SoX - Sound eXchange

--- On Mon, 19/1/09, Klaus Schulz <kls.schlz@...> wrote:
> I did understand that a 2 pole filter if sox
> highpass/lowpass is used
> generates a 12db/oct slope. e.g.
> 
> highpass 30 ( default 2-pole)
> 
> If I want a much sharper response:
> 
> Would I translate it into something like this example:
> 
> filter 30-0 256 1

Yes, something like this. Use --plot to select gnuplot or octave/matlab plotting of the response, then you can fine tune the parameters to get the slope you want.

> I don't know how Windowsize and Beta really affect the
> filter slope. The
> man-page is not really clear on that.

Basically Windowsize affects slope and Beta affects stop-band level.  Use --plot to check.

> Two more Qs:
> Are there phase linear filters available?

`filter' is linear phase

> How about filter
> types
> Butterworth,Linkwitz etc.?

`highpass' and `lowpass' default to Butterworth (non-linear phase).  You can cascade two Butterworth filters to make a Linkwitz-Riley filter.

Cheers,
Rob

That's seems to me a trial and error exercise. Never used Octave. Is this
available under Linux? Thats sounds to me like a project on its own. ;)

I read today that FIR filters are behind "Filter". How many taps are you
using?

THX

On Mon, Jan 19, 2009 at 5:38 PM, robs <aquegg@...> wrote:

> --- On Mon, 19/1/09, Klaus Schulz <kls.schlz@...> wrote:
> > I did understand that a 2 pole filter if sox
> > highpass/lowpass is used
> > generates a 12db/oct slope. e.g.
> >
> > highpass 30 ( default 2-pole)
> >
> > If I want a much sharper response:
> >
> > Would I translate it into something like this example:
> >
> > filter 30-0 256 1
>
> Yes, something like this. Use --plot to select gnuplot or octave/matlab
> plotting of the response, then you can fine tune the parameters to get the
> slope you want.
>
> > I don't know how Windowsize and Beta really affect the
> > filter slope. The
> > man-page is not really clear on that.
>
> Basically Windowsize affects slope and Beta affects stop-band level.  Use
> --plot to check.
>
> > Two more Qs:
> > Are there phase linear filters available?
>
> `filter' is linear phase
>
> > How about filter
> > types
> > Butterworth,Linkwitz etc.?
>
> `highpass' and `lowpass' default to Butterworth (non-linear phase).  You
> can cascade two Butterworth filters to make a Linkwitz-Riley filter.
>
> Cheers,
> Rob
>
>
>
>
>

--- On Mon, 19/1/09, Klaus Schulz <kls.schlz@...> wrote:
> That's seems to me a trial and error exercise.

Yes, it's not great, but many people have found it good enough so far.

FYI, there is a new more user-friendly `filter' effect currently under development; it will also have more features such as minimum phase. 

> Never
> used Octave. Is this
> available under Linux?

Yes, most distros should have it.

> I read today that FIR filters are behind
> "Filter". How many taps are you
> using?

The windowsize parameter is basically the number of taps (or half that number, can't recall offhand).

/Rob

Hi Rob.

Is there an example how to generate that response and get it displayed by
octave?

I need to 1. generate a sweep  2. run it over the filter  3. plot it into
file  and than  4. display it with octave.

It would be very kind of you if you could show directions. That'll save me
probably two days of research.

THX a lot.

Klaus



On Mon, Jan 19, 2009 at 6:41 PM, robs <aquegg@...> wrote:

> --- On Mon, 19/1/09, Klaus Schulz <kls.schlz@...> wrote:
> > That's seems to me a trial and error exercise.
>
> Yes, it's not great, but many people have found it good enough so far.
>
> FYI, there is a new more user-friendly `filter' effect currently under
> development; it will also have more features such as minimum phase.
>
> > Never
> > used Octave. Is this
> > available under Linux?
>
> Yes, most distros should have it.
>
> > I read today that FIR filters are behind
> > "Filter". How many taps are you
> > using?
>
> The windowsize parameter is basically the number of taps (or half that
> number, can't recall offhand).
>
> /Rob
>
>
>
>

On Tue, Jan 20, 2009 at 4:04 AM, Klaus Schulz <kls.schlz@...> wrote:
> Hi Rob.
>
> Is there an example how to generate that response and get it displayed by
> octave?
>
> I need to 1. generate a sweep  2. run it over the filter  3. plot it into
> file  and than  4. display it with octave.
>
> It would be very kind of you if you could show directions. That'll save me
> probably two days of research.


Not sure what you mean by sweep, but i guess the rest would go along
those lines:
sox --plot octave /tmp/1.ogg -n highpass 1320 > plot.m
octave plot.m

-Pascal
-- 
Homepage (http://organact.mine.nu)
Debian GNU/Linux (http://www.debian.org)
LACIME: École de technologie supérieure (http://lacime.etsmtl.ca)

Hi there.

What's in /tmp/1.ogg? A frequency sweep? I was thinking to run a sweep from
10Hz to 20khz over the filter, which would then end up in the filter
response. Am I wrong here?

Cheers

On Tue, Jan 20, 2009 at 3:41 PM, Pascal Giard <evilynux@...> wrote:

> On Tue, Jan 20, 2009 at 4:04 AM, Klaus Schulz <kls.schlz@...>
> wrote:
> > Hi Rob.
> >
> > Is there an example how to generate that response and get it displayed by
> > octave?
> >
> > I need to 1. generate a sweep  2. run it over the filter  3. plot it into
> > file  and than  4. display it with octave.
> >
> > It would be very kind of you if you could show directions. That'll save
> me
> > probably two days of research.
>
>
> Not sure what you mean by sweep, but i guess the rest would go along
> those lines:
> sox --plot octave /tmp/1.ogg -n highpass 1320 > plot.m
> octave plot.m
>
> -Pascal
> --
> Homepage (http://organact.mine.nu)
> Debian GNU/Linux (http://www.debian.org)
> LACIME: École de technologie supérieure (http://lacime.etsmtl.ca)
>
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by:
> SourcForge Community
> SourceForge wants to tell your story.
> http://p.sf.net/sfu/sf-spreadtheword
> _______________________________________________
> Sox-users mailing list
> Sox-users@...
> https://lists.sourceforge.net/lists/listinfo/sox-users
>

On Tue, Jan 20, 2009 at 10:16 AM, Klaus Schulz <kls.schlz@...> wrote:
> Hi there.
>
> What's in /tmp/1.ogg? A frequency sweep? I was thinking to run a sweep from
> 10Hz to 20khz over the filter, which would then end up in the filter
> response. Am I wrong here?

Ah! got you for the sweep.
Here's how to create a 3 second-long sweep from 10Hz to 20kHz saved in
/tmp/out.wav:

sox -n /tmp/out.wav synth 3 sine 10-20000

My /tmp/1.ogg could pretty much be anything. If i understand correctly
the ploted result won't vary with the input as... (beware, i might
translate this wrong) it plots the unit step response of the transfer
function.

Also note that only the first effect supporting plotting will be taken
into account (in the plot).
e.g. if i do something like: "sox --plot octave /tmp/out.wav -n
equalizer 60 0.5q -6 equalizer 1000 0.5q 10 > /tmp/plot.m", only the
first equalization filter will be shown.

HTH,

-Pascal
-- 
Homepage (http://organact.mine.nu)
Debian GNU/Linux (http://www.debian.org)
LACIME: École de technologie supérieure (http://lacime.etsmtl.ca)

I figured the "sweep" out 2 minutes ago. ;) THX

Do you manage to get the X/Y dimensions of the octave  graph adjusted?

THX
Klaus

On Tue, Jan 20, 2009 at 4:38 PM, Pascal Giard <evilynux@...> wrote:

> On Tue, Jan 20, 2009 at 10:16 AM, Klaus Schulz <kls.schlz@...>
> wrote:
> > Hi there.
> >
> > What's in /tmp/1.ogg? A frequency sweep? I was thinking to run a sweep
> from
> > 10Hz to 20khz over the filter, which would then end up in the filter
> > response. Am I wrong here?
>
> Ah! got you for the sweep.
> Here's how to create a 3 second-long sweep from 10Hz to 20kHz saved in
> /tmp/out.wav:
>
> sox -n /tmp/out.wav synth 3 sine 10-20000
>
> My /tmp/1.ogg could pretty much be anything. If i understand correctly
> the ploted result won't vary with the input as... (beware, i might
> translate this wrong) it plots the unit step response of the transfer
> function.
>
> Also note that only the first effect supporting plotting will be taken
> into account (in the plot).
> e.g. if i do something like: "sox --plot octave /tmp/out.wav -n
> equalizer 60 0.5q -6 equalizer 1000 0.5q 10 > /tmp/plot.m", only the
> first equalization filter will be shown.
>
> HTH,
>
> -Pascal
> --
> Homepage (http://organact.mine.nu)
> Debian GNU/Linux (http://www.debian.org)
> LACIME: École de technologie supérieure (http://lacime.etsmtl.ca)
>
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by:
> SourcForge Community
> SourceForge wants to tell your story.
> http://p.sf.net/sfu/sf-spreadtheword
> _______________________________________________
> Sox-users mailing list
> Sox-users@...
> https://lists.sourceforge.net/lists/listinfo/sox-users
>

On Tue, Jan 20, 2009 at 10:46 AM, Klaus Schulz <kls.schlz@...> wrote:
> I figured the "sweep" out 2 minutes ago. ;) THX
>
> Do you manage to get the X/Y dimensions of the octave  graph adjusted?

I haven't played around with this...
Altho i expect sox not to provide much options to the plot function...

Instead i expect to have to play around with the generated plot.m file.

-Pascal
-- 
Homepage (http://organact.mine.nu)
Debian GNU/Linux (http://www.debian.org)
LACIME: École de technologie supérieure (http://lacime.etsmtl.ca)

Hi there.

I tried below: The filter function failed. What am I doing wrong here?

#OK
sox -n output.au synth 20 sine 10-20000
sox --plot octave -n output.au  highpass 80 > plot.m
octave plot.m

#NOK!!!!!!
sox -n output.au synth 20 sine 10-20000
sox --plot octave -n output.au  filter 80-0 256 1 > plot.m
octave plot.m

Error-MESSAGE from sox --plot:
sox filter: filter: adjusted freq 20-24000 is identity

plot-m is not usable. Perhaps a problem with "filter" > octave:

#COMMENT:

I am wondering if it wouldn't be better to generate .png plots with Octave.
This way it is much easier to save and compare results.
Using e.g. display of e.g. imagemagick could easily display even a set of
simulations


THX
Klaus

On Tue, Jan 20, 2009 at 5:03 PM, Pascal Giard <evilynux@...> wrote:

> On Tue, Jan 20, 2009 at 10:46 AM, Klaus Schulz <kls.schlz@...>
> wrote:
> > I figured the "sweep" out 2 minutes ago. ;) THX
> >
> > Do you manage to get the X/Y dimensions of the octave  graph adjusted?
>
> I haven't played around with this...
> Altho i expect sox not to provide much options to the plot function...
>
> Instead i expect to have to play around with the generated plot.m file.
>
> -Pascal
> --
> Homepage (http://organact.mine.nu)
> Debian GNU/Linux (http://www.debian.org)
> LACIME: École de technologie supérieure (http://lacime.etsmtl.ca)
>
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by:
> SourcForge Community
> SourceForge wants to tell your story.
> http://p.sf.net/sfu/sf-spreadtheword
> _______________________________________________
> Sox-users mailing list
> Sox-users@...
> https://lists.sourceforge.net/lists/listinfo/sox-users
>

On Tue, Jan 20, 2009 at 12:24 PM, Klaus Schulz <kls.schlz@...> wrote:
[...]
> I tried below: The filter function failed. What am I doing wrong here?
[...]
> #NOK!!!!!!
> sox -n output.au synth 20 sine 10-20000
> sox --plot octave -n output.au  filter 80-0 256 1 > plot.m
> octave plot.m
>
> Error-MESSAGE from sox --plot:
> sox filter: filter: adjusted freq 20-24000 is identity

Hmmm... i don't know what the problem is.
>From the manpage it seems to me like this should be working...

-Pascal
-- 
Homepage (http://organact.mine.nu)
Debian GNU/Linux (http://www.debian.org)
LACIME: École de technologie supérieure (http://lacime.etsmtl.ca)

--- On Tue, 20/1/09, Pascal Giard <evilynux@...> wrote:
> > I tried below: The filter function failed. What am I
> doing wrong here?
> [...]
> > #NOK!!!!!!
> > sox -n output.au synth 20 sine 10-20000
> > sox --plot octave -n output.au  filter 80-0 256 1 >
> plot.m
> > octave plot.m
> >
> > Error-MESSAGE from sox --plot:
> > sox filter: filter: adjusted freq 20-24000 is identity

The recommended way to plot this would be as follows:

  sox --plot octave -r 8k -n -n filter 80- 256 1 > filter.plt
  octave filter.plt

You need neither input nor output file to calculate a filter & its response (hence the -n's).  However you do need a sampling-rate to specify the filter (hence the -r).

The above commands work and plot the response, however, if you increase the sampling rate or decrease the stop-band you hit the limitations of the current filter effect (the new one when it's ready won't have these limitations).

This also works:

  sox --plot octave -r 44.1k -n -n filter 250- 256 1 > filter.plt

but maybe 250 is too high for you.

BTW, don't ever create a file called plot.m; since plot is an octave/matlab keyword, it will crash and burn.
  
Sorry, best we can do for now.

Cheers,
Rob

On Tue, Jan 20, 2009 at 3:04 PM, robs <aquegg@...> wrote:
[...]
> BTW, don't ever create a file called plot.m; since plot is an octave/matlab keyword, it will crash and burn.

Oh damn! I had taken this filename example in the manpage. Please
rename it if it's a bad idea...

Cheers,

-Pascal
-- 
Homepage (http://organact.mine.nu)
Debian GNU/Linux (http://www.debian.org)
LACIME: École de technologie supérieure (http://lacime.etsmtl.ca)