Bandpass Filter Design for All Audible Frequencies

Thread Starter

jchernak

Joined Jan 28, 2017
2
Hello all,

I'm on a project designing a microphone with built-in filtering. We are simply looking to filter out any inaudible frequencies we record. That is, we're looking to design a bandpass filter for which the passband is 20Hz to 20kHz. The filter will be constructed with discrete components on a PCB.

I've been researching some bandpass design methods and wanted some advice on which to choose. We'd like this filter to be fairly selective, so we're hoping for a filter of about the 4th order. The #1 concern, though, is that the passband is as flat as possible.

  1. One method would be to cascade 2 bandpass filters that are centered at my high and low cutoff frequencies. This sounds good, but it also looks like the passband's flatness becomes increasingly hard to maintain as the cutoff frequencies separate further.

  2. The other method that we're considering is just to cascade a low-pass and a high-pass filter (both 2nd- or higher-order) to achieve the bandpass effect. This sounds good to me too – and likely a lot simpler.
Any advice on which would work better? Or a third method altogether?

Thanks in advance!
 

crutschow

Joined Mar 14, 2008
34,201
I agree with Joey.
Separate low-pass and high-pass filters would be better for flatness.
For maximum flatness in the passband you want to use Butterworth filter types.
The free Filterpro program from Texas Instruments will help you design 4th order active filters.
 

Papabravo

Joined Feb 24, 2006
21,094
Notice that @crutschow suggested active filters, which are made with opamps, resistors and capacitors. This is because trying to design passive filters with inductors, and capacitors leads to unwieldy and potentially unrealizable component values at audio frequencies. My recommendation is to follow that suggestion.

You might also want to give some thought to how much attenuation you want in the stopband(s) and how steep you want the transition between the passband and the stopbands. Unreasonable expectations will increase the cost and complexity of any design.

If it were me, I would start with just the low pass filter with a corner frequency @ 22.05. This will match the typical performance of an audio CD @ 16-44.1 kHz.
 

AnalogKid

Joined Aug 1, 2013
10,971
Note that as the filter order increases, the phase distortions around the cutoff frequencies become worse and extend outward in frequency. A Bessel architecture will have a flat passband and less phase distortion, at the cost of a more rounded attenuation curve near the cutoff frequencies. Also, there are switched capacitor filter blocks that can get you very high order filter functions with way less work.

ak
 

Papabravo

Joined Feb 24, 2006
21,094
Of course once you have and analog low pass anti-aliasing filter on the front end there is all manner of digital magic you can perform with even a very modest processor.
 
Last edited:

Thread Starter

jchernak

Joined Jan 28, 2017
2
Thanks for all the quick and helpful responses!

Sounds like our gut was right – LP/HP is the way to go and that's what we'll do. Most likely Butterworth as well, especially since phase distortion is acceptable for our application.

And yes, I may have been misleading by saying "discrete components" but we were planning on designing active filters.

Thanks again!
 

ian field

Joined Oct 27, 2012
6,536
Hello all,

I'm on a project designing a microphone with built-in filtering. We are simply looking to filter out any inaudible frequencies we record. That is, we're looking to design a bandpass filter for which the passband is 20Hz to 20kHz. The filter will be constructed with discrete components on a PCB.

I've been researching some bandpass design methods and wanted some advice on which to choose. We'd like this filter to be fairly selective, so we're hoping for a filter of about the 4th order. The #1 concern, though, is that the passband is as flat as possible.

  1. One method would be to cascade 2 bandpass filters that are centered at my high and low cutoff frequencies. This sounds good, but it also looks like the passband's flatness becomes increasingly hard to maintain as the cutoff frequencies separate further.

  2. The other method that we're considering is just to cascade a low-pass and a high-pass filter (both 2nd- or higher-order) to achieve the bandpass effect. This sounds good to me too – and likely a lot simpler.
Any advice on which would work better? Or a third method altogether?

Thanks in advance!
Above audio frequencies are likely to be the biggest problem, most people just shunt the mic with a capacitor.

Most mics fall off rapidly at sub audio frequencies, but it depends on the type of mic and how important a sharp cut off is.

An op-amp gyrator may be the simplest solution - if you search the guitar pedal schematic archives; you can find examples of the same thing with a single transistor and handful of discretes.
 
Top