Regulation and Dynamics
Regulation and Dynamics
Frédérick BORDRY
Frédérick BORDRY
3 Regulation
on
Tracking
si
Ses
Ts y(t)
yref(k) Power
T 1/S DAC
converter
Digital k.Ts
÷k
Anti
controller Digital
R Filter
ADC aliasing
filter
Oversampling
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Glossary
– Accuracy
– Accuracy
Long term setting or measuring uncertainty taking into consideration
Long term setting or measuring uncertainty taking into consideration
the full range of permissible changes* of operating and environmental
the full range of permissible changes* of operating and environmental
conditions.
conditions.
Precision
**requires definition
requires definition
– Reproducibility
– Reproducibility
Uncertainty in returning to aaset of previous working values from
Uncertainty in returning to set of previous working values from
cycle to cycle of the machine.
cycle to cycle of the machine.
– Stability
– Stability
Maximum deviation over aaperiod with no changes in operating
Maximum deviation over period with no changes in operating
conditions.
conditions.
Accuracy, reproducibility and stability are defined for aagiven period
Accuracy, reproducibility and stability are defined for given period
Precision is qualitative . Accuracy is quantitative.
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Power Converter Tolerances for LHC
Power Converter Tolerances for LHC
Circuit Nominal Current One Year One day 1/2 hour Resolution
Type Current Polarity Accuracy Reproducibility Stability
(A) (ppm of Inominal) (ppm of Inominal) (ppm of Inominal) (ppm of Inominal)
± 50
Main Bends, Main Quads Unipolar ± 10 ±3 1
13000 ± 20 with calibration
8000/ ± 100
Inner triplet Unipolar ± 20 ± 10 15
6000 ± 20 with calibration?
Dispersion suppressor Unipolar ± 70 ± 10 ±5 15
6000
Insertion quadrupoles Unipolar ± 70 ± 10 ±5 15
6000
Separators (D1,D2,D3,D4) Unipolar ± 70 ± 10 ±5 15
6000
Trim quadrupoles Bipolar ± 200 ± 50 ± 10 30
600
SSS correctors Bipolar ± 200 ± 50 ± 10 30
600
Spool pieces Bipolar ± 200 ± 50 ± 10 30
600
Orbit correctors Bipolar ± 1000 ± 100 ± 50 30
120/60
Precision Control
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Power converter : voltage ripple specification
Power converter 50 HzRipple 300 HzRipple
type pk-pk pk-pk
(mV) (mV)
[13kA, ±180V] 20 130
[13kA, 18V] 50 350 Load identification
[8kA,8V] 5 30
[6kA,8V] 5 30
[4kA,8V] 5 30
[±600A,±10V] 5 20
[±600A,±40V] 5 20
[±120A,±10V] 5 20
[±60A,±8V] 200 1200 20 ; 120
[900A, 550V] 40 250
[900A,1000V] 80 450
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Voltage loop
Current loop
I
Iref B
Vref
+ εΙ V
Reg. G(s)
-
F(s) εV
Imeasured
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Closed Loop
Closed Loop
Disturbances
Perturbations
Feed P
Yref + ε Y
Reg. System
-
F(s) H(s)
Transducer
G(s)
Open loop : system H(s) ; system with controller F(s).H(s)
Closed loop : HCL(s) = F(s).H(s) / [ 1 + F(s). H(s) G(s) ]
Steady state error : F(s) must contains the internal model of the reference (the
transfer function that generates Yref(t) from the Dirac impulse ; e.g. step =
(1/s) * Dirac ; ramp = (1/s2) * Dirac,...
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Choice of the desired performances
Choice of the desired performances
The choice of desired performances in terms of the response time
(bandwidth) is linked to the dynamics of the open-loop system and to
the power availability of the actuator during the transient :
acceleration of the natural response requires control peaks that are
greater than the steady-state values
U Y
Umax/Ustat ≅ desired speed/natural speed
≅ desired bandwidth / natural bandwidth
≅ fCLB / fOLB
Or, the actuators will have to be chosen as a function of the
desired performances and the open-loop response of the system
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Choice of the desired performances (cont’d)
Choice of the desired performances (cont’d)
The robustness of the closed-loop system
is linked to the ratio fCLB / fOLB
(sensitivity to the uncertainty or variation of the
model parameters)
If the ratio is too high => bad robustness and
If the ratio is too high => bad robustness and
need to have a good knowledge of the process
need to have a good knowledge of the process
model (high frequency identification ;; e.g.
model (high frequency identification e.g.
LHC string dipole)
LHC string dipole)
or use of adaptive control
or use of adaptive control
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Main LHC quadrupole
Main LHC quadrupole [13000A,18V]
Ω
Magnet chain ::L=0.27 H ;;R = 0.8 mΩ => ustat = R.13000 ≅ 10.4 V
Ω
Magnet chain L=0.27 H R = 0.8 mΩ => ustat = R.13000 ≅ 10.4 V
T = L/R = 340 ss => ffOLB ≅ 0.47 mHz
T = L/R = 340 => B ≅ 0.47 mHz
OL
Large signal ::umax/ustat ≅ 1.7 => ffCL ≅ 0.8 mHz => tt = 430 ss
Large signal umax/ustat ≅ 1.7 =>
CL
B ≅ 0.8 mHz => R = 430
B R
Small signal ::ffCL = 1 Hz => ffCL //ffOL ≅ 2200 !!!!…
Small signal
CL
BB = 1 Hz =>
CL
BB
OL
BB ≅ 2200 !!!!…
Then ustat = 7.6 //2’200 ≅ 3.5 mV
Then ustat = 7.6 2’200 ≅ 3.5 mV
Ω
=> ∆II= 3.5 mV/0.8 mΩ = 4.4 A = 0.035 % Inom = 350 ppm Inom
Ω
=> ∆ = 3.5 mV/0.8 mΩ = 4.4 A = 0.035 % Inom = 350 ppm Inom
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
LHC orbit corrector
LHC orbit corrector [±60A,±8V] (arc corrector)
Ω
Magnet : L=7 H ; R = 30 mΩ (60m of 35 mm2)
T = L/R = 300 s => fOLB ≅ 0.5 mHz
Ustat = R.I = 1.8V
Large signal : umax/ustat ≅ 4 => fCLB ≅ 2 mHz => tR = 175 s
(dI/dtmax ≅ 1A/s)
Small signal : fCLB ≅ 1 Hz => umax/ustat ≅ 1/ 0.5 10-3 = 2000 !
Ω
Then ustat = 6 / 2000 = 3mV => ∆I = 3mV/30 mΩ = 0.1 A = 0.15 % Imax
Is it enough ? fCLB ≅ 0.1 Hz ?
LHC Project Note 183 : “The power converters involved in analogue feedback
of the local orbit may need to deal with correction rates between 10 and 500 Hz”;
fs (rate ?) = 500 Hz => fCLB ≅ 50Hz (∆I = 1% : Umax = 2400 V ?????...)
(Umax= 8V => ∆I = 30 ppm Imax)
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Closed-loop bandwidth : small signal amplitude
Chromaticity sextupoles SF1 and SF2
SF1, SD1, SF2, SD2 64 circuits
SF1 : beam 1 and beam 2 ; one per arc => 16 circuits
SD1 : beam 1 and beam 2 ; one per arc => 16 circuits
SF2 : beam 1 and beam 2 ; one per arc => 16 circuits
Example : SD2 : beam 1 and beam 2 ; one per arc => 16 circuits
Each circuit: 12 magnets in series for SD families and 10 magnets for SF families
Converter ± 600 Imax (A) ± 10 Umax (Volt)
Lcircuit 0.36 H SF1 and SF2 families
Rcircuit 0.0083 Ohm Ucollision 4.98 V
Time_constant 43.37 s
f open_loop_B 0.0037 Hz 3.7 mHz
Large signal
dI/dtmax 14 A/s
Umax/Ucollision 2.0 = f closed_loop_B/ f open_loop_B
f closed_loop_B 0.007 Hz 7.4 mHz tr 48 s
Small signal
f closed_loop_B 1 Hz
Uavailable/Ucollision 273
Uavailable 5.02 V (= Ummax - Ucollision)
Ustat 0.018 V 18 mV
∆I 2.2 A 0.4 % of Imax
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Voltage and current loop bandwidth
Voltage and current loop bandwidth
Cold Circuits Current Converter Inductance Resistance TC max. Ramp U max. Voltage Current
0 to 100% loop loop
(kA) (mH) (mohms) (sec) (sec) (V) (Hz) (kHz)
Main Bend 13.00 [13kA,±180V] 16632 0.8 20790 1300 176.7 500 0.1 to 1.0
Main Quads 13.00 [13kA,18V] 263 to 285 0.8 to 0.97 356 1300 15.2 500 0.1 to 1.0
Inner Triplet 7.00 [8kA,8V] 220 0.33 667 1300 3.5 500 0.1 to 1.0
" " Trim 4.40 [6kA,8V] 38 0.45 84 1300 2.1 500 0.1 to 1.0
Disper. Quads 5.82 [6kA,8V] 21 to 26 0.49 to 0.95 53 360 5.9 500 0.1 to 1.0
Insert. Quads 4.65 to 5.82 [6kA,8V] 15 to 30 0.49 to 0.9 61 360 5.7 500 0.1 to 1.0
Insert. Quads 3.90 [4kA,8V] 74 to 148 0.67 to 0.9 221 360 5.1 500 0.1 to 1.0
Separators 6.00 [6kA,8V] 28 to 50 0.52 to 0.68 96 360 4.9 500 0.1 to 1.0
Q6,pts 3 & 7 0.60 [±600A,±10V] 600 1.8 to 7.02 333 360 5.2 1000 0.1 to 1.0
Q6,pts 3 & 7 0.60 [±600A,±10V] 600 23 26 360 14.8 1000 1 to 2
Spool b3,b5 0.60 [±600A,±10V] 31 to 123 8.28 15 120 5.6 1000 1 to 2
Trim Quads 0.60 [±600A,±10V] 31 to 248 1.8 to 8.28 138 120 6.2 1000 0.1 to 1.0
Trim Quads 0.60 [±600A,±10V] 31 to 248 25 10 120 16.2 1000 1 to 2
SSS Correctors 0.60 [±600A,±10V] 72 to 432 2.34 to 8.28 185 120 7.1 1000 0.1 to 1.0
SSS Correctors 0.60 [±600A,±10V] 72 to 144 26 to 47 5.5 120 28.9 1000 1 to 2
Octupoles 0.60 [±600A,±10V] 13.5 to 18 1.8 to 8.28 10.0 120 5.1 1000 1 to 2
Octupoles 0.60 [±600A,±40V] 13.5 to 18 49 0.4 120 29.5 1000 1 to 2
Spool b4 0.12 [±120A,±10V] 31 35 0.9 120 4.2 1000 1 to 2
Orbit Correctors 0.12 [±120A,±10V] 2240 to 4100 11 to 49 373 120 8.0 1000 0.1 to 1.0
Low B Corr. 0.12 [±120A,±10V] 3.7 to 21 9.4 to 36 2 120 4.341 1000 1 to 2
Orbit Correctors 0.06 [±60A,±8V] 7000 31 226 120 5.4 2000 0.1 to 1.0
Warm Circuits Current Converter Inductance Resistance TC max. Ramp U max. Voltage Current
loop loop
(kA) (mH) (mohms) (sec) (sec) (V) (Hz) (kHz)
Quads 0.81 [900A,550V] 640 410 to 534 1.6 120 436.9 70 5 to 10
Separators 0.81 [1000A,1000V] 1200 1800 1000 1.8 120 822.2 70 5 to 10
Dump Septum 0.88 [900A,550V] 1060 569 1.9 120 508.5 70 5 to 10
Warm Trim/Corr 0.60 [±600A,±40V] 25 64 26 59 2.5 120 35.7 1000 5 to 10
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
R.S.T control :: generic control
R.S.T control generic control
yref(k) H(z-1)
T 1/S =z -d B(z -1)/A(z -1)
Tracking Regulation
Plant
R
Tracking and Regulation with independent objectives
Tracking and Regulation with independent objectives
This control strategy permits a R.S.T digital controller to be designed for
both stable and unstable systems:
- without restriction on the degrees of the polynomials A(z-1) and B(z -1) of
the plant transfer function
- without restriction on the sampled time delay of the system (d)
This strategy can only be applied to discrete-time models with stable zeroes
(minimum phase systems):
- high sampling period => unstable zeroes
- fractional time delay > 0.5 x Ts
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Summary
Summary
Ts
yref(k) y(t)
T 1/S DAC System
k.Ts
Digital Anti
÷k
Digital
controller R Filter
ADC aliasing
Frequency Oversampling filter
Divider
Based on fOLB and power of the actuator : choice of the closed-
loop performances [fCLB (tr ) and Q (M) ]
Robustness ; fCLB / fOLB (Internal saturation : controlability)
fs (sampling period) : choice based on the fCLB
fs = 1/Ts = (6 to 25) * fCLB
Discrete model H(z-1) at Ts
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Ramp from to 500A
500 A
[±600A,±10V]
[±600A,±10V]
Superconducting magnet
Superconducting magnet
L= 1 mH
L= 1 mH
Rcircuit = 3.3 mΩ
Rcircuit = 3.3 mΩ
ττ= L/R = ~ 300 ms
= L/R = ~ 300 ms
ADC 16 bits (15 ppm)
ADC 16 bits (15 ppm)
FCLB = 4 Hz
±5mA
500 A (±8 ppm)
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
4 4
x 10 x 10
1.3 1.3
25 ppm
1.28 1.3
1.26 1.3
1A
1.2999
1.24
1.2999
1.22
1.2999
1.2
1000 2000 3000 4000 5000 6000 7000 8000 9000 8550 8600 8650 8700 8750 8800 8850 8900 8950 9000
4
x 10
Ramp from 200 A to 13000A 1.3
1.3
10 ppm 1.3
[13kA,18V] converter
[13kA,18V] converter 1.3
1mH inductance ;;
1mH inductance 1.3
0.8 mΩ resistance
0.8 mΩ resistance 1.3
(τ = 1.5 s)
(τ = 1.5 s) 1.3
dI/dt = 200 A/s
dI/dt = 200 A/s 8700 8750 8800 8850 8900 8950 9000
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Error =
Imeas - Iref
30 ppm
4
x 1 0
1 . 3
1 . 3
Iref 1 . 3
10 ppm
1 . 3
ADC 1 . 3
1 . 3
1 . 3
8 7 0 0 8 7 5 0 8 8 0 0 8 8 5 0 8 9 0 0 8 9 5 0 9 0 0 0
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Start of the ramp (from 200 A)
225
dI/dt = 50 A/s
220
215
Iref
210
205
ADC
No lagging error !
200
1200 1400 1600 1800 2000 2200 2400
212.5
215
212.4
214 212.3
212.2
213
212.1
212
212
211.9
211.8
211
211.7
210 211.6
211.5
1950 2000 2050 2100 2150 2200 2060 2070 2080 2090 2100 2110
10 ms
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
I1
Vcv1
Vcv2
8kA 6kA
I2
1 2 2 3
KEK Fermilab Fermilab KEK
IF = I1 + I2 IK = I1
r1 r2 1 1
− − −
d i1 L1 L1 i1 L1 L1 .Vcv1
i 2 = r1
dt 1
+
1 i 2 1 1
.
1 Vc 2
− r 2. + − +
L1 L1 L 2 L1 L1 L 2
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Decoupling Principle ::
Decoupling Principle
W U dX X
dX D = A. X + B.U
= A. X + B.U dt
dt
dX K
= ( A + B.K ).X + B.D.W
dt
Ad Bd Choice :
Ad and Bd : diagonal matrices − r1 0
L1 + L2
Ad =
Decoupling matrices : r2
0 −
K = B-1 . (Ad -A) L2
D = B -1 . B 1
L1 + L 2 0
Digital : Bd =
X*(k+1) = F*.X*(k) + H*.U*(k) 1
0
with F*= e A. ts ; H* = A-1. (e A. ts - I).B L2
ts : sampling period
X*(k+1) = Fd*.X*(k) + Hd*.W*(k)
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Dcct 1
DSP 1 ADC
Vconv1
Converter 1
Ikref = I1ref 8kA
Reg. 1 Voltage source
DAC
Vref1
Analog
Digital decouplind card
Vconv1 = V1ref + k1v. V2ref +
K1i. I2
Vconv2 = V2ref + k2v. V1ref +
K2i. i1
DSP 2
I2ref Vref2 Converter 2
-
IFref +
Reg. 2 DAC
4kA
Vconv2
Voltage source
ADC
Dcct 2
Digital
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
MQM connection ::3 busbars and 3 current leads
MQM connection 3 busbars and 3 current leads
r1
L1
Vcv1
di1
Vcv1 = r1.i1 + L1. + rc.(i1 − i2)
rc dt
di2
Vcv 2 = r 2.i 2 + L 2. + rc.(i1 − i 2)
dt
Vcv2 L2
r2 With r1 = r2 = rc = r
− r1 + rc rc 1 0 Vcv1
d i1 L1 L1 . i1 + L1
i 2 = .
dt rc r 2 − rc i2 1 Vc 2
− − 0
L2 L2 L2
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
MQM connection ::3 busbars and 3 current leads
MQM connection 3 busbars and 3 current leads MQM
Dcct 1
DSP 1 ADC
Vconv1
Converter 1
I1ref 6kA
Reg. 1 Voltage source
DAC
Vref1
Analog
Digital decouplind card
Vconv1 = V1ref + k1v. V2ref +
K1i. I2
Vconv2 = V2ref + k2v. V1ref +
K2i. i1
DSP 2
I2ref Vref2 Converter 2
Reg. 2 DAC
6kA
Vconv2
Voltage source
ADC
Dcct 2
Digital
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Power converter control and high precision issues
Power converter control and high precision issues
3.1 Regulation and dynamics - F. Bordry
3.2 Hardware implementation - J. Pett
3.3 Software implementation - Q. King
3.4 DCCT and Standards Lab infrastructure - F. Power
Tracking Regulation
3.5 On site calibration system - G. Fernqvist
ns
ssio 6
Se & Output current
5 Power
Load
Power input processor
Control Current
signals Measurement
Controller Gateway
I(t)
time
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO
Powering Workshop - 28th November 2000 Fk. Bordry SL/PO