Butterworth vs. Bessel vs. 2nd order...

[squidbait]

I realize there are books written entirely on this subject, but I was hoping to get a brief explanation of how to make the choice in which type of crossover to use. I understand that second order rolls off differently than 4th order. But how do I decide which order I want to use, and which type I want to use?

[sreten]

Your right it needs a book......

But first off you decide what acoustic target slopes you want,
Linkwitz / Riley 2nd order or fourth order are a good as any.

But you may have phase problems due to driver offset so
you can go asymmetrical, 1st with 2nd, 2nd with 3rd etc.

You may also need to consider baffle step compensation and EQing
response anomalies in the drivers to arrive at your target acoustic
response.

sreten.

[azira]

In short, with filters you have 3 key parameters that you have to work with and trade off.

First: the "cutoff" frequency. This is the frequency that your filter operates around, if it's a lowpass filter then frequencies below this point are allowed to pass (hopefully at full strength not more or less) and frequencies above this point are attenuated (ideally as much as possible).

Second: A roll-off characteristic, this is how quickly do you want frequencies above the cutoff to be attenuated. This generally relates to the order of the filter. 1st order is -3db/octave 2nd order is -6db/octave etc..

Third: How much ripple you are willing to tolerate in the pass band. The pass band is the frequency range that is allowed to pass in your filter. This parameter has to trade off with the second because one characteristic of analog filters is that the faster the roll-off, the more ripple or peaking there is near the cutoff.


Butterworth filters are a classification of filters with special mathematical properties. Their characterisitic is that the passband has a maximally flat response. This is a good trait. The downside is that it doesn't roll-off as fast as other types of filters

Chebychev filters are another type of filter also with special mathematical properties. These filters tend to have a fairly steep initial rolloff with the "ringing" (peaking + ripple) as a controllable parameter.

Bessel filters are more of a filter response rather than a filter type. The good characteristic of a bessel type filter is that it has a constant group delay or phase shift for all frequencies. This makes it harder to design but it does not exhibit any ringing for a step response. That is a very good trait for an analog filter especially for digital logic.

[CeramicMan]

Originally posted by azira
...Bessel filters are more of a filter response rather than a filter type. The good characteristic of a bessel type filter is that it has a constant group delay or phase shift for all frequencies. This makes it harder to design but it does not exhibit any ringing for a step response. That is a very good trait for an analog filter especially for digital logic.

The group delay isn't constant across all frequencies, but rather, the group delay of a Bessel filter is a step without any peak. Importantly, a Bessel filter has the sharpest possible filtering without any ringing in the time domain (ie: no resonance). A compromise type of filter is a Gaussian filter, which has a Bessel response in the pass-band and up to a certain point past the cut-off frequency. Beyond that point the roll-off is sharpened, thereby improving signal rejection, but also producing ringing. However, that ringing is already attenuated and is therefore less noticeable than with other filters such as Butterworth or Chebyshev filters.

CM

[Bill Fitzpatrick]

Originally posted by azira


Second: A roll-off characteristic, this is how quickly do you want frequencies above the cutoff to be attenuated. This generally relates to the order of the filter. 1st order is -3db/octave 2nd order is -6db/octave etc..

Ahem. You wan't to try that again?

[azira]

Originally posted by Bill Fitzpatrick


Ahem. You wan't to try that again?

Heh yeah I do, geez bad mistake too.

Sorry, your first order filters are -6db/oct and your 2nd order filters are -12db/oct and ...

[pooge]

Bessel filters have constant group delay only for low pass versions. High pass Bessel filters do not have constant group delay for all frequencies.

[jackinnj]

when I first downloaded FilterPro from Texas Instruments it was only Low Pass, now it has High Pass, you can also plug in different Q's -- so if you have an older version you might want to upgrade. The Application Report also gives some tips on the minimum GBW for the opamps in use.