[SOLVED] MFB narrow bandpass review and other active filter related questions

[d123]

Hi,

I'm trying to learn a little about active filters. Chose specs of high Q, passband centred at ~10.5kHz, gain of 10 (for no special reason).

Before getting onto the MFB I copied, some questions would be:

For flat passband and sharp stopband transition and minimum parts count, which is the most appropriate topology?

Not Tschebyshev, clearly; Butterworth seem far too slow in stopband and Bessel seem worse than Butterworth in that respect. Or are these assumptions incorrect?

MFB have drawback of inverting (from what I've learnt so far) - does that mean an allpass is required after MFB to get in-phase output, and is that horrendous to put together for a person who still hasn't found a suitable explanation of how to implement "s" and "√-1" in formulas I see, nor how to assess phase response in AC transfer results?

I copied the MFB narrow bandpass from this pdf:

View attachment Band_Pass_Filters.pdf

This is the schematic and the results:



The resistor values are calculator-based results, not standard values, so as to see what to expect in an ideal world.

From the tiny graph, do the results, centred at 10.5kHz seem ~realistic and what one would hope to see? How far does a real implementation diverge from this, if it's possible to describe?

If the input signal is centred at 2V and rises/falls 0.5V, what kind of output voltage would I expect with a gain of 10, and how do I extrapolate this from the gain graph results, please?

I think I have quite a few more questions but I'll try not to be a pita. A problem I have is that I'm beginning to think a lot of the didactic material/app notes seem to show somewhat idealised/exaggerated response curves to help understand the premise more than the reality, maybe I'm wrong.

Thanks

 

[FvM]

The requested "flat passband", similar to "wide band pass filter" in the link seems to refer to a high-pass/low-pass combination rather than a standard bandpass prototype.

In any case, you should specify the intended stop- and passband characteristic and then select an appropriate filter design.

 

[d123]

Hi,

I was reading about wire tracers and one that transmits 1.9kHz to 2.1kHz. I'd like to feel confident with filters so started at ~10kHz (9.5kHz to 10.5kHz) as a learning point.

Before looking at an MFB filter I simulated a cascaded low pass and high pass Butterworth and found the transition from pass to stop was too slow (I think, which is why I'm asking questions here) but the passband was acceptable. Below is the schematic and the results.



Should I only be looking at the point at -3dB from the peak of the passband to evaluate if the design is appropriate to requirements? What can be interpreted from the gain graph and the phase graph, respectively? Input is same 2V centred +-0.5V squarewave.

So, to define specifications: if for example, the passband were 9.5kHz to 10.5kHz and the required transition were intended to not pass anything outside this range at the output, which topology would give suitable results using few(-est) parts? Would Sallen-Key be more appropriate for what is intended?

Lastly, if the peak input voltage is 2.5V, would the output voltage (with a gain of 1) transmit the input voltage faithfully?

 

[FvM]

It's not a matter of topology, it's a question of filter type and parameters. The same filter type can be implemented in different ways, e.g. passive LC, various active RC filter topologies. Stop band transition steepness mainly depends on filter order, what do you require?

 

[LvW]

Hi,
Before looking at an MFB filter I simulated a cascaded low pass and high pass Butterworth and found the transition from pass to stop was too slow

Don`t ovelook the most important degree of freedom: Filter order !!
In your example, you were using two first order filters only .

 

[BradtheRad]

active filters. Chose specs of high Q...

For flat passband and sharp stopband transition and minimum parts count, which is the most appropriate topology?

The simple LC network is hard to beat. Its resonant behavior rejects all frequencies except its center frequency.
Higher Q is associated with less resistance, because that is how you get a response based more 'purely' on the reactive components.

L:C ratio is important. This is related to Ampere level. In some situations you can make an LC network provide gain (even greater than 10), although it's less easy with real components compared to a simulator which uses ideal components.

We generally find it inconvenient to work with inductors, so the active filter is usually built with capacitor-resistor networks. I suspect that these contain an active inductor somewhere, to substitute for a large Henry value.

You can get sharper rolloff with multiple LC networks. Similarly you can do it with multiple active filters. I used a biquad filter (3 op amps & 3 RC networks), to extract a particular morse code broadcast, when one or two others were audible a few tone-steps away. Another type I considered was a state variable filter. They were one of many I saw in the Active filter cookbook.

 

[FvM]

May we finally hear a quantitative filter specification?

 

[LvW]

You can get sharper rolloff with multiple LC networks. Similarly you can do it with multiple active filters. I used a biquad filter (3 op amps & 3 RC networks), to extract a particular morse code broadcast, when one or two others were audible a few tone-steps away. Another type I considered was a state variable filter. They were one of many I saw in the Active filter cookbook.

Selecting the most appropriate filter topology for a specific application is a rather challenging task. You must consider technical as well as economical aspects:
Filter order (damping requirements), Filter type (Butterworth, Chebyshev,...), Frequency range, tuning capabilities, amplifier properties, number of amplifier units, sensitivity aspects (active/passive tolerances), cost, power consumption, .....

 

[d123]

May we finally hear a quantitative filter specification?

I'm sorry, what do you mean? I think I stated where I am with filters from the word go. Presumably I'm not up to answering your question or you haven't understood my posts, just as my original questions in post #1 haven't been answered so by now I consider this thread a waste of time and generally unhelpful. I asked the questions to learn about this subject, not to be patronised,I'm afraid. If with two schematics and response graphs and a beginners description of what they are attempting to achieve or interested in discussing it's not yet clear, I guess it will never be.

I'm closing this thread as it really does seem pointless to go round in circles.

 

[FvM]

Referring to your previous post

So, to define specifications: if for example, the passband were 9.5kHz to 10.5kHz and the required transition were intended to not pass anything outside this range at the output, which topology would give suitable results using few(-est) parts? Would Sallen-Key be more appropriate for what is intended?

You are describing an ideal filter. A real filter has limited flat pass band, finite steep transitions and still passes some signal residuals in the stop band.

The filter performance, e.g. transition steepness doesn't depend on the circuit topology, mainly on the filter order, in other words number of frequency selective elements, either LC or active RC blocks. The circuits in post #1 and #3 are both first order filters and have 6 dB/octave asymptotic slope, in so far they have despite of different pass band shape similar stop band performance.

Besides asking for better filters, you didn't tell how good you need it. But without deciding this point, there's no answer.

- - - Updated - - -

I don't yet understand how the imagined square wave signal is related to the filter problem. Square wave is a wide band signal, so surely it won't keep the shape when passing a narrow band pass. But what kind of square wave do you consider? Fundamental near the BP center or much below it? What's the intended output waveform?

 

[BradtheRad]

sharp stopband transition

The twin-T makes a good notch filter. Evidently it is able to split the signal into two different phase relationships, thus cancelling the center frequency because it achieves 180 degree phase opposition.

It's the focus of this article:

https://www.electronics-notes.com/a...lifier-op-amp/notch-filter-active-circuit.php

 

[LvW]

I'm sorry, what do you mean? I think I stated where I am with filters from the word go. Presumably I'm not up to answering your question or you haven't understood my posts, just as my original questions in post #1 haven't been answered so by now I consider this thread a waste of time and generally unhelpful. I asked the questions to learn about this subject, not to be patronised,I'm afraid. If with two schematics and response graphs and a beginners description of what they are attempting to achieve or interested in discussing it's not yet clear, I guess it will never be.
I'm closing this thread as it really does seem pointless to go round in circles.

A good and comprehensive answer requires a clear and precise question.
Some examples:

Quote: "For flat passband and sharp stopband transition and minimum parts count, which is the most appropriate topology?"
What do you mean with "sharp"? That´s what FvM was referring to while asking for a clear "specification".
More than that, in my former answer I have tried to explain to you that it is not possible to suggest a "most appropriate" topology - without knowing something about additional requirements (at least a clear specification).

Quote: "Not Tschebyshev, clearly; Butterworth seem far too slow in stopband ...."

"Far too slow..." what does this mean? Didn`t you read the answer in which the order of the filter was mentioned? With Butterworth approximation you can have each stopband attenuation you like - depending on the filter order.

Quote: "MFB have drawback of inverting (from what I've learnt so far) - does that mean an allpass is required after MFB to get in-phase output, and is that horrendous to put together for a person who still hasn't found a suitable explanation of how to implement "s" and "√-1" in formulas I see, nor how to assess phase response in AC transfer results?"

Don`t you think it would be important for somebody (who is willing to help you) to know if you know the meaning of "s"? Or did you expect to get in this forum a complete lesson on sysytem theory and the background for using a complex frequency variable?

Perhaps, you should try to be a bit "self-critical"?