6
\$\begingroup\$

enter image description here

Is there an explicit formulation for n-cascaded RC filters?

The resistor and capacitor values are not necessarily same. Assume that:
\$ R_i \neq R_j \$ and \$ C_i \neq C_j \$ for \$ i \neq j\$ and \$ 1 \leq i,j \leq n \$.

I built a 5th degree filter with \$R_i=1k\Omega\$, \$C_i=1\mu F\$ and \$V_{in}=1\sin(2\pi \times 1kHz) \; V\$. The output signal was nearly smooth for a \$1k\Omega\$ lead. And rise of the output signal was visible by eye when the input was a step signal.

I tried to find transfer response of this 5th order filter, but after the 3rd order, the size of A4 paper became insufficient for writing the expressions.

\$\endgroup\$
  • \$\begingroup\$ I think your image may be missing \$\endgroup\$ – Andy aka Nov 11 '13 at 8:19
  • \$\begingroup\$ It is showing up to me... Anyway: there is no analytical expression for an n-stage RC filter. I tried to crack this problem a couple of months ago and it just gets very unwieldy, even when you can generalize R1=R2=...=Rn. The best way to go is just to regard each stage as a shaped attenuator with finite input and output impedance and multiply them together numerically. \$\endgroup\$ – user36129 Nov 11 '13 at 8:25
  • \$\begingroup\$ I can see it now!! \$\endgroup\$ – Andy aka Nov 11 '13 at 9:33
2
\$\begingroup\$

There isn't an explicit formula for n cascaded RC low pass filters (as you have shown) because as you move from left to right, R2 and C2 load the output of R1 and C1 thus changing R1&C1's response and because R3 and C3 are doing the same to R2 and C2 this in turn reflects on the loading of R1 and C1.

Using 2 port networks and matrix maths does help you solve more easily for several sections but the formulas do tend to get long after a few sections and solving for the cut-off frequency and Q is very tiresome.

The best solution is to use a simulator like LTSpice.

Given that you have an op-amp in your circuit acting as a buffer it wouldn't be unreasonable to use this to help. This sallen-key calculator is very good and gives you the response and formulas for a 2nd order low pass filter: -

enter image description here

Cascading 2 of these will probably get you a better response than cascading five passive RC filters and the beauty of this method is that you can cascade them without interaction between components because the op-amp acts as a buffer.

\$\endgroup\$
  • \$\begingroup\$ I've sometimes wondered if the subdivision of filters into 2-pole sections primarily for ease of figuring out what a filter is doing? I would expect that adding an extra pole using a passive RC combos won't quite yield a full "pole"'s worth of performance, but if the impedance of the second RC is large relative to the first, the performance should be comparable, and I would think that a 9-pole filter using 3 op amps should be able to yield cutoff-frequency performance comparable to an 8-pole filter using four, though I'm not sure about phase performance. \$\endgroup\$ – supercat Nov 11 '13 at 17:22
  • \$\begingroup\$ @supercat I did the very same last week; three cascaded LCR low pass filters with approx 3 times the impedance for subsequent stages. You can only push it so far though! \$\endgroup\$ – Andy aka Nov 11 '13 at 18:10
  • \$\begingroup\$ Certainly I wouldn't expect a 16-pole cascaded RC filter to work very well, but I would think that getting an extra pole or two on each stage of a filter should be entirely practical since in some filter topologies a low-impedance op-amp output will be fed into a high-impedance input so it would seem one could replace an RC stage with either two stages that have a 100:1 impedance variation between them, or three stages with a 10:1 impedance separation. Even just getting three poles per op amp instead of four would seem a pretty big win. \$\endgroup\$ – supercat Nov 11 '13 at 18:23
  • \$\begingroup\$ It's easier with LCRs to do that because losses are smaller.... just small value resistors to control Q. RC it's harder or at least seems to be. \$\endgroup\$ – Andy aka Nov 11 '13 at 18:27
0
\$\begingroup\$

The behavior of any series-parallel combination of resistors, capacitors, and inductors may be modeled at any given frequency using complex arithmetic by regarding each capacitor and inductor as though it were a resistor whose "resistance" [actually impedance] is an imaginary number; a resistor in series with a capacitor thus behaves like a "resistor" with a complex impedance whose real part comes from the resistor and whose imaginary part comes from the capacitor. The normal formula for parallel resistors (1/(1/R1+1/R2)) works just as well for impedances as for simple resistances. Using complex maths, one could compute the voltage output of a cascaded RC filter in terms of the input using a somewhat-reasonable-looking formula. Unfortunately, those forumulae become very messy if one wants to use real numbers to compute the real (in-phase) and imaginary (out-of-phase) components of the final signal, since division by a complex number generally requires normalizing it, so something like (a+bj)/(c+dj) becomes ((ac-bd)+(bc+ad)j)/(c^2-d^2). Nesting such expressions very deeply causes them to become unwieldy very quickly.

Still, if you have a package which can easily handle math with complex numbers, you should be able to come up with a complex transfer function and then plot it. Alternatively, if you have a circuit simulator with a Bode plotter function (which stimulates a circuit with various frequencies and plots the response) you could use that to plot the behavior of your circuit numerically.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.