance weighted according to the method of the smallestsquares. A process is sufficiently characterized by this information so thatthe controller algorithm can be applied. Equation 6.9 shows the Laplacetransformation for the very often used PID controller:M(s) = Kc · 1 +S+ Td · S · E (s)Ti(6.9)withM(s) =Kc =Ti =Td =E(s) =controlled variable or powerControl amplification (the proportional term)integration timedifferentiation timeprocess deviationFig.
6.9 shows the control algorithm and a process with a phase shift of thefirst order and a dead time. The dynamics of the measuring device and thecontrol elements (in our case the evaporator and the power supply) are1.00 K p0.0632 K ppoint ofmaximumrisesetpoint deviationR(s)+ E(s)0Lt (0.632)(Σ)–Time tKc (1 +s+ Td * s)Ti[process]–LK p · eaaasT1s + 1precipitation rateC(s)[controller]T1 = t(0.632) – LKp = (change in output signal)/(change in control signal)Fig. 6.8Process response to a step change with t = 0 (open loop, control signal amplified)Fig.
6.9Block diagram of the PID controller130HomeThin film controllers/control unitsimplicitly contained in the process block. R(s) represents the rate setpoint.The return mechanism is the deviation created between the measuredprecipitation rate C(s) and the rate setpoint R(s).Kc = The key to use of any control system is to select the correct values for Kc,Td and Ti. The Òoptimum controlÓ is a somewhat subjective term that ismade clear by the presence of different mathematical definitions:Ti = Usually the smallest square error ISE (Integral Square Error) is used as ameasure of the quality of the control: LTd = (0.381⋅ T1) ⋅ T1ISE = ∫ e2(t) ⋅ dtThe integral of the absolute value of the deviation IAE (Integral AbsoluteError) was also proposed as a measure for control quality:(6.11)This is more sensitive for small deviations, but less sensitive for largedeviations than ISE.Graham and Lanthrop introduced the integral over time, multiplied by theabsolute error ITAE (Integral Time Absolute Error), as a measure for controlquality:ITAE = ∫ t ⋅ e(t) ⋅ dt 119.
L ⋅ T1 T1(6.13)0.738(6.14)0995.(6.15)(6.10)Here e is the error (the deviation): e = rate setpoint minus measured rate.ISE is relatively insensitive to small deviations, but large deviationscontribute substantially to the value of the integral. The result is smallÒovershootsÓ, but long ripple times because deviations occurring latecontribute little to the integral.IAE = ∫ e(t) ⋅ dt– 0947. 136.
L ⋅ K p T1(6.12)The ITAE is sensitive to initial and, to a certain extent, unavoidabledeviations. Optimum control responses defined through ITAE consequentlyhave short response times and larger ÒovershootsÓ than in the case of theother two criteria. However, ITAE has proven to be very useful forevaluating the regulation of coating processes.INFICONÕs Auto Control Tune is based on measurements of the systemresponse with an open loop.
The characteristic of the system response iscalculated on the basis of a step change in the control signal. It isdetermined experimentally through two kinds of curve accordance at twopoints. This can be done either quickly with a random rate or more preciselywith a rate close to the desired setpoint. Since the process responsedepends on the position of the system (in our case the coating growth rate),it is best measured near the desired work point. The process informationmeasured in this way (process amplification Kp, time constant T1 and deadtime L) are used to generate the most appropriate PID control parameters.The best results in evaluating coating control units are achieved with ITAE.There are overshoots, but the reaction is fast and the ripple time short.Controller setting conditions have been worked up for all integral evaluationcriteria just mentioned so as to minimize the related deviations.
With amanual input as well as with experimental determination of the processresponse coefficients, the ideal PID coefficients for the ITAE evaluation caneasily be calculated from equations 6.13, 6.14 and 6.15:For slow systems the time interval between the forced changes in controlvoltage is extended to avoid ÒhangingÓ the controller (hanging = rapidgrowth of the control signal without the system being able to respond to thealtered signal).
This makes a response to the previous change in thecontroller setting and ÒpowerfulÓ controller settings possible. Anotheradvantage is the greater insensitivity to process noise because the dataused for control do not come from merely one measurement, but fromseveral, so that the mass-integrating nature of the quartz crystal is utilized.In processes with short response times (short time constants) and small tounmeasurable dead times, the PID controller often has difficulties with thenoise of the coating process (beam deflection, rapid thermal short-circuitsbetween melt and evaporator, etc.).
In these cases a control algorithm ofthe integral reset type is used with success. This controller alwaysintegrates the deviation and presses the system towards zero deviation.This technique works well with small or completely imperceptible deadtimes. However, if it is used with a noticeable phase shift or dead time, thecontroller tends to generate oscillations because it overcompensates thecontroller signal before the system has a chance to respond.
Auto ControlTune recognizes the properties of these fast systems during themeasurement of a step response and utilizes the information to calculatethe control amplification for a non-PID control algorithm.6.10 INFICON instrument variantsThe instrument models available differ both in hardware and softwareequipment: the simplest unit, the XTM/2, is purely a measuring or displaydevice that cannot control vacuum coating.The XTC/2 and XTC/C group can control vacuum coating sources and upto three different coatings of a process (not to be confused with ninedifferent coating programs).
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
