AudioMasters
 
 User Info & Key Stats   
Welcome, Guest. Please login or register.

Login with username, password and session length
July 30, 2010, 11:37:24 AM
70766 Posts in 7409 Topics by 2232 Members
Latest Member: Beriganfan
News:       Buy Adobe Audition:
+  AudioMasters
|-+  Audio Related
| |-+  General Audio
| | |-+  Band-pass filter
  « previous next »
Pages: [1] Print
Author
Topic: Band-pass filter  (Read 1141 times)
« on: December 03, 2009, 02:12:21 PM »
jambalaya Offline
New Member
*
Posts: 2



I'd really like to have an audio processing software with many specific functions. For example in this case I want to select a very small range of frequency (say, 150 to 152 Hz) and lower all the ones outside the range. Do you have any suggestions?
Logged
Reply #1
« on: December 03, 2009, 03:18:50 PM »
SteveG Offline
Administrator
Member
*****
Posts: 9584



Do you have any suggestions?

Yeah. Learn a bit more about what you're doing, for a start. If you could actually achieve a filter that narrow, the only thing you'd get out of it was a sine wave which was either there or not there, depending on whether that particular part of the spectrum was excited or not. So pretty much useless for any practical purposes. Can Audition do this as it stands? Probably yes, using the parametric EQ and high Q values. But why?
Logged

Reply #2
« on: December 03, 2009, 06:16:03 PM »
oretez Offline
Member
*****
Posts: 654



probably one of the easiest ways to test some of what Steve's talking about is with the FFT filter

There can be all sorts of reasons for manipulating audio files that have little to do with conventional processing  but as Steve indicated with that narrow a band it is kind of a crap shoot whether you get anything resembling your expectations . . . steep (and/or narrow) filters can introduce artifacts that obviate the thing trying to be accomplished  there is  a reason why in pre digital days steep filters intended for conventionally musical results tended to employ hybrid strategies (e.g. Butterworth + Chebychev) less precise but fewer unwanted artifacts

threw in some pics . . . they do not necessarily demonstrate anything (other then playfulness) but hey

first two use a single FFT bandpass centered @ 152 with 149 low and 156 high reduction is only 30 dB and while the audio was effected you could still tell, from just a casual listen, what the tune was.  third pic employed a bessel band pass on the same material with a 36 dB cut, fourth file used a Chebychev with 96 dB cut.  Without extreme make up gain (amplifying something somewhere), with vertical slope and cut outside the band down to noise floor  you had you still had significant 'information', @ -30 dB across a wide spectrum (with regard to narrowness of the band) 100-500 Hz.  You don't, necessarily 'hear' anything (except an occasional blip) until you apply make up gain then you hear the system self noise

in any case though there are tools in AA that let you play around with steep narrow filters. 
Logged
Reply #3
« on: December 03, 2009, 08:10:15 PM »
SteveG Offline
Administrator
Member
*****
Posts: 9584



That's the problem - it really doesn't work too well, simply because awful things happen. Filters either ring like mad because of the steep skirts, so you get no results that are recognisable as what you wanted or, if you try to do this with filters that won't ring, you don't get anything like the attenuation you thought. If you try it with noise, and actually measure the real attenuation you get with a 3Hz bandwidth, it's very small whatever the system was set to - even if you try to specify something alarming, like -80dB. Or, if you try it with a properly recursive filter like the FFT, you end up with a boost, not a cut (the ringing skirt issue).
Logged

Reply #4
« on: December 06, 2009, 06:59:04 PM »
jambalaya Offline
New Member
*
Posts: 2



Yeah I see.. Actually, it's obvious that with such a narrow range I'll get nothing useful. I should have thought of it just one second after putting myself that question! So sorry  grin
I was wondering if I could get to isolate different voices in a recording, but I think I can only approximately divide men's from women's voices, and that's it..

Still, if it's not that bad going a little OT.... I have a song, with a sort of intro and then a powerful attack with distorted guitars, bass, and imposing strings.. What do I have to concentrate on to get a strong and very powerful attack? I mean, for example, each instrument has to start playing at the same moment, "superbassed" kick, bass sound rich in frequencies about 40-100 Hz, etc...

Thanks for everything and sorry for my English  smiley
Logged
Pages: [1] Print 
« previous next »
Jump to:  

Powered by MySQL Powered by PHP Valid XHTML 1.0! Valid CSS! Ig-Oh Theme by koni.