Hi,
I am searching for an adaptive masking method,
like pvstencil, but with gain modification only if the signal is
*above* the threshold set in the masking table.
I had thought I could do it with pvstencil, but keep failing to see how...
Any ideas?
I would like to use it to control the resulting amplitude when
convolving two signals and they accidentally coincide on the same
resonant frequencies.
I'd like to leave the signal unchanged whenever the two signals
contain (more or less) different spectra, and only apply filtering
when the signals have somewhat similar/matching spectra.

best
Oeyvind

Thanks Peiman, it looks like an interesting effect,
but I can't seem to make it work properly (not having the expected
effect on the audio output when adjusting the parameters)
Do you have some example parameter settings?
best
Oeyvind

2012/3/28 peiman khosravi :
> Hi Oeyvind,
>
> I have something like this in FFTools that mimics the GRM tools contrast
> plug-in. It basically devides the bins into three groups according to two
> threshold values. You can then do what you want to to any of the groups.
>
> Here is the CSD (the magic happens in the UDO). It's a bit messy and not
> that straightforward. Let me know if it doesn't make sense!
>
> Best,
>
> Peiman
>
>
>  
> 
>
> 
> 
>
>
> sr=96000
> ksmps=100
> nchnls=1
> 0dbfs = 1
>
>
>
>     opcode thresh, 0, kkkkkkiii
>
>
> klowgain,kmidgain,khighgain,kthreshlow, kthreshhigh, kband, inumbins,
> iampout, iampinsmooth xin
>
>
> iclear     ftgen 0, 0, inumbins, 2, 0
>     tablecopy iampout, iclear
>
> kcount = 0
> kvol = 0
>
> kbandfixed = kband
>
> loop:
> kamp     table kcount, iampinsmooth
>
> kvol = kvol + kamp
> kcount = kcount + 1
>
> if (kcount == kband) kgoto thresh
>         kgoto loop
>
> thresh:
>
>     kband2 = kband
>     kband = kband + kbandfixed
>     kvol = kvol / kbandfixed
>     if (kvol < kthreshlow) kgoto low
>         kgoto high
>
>     low:
>         kband3 = kband2-kbandfixed
>
>         vadd  iampout, klowgain, kbandfixed, kband3
>         kgoto contin
>
>     high:
>              ;printk2 kvol
>
>     kband3 = kband2-kbandfixed
>     if (kvol < kthreshhigh) then
>         vadd  iampout, kmidgain, kbandfixed, kband3
>
>     else
>         vadd  iampout, khighgain, kbandfixed, kband3
>
>     endif
>
>         kgoto contin
>
> contin:
> kvol = 0
>
>     if (kcount < inumbins) kgoto loop
>
> endop
>
> instr 1
>
> iampout ftgen 0,0,32769,2,0
> iampinsmooth ftgen 0,0,32769,2,0
>
>
> kfftsize     init 4096
> kol         init 4
> kwindmult     init 2
> kwindtype     init 0
> kBand         init 256
>
> kfftsize    chnget "fftsize"
> kol        chnget "ol"
> kwindmult    chnget "window"
> kwindtype    chnget "windtype"
> kBand        chnget "bands"
>
> ktrig           changed     kfftsize,kol, kwindmult, kwindtype, kBand
> if ktrig = 1 then
>                 reinit      pass
> endif
>
> pass:
>
> ifftsize =  i(kfftsize)
> iol =  i(kol)
> iNumBins = ifftsize/2 + 1
> ihop = ifftsize/iol
> iwindow = ifftsize*i(kwindmult)
> iwindtype = i(kwindtype)
> ibands = i(kBand)
> ;
>
> if (ihop >= ifftsize) then
> ihop          =           ihop / 2
> endif
>
> if (ihop < 16) then
> ihop          =           16
> endif
>
>
> kdepth    chnget "depth"
> kdepth    port kdepth, 0.01
>
> kaver chnget    "aver"
> ksmooth chnget "smooth"
>
> kthreshHigh chnget "threshH"
> kthreshLow chnget "threshL"
>
> khighGain chnget "gainH"
> khighGain    port khighGain, 0.01
> khighGain = khighGain+0.000000001
>
> kmidGain chnget "gainM"
> kmidGain    port kmidGain, 0.01
> kmidGain = kmidGain+0.000000001
>
> klowGain chnget "gainL"
> klowGain    port klowGain, 0.01
> klowGain = klowGain+0.000000001
>
>
>
> ain1    in
>
> fsig1 pvsanal ain1, ifftsize, ihop, iwindow, iwindtype
>
> fsigblur        pvsblur fsig1, ,0, 2
>
> kflag pvsftw  fsigblur, iampinsmooth
>
> if (kflag == 1) then ; only proc when frame is ready
>
>     thresh  klowGain,kmidGain,khighGain,kthreshLow, kthreshHigh, ibands,
> iNumBins, iampout, iampinsmooth
>
>     pvsftr fsigblur, iampout
>
> endif
>
> fsigsmooth pvsmooth fsigblur, ksmooth, ksmooth
>
> fsigout pvsfilter fsig1, fsigsmooth, kdepth
>
>
> aout1 pvsynth fsigout
>
>
> out aout1
>
> endin
>
>
>
>
> 
> 
>
> i1 0 36000
>
> e
>
> 
> 
>
>
> On 27 March 2012 23:03, Oeyvind Brandtsegg 
> wrote:
>>
>> Hi,
>> I am searching for an adaptive masking method,
>> like pvstencil, but with gain modification only if the signal is
>> *above* the threshold set in the masking table.
>> I had thought I could do it with pvstencil, but keep failing to see how...
>> Any ideas?
>> I would like to use it to control the resulting amplitude when
>> convolving two signals and they accidentally coincide on the same
>> resonant frequencies.
>> I'd like to leave the signal unchanged whenever the two signals
>> contain (more or less) different spectra, and only apply filtering
>> when the signals have somewhat similar/matching spectra.
>>
>> best
>> Oeyvind
>>
>>
>> Send bugs reports to the Sourceforge bug tracker
>>            https://sourceforge.net/tracker/?group_id=81968&atid=564599
>> Discussions of bugs and features can be posted here
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe
>> csound"
>>
>



-- 

Oeyvind Brandtsegg
Professor of Music Technology
NTNU
7491 Trondheim
Norway
Cell: +47 92 203 205

http://flyndresang.no/
http://www.partikkelaudio.com/
http://soundcloud.com/brandtsegg
http://soundcloud.com/t-emp

Thanks Peiman and Tito for the nice suggestions,
I can adapt these ideas to fit my purposes.
best
Oeyvind

2012/3/29 peiman khosravi :
> Hi Oeyvind,
>
> Try these settings. I've also commented the parameters to make some sense.
>
> Best,
>
> Peiman
>
>
> 
> 
> -odac
> 
> 
>
>
> sr=44100
>
> ksmps=100
> nchnls=1
> 0dbfs = 1
>
>
>
>     opcode thresh, 0, kkkkkkiii
>
>
> klowgain,kmidgain,khighgain,kthreshlow, kthreshhigh, kband, inumbins,
> iampout, iampinsmooth xin
>
>
> iclear     ftgen 0, 0, inumbins, 2, 0
>     tablecopy iampout, iclear
>
> kcount = 0
> kvol = 0
>
> kbandfixed = kband
>
> loop:
> kamp     table kcount, iampinsmooth
>
> kvol = kvol + kamp
> kcount = kcount + 1
>
> if (kcount == kband) kgoto thresh
>         kgoto loop
>
> thresh:
>
>     kband2 = kband
>     kband = kband + kbandfixed
>     kvol = kvol / kbandfixed
>     if (kvol < kthreshlow) kgoto low
>         kgoto high
>
>     low:
>         kband3 = kband2-kbandfixed
>
>         vadd  iampout, klowgain, kbandfixed, kband3
>         kgoto contin
>
>     high:
>              ;printk2 kvol
>
>     kband3 = kband2-kbandfixed
>     if (kvol < kthreshhigh) then
>         vadd  iampout, kmidgain, kbandfixed, kband3
>
>     else
>         vadd  iampout, khighgain, kbandfixed, kband3
>
>     endif
>
>         kgoto contin
>
> contin:
> kvol = 0
>
>     if (kcount < inumbins) kgoto loop
>
> endop
>
> instr 1
>
> Sfile = "/Applications/Max5/examples/sounds/cherokee.aif"
>
> kBand         init     1    ;number of frequency bands into which the fft
> bins are devided. 1 means one bin per band.
>
> kdepth        init     1     ;0 is 100 dry signal
> kaver         init    0    ;time in seconds for averaging the spectral
> changes in the analysed signal. Gives a smoother result
> ksmooth     init    1    ;smoothing the dynamic changes applied to the input
> signal
> kthreshHigh     init    .05     ;value for the upper amplitude threshold
> kthreshLow     init    .03    ;value for the lower amplitude threshold
>
> ;This setting will zero all the bins above kmidGain
> khighGain     init    1    ;gain applied to bins whose amplitude falls above
> kthreshHigh
> kmidGain     init    0    ;gain applied to bins whose amplitude falls below
> kthreshHigh and above kthreshLow
> klowGain     init    0    ;gain applied to bins whose amplitude falls below
> kthreshLow
>
>
>
>
> kfftsize     init 4096
> kol         init 4
> kwindmult     init 2
> kwindtype     init 0
>
> kfftsize    chnget "fftsize"
> kol        chnget "ol"
> kwindmult    chnget "window"
> kwindtype    chnget "windtype"
> kBand        chnget "bands"
>
>
> iampout ftgen 0,0,32769,2,0
> iampinsmooth ftgen 0,0,32769,2,0
>
> ktrig           changed     kfftsize,kol, kwindmult, kwindtype, kBand
> if ktrig = 1 then
>                 reinit      pass
> endif
>
> pass:
>
> ifftsize =  i(kfftsize)
> iol =  i(kol)
> iNumBins = ifftsize/2 + 1
> ihop = ifftsize/iol
> iwindow = ifftsize*i(kwindmult)
> iwindtype = i(kwindtype)
> ibands = i(kBand)
> ;
>
> if (ihop >= ifftsize) then
> ihop          =           ihop / 2
> endif
>
> if (ihop < 16) then
> ihop          =           16
> endif
>
> ;avoide denormalisation
> khighGain = khighGain+0.000000001
> kmidGain = kmidGain+0.000000001
> klowGain = klowGain+0.000000001
>
>
>
> ain1    diskin     Sfile, 1, 0, 1
>
>
> fsig1 pvsanal ain1, ifftsize, ihop, iwindow, iwindtype
>
> fsig1copy    pvsmix    fsig1, fsig1
> ;fsigblur pvsmooth fsig1, ksmooth, ksmooth
>
> kflag pvsftw  fsig1copy, iampinsmooth
>
>
> if (kflag == 1) then ; only proc when frame is ready
>
>     thresh  klowGain,kmidGain,khighGain,kthreshLow, kthreshHigh, ibands,
> iNumBins, iampout, iampinsmooth
>
>     pvsftr fsig1copy, iampout
>
> endif
>
> fsigsmooth pvsmooth fsig1copy, ksmooth, 1
>
>
> fsigout pvsfilter fsig1, fsigsmooth, kdepth
>
>
> aout1 pvsynth fsigout
>
>
> out aout1
>
> endin
>
>
>
>
> 
> 
>
> i1 0 36000
>
> e
>
> 
> 
>



-- 

Oeyvind Brandtsegg
Professor of Music Technology
NTNU
7491 Trondheim
Norway
Cell: +47 92 203 205

http://flyndresang.no/
http://www.partikkelaudio.com/
http://soundcloud.com/brandtsegg
http://soundcloud.com/t-emp

got it,
thanks :-)


2012/3/30 Tito Latini :
> Sorry, there is an error in my example; it is correct
>
> fsig  pvsgain   fsrc, kgain
> fres  pvstencil fsig, 1/kgain, kgain, 1
> aout  pvsynth   fres
>
> tito
>
>
> Send bugs reports to the Sourceforge bug tracker
>            https://sourceforge.net/tracker/?group_id=81968&atid=564599
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>



-- 

Oeyvind Brandtsegg
Professor of Music Technology
NTNU
7491 Trondheim
Norway
Cell: +47 92 203 205

http://flyndresang.no/
http://www.partikkelaudio.com/
http://soundcloud.com/brandtsegg
http://soundcloud.com/t-emp

Hi Emil,

I did not look at the code in detail, but I did notice:

idur	= 	p3

If you are using tied notes for legato then you will need to change that to:

idur = abs(p3)

because p3 will be negative.  That would certainly cause issues as
idur is used as-is elsewhere in that instrument.

As for flute sounds, have you looked at Lee Zakian's flute instruments?

http://www.jlpublishing.com/Csound.htm

The zip for Transversal Meditiation has orc/sco files. I've always
enjoyed the sounds in that piece.

steven



On Wed, Mar 28, 2012 at 8:09 AM, Emil Barton  wrote:
>
> Hi,
>
> I've included the Mykelson-Cook's slide flute in my program because it's a beautiful instrument that represents well the power of Csound in my opinion, (I hope I didn't break any copyright, I didn't check that, it probably comes from the Csound book, I'm not sure).
>
> However I've not been able to work out a fast legato with this instrument. One can see that in the little piece I sculpted in my program's output (cf.http://www.emilbarton.net76.net/data/ballad.csd) : as soon as I tried to go fast it got messy and I stopped because the goal was to illustrate the program, not really to write a ballad.
>
> Does anyone know how to get the best ratio in parameters tuning for legato purpose in this difficult instrument?
>
> Thank you in advance,
>
> Emil
>
>

Emil,

You've surely already considered this, but just in case:
does your algorithm for legato notes -- i.e., the notes
that are to be connected *to* -- omit the attack phase?

Peter


Steven Yi wrote:
> Hi Emil,
>
> I did not look at the code in detail, but I did notice:
>
> idur	= 	p3
>
> If you are using tied notes for legato then you will need to change that to:
>
> idur = abs(p3)
>
> because p3 will be negative.  That would certainly cause issues as
> idur is used as-is elsewhere in that instrument.
>
> As for flute sounds, have you looked at Lee Zakian's flute instruments?
>
> http://www.jlpublishing.com/Csound.htm
>
> The zip for Transversal Meditiation has orc/sco files. I've always
> enjoyed the sounds in that piece.
>
> steven
>
>
>
> On Wed, Mar 28, 2012 at 8:09 AM, Emil Barton  wrote:
>>
>> Hi,
>>
>> I've included the Mykelson-Cook's slide flute in my program because it's a beautiful instrument that represents well the power of Csound in my opinion, (I hope I didn't break any copyright, I didn't check that, it probably comes from the Csound book, I'm not sure).
>>
>> However I've not been able to work out a fast legato with this instrument. One can see that in the little piece I sculpted in my program's output (cf.http://www.emilbarton.net76.net/data/ballad.csd) : as soon as I tried to go fast it got messy and I stopped because the goal was to illustrate the program, not really to write a ballad.
>>
>> Does anyone know how to get the best ratio in parameters tuning for legato purpose in this difficult instrument?
>>
>> Thank you in advance,
>>
>> Emil
>>
>>
>
>
> Send bugs reports to the Sourceforge bug tracker
>              https://sourceforge.net/tracker/?group_id=81968&atid=564599
> Discussions of bugs and features can be posted here
> To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"
>
>
>

P.S.  I suppose that's what tied notes are for, but I
         don't know how well they "do" legato per se.

PMA wrote:
> Emil,
>
> You've surely already considered this, but just in case:
> does your algorithm for legato notes -- i.e., the notes
> that are to be connected *to* -- omit the attack phase?
>
> Peter
>
>
> Steven Yi wrote:
>> Hi Emil,
>>
>> I did not look at the code in detail, but I did notice:
>>
>> idur = p3
>>
>> If you are using tied notes for legato then you will need to change
>> that to:
>>
>> idur = abs(p3)
>>
>> because p3 will be negative. That would certainly cause issues as
>> idur is used as-is elsewhere in that instrument.
>>
>> As for flute sounds, have you looked at Lee Zakian's flute instruments?
>>
>> http://www.jlpublishing.com/Csound.htm
>>
>> The zip for Transversal Meditiation has orc/sco files. I've always
>> enjoyed the sounds in that piece.
>>
>> steven
>>
>>
>>
>> On Wed, Mar 28, 2012 at 8:09 AM, Emil Barton wrote:
>>>
>>> Hi,
>>>
>>> I've included the Mykelson-Cook's slide flute in my program because
>>> it's a beautiful instrument that represents well the power of Csound
>>> in my opinion, (I hope I didn't break any copyright, I didn't check
>>> that, it probably comes from the Csound book, I'm not sure).
>>>
>>> However I've not been able to work out a fast legato with this
>>> instrument. One can see that in the little piece I sculpted in my
>>> program's output (cf.http://www.emilbarton.net76.net/data/ballad.csd)
>>> : as soon as I tried to go fast it got messy and I stopped because
>>> the goal was to illustrate the program, not really to write a ballad.
>>>
>>> Does anyone know how to get the best ratio in parameters tuning for
>>> legato purpose in this difficult instrument?
>>>
>>> Thank you in advance,
>>>
>>> Emil
>>>
>>>
>>
>>
>> Send bugs reports to the Sourceforge bug tracker
>> https://sourceforge.net/tracker/?group_id=81968&atid=564599
>> Discussions of bugs and features can be posted here
>> To unsubscribe, send email sympa@lists.bath.ac.uk with body
>> "unsubscribe csound"
>>
>>
>>