Upload
ahmed58seribegawan
View
217
Download
0
Embed Size (px)
Citation preview
8/12/2019 KuenziFilter.pdf
1/38
Wir schaffen Wissen heute fr morgen
Paul Scherrer Institut
10.05.2014
Ren Knzi
Filter Design - Passive Power Filters
CERN ccelerator School !"#$% &a'en% Swit(erlan'
8/12/2019 KuenziFilter.pdf
2/38
Suitable Filter Structures
R. Knzi Power Filter Design CAS 2014 10.05.2014 2
8/12/2019 KuenziFilter.pdf
3/38
Suitable Filter Structures
R. Knzi Power Filter Design CAS 2014 10.05.2014 3
8/12/2019 KuenziFilter.pdf
4/38
H-Bridge
R. Knzi Power Filter Design CAS 2014 10.05.2014 4
8/12/2019 KuenziFilter.pdf
5/38
H-Bridge with CM filter
R. Knzi Power Filter Design CAS 2014 10.05.2014 5
8/12/2019 KuenziFilter.pdf
6/38
2nd order filter - Transferfunction
R. Knzi Power Filter Design CAS 2014 10.05.2014 6
1
1
1
1
R
C
s 1
LCRC s L C C s
RCs 1
1
1
with
3rd order PT
(1)
1st order PD
8/12/2019 KuenziFilter.pdf
7/38
2nd order filter - Transferfunction
R. Knzi Power Filter Design CAS 2014 10.05.2014 7
1
1
1
3rd order PT in itsnormalized form:
(2)
8/12/2019 KuenziFilter.pdf
8/38
2nd order filter - Transferfunction
R. Knzi Power Filter Design CAS 2014 10.05.2014 8
By expanding (2) and comparing the coefficients with (1) we get:
1
1
(3a)
(3b)
(3c)
The 3 independent equations (3a.3c) contain 5 unknowns (L1, C1, RD, CD and0). Therefore we have the choice to select 2 of them and the remaining 3 depend
on that selection.
8/12/2019 KuenziFilter.pdf
9/38
2nd order filter - Selection of 0
R. Knzi Power Filter Design CAS 2014 10.05.2014 9
For a given frequency
B well in the blocking area (
B >>
0) we can define thedesired attenuation GB. In the blocking area the highest order terms of both the nu-merator and denominator in equation (1) dominate, therefore (1) can be simplified to:
! 1
(4)
8/12/2019 KuenziFilter.pdf
10/38
2nd order filter - Definition of L1
R. Knzi Power Filter Design CAS 2014 10.05.2014 10
For cost reasons L1 shouldbe as small as possible,but a too small inductancewill result in an excessive
ripple current!
The DC-voltage across C1 is m*VDC. When the IGBT is on, the current in L1 increases
and the peak-peak ripple current
IL1 can be calculated:
"# $%#
$& "' "' "' 1 (
)*# ( + $%#
$& ( 1
,- "
' 1 (
"' 1 ( (
,-
"' ./0
,- )*#
Maximum 0.25for m = 0.5
(5)
8/12/2019 KuenziFilter.pdf
11/38
2nd order filter - Definition of L1
R. Knzi Power Filter Design CAS 2014 10.05.2014 11
Alternative approach to determine L1:
*#2345567255
2345567255
0 8 ,345567
2345567255
0 8 ,345567 *#2345567255(6)
8/12/2019 KuenziFilter.pdf
12/38
2nd order filter Calcualting filter elements
R. Knzi Power Filter Design CAS 2014 10.05.2014 12
Substitute (3a) in (3c) and we receive:
Selection: L1 and 0 Selection: C1 and 0 Selection: L1 and C1
Solve (3b) for CD:
Solve (3a) for RD:
(7a) (7b) (7c)
(8)
(9)
8/12/2019 KuenziFilter.pdf
13/38
2nd order filter Example 1
R. Knzi Power Filter Design CAS 2014 10.05.2014 13
DC-link voltage: 200V DC-link current: 500A IL1 50App. C1 must be 22mF (because of high ripple current)
Design a 2nd order filter for all three given optimization methods and compare the results.
8/12/2019 KuenziFilter.pdf
14/38
2nd order filter Example 1
R. Knzi Power Filter Design CAS 2014 10.05.2014 14
Select L1 to meet the ripplecurrent requirement:
Select C1:
Calculate the remaining
filter elements
(6)
(7c, 8, 9)
2345567255
0 8 ,345567 *#2345567255
0.. ./19 "::
0 8 9..; .
00(?
8/12/2019 KuenziFilter.pdf
15/38
2nd order filter Example 1
R. Knzi Power Filter Design CAS 2014 10.05.2014 15
x3 x5 x8
Results:
8/12/2019 KuenziFilter.pdf
16/38
2nd order filter Example 1
R. Knzi Power Filter Design CAS 2014 10.05.2014 16
Maximum amplitude of resonance:Butterworth 4.5 dBBessel 3.1 dBCritical damping 2.3 dB
Frequency, for -3 dB attenuation:Butterworth 74 Hz
Bessel 67 HzCritical damping 59 Hz
Attenuation: -28dB @ 300Hz
8/12/2019 KuenziFilter.pdf
17/38
2nd order filter Example 2
R. Knzi Power Filter Design CAS 2014 10.05.2014 17
DC-link voltage: 120V fs = 20kHz IOut_max = 500A IL1 50App.
Attentuation: GB = 250 @
B = 2*
*20kHz
Same premises asfor example 3
Design a 2nd order filter for all three given optimization methods and compare the results.
8/12/2019 KuenziFilter.pdf
18/38
2nd order filter Example 2
R. Knzi Power Filter Design CAS 2014 10.05.2014 18
"' ./0
,- )*#
10." ./0
0.>@ .< 9.=>
Select L1 to meet the ripplecurrent requirement:
Select 0 to meet theattenuation requirement:
Calculate the remaining
filter elements
(5)
(7a, 8, 9)
(4)
d
8/12/2019 KuenziFilter.pdf
19/38
2nd order filter Example 2
R. Knzi Power Filter Design CAS 2014 10.05.2014 19
x3 x5 x8
Results:
2 d d fil E l 2
8/12/2019 KuenziFilter.pdf
20/38
2nd order filter Example 2
R. Knzi Power Filter Design CAS 2014 10.05.2014 20
Maximum amplitude of resonance:Butterworth 4.5 dBBessel 3.1 dB
Critical damping 2.3 dB
Frequency, for -3 dB attenuation:Butterworth 1.5 kHz
Bessel 1.4 kHzCritical damping 1.2 kHz
Attenuation: -48dB @ 20kHz
4th d filt T f f ti
8/12/2019 KuenziFilter.pdf
21/38
4th order filter - Transferfunction
R. Knzi Power Filter Design CAS 2014 10.05.2014 21
1
1
1
1
1
1
1
1
4th d filt T f f ti
8/12/2019 KuenziFilter.pdf
22/38
4th order filter - Transferfunction
R. Knzi Power Filter Design CAS 2014 10.05.2014 22
1
A B
1
1
AA B
B
1
B
A
with
1st order PD
5th order PT
(10)
4th d filt T f f ti
8/12/2019 KuenziFilter.pdf
23/38
4th order filter - Transferfunction
R. Knzi Power Filter Design CAS 2014 10.05.2014 23
5th order PT in its normalized form:
(11) 1
1 1
1
Optimisation methods:
4th order filter Transferfunction
8/12/2019 KuenziFilter.pdf
24/38
4th order filter - Transferfunction
R. Knzi Power Filter Design CAS 2014 10.05.2014 24
1
A A
B
B
1
By expanding (11) and comparing the coefficients with (10) we get:
B
B
A
A
(12a)
(12b)
(12c)
(12d)
(12e)
4th order filter selection of and L
8/12/2019 KuenziFilter.pdf
25/38
4th order filter selection of 0 and L1
R. Knzi Power Filter Design CAS 2014 10.05.2014 25
The 5 independent equations (12a.12e) contain 7 unknowns (L1, C1, L2, C2, RD, CDand 0). Therefore we have the choice to select 2 of them (0 and L1) the remaining 5depend on that selection.
For a given frequency B well in the blocking area (B >> 0) we can define thedesired attenuation GB. In the blocking area the highest order terms of both the
numerator and denominator in equation (10) dominate, therefore (10) can besimplified to:
A
A A
B
B
B
B
D(13)
! B 1
Select L1 according to ripple current requirements with (5) or (6)
4th order filter Calculating filter elements
8/12/2019 KuenziFilter.pdf
26/38
4th order filter Calculating filter elements
R. Knzi Power Filter Design CAS 2014 10.05.2014 26
By solving the equation system (12a12e) we get:
B
B
A
A
B A
BA 1
A
BA
A
A
B A
with
(14a)
(14b)
(14c)
(14d)
(14e)
4th order filter Example 3
8/12/2019 KuenziFilter.pdf
27/38
4th order filter Example 3
R. Knzi Power Filter Design CAS 2014 10.05.2014 27
Design a 4th order filter for all three given optimization methods and compare the results.
DC-link voltage: 120V
fs = 20kHz IOut_max = 500A IL1 50App. Attentuation: GB = 250 @ B = 2**20kHz
Same premises asfor example 2
4th order filter Example 3
8/12/2019 KuenziFilter.pdf
28/38
4th order filter Example 3
R. Knzi Power Filter Design CAS 2014 10.05.2014 28
"' ./0
,- )*#
10." ./0
0.>@ .< 9.=>
Select L1 to meet the ripplecurrent requirement:
Select 0 to meet theattenuation requirement:
D
Calculate the remaining
filter elements
(5)
(14a.e)
(13)
4th order filter Example 3
8/12/2019 KuenziFilter.pdf
29/38
4 order filter Example 3
R. Knzi Power Filter Design CAS 2014 10.05.2014 29
Results:
4th order filter Example 3
8/12/2019 KuenziFilter.pdf
30/38
4 order filter Example 3
R. Knzi Power Filter Design CAS 2014 10.05.2014 30
Maximum amplitude of resonance:Butterworth 8.6 dBBessel 5.4 dB
Critical damping 3.8 dB
Frequency, for -3 dB attenuation:Butterworth 5.5 kHz
Bessel 5.0 kHzCritical damping 3.9 kHz
Attenuation: -48dB @ 20kHz
Comparison of different filter designs
8/12/2019 KuenziFilter.pdf
31/38
Comparison of different filter designs
R. Knzi Power Filter Design CAS 2014 10.05.2014 31
+64%
+42%
Comparison of different filter designs
8/12/2019 KuenziFilter.pdf
32/38
Comparison of different filter designs
R. Knzi Power Filter Design CAS 2014 10.05.2014 32
Practical Aspects - Wiring
8/12/2019 KuenziFilter.pdf
33/38
Practical Aspects Wiring
Effect of a 0.5m longwire 16mm2 (wiring of C1and C2)
Skin Effect Skin depth in Cu @
20kHz: 0.5mm
Reduces the effective
cross section to 6.3 mm2
Wire resistance @20kHz: 1.4m
Wire inductance is
approx. 0.5H
R. Knzi Power Filter Design CAS 2014 10.05.2014 33
To be avoided !
Practical Aspects low inductive setup
8/12/2019 KuenziFilter.pdf
34/38
Practical Aspects low inductive setup
R. Knzi Power Filter Design CAS 2014 10.05.2014 34
C1
L1
L2
C2
RD
CD
Life time of electrolytic capacitors
8/12/2019 KuenziFilter.pdf
35/38
y p
R. Knzi Power Filter Design CAS 2014 10.05.2014 35
The useful life time of an electrolytic capacitor depends verymuch on the ripple current and the ambient temperature.
Nominal ripple current at nominal frequency (100Hz) and nominal capacitor temperature (85C).
17.4A in our example
8/12/2019 KuenziFilter.pdf
36/38
Life time of electrolytic capacitors
8/12/2019 KuenziFilter.pdf
37/38
y p
R. Knzi Power Filter Design CAS 2014 10.05.2014 37
Determine the allowed ripple
current for a desired usefullife and Ta:
For 25000h (3 years)
and Ta = 50C,2.6 * 23.5A = 61A
For 250000h (30years)and Ta = 50C
0.85 * 23.5A = 20A
The useful life time dramatically decreases at higher ambient temperatures!
8/12/2019 KuenziFilter.pdf
38/38
References
U. Tietze, Ch. Schenk; Halbleiter-Schaltungs-Technik, 12. Auflage, Pages 815ff
Epcos, Datasheet, Capacitors with screwterminals Type B43564, B43584, November2012
R. Knzi Power Filter Design CAS 2014 10.05.2014 38