3

I have the next Sallen Key type filter
enter image description here

I have determined the transfer function and the coefficients of \$a_{0}\$ and \$a_{1̣̣}\$, based on

\$N(s)=\frac{Ka_{0}}{s^{n}+a_{n-1}s^{n-1}+....a_{1}s+a_{0}}\$

But then how do I use this to determine (or approximate) analytically what kind of response the filter has? (Butterworth, Chebyshev etc.)

This is besides making a frequency sweep using a simulator.
UPDATE
Derivating the transfer function. I have named the node between the 47k resistances as \$V_{x}\$ and the node of the negative bias \$V_{y}\$
\$V_{out}=(1+\frac{R_{A}}{R_{B}})V_{y}\$

\$G=1+\frac{R_{A}}{R_{B}}=\frac{R_{A}+R_{B}}{R_{B}}\$
\$V_{y}=\frac{R_{B}}{R_{A}+R_{B}}V_{out}\$
Simplifying
\$V_{y}=\frac{V_{out}}{G}\$
using a voltage divider
\$V_{y}=\left(\frac{V_{x}}{R_{2}+\frac{1}{C_{2}s}}\right)\frac{1}{C_{2}s}=V_{x}\frac{1}{1+R_{2}C_{2}s}\$
\$V_{x}=(1+R_{2}C_{2}s)V_{y}\$
\$0=\frac{V_{x}-V_{in}}{R_{1}}+\frac{V_{x}-V_{y}}{R_{2}}+\frac{V_{x}-V_{out}}{\frac{1}{sC1}}\$
\$\frac{V_{x}-(1+R_{2}C_{2}s)V_{y}-V_{in}}{R_{1}}+\frac{(1+R_{2}C_{2}s)V_{y}-V_{y}}{R_{2}}+\frac{(1+R_{2}C_{2}s)V_{y}-V_{out}}{\frac{1}{sC1}}=0\$
\$\frac{(1+R_{2}C_{2}s)V_{y}}{R_{1}}-\frac{V_{in}}{R_{1}}+\frac{(1+R_{2}C_{2}s)V_{y}}{R_{2}}-\frac{V_{y}}{R_{2}}+\frac{(1+R_{2}C_{2}s)V_{y}}{\frac{1}{sC1}}-\frac{V_{o}}{\frac{1}{sC1}}=0\$
\$\frac{(1+R_{2}C_{2}s)V_{out}}{GR_{1}}-\frac{V_{in}}{R_{1}}+\frac{(1+R_{2}C_{2}s)V_{out}}{GR_{2}}-\frac{V_{out}}{GR_{2}}+\frac{(1+R_{2}C_{2}s)V_{out}}{\frac{G}{sC1}}-\frac{V_{o}}{\frac{1}{sC1}}=0\$
\$V_{out}\left[\frac{(1+R_{2}C_{2}s)}{GR_{1}}+\frac{(1+R_{2}C_{2}s)}{GR_{2}}-\frac{1}{GR_{2}}+\frac{(1+R_{2}C_{2}s)}{\frac{G}{sC1}}-\frac{1}{\frac{1}{sC1}}\right]=\frac{V_{in}}{R_{1}}\$
\$\frac{V_{out}}{G}\left[\frac{(1+R_{2}C_{2}s)}{R_{1}}+\frac{(1+R_{2}C_{2}s)}{R_{2}}-\frac{1}{R_{2}}+\frac{(1+R_{2}C_{2}s)}{\frac{1}{sC1}}-\frac{1}{\frac{1}{GsC1}}\right]=\frac{V_{in}}{R_{1}}\$
\$\frac{V_{out}}{G}\left[\frac{1}{R_{1}}+\frac{(R_{2}C_{2}s)}{R_{1}}+\frac{1}{R_{2}}+\frac{(R_{2}C_{2}s)}{R_{2}}-\frac{1}{R_{2}}+\frac{1}{\frac{1}{sC1}}+\frac{(R_{2}C_{2}s)}{\frac{1}{sC1}}-\frac{1}{\frac{1}{GsC1}}\right]=\frac{V_{in}}{R_{1}}\$
\$\frac{V_{out}}{G}\left[\frac{1}{R_{1}}+\frac{(R_{2}C_{2}s)}{R_{1}}+C_{2}s+sC1+sC1R_{2}C_{2}s-GsC_{1}\right]=\frac{V_{in}}{R_{1}}\$
\$\frac{V_{out}}{G}\left[\frac{1+R_{2}C_{2}s+R_{1}C_{2}s+R_{1}sC1+R_{1}sC1R_{2}C_{2}s-R_{1}GsC_{1}}{R_{1}}\right]=\frac{V_{in}}{R_{1}}\$
\$\frac{V_{out}}{G}\left[1+R_{2}C_{2}s+R_{1}C_{2}s+R_{1}sC1+R_{1}sC1R_{2}C_{2}s-R_{1}GsC_{1}\right]=V_{in}\$
\$\frac{V_{out}}{G}\left[s^{2}R_{1}R_{2}C1C_{2}+s(R_{1}C_{1}+R_{2}C_{2}+R_{1}C_{2}-GR_{1}C_{1})+1\right]=V_{in}\$
\$\frac{V_{out}}{G}\left[s^{2}R_{1}R_{2}C1C_{2}+s(R_{1}C_{1}-GR_{1}C_{1}+R_{2}C_{2}+R_{1}C_{2})+1\right]=V_{in}\$
\$\frac{V_{out}}{G}\left[s^{2}R_{1}R_{2}C1C_{2}+s(R_{1}C_{1}(1-G)+R_{2}C_{2}+R_{1}C_{2})+1\right]=V_{in}\$
\$\frac{V_{out}}{V_{i}}=\frac{G}{s^{2}R_{1}R_{2}C1C_{2}+s(R_{1}C_{1}(1-G)+R_{1}C_{2}+R_{2}C_{2})+1}\$
\$\frac{V_{out}}{V_{i}}=\frac{G}{s^{2}R_{1}R_{2}C1C_{2}+s(R_{1}C_{1}(1-G)+R_{1}C_{2}+R_{2}C_{2})+1}\$
\$\frac{V_{out}}{V_{i}}=\frac{\frac{G}{R_{1}R_{2}C1C_{2}}}{s^{2}+s(\frac{1}{R_{1}C_{1}}+\frac{1}{R_{1}C_{2}}+\frac{(1-G)}{R_{2}C_{2}})+\frac{1}{R_{1}R_{2}C1C_{2}}}\$
if \$R_{1}=R_{2}=R\$ and \$C_{1}=C_{2}=C\$
\$\frac{V_{out}}{V_{i}}=\frac{\frac{G}{R^{2}C^{2}}}{s^{2}+s(\frac{1}{RC}+\frac{1}{RC}+\frac{(1-G)}{RC})+\frac{1}{R^{2}C^{2}}}\$
\$\frac{V_{out}}{V_{i}}=\frac{\frac{G}{R^{2}C^{2}}}{s^{2}+s(\frac{(3-G)}{RC})+\frac{1}{R^{2}C^{2}}}\$
\$G=K\$

\$a_{0}=\frac{1}{R^{2}C^{2}}\$

\$a_{1}=\frac{3-G}{RC}\$

avelardo
  • 495
  • 3
  • 10
  • 4
    Why don't you write out the transfer function that you've determined for this circuit? That alone will move me better forward towards suggesting an answer for you. (If I find an error in what you've developed, then I may ask you to show your work, as well.) – jonk Jun 20 '21 at 19:26
  • 1
    Ok, give me a couple a minutes its a bit long to write down,thanks in advance! – avelardo Jun 20 '21 at 19:41
  • 3
    That's an equal value design with a gain of 1.587 which gives a Butterworth response with a Q of 0.707 and a -3dB cut-off frequency of 1/(2*pi*RC) = 10.261 kHz –  Jun 20 '21 at 20:21
  • 1
    @jonk I have added the derivation of the transfer function – avelardo Jun 20 '21 at 20:44
  • 5
    @avelardo Thanks!! And Wow! It sure shows that you put some effort into this. +1 – jonk Jun 20 '21 at 20:45
  • 1
    @James, and how can be said its a Buttewroth? I mean, can it be done viewing a table or so? – avelardo Jun 20 '21 at 20:45
  • 1
    @jonk, thanks to you for taking your time, you are very kind. Sorry for my ignorance. – avelardo Jun 20 '21 at 20:49
  • 3
    @avelardo Solving, I get \$\omega_{_0}=\frac1{R \, C}\$, \$\zeta=1-\frac12\frac{R_2}{R_1}\$, and \$A=1+\frac{R_2}{R_1}\$ for the standard 2nd order low-pass form of \$\mathcal{H}=\frac{V_O}{V_I}=A\cdot \frac{\omega_{_0}^2}{s^2+2\zeta\,\omega_{_0}\,s+\omega_{_0}^2}\$. Plugging in values for \$R_1\$ and \$R_2\$, I get \$\zeta\approx 0.706\$. This is very close to a Butterworth, which for a 2nd order uses \$\zeta=\frac12\sqrt{2}\$. (The damping is just slightly less, so it will be just slightly towards the under-damped side of things.) If you need me to write this up in more detail, I will. – jonk Jun 20 '21 at 20:54
  • 3
    @avelardo Yes, there is a table in The Art of Electronics by Horowitz & Hill (the chapter on filters) which details what the gain needs to be set to, for equal component value filters, in order to achieve a certain response (Butterworth, Bessel or Chebyshev). The table gives a gain value of 1.586 for a 2 pole Butterworth response but I have seen that gain value detailed from other sources. –  Jun 20 '21 at 20:56
  • 2
    @avelardo I'd call this a Butterworth, though. It's too close to it to bother calling it anything else. – jonk Jun 20 '21 at 20:58
  • @jonk I will go with that. – avelardo Jun 20 '21 at 21:12
  • 1
    @avelardo So no need to add an answer, then?$$$$By the way, yes, there are tables. For example, a 2nd order Butterworth uses the analytic:$$s^2 + \sqrt{2}\,s+1$$ A 4th order Butterworth uses this analytic: $$\left(s^2+\frac14\left(\sqrt[\frac32]{2+\sqrt{2}}+\sqrt{2}\sqrt{2-\sqrt{2}}\right)s+1\right)\cdot\left(s^2+\frac14\left(\sqrt[\frac32]{2-\sqrt{2}}-\sqrt{2}\sqrt{2+\sqrt{2}}\right)s+1\right)$$In the 4th order, \$\zeta_{_1}\approx 1.84776\$ and \$\zeta_{_2}\approx 0.76537\$. – jonk Jun 20 '21 at 21:16
  • @jonk, Im not sure if it it or not, because using the values of the components it gives a bit up for a Butterworth but not enough for Chebyshev- – avelardo Jun 20 '21 at 21:48

1 Answers1

3

Because you have the transfer function with all parts values you can compare this function with the general function - expressed NOT with coefficients (which have no direct relation to the kind of response) but with the pole data wp (pole frequency) and \$Q_p\$ (pole quality factor).

It is in particular the Qp value that gives you information on the kind of response.

Example: \$Q_p=0.7071\$ (Butterworth) and \$Q_p>0.7071\$ (Chebyshev).

Knowing the kind of filter response you can use filter tables to find the relation between wp and the corresponding cut-off frequency wc of the lowpass.

Transfer function (2nd order):

$$H(s)=\dfrac{A_o}{1+\dfrac{s}{\omega_pQ_p}+\dfrac{s^2}{\omega_p^2}}$$

Comment 1: In your circuit there are two equal \$R\$ and to equal \$C\$ values. In this special case ("equal component design") the pole \$Q\$ is \$Q_p=\dfrac{1}{3-v}\$ with \$v\$ = closed-loop opamp gain.

Comment 2: This "equal component design" has some disadvantages: It requires an "odd" gain value and the Qp value is very sensitive to gain variations. For this reason two other approaches are used very often (however, with "odd" and unequal values for R or C): Unity gain or gain of two (feedback with two equal resistors).

LvW
  • 24,857
  • 2
  • 23
  • 52
  • 3
    I have made many Sallen-Key second-order lowpass filters using equal-value R and C and an opamp gain of about 1.6 to produce a Butterworth response. – Audioguru Jun 20 '21 at 20:53
  • 2
    @LvW, and how can be calculated the Pole frequency? is the same that the cut-off frequency? – avelardo Jun 20 '21 at 20:59
  • 1
    The pole frequency wp can be derived directly from the transfer function (in your case: wp=1/RC). For Butterworth only we have wp=wc. For Chebyshev responses the ratio wp/wc is given in filter tables for different orders. – LvW Jun 21 '21 at 07:03
  • 3
    @a concerned citizen, thank you for editing. – LvW Jun 21 '21 at 07:05