Sie sind auf Seite 1von 6

A Quality Factor QN for a Digital Band-Pass and Band-Stop Filter

Phuoc Si Nguyen
Victoria University of Technology, Electrical and Electronics Department
Melbourne, Victoria, Australia
nguyensiphuoc@outlook.com

Abstract—As defined for an analog band-pass or band-stop the quality factor Q with f0 and bw in the designed digital
filter, a quality factor Q is a ratio between the center frequency filter will give a different value as required of the quality
f0 and the −3dB bandwidth, which is not valid for designing a factor [7].
digital band-pass and band-stop filter from an analog filter by This paper introduces a formula to calculate the notch
using bilinear z-transformation because the center frequency f0 frequency fN from the upper frequency fU and the lower
is not a notch frequency. This paper introduces a new quality frequency fL; this formula is derived from bilinear z-
factor QN that is used to control the notch frequency fN and the transformation with pre-warping frequency for transforming
−3dB bandwidth for a digital band-pass and band-stop filter. an analog low-pass prototype filter into a digital band-pass
Two parameters QN and fN are used to derive a new formula and band-stop filter.
for bilinear z-transformation with pre-warping frequency When the notch frequency fN is determined by using
when transforming an analog low-pass prototype filter into a inverse bilinear z-transformation with pre-warping
digital band-pass or band-stop filter. Also in this paper, the frequency, the new quality factor QN is defined, calculated
poles of the digital band-pass and band-stop filter can be
from fN and the −3dB bandwidth bw. The two parameters fN
described as a function of Q N and fN and, depending on the
value of QN, the poles of the digital filter can be a real or
and QN are used as a new method for designing digital band-
complex number that is represented for the digital filter as pass and band-stop filters that are easy to design for hand-
overdamped, underdamped or critical-damped. calculating and computing, and satisfy the requirements of
the design.
Keywords: quality factor; damping ratio; digital band-stop The unit step response of a digital band-stop filter with
filter; digital band-pass filter; bilinear z-transformation. different value QN is also analyzed in this paper.
II. DIGITAL BAND-STOP FILTER DESIGN WITH A CENTER
I. INTRODUCTION FREQUENCY F0 AND QUALITY FACTOR Q
There are several methods for designing a digital band- This section analyzes previous works for the design of a
pass and band-stop filter; all design methods must satisfy the digital Butterworth band-stop filter from an analog filter by
three most common requirements of the filter’s specification: applying bilinear z-transformation with pre-warping
the −3dB bandwidth, the center frequency, and the maximum frequency with two given parameters, f0 and Q. The analog
attenuation pass band or minimum attenuation stop [1]. In an filter can be an analog low-pass prototype [8] or an analog
analog band-pass and band-stop filter, the −3dB bandwidth band-stop filter [7][9].
bw is the difference of the upper frequency fU and the lower In analog filter design, the center frequency f0, the −3dB
frequency fL, while the center frequency f0 is where the bandwidth bw, and the quality factor Q are defined by:
maximum attenuation in a band-pass filter or the minimum
attenuation in a band-stop filter that occurs in magnitude f0  fU f L (1)
frequency response of the filter is a geometric mean of fU and bw  fU  f L (2)
fL. The quality factor Q is defined as a ratio of f0 and the
−3dB bandwidth and it is described if the filter is f0
Q (3)
overdamped (Q < 0.5), underdamped (Q > 0.5) and critical- bw
damped (Q = 0.5) in step response [2]. From (1), (2), and (3), the upper frequency fU and the
lower frequency fL can be obtained from f0 and Q as below:
In digital filters, let fN be the notch frequency where the   1 1 
frequency at which the maximum attenuation in band-pass  f L  f0  1   
  4Q 2
2Q 
filter or the minimum attenuation in band-stop filter occurs 
 (4)
in magnitude frequency response of the filter. As with   1 1 
previous band-pass filter and band-stop filter designs  fU  f0  1   
[1][3][4][5], the magnitude frequency response shows that 
  4 Q 2
2 Q 
the notch frequency fN is not the center frequency f0 and
always remains unknown [6]. Therefore, the maximum A. Method 1: Transforming an Analog Low-Pass
attenuation in band-pass filters or the minimum attenuation Prototype Filter into a Digital Band-Stop Filter
in band-stop filters cannot be controlled by the center A coefficients transfer function H(z) of a 2nth-order
frequency f0 in the digital domain; also, the computation of digital band-stop filter can be designed from a coefficients
transfer function H(s) of an nth-order analog low-pass In this design, by using f0 and Q to calculate fL and fU, the
prototype filter by using bilinear z-transformation with pre- bandwidth bw is satisfied but the center frequency f0 is not
warping frequency [8] by replacing s by a function fbs(z) as the notch frequency fN; this frequency fN remains unknown,
below which causes an error for computing the quality factor Q.
1
s  fbs ( z )  (5) B. Method 2: Transforming an Analog Band-Stop Filter
z 1 z 1 into a Digital Band-Stop Filter
U L
z 1 z 1 This section analyses another method to design a digital
Where parameters U and L can be calculated from the band-stop filter from an analog band-stop filter by using
lower frequency fL and the upper frequency fU as bilinear z-transformation [9] described as
 fU  U    
U  2 f  c  cot  2  U 
c
 
    1  ct 0 z 1
H( z )  H  s 
s
  (6) 
(8)
  L   L  t  f0  z  1 
fL  tan    
  L  2 f  t  tan  2   1  ct   fs  
 s  
Example 1: Design a 2nd-order digital Butterworth band- The coefficients transfer function H(s) of a 2nd-order
stop filter to remove an input signal that has a frequency analog band-stop filter can be written as
f = 100Hz from a 1st-order analog Butterworth low-pass s 2  02
H( s ) 
prototype filter with a quality factor Q = 2 and a sampling 
frequency fs = 1000Hz. s 2  0 s  02
Q
If chosen f0 = 100Hz is the center frequency, then the
Where ω0 = 2πf0 and Q is a quality factor as defined in
bandwidth bw that can be calculated from (3) is 50Hz. From
(3). From (8), the coefficients transfer function H(z) of a 2 nd-
(4), fL = 78.08Hz and fU = 128.08Hz, and from (6)
order digital band-stop filter can be written as
U = 5.7059 and L = 0.6079. The coefficients transfer
function H(s) of a 1st-order analog Butterworth low-pass 2Q z 2  2 cos  0  z  1
H( z )  (9)
prototype filter can written as 2Q  sin  0  2Q cos  0  2Q  sin  0 
z2  2 z
1 2Q  sin  0  2Q  sin  0 
H( s )  , substitute s = fbs(z) into Hs)
1 s f0
where  0  2
The coefficients transfer function H(z) of a 2nd-order fs
digital band-stop filter can be found as Example 2: Design a 2nd-order digital Butterworth band-
(U  L )z 2  2(U  L )z  U  L stop filter as in Example 1.
H( z )  (7)
(U  L  1 )z 2  2(U  L )z  (U  L  1 ) The coefficients transfer function H(z) of a 2nd-order
digital band-stop filter can be found as
6.3138 z 2  10.196 z  6.3138
 H( z )  z 2  1.618 z  1
7.3138 z 2  10.196 z  5.3138 H( z )  0.8719 2
The magnitude frequency response of H(z) is shown in z  1.4107 z  0.7438
Fig. 1. The magnitude frequency response of H(z) is shown in
Fig. 2.

Fig. 1. The magnitude frequency response of 2nd-order digital Butterworth


band-stop filter with f0 = 100Hz, Q = 2 and fs = 1000Hz. Fig. 2. The magnitude frequency response of 2nd-order digital Butterworth
In Fig. 1, fU = 128.08Hz and fL = 78.08Hz occur at −3dB, band-stop filter with f0 = 100Hz, Q = 2 and fs = 1000Hz.
bw = 50Hz, which satisfy the requirements of the design, but In Fig. 2, the notch located at fN = 100Hz where the
the center frequency f0 = 100Hz occurs at −35.43dB where desired input signal can be removed with the minimum
the desired signal is rejected, while the minimum magnitude magnitude of −295.7dB satisfied the requirements of the
is −76.48Hz at fN = 100.4Hz where the notch is located. design. The lower frequency fL = 79.08Hz and the upper
From (3), the quality factor Q = 100.4/50 = 2.008 is bigger frequency fU = 125.5Hz cannot be calculated from (4). From
than the quality factor Q = 2, as required. (1), the center frequency f0 = 99.622Hz is not the notch
frequency fN = 100Hz; from (2), the bandwidth bw = fU- 
fs  f   f 
fL = 46.4Hz; from (3), the quality factor fN a tan  tan   L  tan   U   ( 13 )
Q = 100/46.4 = 2.1552 is bigger than the quality factor    fs   fs  

Q = 2, as required.
Substitute fL, fU and fs into (13); fN = 100.4283Hz is the
As discussed in Examples 1 and 2, the design of a digital
same as the notch frequency where the notch is located as
band-stop filter with Method 1 will satisfy the required
shown Fig. 1.
bandwidth bw, but the notch filter does not locate at the
In Example 2, fL = 78.08Hz and fU = 125.5Hz; from (13),
center frequency f0 and locates at an unknown frequency fN.
fN = 100Hz where the notch is located as shown in Fig. 2.
While Method 2 satisfies the requirement of the notch filter
The next section introduces a new formula for bilinear z-
located at a desired frequency fN, the lower frequency fL and
transformation with pre-warping frequency when
the upper frequency fU cannot be determined; therefore, the
transforming an analog low-pass prototype filter into a
bandwidth bw does not satisfy the requirement of the design.
digital band-stop filter.
In conclusion, for both methods the center frequency f0 is not
the notch frequency fN and so the definition of the quality IV. NEW FORMULA FOR DESIGNING A DIGITAL BAND-
factor Q, as in (3), cannot apply to design a digital band-stop STOP FILTER
filter.
From (12), the parameters U and L in (5) and (6) can be
III. THE NOTCH FREQUENCY FN rewritten as
This section derives a formula that is described by the  c  f 
relationship among the notch frequency fN, the lower U  QN cot   N   QN cq
 1  ct  fs 
frequency fL and the upper frequency fU in a digital band-  ( 14 )
stop filter by using inverse bilinear z-transformation with  t  f 
pre-warping frequency. L   QN tan   N   QN tq
 1  ct  fs 
Inverse bilinear z-transformation with pre-warping
frequency is a method to map the zeros and poles of an  f   f 
analog low-pass prototype filter into zeros and poles of a where cq  cot   N  , tq  tan   N  and QN is a
digital filter [10] [11] [12].  f s   fs 
If Zk is a zero and Pk is a pole of an analog low-pass constant number that can be found as
prototype filter, and zk is a zero, pk is a pole of a digital band-  f   f 
stop filter, then from (5) zk and pk can be calculated as tan   N  sin  2 N 
QN  UL   f s 
  fs 
( 15 )
 1  fU   fL   bw 
 U  L 2
 4UL tan     tan    2 tan   
z  Z k  fs   fs   fs 
 k 1
 U  L
 Z k The bilinear z-transformation with pre-warping
 ( 10 ) frequency fbs(z) in (5) can be rewritten as
 1 1
 U  L  4UL s  fbs ( z )  ( 16 )
p  Pk2 z 1 z 1
QN cq  QN tq
 k 1 z 1 z 1
 U  L
Substitute (16) into the coefficients transfer function H(s)
 P k
In any filters, the zeros and poles can be real, a complex of a 1st-order analog Butterworth low-pass prototype filter;
number or infinite. The number of zeros is the same as the the coefficients transfer function H(z) of a 2nd-order digital
number of poles and they are equal to the order of the filter Butterworth band-stop filter in terms of QN and fN can be
[12]. So, the zeros zk for the digital Butterworth band-stop written as
filter can be found as 2QN z 2  2 cos  N  z  1
H( z )  ( 17 )
 L 2QN  sin  N  2QN cos  N  2Q  sin  N 
U  L  i 4UL 2 a tan
 U  z2  2 z N
zk  e  
 e  i N ( 11 ) 2QN  sin  N  2QN  sin  N 
UL From (12) and QN, the lower frequency fL and the upper
Where θN = 2πfN/fs, from (6) and (11), the relationship frequency fU can be calculated as
among fN, fL and fU can be found as
  fL   fN   1 1 
 f   f   f   tan     tan     1   
tan   N   tan   L  tan   U  ( 12 )   fs   fs  
2
4QN 2QN 
 f s   f s   fs   ( 18 )
In Example 1, fL = 78.08Hz, fU = 128.08Hz, and the   fU   fN   1 1 
sampling frequency fs = 1000Hz; from (12) the Notch  tan     tan     1  2
 
frequency fN can be calculated as   fs   f s  4Q N 2Q N 
A. Designing a Digital Band-Stop Filter with Given QN From Fig. 4, the notch filter located at fN = 100Hz,
and fN fL = 81.7Hz, fU = 121.7Hz, and the bandwidth bw = 121.7-
Example 3: Design a 2nd-order digital Butterworth band- 81.7 = 40Hz.
stop filter to remove an input signal that has a frequency C. Designing a Digital Band-Stop Filter with Given fL or
f = 100Hz with given QN = 0.5 and fs = 1000Hz. fU and fN
Let fN = 100Hz, θN = 2πfN/fs = 0.2π and QN = 0.5. From
(15), the required bandwidth bw = 169.1466Hz. From (18), Example 5: Design a 2nd-order digital Butterworth band-
stop filter to remove an input signal that has a frequency
fU = 211.7308Hz and fL = 42.5842Hz. The coefficient
f = 100Hz with a given lower frequency fL = 20Hz and a
transfer function H(z) can be found from (17) as
sampling frequency fs = 1000Hz.
z 2  1.618 z  1 From (12), fU = 328.932Hz; from (15), QN = 0.2012 and
 H( z )  0.6298 2
z  1.019 z  0.2596 the bandwidth bw = 308.932Hz. The coefficient transfer
The magnitude frequency response of H(z) is shown in function H(z) can be found from (17) as
Fig. 3. z 2  1.618 z  1
 H( z )  0.4064 2
z  0.6575 z  0.1873
The magnitude frequency response of H(z) is shown in
Fig. 5.

Fig. 3. The magnitude frequency response of a 2nd-order digital Butterworth


band-stop filter with fN = 100Hz, QN = 0.5 and fs = 1000Hz.
In Fig. 3, the notch filter is located at fN = 100Hz. From
(18), the lower frequency fL = 42.58Hz and the upper
Fig. 5. The magnitude frequency response of a 2nd-order digital Butterworth
frequency fU = 211.7Hz; the bandwidth bw = fU- band-stop filter with fN = 100Hz, fL = 20Hz and fs = 1000Hz.
fL = 169.12Hz. This satisfied the requirements of the design. In Fig. 5, the notch filter located at fN = 100Hz,
B. Design a Digital Band-Stop Filter with Bandwidth bw fL = 20Hz, and fU = 328.9Hz, and the bandwidth bw = fU-
and fN fL = 308.9Hz, satisfied the requirements of the design.
Example 4: Design a 2nd-order digital Butterworth band- V. THE NEW QUALITY FACTOR QN
stop filter to remove an input signal that has a frequency
f = 100Hz with a given bandwidth bw = 40Hz and The poles of a digital filter can be derived from the poles
fs = 1000Hz. of an analog low-pass prototype filter, as in (10). The pole of
From (15), QN = 2.3266. From (18), fL = 81.7147Hz and a 1st-order Butterworth low-pass prototype filter is PK = −1;
fU = 121.7147Hz. The coefficient transfer function H(z) can the poles pk of a 2nd-order digital Butterworth band-stop filter
be found from (17) as can be calculated as
z 2  1.618 z  1 U  L  1  4UL U  L  1  4QN
2

 H( z )  0.8878 2 pk   ( 19 )
z  1.437 z  0.7757 U  L1 U  L1
The magnitude frequency response of H(z) is shown in Depending on the values of QN, the poles pk can be a
Fig. 4. complex conjugate pair, two unequal real numbers, or two
equal real numbers; QN as defined in (15) is called the
quality factor for a digital band-stop filter.
 bw 
tan   
Let  
1
  fs  is a damping factor.
2QN sin  N 
If given the notch frequency fN, increasing QN, the
damping factor ζ is decreasing and the bandwidth of the
digital band-stop is also decreasing.
Case 1: 0 < QN < 0.5  ζ > 1, this is called the
overdamped case and the poles pk are two unequal real
Fig. 4 The magnitude frequency response of a 2nd-order digital Butterworth numbers; they can be calculated as
band-stop filter with fN = 100Hz, Bw = 40Hz and fs = 1000Hz.
cos  N   sin  N   2  1 damping factor ζ is always chosen to be greater than
pk  ( 20 ) zero when designing a digital band-stop filter.
1   sin  N  The unit step response of the 2nd-order digital
Case 2: QN = 0.5  ζ = 1, this is called critically damped Butterworth band-stop filter in Examples 3, 4, and 5
and the poles are two equal real numbers; they can be corresponding to the quality factor QN = 0.5, 2.3266, and
calculated as 0.2012 are shown in Fig. 7.
cos  N 
pk  ( 21 )
1  sin  N 
Case 3: QN > 0.5  0 < ζ < 1, this is called the
underdamped case and the poles pk are a complex conjugate
pair; they can be calculated as
cos  N   i sin  N  1   2  i p
pk   re ( 22 )
1   sin  N 
Where
1   sin  N 
r
1   sin  N 

and  p  a tan tan  N  1   2  Fig. 7. The unit step response of a digital Butterworth band-stop filter with
QN = 0.2012, 0.5 and 2.3266.
- In Example 3, QN = 0.5 = > ζ = 1 (critically
The poles pk of a 2nd-order digital Butterworth band-stop
filter are distributed in z-domain with different values of damped), and there are two equal real poles
damping factor ζ, as shown in Fig. 6. pk = 0.5095. At time t = 0, the amplitude equals to
filter gain g2 = 0.6298. The unit step response of the
filter returns to 1 amplitude faster without
oscillating.
- In Example 4, QN = 2.3266 = > ζ = 0.2149 (ζ<1,
underdamped), and there is a complex conjugate
pair pole pk = 0.7183 ± i0.5097. The amplitude
equals to filter gain g1 = 0.8878 when t = 0. The
filter oscillates with the amplitude reduced to 1. The
filter reached 1 amplitude faster than the critically
damped case, but introduced overshoots and
undershoots.
- In Example 5, QN = 0.2012 = > ζ = 2.4851 (ζ > 1,
overdamped), and there are two unequal real poles
pk = 0.8722 and −0.2147. At time t = 0, the
amplitude equals to the filter gain g3 = 0.4048. The
filter returns slower to 1 amplitude without
Fig. 6. The poles pk of a 2nd-order digital Butterworth band-stop filter are
oscillating.
distributed in z-domain with different values of damping factor ζ. In a special case that is applied for designing a half-band
Fig. 6 illustrates the distribution in digital domain of the digital Butterworth band-stop filter for a multi-rate system
poles pk varying with damping factor ζ. [13], when the notch frequency fN equals one-quarter of the
- When ζ = 1 (critically damped), there are two equal sampling frequency fs, the coefficients transfer H(z) in (17)
real poles located on the real axis at point D. can be rewritten as
- When ζ > 1 (overdamped), there are two unequal 1 z2  1
real poles located on the real axis and −1 < pk < 1. If H( z )  ( 23 )
1   z2  1  
ζ = ∞, the poles pk = ± 1.
1
- When 0 < ζ < 1 (underdamped), there is a complex
conjugate pair pole distributed on a circle with the where the damping factor ζ can be calculated from the
center C(1/cos(θN),0) and a radius R = tan(θN). This bandwidth bw of the filter as
circle and the unit circle are orthogonal at point A  bw 
and B where the zeros of the digital band-stop filter   tan    ( 24 )
are located.  fs 
- When ζ = 0, the complex poles are equal to the zeros - ζ = 1, there are two poles located at the original axis
and the coefficients transfer function H(z) in (17) 0(0,0), pk = 0.
equals 1 and is not a band-stop filter; therefore, the - ζ > 1, there are two real poles located on the real axis
and symmetric to the imaging axis.
 1 frequency fN, the lower frequency fL and the upper frequency
pk   ( 25 ) fU were also formulated. Two parameters QN or ζ and fN can
 1 be used to design a digital band-stop filter and also a digital
- 0 < ζ < 1, there are two poles located on the imaging band-pass filter. The steps of the design demonstrated in the
axis and symmetric to the real axis examples are simple and easy for calculation and
1 programming. The unit step response of the filters for
pk   i ( 26 ) overdamped, critically damped and underdamped cases were
1
also discussed. For further research, the three major
parameters of the filter, overshoot, and the rise time and
VI. DIGITAL BAND-PASS FILTER
delay time associated with the unit step response will be
The coefficients transfer function H(z) of a 2nd-order studied.
digital Butterworth band-pass filter can be written in terms of
the damping factor ζ and fN as ACKNOWLEDGMENT
 sin  N  z2  1 This article comprises independent research by the
H( z )  ( 27 )
1   sin  N  2 cos  N  1   sin  N  author; he would like to share his idea about designing a
z2  z
1   sin  N  1   sin  N  digital filter with a new definition of the quality factor in the
digital domain with those who have the same interest. He
From (27), the zeros of the filter are located at zk = ± 1,
would greatly appreciate all feedback from readers to
while the poles pk can be found as the poles of the digital
improve this research!
band top filter as discussed in Section V. In this case, the fN
equals one-quarter fs; H(z) in (27) can be rewritten as REFERENCES
 z2  1 Chakraborty, S, Dey, A, Saha, K K, and Dutta, N 2015, “Design and
H( z )  ( 28 ) [1]
1   z2  1   determination of optimum coefficients of high-Q IIR notch filter”,
IJIREEICE, vol. 3, no. 9, pp. 33–39.
1
[2] https://en.wikipedia.org/wiki/Q_factor.
nd
Example 6: Design a 2 -order Digital Butterworth half- [3] Nikolić, S and Stančić, G 2012, “Design of IIR notch filter with
band band-pass filter with a bandwidth of 2Hz at a sampling approximately linear phase”, Circuits, Systems, and Signal
frequency 1000Hz. Processing, vol. 31, no. 6, pp. 2119–2131.
From (24), ζ = 0.0063, the zeros zk = ± 1, and the poles [4] Chakraborty, S and Patra, A 2013, “Performance analysis of IIR
pk = ± i0.9937. From (18), fL = 249Hz and fU = 251Hz. The digital band stop filter”, International Journal of Advanced Research
notch frequency fN = 250Hz. From (28), the coefficients in Computer Engineering & Technology, vol. 2, no. 5, pp. 1749.
transfer function H(z) can be found as [5] Wang, C M and Xiao, W C 2013, “Second-order IIR notch filter
design and implementation of digital signal processing system”,
z2  1 Applied Mechanics and Materials, vol. 347, pp. 729–732.
H( z )  0.0062 2
z  0.9875 [6] Stančić, G and Nikolić, S 2011, “Design of narrow stopband
recursive digital filter”, Facta Universitatis-series: Electronics and
The magnitude frequency response of H(z) is shown in Energetics, vol. 24, no. 1, pp. 119–130.
Fig. 8.
[7] Petrovsky, A A 1996, “The synthesis of a high order digital bandpass
filters with tunable centre frequency and bandwidth,” in European
Signal Processing Conference, 1996. EUSIPCO 1996. 8th, pp. 1–4.
[8] Phuoc S N 2014, “Design of digital filters using Pascal’s triangle”,
paper presented at 38th Vibration Institute Annual Training
Conference, Santiano, Texas, USA, June 2014.
[9] Shah, V and Patil, R S 2008, “IIR bandpass filter design using
modified transfer function implementation algorithm”, in
Proceedings of the 7th WSEAS International Conference on Signal
Processing, Robotics and Automation, pp. 265–269.
[10] Garcia-Ugalde, F J 2011, “Z-transform by Pascal matrix and its
applications in the design of IIR Filters”, Journal of Applied Research
of Technology, vol. 9, no. 3, pp. 355–366.
Fig. 8. The magnitude frequency response of a half-band digital [11] Chivapreecha, S 2005, “Bilinear s–z frequency transformation using
Butterworth band-pass filter with the bandwidth 2Hz at a sampling matrix Pascal operation,” IEEE Communications and Information
frequency fs = 1000Hz. Technology, vol. 1, pp. 764–767.
In Fig. 8, the notch filter located at fN = 250Hz, [12] Nguyen, P S 2016, “Frequency transformation in digital domain”,
fL = 249Hz, fU = 251Hz, and the bandwidth bw = fU-fL = 2Hz International Journal of Signal Processing Systems, vol. 4, no. 5, pp.
389–393. doi:10.18178/ijsps.4.5.389-393
satisfied the requirements of the design.
[13] Nguyen, P S, “A general method design for digital half-band iir band-
pass and band-stop filters”, www.researchgate.
VII. CONCLUSION
The new definition for quality factor QN and the damping
factor ζ were studied and the relationship among the notch

Das könnte Ihnen auch gefallen