In litho machine zapping process, revoluting motor remains the control method that concussion suppresses
Technical field
The invention belongs to the technical field of the high precision quick position control that litho machine double-workpiece-table frame movement controls.
Background technology
Litho machine is superhigh precision control system, needs to realize nano level motion and positioning precision.As the subsystem of litho machine, double-workpiece-table system is responsible for silicon chip and mask sheet to be transported to assigned address, completes and is synchronized with the movement, and its kinematic accuracy will directly affect the effect of photoetching.The key that double-workpiece-table system is different from single workpiece table system is just zapping system, a complete silicon wafer exposure process mainly comprises six steps, in single workpiece table system, these six steps are that single work stage completes in order successively, and therefore the processing time of silicon chip is six step sums; And in double-workpiece-table system, complete this six steps by two work stage simultaneously.In double-workpiece-table litho machine, the time of a process silicon chip comprises time shutter and zapping time, shortens the T.T. that the zapping time can reduce process silicon chip, thus improves photoetching productive rate.
The core of double-workpiece-table zapping system is a revoluting motor, and it is in the centre position of whole device, and two spinning motors are arranged in work stage.Zapping the first two work stage lays respectively at exposure region and pretreating zone, after zapping starts, two work stage are moved horizontally to zapping district, mounting revoluting motor, location swap is realized under revoluting motor drives, following two work stage move to pre-service position and exposure position respectively, achieve exchange function.Whole zapping process accurately goes to assigned address by desired trajectory, converted specific angle, and after can ensureing zapping like this, work stage can be parked on target location more accurately.But in existing litho machine double-workpiece-table zapping process, revoluting motor there will be the problem of residual concussion when quick position.In existing zapping, given to accurately navigating to target location from applying, the movement locus of revoluting motor is 3 rank S curve, and require the track and localization realized completely in whole motion process track, positioning precision will reach nanoscale.6S positioning time of revoluting motor during owing to not adding input shaper technology, tracking accuracy does not reach requirement.
Summary of the invention
The object of this invention is to provide the control method that revoluting motor in a kind of litho machine zapping process remains concussion suppression, is to solve in litho machine double-workpiece-table zapping process, the problem of the residual concussion that revoluting motor occurs when quick position.
Described object is realized by following scheme: in described a kind of litho machine zapping process, revoluting motor remains the control method that concussion suppresses, and its method step is:
Step one: apply Step reference to revoluting motor under presence, amplitude is 10 degree, preserves revoluting motor at this to the operation curve data of fixing by host computer;
Step 2: by the position of reshaper and the location swap of controlled device, carry out shaping to output, the curve data that host computer is preserved, as the Y_ in simulink emulation, carries out off-line shaping;
Step 3: in off-line case, uses the ant group algorithm improved to obtain optimum input shaper parameter; The process of the ant group algorithm of described improvement is as follows: the span x estimating each variable
il≤ x
i≤ x
iu, i=1,2 ... N, total N number of variable, is carried out n decile respectively,
knit a net lattice in the N dimension space region that N number of variable forms, and a net point in space corresponds to a state, and this state is that a N ties up position vector, comprises the information of N number of variable; By human oasis exploited throughout all trellis state, obtain the target function value that each state is corresponding, determine the net point that target function value is minimum, the value of the N number of variable corresponding to this point can be obtained; Then reduce the span of each variable, new span depends on the net point that target function value is minimum, then the lattice that knit a net in new N dimension space region, and so repeatedly, until the spacing of grid is less than precision given in advance, algorithm stops; Described variable is A
1and t
2, span is 0≤A
1≤ 1,0≤t
2≤ 5, by A
1and t
2be set to x respectively
1and x
2, and all carry out 10 deciles, then
X
2l=0, x
2u=5,
corresponding 11 nodes of each variable, make x
1corresponding node is i, i=1,2 ... 11, x
2corresponding each node is j, j=1,2 ... 11, therefore can obtain 121 trellis state, namely obtain 121 position vectors; Choose objective function to be taken as:
wherein e is error, and y is the output after shaping, and max (y) is maximum overshoot, so quick and precisely locate to make revoluting motor reach, the less performance of target function value is more excellent; Successively by each ant throughout all grids, obtain the target function value that each state is corresponding, determine the state that minimum target function value is corresponding; Therefore the Two Variables nodes that just can obtain corresponding to this optimum state is set to i respectively
best, j
best, the span then reducing two variablees is x
1l=x
1l+ (i
best-2) * h
1, x
1u=x
1l+ (i
best+ 2) * h
1; x
2l=x
2l+ (j
best-2) * h
2, x
2u=x
2l+ (j
best+ 2) * h
2; 10 deciles are carried out respectively to the variable after reducing the scope again and can obtain new h
1and h
2, knit a net lattice in new region, until max (h
1, h
2) stopping of < 0.00001 algorithm, then optimum solution is
Step 4: apply S curve to revoluting motor given, target location is 30 degree, the optimum input shaper parameter obtained is utilized to carry out input shaper under presence, what now adopt is traditional reshaper control technology, and directly shaping is carried out to input, open loop feedforward control is carried out to revoluting motor.
The present invention can carry out input shaper to the S curve of revoluting motor is given, and revoluting motor can be made to realize quick position, and positioning time reduces to 2S by 6S.Due in whole zapping process, revoluting motor needs to drive double-workpiece-table to realize position and exchange function, so after the locating speed raising of revoluting motor, the time of whole zapping just decreases, and improves the productive rate of litho machine.
Accompanying drawing explanation
Fig. 1 is the simulink off-line simulation model schematic related in the present invention;
Fig. 2 is under off-line case, response results comparison diagram before and after the shaping of Step reference revoluting motor;
Fig. 3 is in online situation, does not add 3 rank S curve of input shaper given and do not add 3 rank S curve response results figure of reshaper;
Fig. 4 is in online situation, adds the given 3 rank S curve response results figure with adding after reshaper of 3 rank S curve of input shaper.
Embodiment
Embodiment one: shown in composition graphs 1, Fig. 2, Fig. 3, Fig. 4, in Fig. 1, Y_: the input of reshaper; A1: the amplitude of input shaper first pulse; A2: the amplitude of input shaper second pulse; T2: the Slack time of input shaper second pulse; Y: export after shaping; Scope: oscillograph; Its method step is:
Step one: apply Step reference to revoluting motor under presence, amplitude is 10 degree, preserves revoluting motor at this to the operation curve data of fixing by host computer;
Step 2: by the position of reshaper and the location swap of controlled device, carry out shaping to output, the curve data that host computer is preserved, as the Y_ in simulink emulation, carries out off-line shaping;
Step 3: in off-line case, uses the ant group algorithm improved to obtain optimum input shaper parameter; The process of the ant group algorithm of described improvement is as follows: the span x estimating each variable
il≤ x
i≤ x
iu, i=1,2 ... N, total N number of variable, is carried out n decile respectively,
knit a net lattice in the N dimension space region that N number of variable forms, and a net point in space corresponds to a state, and this state is that a N ties up position vector, comprises the information of N number of variable; By human oasis exploited throughout all trellis state, obtain the target function value that each state is corresponding, determine the net point that target function value is minimum, the value of the N number of variable corresponding to this point can be obtained; Then reduce the span of each variable, new span depends on the net point that target function value is minimum, then the lattice that knit a net in new N dimension space region, and so repeatedly, until the spacing of grid is less than precision given in advance, algorithm stops; Described variable is A
1and t
2, span is 0≤A
1≤ 1,0≤t
2≤ 5, by A
1and t
2be set to x respectively
1and x
2, and all carry out 10 deciles, then
X
2l=0, x
2u=5,
corresponding 11 nodes of each variable, make x
1corresponding node is i, i=1,2 ... 11, x
2corresponding each node is j, j=1,2 ... 11, therefore can obtain 121 trellis state, namely obtain 121 position vectors; Choose objective function to be taken as:
wherein e is error, and y is the output after shaping, and max (y) is maximum overshoot, so quick and precisely locate to make revoluting motor reach, the less performance of target function value is more excellent; Successively by each ant throughout all grids, obtain the target function value that each state is corresponding, determine the state that minimum target function value is corresponding; Therefore the Two Variables nodes that just can obtain corresponding to this optimum state is set to i respectively
best, j
best, the span then reducing two variablees is x
1l=x
1l+ (i
best-2) * h
1, x
1u=x
1l+ (i
best+ 2) * h
1; x
2l=x
2l+ (j
best-2) * h
2, x
2u=x
2l+ (j
best+ 2) * h
2; 10 deciles are carried out respectively to the variable after reducing the scope again and can obtain new h
1and h
2, knit a net lattice in new region, until max (h
1, h
2) stopping of < 0.00001 algorithm, then optimum solution is
Step 4: apply S curve to revoluting motor given, target location is 30 degree, the optimum input shaper parameter obtained is utilized to carry out input shaper under presence, what now adopt is traditional reshaper control technology, and directly shaping is carried out to input, open loop feedforward control is carried out to revoluting motor.
Principle: in actual zapping, given to accurately navigating to target location from applying, the movement locus of revoluting motor is 3 rank S curve, and require the track and localization realized completely in whole motion process track, positioning precision will reach nanoscale.6S positioning time of revoluting motor during owing to not adding input shaper technology, tracking accuracy does not reach requirement.In order to improve positioning precision, realize fast channel switching, scheme of the present invention adopts input shaper open loop feedforward control technology, only needs to carry out shaping to the given 3 rank S curve signals of revoluting motor, just can eliminate the residual concussion of motor, realize quick position.
Traditional control system with input shaper is all feedforward control, directly carries out shaping to input, and then the input after shaping is applied to controlled device.And in actual applications owing to needing to use optimized algorithm to carry out optimization to the parameter of input shaper, just need constantly to change parameter, easily evoke the concussion of system in this process, so change the position of input shaper in this invention.The position of input shaper and controlled device is exchanged, namely shaping is carried out to output, emulate under realizing off-line case, so just conveniently determine optimized parameter.After parameter is determined, still by traditional input shaper technology control system.
Because the condition of work of General System under step function input action is severeer, if its performance index under step function effect can meet the demands, think that the performance under other forms also can meet the demands, therefore the present invention is when determining input shaper parameter, be captured in the system responses data under step function input action, shaping is carried out to Stepped Impedance Resonators, determine optimized parameter, again this optimized parameter is substituted in 3 rank S curve input shapers, feedforward control is carried out to revoluting motor, just can meet the needs of quick position.Emulate under realizing off-line case, need to gather the operation curve data of motor before not adding shaping, the data of adopting, as the input of off-line simulation, namely carry out shaping to these data.First apply Stepped Impedance Resonators to revoluting motor, be given as 10 degree, the actual location data of motor in operational process can be obtained by host computer, preserve this data file.
As a kind of feedforward control, input shaper carries out design according to system performance and solves, and obtains the pulse train of different amplitude and Slack time, then input signal and not reshaper spike train is carried out convolution, and the control signal produced after shaping carrys out control system.In the present invention is the simplest input shaper, is namely made up of two pulses
in order to make training time the shortest, make t
1=0, then
in order to ensure that system can reach original output point, the equation of constraint of gain also should be increased: A
1+ A
2=1, A
i> 0 (i=1,2).Can be found out by the frequency-domain expression of input shaper, the place of the most critical of design input shaper is exactly how to determine the amplitude of each pulse and the Slack time corresponding to it according to the requirement of system.