[Clam-devel] building 3 bands equalizer or using ladspa EQ

globot gglobot at gmail.com
Thu Aug 16 04:26:23 PDT 2007


>Message: 1
>Date: Tue, 14 Aug 2007 23:21:00 -0700
>From: Xavier Amatriain <xavier at amatriain.net>
>Subject: Re: [Clam-devel] building 3 bands equalizer
>To: clam-devel at llistes.projectes.lafarga.org
>Message-ID: <46C29B4C.7030405 at amatriain.net>
>Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
>(On holidays but still with cellphone internet :)
>
>Goblot, the 3BandFilter does (or should do) exactly what you ask for. If 
>you have glitches when you use gains different from 0 this might be due
>to the settings you have in the SFFT process. You have to make sure that 
>the window size is 4 times the hop size.
>
>  
>
Xavier, untile now i was just testing with FFT (rfftw, fftw3, ooura) but 
there is only the audioSize in configuration params (cf FFTConfig ) ... 
so how to define a hope for them ?
today i also tried with the SpectralAnalysis/Synthesis, but can't get 
what i want, When testing with a perfect low frequency sinusoidal wave, 
i got a Hi freq sound of windowsize length (in frame) that is added to 
my first wave. (i want realtime, so i tried to define small buff). Now i 
can feel my eers are tired of listenning hard noise :(

so how to just use a simple equalizer, that will add a gain according to 
the frequency to my sound in realtime (less than 2ms) ?

I have seen some ladspa EQ that might be good enought for me, but i 
can't make them as fast as i want, got arround 34ms for a 15 bands EQ ( 
cf: http://plugin.org.uk/ , multiband EQ) and around 17ms for the 8 
Bands EQ (my mesurment include mixing and amplification).
Is it possible to reduce the buffer size for ladspa plugins ? (something 
like 64 or 32 frames for a 96000Hz sample rate would be great).

PS: some ladspa plugin must be modified to work with NE, some have names 
containing '%d' (port num) that doesn't seem to be supported with NE, 
several port will have the same name. (i don't know much about ladspa 
standard, it might be a mistake)

PPS: about ladspa, i have seen no problem exepted the one i mension 
before. It look good 8)


>XA
>
>Pau Arumi wrote:
>  
>
>>En/na globot ha escrit:
>>    
>>
>>>hi,
>>>
>>>here is what i wanted to do :
>>>a simple equalizer with 3 section :
>>>   - from 0 to lowFreq, apply lowGain
>>>   - upper than highFreq, apply highGain
>>>   - between lowFreq and highFreq apply a progressive gain
>>>
>>>with lowFreq, highFreq user defined frequencies.
>>>with lowGain, highGain user defined gain in db (from 0 to -100)
>>>
>>>I was thinking that the threeBandFilter might have been usefull, for 
>>>this task. But each time i use it i have some very strange glitch 
>>>(look like some noise is generated when highGain, midGain or lowGain 
>>>are different from 0)
>>>
>>>i did my test with a generated perfect sinusoidal signal from 20 to 
>>>20000Hz (with a constant amplitude) using CLAM 1.1.0 (not svn)
>>>
>>>I probably miss used it, but can't figure out how to do. so can i 
>>>have some explaination about :
>>>-threeBandFilter
>>>-threeBandAM
>>>-vocoder
>>>-threebandcompressor
>>>
>>>i want to know what exactly do those class, and what are the 
>>>parameter (range of value, signification ...) because i am not realy 
>>>an expert in sound processing and now i got a little confuse.
>>>
>>>how can i make an equalizer ? with the vocoder ? (it look like it 
>>>generate the same kind of noise as the 3 band filter)
>>>
>>>thanks
>>>
>>>ps: it would be cool to add details in the doc too.
>>>      
>>>
>>Agree, they need some documentation.
>>I'm afraid Xavier, the author of these classes, is on holidays till
>>September. So ask again then in case you still have questions.
>>Till then I suggest a little of trial-and-error alternated with
>>code reading (and why not? post your findings)
>>
>>Cheers
>>Pau
>>
>>_______________________________________________
>>Clam-devel mailing list
>>Clam-devel at llistes.projectes.lafarga.org
>>https://llistes.projectes.lafarga.org/cgi-bin/mailman/listinfo/clam-devel
>>    
>>




More information about the clam-devel mailing list