[Porksoda]

Hey guys and gals,

So I am tearing apart my AV-710 to upgrade the power caps and put in a new output stage. I noticed in reading art of electronics that among hte filters it mention is the bessel filter, and that it exhibits nearly constant delay across the passband, or almost no phase distortion. I was wondering why this design is never implemented in any DIY projects around here. Other than the fact that it requires an op-amp, it seems like an absolutely ideal solution to the phase distortion problems associated with the traditional RC filter I see used here.

Does anybody know of any problems associated with Bessel filters to explain why they are not more common?

 

[jefemeister]

Bessel filters have very slow roll off which means that the filter order must be very high in order to acheive stopband attenuation on par wither other filter implementations. This would be less of a problem with high-upsampling rates since the mirror images would be placed very far out in the frequency spectrum by definition. It all goes back to the classic Time-Domain vs Frequency Domain design consideration: increased time-domain performance (what I, perhaps improperly, refer to as phase coherence) traded off with decreased frequency-performance (high-freq rolloff)

edit: more freq-performance issues: it's more than just high-freq roll off. There also ringing, ripple, etc...

 

[Glassman]

bessel filters are nothing special really, just one of a few filter types used in analog domain.. don't think it's something completely different or the like.. Bessel filters are really the best when it comes to phase coherency and the worst in terms of filter slope, I bet they are used in many audio designs.. probably the most widely used are Butterworth filters, which are a good compromise, I bet these are used by the 'measurement is everything' guys, can be most likely found on soundcards of all kinds and many audio components too, not the best for audio, nothing terrible either.. then there are Chebyshev filters, these have the steepest slope, but lots of passband ripple and phase performance is bad, not used anywhere in audio world, good for instrumentation.. there is no difference in topology amongst these filter types, just different formulas to get the appropriate component values..

 

[nikongod]

another fun point:getting caps or inductors of good quality that are of appropriate size can be dificult. its often dificult. you usually wind up putting a monsterous electrolytic cap in parallel with a not quite as huge film cap... the sound is better than an electrolytic cap, but not as good as a straight film cap.

you also have to contend with drivers that change impedance... and the fact that a filter for a low impedance headphone is nowehere near the same as one for a higher impedance phone.

build an active filter if you want to do that. value/slope changes are inexpensive, good quality caps in the sizes comonly used are CHEAP. the quality of the cap is not as absoutley important as in a passive system. the same active 18khz low pass 18db/octave filter with variable bass boost and switch selectable "boost threshold" can be used on your grado rs1's and your senn hd-650's perhaps even your stax if you set it up right. their only downside is that they need an amp. but you should have one anyways, and EVERYTHING will fit into a moderatly sized enclosure.