audacity-devel

[Audacity-devel] Classic Filters status

[Norm C <n_b...@ho...>]

I've been out of touch for a few months but I gather there's been some rather
polarized discussions regarding the Classic Filters and they're now out of
the build. In my view that's unfortunate. I'll summarize what I see as the
issues:

- badly named: I agree, that's why I called it Scientific Filters when I
submitted it. There's ample justification for this name:
CoolEditPro/Audition calls the same effect Scientific Filters, and Matlab
(certainly a major scientific application) uses these same filters (butter,
cheby1 and cheby2) as its workhorse filters
- the filters are "unorthodox": Certainly not - as I pointed out,
CEP/Audition and Matlab use these same filters (with the same behaviour near
the Nyquist frequency), as do many other audio applications (eg. Logic,
Goldwave and Wavosaur). Yes, the high frequency response is different from
the other filters in Audacity, but it's good to have choices.
- I thought I saw that someone had complained the first order lowpass is
different from the other orders. It's not, they're all completely
consistent. I'll explain further if anyone still thinks this is an issue.
- documentation is inadequate: OK. CEP's Scientific Filter documentation
would be a good place to start. And I provided a fair amount of commentary
on the filter when I first submitted it. I'd be glad to provide more - just
let me know exactly what needs to be addressed.
- doesn't work properly on tracks with multiple sample rates (same problem
as Equalization has, I think): True, and I don't think there's a simple
consistent way to make it work, as the filter response varies with sample
rate. At a minimum we'd need to display two different frequency responses on
the same graph and I think that would be both complex and confusing to the
user. In my view the solution is to prevent the user from applying the same
filter simultaneously to tracks with multiple sample rates and explain what
he needs to do.
- Chebyshev filters are too uncommon to be used here: Certainly they're less
common than Butterworth but it's far from unheard of to find them in audio
gear, both analog and digital. I've got lots of examples if anyone cares.

Steve, Gale: you may still have concerns that I've missed here. Let me know
if so, I'll try to address them.

Norm




--
View this message in context: http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152.html
Sent from the audacity-devel mailing list archive at Nabble.com.

Re: [Audacity-devel] Classic Filters status

[Federico M. <fm...@fc...>]

Norm,

> - badly named: I agree, that's why I called it Scientific Filters when 
> I submitted it. There's ample justification for this name: 
> CoolEditPro/Audition calls the same effect Scientific Filters, and 
> Matlab (certainly a major scientific application) uses these same 
> filters (butter, cheby1 and cheby2) as its workhorse filters

I think this had been already discussed. Scientific Filters is just a 
fantasy name. These filters are no more (and no less) scientific that 
any other, including fft filters implemented in the EQ effect.

> - the filters are "unorthodox": Certainly not - as I pointed out, 
> CEP/Audition and Matlab use these same filters (with the same 
> behaviour near the Nyquist frequency), as do many other audio 
> applications (eg. Logic, Goldwave and Wavosaur). Yes, the high 
> frequency response is different from the other filters in Audacity, 
> but it's good to have choices.

I think that when somebody said that, he was meaning that low-order 
Butterworth or Chebyshev are not truely so, but have discrepancies at 
frequencies close to the Nyquist limit, as you mentioon here:

> - I thought I saw that someone had complained the first order lowpass 
> is different from the other orders. It's not, they're all completely 
> consistent. I'll explain further if anyone still thinks this is an issue.

It is impossible that an IIR filter can match exactly the amplitude and 
phase frequency response along the full Nyquist range because the 
bilinear transformation on which it lies (I love this pun) is only an 
approximation that fails close to the Nyquist limit. May be there is a 
more close transformation (and of course there may be ad-hoc 
corrections), but bilinear is a standard conversion from analog filrters 
to some digital "equivalent".

> - documentation is inadequate: OK. CEP's Scientific Filter 
> documentation would be a good place to start. And I provided a fair 
> amount of commentary on the filter when I first submitted it. I'd be 
> glad to provide more - just let me know exactly what needs to be 
> addressed.

Possibly, some hints to its applications, for instance simulation of 
analog filters, particularly their transient response, perhaps emulation 
of some vintage effects that may have used these filters, simulation of 
analog antialiasing filters. Another, very low-latency real-time 
filtering, but I don't know if the idea was to implement real-time 
operation.

> - doesn't work properly on tracks with multiple sample rates (same 
> problem as Equalization has, I think): True, and I don't think there's 
> a simple consistent way to make it work, as the filter response varies 
> with sample rate. 

There is, provided the track sample rate is known (and I believe it is 
known), since there is a question of applying a scale factor for the 
frequency arguments that depends on the sample rate.

> - Chebyshev filters are too uncommon to be used here: Certainly 
> they're less common than Butterworth but it's far from unheard of to 
> find them in audio gear, both analog and digital. I've got lots of 
> examples if anyone cares.

Please, comment on them.

Regards,

Federico

Re: [Audacity-devel] Classic Filters status

[Norm C <n_b...@ho...>]

Federico Miyara wrote
> Norm,
> 
>> - badly named: I agree, that's why I called it Scientific Filters when 
>> I submitted it. There's ample justification for this name: 
>> CoolEditPro/Audition calls the same effect Scientific Filters, and 
>> Matlab (certainly a major scientific application) uses these same 
>> filters (butter, cheby1 and cheby2) as its workhorse filters
> 
> I think this had been already discussed. Scientific Filters is just a 
> fantasy name. These filters are no more (and no less) scientific that 
> any other, including fft filters implemented in the EQ effect.

Fantasy maybe, but the many thousands of CEP/Audition users will recognize
it immediately. And I do think it's better than Classic Filters, but
whatever, I'm easy.


>> - the filters are "unorthodox": Certainly not - as I pointed out, 
>> CEP/Audition and Matlab use these same filters (with the same 
>> behaviour near the Nyquist frequency), as do many other audio 
>> applications (eg. Logic, Goldwave and Wavosaur). Yes, the high 
>> frequency response is different from the other filters in Audacity, 
>> but it's good to have choices.
> 
> I think that when somebody said that, he was meaning that low-order 
> Butterworth or Chebyshev are not truely so, but have discrepancies at 
> frequencies close to the Nyquist limit, as you mentioon here:

All orders have that issue. It's just more obvious on the lower ones.


>> - I thought I saw that someone had complained the first order lowpass 
>> is different from the other orders. It's not, they're all completely 
>> consistent. I'll explain further if anyone still thinks this is an issue.
> 
> It is impossible that an IIR filter can match exactly the amplitude and 
> phase frequency response along the full Nyquist range because the 
> bilinear transformation on which it lies (I love this pun) is only an 
> approximation that fails close to the Nyquist limit. May be there is a 
> more close transformation (and of course there may be ad-hoc 
> corrections), but bilinear is a standard conversion from analog filrters 
> to some digital "equivalent".

Indeed it is.


>> - documentation is inadequate: OK. CEP's Scientific Filter 
>> documentation would be a good place to start. And I provided a fair 
>> amount of commentary on the filter when I first submitted it. I'd be 
>> glad to provide more - just let me know exactly what needs to be 
>> addressed.
> 
> Possibly, some hints to its applications, for instance simulation of 
> analog filters, particularly their transient response, perhaps emulation 
> of some vintage effects that may have used these filters, simulation of 
> analog antialiasing filters. Another, very low-latency real-time 
> filtering, but I don't know if the idea was to implement real-time 
> operation.

Good points, except I'm not sure what the relevance of real-time operation
is...?


>> - doesn't work properly on tracks with multiple sample rates (same 
>> problem as Equalization has, I think): True, and I don't think there's 
>> a simple consistent way to make it work, as the filter response varies 
>> with sample rate. 
> 
> There is, provided the track sample rate is known (and I believe it is 
> known), since there is a question of applying a scale factor for the 
> frequency arguments that depends on the sample rate.

Of course, there's no problem applying the filter to multiple tracks
correctly; the consistency problem is that the UI displays a frequency
response graph which (with multiple partly-overlapping lines) will be
confusing to users, or (with a single line) wrong.

>> - Chebyshev filters are too uncommon to be used here: Certainly 
>> they're less common than Butterworth but it's far from unheard of to 
>> find them in audio gear, both analog and digital. I've got lots of 
>> examples if anyone cares.
> 
> Please, comment on them.

Chebyshev filters have long been used in crossovers (esp for subwoofers),
low-pass filtering for BBD devices (eg. in chorus/flanger/vibrato gear),
low-pass filtering for downsampling in digital audio gear, radar signal
processing etc. etc. A quick google search will show that there's much
interest in calculators for Cheby designs, and they are discussed in all the
major signal processing and audio engineering handbooks, plus you'll find
them in app notes from the big DSP chip vendors (TI, ADI...). They're
certainly not obscure.

Anyway, I'd be happy to put in the effort to document these filters
satifactorily, and to fix the multiple-sample-rate issue... just need some
indication that no one has some whopping big objection to reinstating the
effect.

Norm




--
View this message in context: http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152p7565203.html
Sent from the audacity-devel mailing list archive at Nabble.com.

Re: [Audacity-devel] Classic Filters status

[Steve t. F. <ste...@gm...>]

On 14 November 2014 06:43, Norm C <n_b...@ho...> wrote:

> Federico Miyara wrote
> > Norm,
> >
> >> - badly named: I agree, that's why I called it Scientific Filters when
> >> I submitted it. There's ample justification for this name:
> >> CoolEditPro/Audition calls the same effect Scientific Filters, and
> >> Matlab (certainly a major scientific application) uses these same
> >> filters (butter, cheby1 and cheby2) as its workhorse filters
> >
> > I think this had been already discussed. Scientific Filters is just a
> > fantasy name. These filters are no more (and no less) scientific that
> > any other, including fft filters implemented in the EQ effect.
>
> Fantasy maybe, but the many thousands of CEP/Audition users will recognize
> it immediately. And I do think it's better than Classic Filters, but
> whatever, I'm easy.
>
>
> >> - the filters are "unorthodox": Certainly not - as I pointed out,
> >> CEP/Audition and Matlab use these same filters (with the same
> >> behaviour near the Nyquist frequency), as do many other audio
> >> applications (eg. Logic, Goldwave and Wavosaur). Yes, the high
> >> frequency response is different from the other filters in Audacity,
> >> but it's good to have choices.
> >
> > I think that when somebody said that, he was meaning that low-order
> > Butterworth or Chebyshev are not truely so, but have discrepancies at
> > frequencies close to the Nyquist limit, as you mentioon here:
>
> All orders have that issue. It's just more obvious on the lower ones.
>
>
> >> - I thought I saw that someone had complained the first order lowpass
> >> is different from the other orders. It's not, they're all completely
> >> consistent. I'll explain further if anyone still thinks this is an
> issue.
> >
> > It is impossible that an IIR filter can match exactly the amplitude and
> > phase frequency response along the full Nyquist range because the
> > bilinear transformation on which it lies (I love this pun) is only an
> > approximation that fails close to the Nyquist limit. May be there is a
> > more close transformation (and of course there may be ad-hoc
> > corrections), but bilinear is a standard conversion from analog filrters
> > to some digital "equivalent".
>
> Indeed it is.
>
>
> >> - documentation is inadequate: OK. CEP's Scientific Filter
> >> documentation would be a good place to start. And I provided a fair
> >> amount of commentary on the filter when I first submitted it. I'd be
> >> glad to provide more - just let me know exactly what needs to be
> >> addressed.
> >
> > Possibly, some hints to its applications, for instance simulation of
> > analog filters, particularly their transient response, perhaps emulation
> > of some vintage effects that may have used these filters, simulation of
> > analog antialiasing filters. Another, very low-latency real-time
> > filtering, but I don't know if the idea was to implement real-time
> > operation.
>
> Good points, except I'm not sure what the relevance of real-time operation
> is...?
>
>
> >> - doesn't work properly on tracks with multiple sample rates (same
> >> problem as Equalization has, I think): True, and I don't think there's
> >> a simple consistent way to make it work, as the filter response varies
> >> with sample rate.
> >
> > There is, provided the track sample rate is known (and I believe it is
> > known), since there is a question of applying a scale factor for the
> > frequency arguments that depends on the sample rate.
>
> Of course, there's no problem applying the filter to multiple tracks
> correctly; the consistency problem is that the UI displays a frequency
> response graph which (with multiple partly-overlapping lines) will be
> confusing to users, or (with a single line) wrong.
>
> >> - Chebyshev filters are too uncommon to be used here: Certainly
> >> they're less common than Butterworth but it's far from unheard of to
> >> find them in audio gear, both analog and digital. I've got lots of
> >> examples if anyone cares.
> >
> > Please, comment on them.
>
> Chebyshev filters have long been used in crossovers (esp for subwoofers),
> low-pass filtering for BBD devices (eg. in chorus/flanger/vibrato gear),
> low-pass filtering for downsampling in digital audio gear, radar signal
> processing etc. etc. A quick google search will show that there's much
> interest in calculators for Cheby designs, and they are discussed in all
> the
> major signal processing and audio engineering handbooks, plus you'll find
> them in app notes from the big DSP chip vendors (TI, ADI...). They're
> certainly not obscure.
>

What would be typical applications of Chebyshev type I and type II filters
for end users of Audacity?

Steve



>
> Anyway, I'd be happy to put in the effort to document these filters
> satifactorily, and to fix the multiple-sample-rate issue... just need some
> indication that no one has some whopping big objection to reinstating the
> effect.
>
> Norm
>
>
>
>
> --
> View this message in context:
> http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152p7565203.html
> Sent from the audacity-devel mailing list archive at Nabble.com.
>
>
> ------------------------------------------------------------------------------
> Comprehensive Server Monitoring with Site24x7.
> Monitor 10 servers for $9/Month.
> Get alerted through email, SMS, voice calls or mobile push notifications.
> Take corrective actions from your mobile device.
>
> http://pubads.g.doubleclick.net/gampad/clk?id=154624111&iu=/4140/ostg.clktrk
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>

Re: [Audacity-devel] Classic Filters status

[Norm C <n_b...@ho...>]

Stevethefiddle wrote
> What would be typical applications of Chebyshev type I and type II filters
> for end users of Audacity?
> 
> Steve

Probably the biggest use would be for modelling hardware or software that
uses Chebyshev filters (modelling is a topic of considerable research these
days). Other reasons for selecting Chebyshev over Butterworth would include
reducing the worst-case group delay by being able to use a lower filter
order to meet a specified magnitude response, and for designing your own
hardware-based or DSP-based filters, in which filter order is a significant
cost factor.

Norm




--
View this message in context: http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152p7566419.html
Sent from the audacity-devel mailing list archive at Nabble.com.

Re: [Audacity-devel] Classic Filters status

[Steve t. F. <ste...@gm...>]

On 26 December 2014 at 21:57, Norm C <n_b...@ho...> wrote:
> Stevethefiddle wrote
>> What would be typical applications of Chebyshev type I and type II filters
>> for end users of Audacity?
>>
>> Steve
>
> Probably the biggest use would be for modelling hardware or software that
> uses Chebyshev filters (modelling is a topic of considerable research these
> days). Other reasons for selecting Chebyshev over Butterworth would include
> reducing the worst-case group delay by being able to use a lower filter
> order to meet a specified magnitude response, and for designing your own
> hardware-based or DSP-based filters, in which filter order is a significant
> cost factor.

Thanks Norm. I can see how that is important in SPICE software
(http://en.wikipedia.org/wiki/SPICE).

Steve

>
> Norm
>
>
>
>
> --
> View this message in context: http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152p7566419.html
> Sent from the audacity-devel mailing list archive at Nabble.com.
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming! The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel

Re: [Audacity-devel] Classic Filters status

[Peter S. <pet...@ya...>]

Norm,
one of the things that you really need to do to obtain support forthe inclusion of this effect of yours is to actively engage with thedocumentation for it in the Manual.  This will not necessarilyguarantee its acceptance into the code bas but it would improveits chances.  
A couple of the Manual authors have made efforts to documentthe functionality but to some extent are flying blind.  It needs inputfrom someone who really understands the whys and whereforesof the effect - particularly the "whys" as in what would I use thisfor - why would I prefer it over other similar effects in Audacity
See this page for the draft documentation:http://manual.audacityteam.org/man/Classic_FiltersIt is in the draft alpha manual but is effectively an orphan page asthe links to it are currently commented out.
In order to engage with that page you would need to apply for an editors account on the Manual. There is a link on the front page of the Manual for that.  
Alternatively you could just supply a draft of an update to that page and we can try to implement the changes fir you - but that wouldbe harder work for both parties.  And you would still have to be involvedin the editorial page reviewing any changes.
Hope that helps,
Peter Sampson
Tel: +44 (0)1625 524 780

      From: Norm C <n_b...@ho...>
 To: aud...@li... 
 Sent: Friday, December 26, 2014 9:57 PM
 Subject: Re: [Audacity-devel] Classic Filters status
   
Stevethefiddle wrote
> What would be typical applications of Chebyshev type I and type II filters
> for end users of Audacity?
> 
> Steve

Probably the biggest use would be for modelling hardware or software that
uses Chebyshev filters (modelling is a topic of considerable research these
days). Other reasons for selecting Chebyshev over Butterworth would include
reducing the worst-case group delay by being able to use a lower filter
order to meet a specified magnitude response, and for designing your own
hardware-based or DSP-based filters, in which filter order is a significant
cost factor.

Norm




--
View this message in context: http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152p7566419.html
Sent from the audacity-devel mailing list archive at Nabble.com.

------------------------------------------------------------------------------
Dive into the World of Parallel Programming! The Go Parallel Website,
sponsored by Intel and developed in partnership with Slashdot Media, is your
hub for all things parallel software development, from weekly thought
leadership blogs to news, videos, case studies, tutorials and more. Take a
look and join the conversation now. http://goparallel.sourceforge.net
_______________________________________________
audacity-devel mailing list
aud...@li...
https://lists.sourceforge.net/lists/listinfo/audacity-devel


  

Re: [Audacity-devel] Classic Filters status

[Gale <ga...@au...>]

Hi Norm,

I'm not strongly opposed but do see docs and use cases as an outstanding 
issue.   


Norm C wrote
> I've been out of touch for a few months but I gather there's been some
> rather polarized discussions regarding the Classic Filters and they're now
> out of the build. In my view that's unfortunate. I'll summarize what I see
> as the issues:
> 
> - badly named: I agree, that's why I called it Scientific Filters when I
> submitted it. There's ample justification for this name:
> CoolEditPro/Audition calls the same effect Scientific Filters, and Matlab
> (certainly a major scientific application) uses these same filters
> (butter, cheby1 and cheby2) as its workhorse filters

I always felt to justify "Scientific" moniker there ought to be more
esoteric
filters e.g. Elliptic, Gaussian and any others that could be useful for
modelling
or educational/research interests (which you give as a use case).

Do you or anyone buy into that? Even if you do, I guess new filters would 
have to wait until release after next. 


Norm C wrote
>    
> - the filters are "unorthodox": Certainly not - as I pointed out,
> CEP/Audition and Matlab use these same filters (with the same behaviour
> near the Nyquist frequency), as do many other audio applications (eg.
> Logic, Goldwave and Wavosaur). Yes, the high frequency response is
> different from the other filters in Audacity, but it's good to have
> choices.

Yes but the greater HF attenuation in your first order Low Pass compared 
to docs that most users will find or understand is a drawback if we're 
presenting these as "Classic". 

Alternatives aren't usually "classic" though I understand the sense in 
which these are "classic".   
  

Norm C wrote
> - documentation is inadequate: OK. CEP's Scientific Filter documentation
> would be a good place to start. And I provided a fair amount of commentary
> on the filter when I first submitted it. I'd be glad to provide more -
> just let me know exactly what needs to be addressed.

Yes this is the main problem (and the first order Low Pass question above 
would I think need a sentence in the docs).  

This is the page in the Manual as it stands:
http://manual.audacityteam.org/man/Classic_Filters .

Not at all ready. Being quite close to 2.1.0, I feel you would need to work 
directly on that page yourself (I can sign you up there). You would need to 
get it passed by someone on the Manual team who only had average
competence in filtering, to verify that it was written in basic enough
fashion 
and said why one would want to use these. 

I know quite a bit of information has been given already about "why" but 
it's quite scattered around in different threads.   


Norm C wrote
> - doesn't work properly on tracks with multiple sample rates (same problem
> as Equalization has, I think): True, and I don't think there's a simple
> consistent way to make it work, as the filter response varies with sample
> rate. At a minimum we'd need to display two different frequency responses
> on the same graph and I think that would be both complex and confusing to
> the user. In my view the solution is to prevent the user from applying the
> same filter simultaneously to tracks with multiple sample rates and
> explain what he needs to do.

That is already addressed so isn't an obstacle. There is an error message
that "all selected tracks must be the same sample rate."


Gale 


Norm C wrote
> - Chebyshev filters are too uncommon to be used here: Certainly they're
> less common than Butterworth but it's far from unheard of to find them in
> audio gear, both analog and digital. I've got lots of examples if anyone
> cares.
> 
> Steve, Gale: you may still have concerns that I've missed here. Let me
> know if so, I'll try to address them.
> 
> Norm





--
View this message in context: http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152p7566434.html
Sent from the audacity-devel mailing list archive at Nabble.com.

Re: [Audacity-devel] Classic Filters status

[Martyn S. <mar...@gm...>]

Hi. I guess Gale is where this is aimed primarily...

On 28/12/2014 15:28, Gale wrote:
> Hi Norm,
>
> I'm not strongly opposed but do see docs and use cases as an outstanding
> issue.

'Classic Filters' (that is, traditional analog IIR filters implemented 
in the digital domain) are just as valid filters, to a user, as the 
rather crude pre-sets available in the current Equalization effect. 
But read on.

I do not agree that documentation is an issue for this new effect.

> Norm C wrote
>> I've been out of touch for a few months but I gather there's been some
>> rather polarized discussions regarding the Classic Filters and they're now
>> out of the build. In my view that's unfortunate. I'll summarize what I see
>> as the issues:
>>
>> - badly named: I agree, that's why I called it Scientific Filters when I
>> submitted it. There's ample justification for this name:
>> CoolEditPro/Audition calls the same effect Scientific Filters, and Matlab
>> (certainly a major scientific application) uses these same filters
>> (butter, cheby1 and cheby2) as its workhorse filters

Agreed, workhorse filters.  Should be in the build.

> I always felt to justify "Scientific" moniker there ought to be more
> esoteric
> filters e.g. Elliptic, Gaussian and any others that could be useful for
> modelling
> or educational/research interests (which you give as a use case).

Well, Gale, I don't think you have the background to discuss Elliptic 
filters, or any others.  The ones selected here are the most common by 
a long way, I think.

> Do you or anyone buy into that? Even if you do, I guess new filters would
> have to wait until release after next.

Why would it have to wait?  More esoteric filters could, obviously, be 
added.  Maybe you could find yet another word on wikipedia and hold 
that up longer, making the effect more complex and documentation 
demands higher.  You say that you are "not strongly opposed", so still 
opposed to this being released, and making the barrier still higher.

> Norm C wrote
>>
>> - the filters are "unorthodox": Certainly not - as I pointed out,
>> CEP/Audition and Matlab use these same filters (with the same behaviour
>> near the Nyquist frequency), as do many other audio applications (eg.
>> Logic, Goldwave and Wavosaur). Yes, the high frequency response is
>> different from the other filters in Audacity, but it's good to have
>> choices.
>
> Yes but the greater HF attenuation in your first order Low Pass compared
> to docs that most users will find or understand is a drawback if we're
> presenting these as "Classic".

I disagree, as I have before.  Just because you, Gale, don't have a 
thorough understanding of DSP, it does not mean that users won't find 
these filters useful.

> Alternatives aren't usually "classic" though I understand the sense in
> which these are "classic".

No, I don't think you do understand Gale.  Please read, for example,
http://web.mit.edu/~cjoseph/Public/Discrete-Time_Signal_Processing_-_Oppenheim_-_2nd_Edition.pdf
especially chapter 7 on 'Filter Design Techniques', or trust that 
people who do understand that stuff know what they are doing.

> Norm C wrote
>> - documentation is inadequate: OK. CEP's Scientific Filter documentation
>> would be a good place to start. And I provided a fair amount of commentary
>> on the filter when I first submitted it. I'd be glad to provide more -
>> just let me know exactly what needs to be addressed.
>
> Yes this is the main problem (and the first order Low Pass question above
> would I think need a sentence in the docs).
>
> This is the page in the Manual as it stands:
> http://manual.audacityteam.org/man/Classic_Filters .
>
> Not at all ready.

I disagree.  That page isn't too bad for a new effect.  Somebody has 
put a lot of useful effort into that.

Being quite close to 2.1.0, I feel you would need to work
> directly on that page yourself (I can sign you up there). You would need to
> get it passed by someone on the Manual team who only had average
> competence in filtering, to verify that it was written in basic enough
> fashion
> and said why one would want to use these.

That sounds fair enough.  Make sure that the manual page is 
understandable by the people who usually write those pages.  It looks 
OK to me.

TTFN
Martyn

> I know quite a bit of information has been given already about "why" but
> it's quite scattered around in different threads.
>
>
> Norm C wrote
>> - doesn't work properly on tracks with multiple sample rates (same problem
>> as Equalization has, I think): True, and I don't think there's a simple
>> consistent way to make it work, as the filter response varies with sample
>> rate. At a minimum we'd need to display two different frequency responses
>> on the same graph and I think that would be both complex and confusing to
>> the user. In my view the solution is to prevent the user from applying the
>> same filter simultaneously to tracks with multiple sample rates and
>> explain what he needs to do.
>
> That is already addressed so isn't an obstacle. There is an error message
> that "all selected tracks must be the same sample rate."
>
>
> Gale
>
>
> Norm C wrote
>> - Chebyshev filters are too uncommon to be used here: Certainly they're
>> less common than Butterworth but it's far from unheard of to find them in
>> audio gear, both analog and digital. I've got lots of examples if anyone
>> cares.
>>
>> Steve, Gale: you may still have concerns that I've missed here. Let me
>> know if so, I'll try to address them.
>>
>> Norm
>
>
>
>
>
> --
> View this message in context: http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152p7566434.html
> Sent from the audacity-devel mailing list archive at Nabble.com.
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming! The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>

Re: [Audacity-devel] Classic Filters status

[Peter S. <pet...@ya...>]

Martyn responded to Gale:>> This is the page in the Manual as it stands:
>> http://manual.audacityteam.org/man/Classic_Filters .
>>
>> Not at all ready.
>
>I disagree.  That page isn't too bad for a new effect.  Somebody has 
>put a lot of useful effort into that. 
Your feedback as it being "not too bad" Martyn as someone with good experiencein this area of expertise is useful to hear.   
I think that what we on the Manual team have always felt lacking on thatpage was some editorial input by the effect's author - which is why both I and Gale have recently written, inviting Norm to participate. 
In particular what we feel is missing is some description of where and why you might choose to deploy the filers in this effect - and some use cases or examplesthat can be included to aid the non-expert reader.  
Perhaps Martyn, if Norm cannot be persuaded to engage, we could persuade youto assist with the improvement of this page?  
None of us on the Manual team afaik has the necessary expertise to provide this,we've already taken it as far as we can with our limited knowledge.
Cheers,Peter.
Peter Sampson
Tel: +44 (0)1625 524 780

  

Re: [Audacity-devel] Classic Filters status

[Martyn S. <mar...@gm...>]

Hi

Sorry I have not got to this sooner.

I have read all the rest of this thread and see that there is 
confusion over when to use the different filter types in an audio 
production context, and a desire to document that.  I cannot offer any 
advice on when to use which for audio production I'm afraid, other 
than listening to the result.  I think Roger said the same.

Obviously the different 'classic' types were originally optimised for 
different things in the world of analogue (and then transferred into 
the digital world) as others have said.  I won't repeat what the 
optimisations were here, but to address the 'editor comments' on
http://manual.audacityteam.org/man/Classic_Filters
the ripple that you allow in the passband for Type 1 is traded off 
against the attenuation that you get in the stopband.  For Type 2 the 
'Minimum Stopband Attenuation' that you specify is traded off against 
the attenuation at the end of the passband.  You can see that in the 
graph if you try different values.  Some of the other comments are no 
longer valid as the boxes are invisible.

In terms of the filter 'order' (or 'Rolloff (dB per octave)'/6) in the 
case of the Nyquist ones), it may be worth noting that the Nyquist 
ones go to 8 (1, 2, 4, 6, 8), 'Classic Filters' go 1 to 10, whereas 
the Equalization effect defaults to 4001 and can be pushed to 8191; 
OK, not quite comparing like with like but you get the idea of how 
powerful the Eq is compared to the others.

Use case:
Say a user wanted to recover speech from a degraded source that had 
lots of interference, particularly at high frequencies (and they 
recognised this).  The easiest option would be the 'Low Pass 
Filter...' Nyquist effect (fewest parameters), then the 'Classic 
Filters...' where they could play with the different options to reduce 
the unwanted HF noise, then the 'Equalization...' where they have 
near-unlimited stuff to play with.  That is a situation that users may 
be faced with, and 'Classic' gives an extra option that could be 
educational.

The Classic's GUI has a similar level of complexity as the EQ (in 
'Draw Curves' mode) in terms of interpreting what they see in the 
graph, but they need some knowledge to use that.  They need to know 
how to use the 'Plot Spectrum...' tool and/or the Spectrogram view to 
make best use of it, but that is surely the purpose of a tutorial, 
rather than the manual.

I do not see any explanation in
http://manual.audacityteam.org/man/Low_Pass_Filter
or
http://manual.audacityteam.org/man/Equalization
of which we should use when and so do not really understand the 
insistence that
http://manual.audacityteam.org/man/Classic_Filters
should have the same. :-)

So, just a few thoughts here.  I'd like Classic to be in this release 
for further comment, as you know.

TTFN
Martyn


On 30/12/2014 09:54, Peter Sampson wrote:
> Martyn responded to Gale:
>  >> This is the page in the Manual as it stands:
>  >> http://manual.audacityteam.org/man/Classic_Filters
> <http://manual.audacityteam.org/man/Classic_Filters>.
>  >>
>  >> Not at all ready.
>  >
>  >I disagree.  That page isn't too bad for a new effect.  Somebody has
>  >put a lot of useful effort into that.
>
> Your feedback as it being "not too bad" Martyn as someone with good
> experience
> in this area of expertise is useful to hear.
>
> I think that what we on the Manual team have always felt lacking on that
> page was some editorial input by the effect's author - which is why
> both I and Gale
> have recently written, inviting Norm to participate.
>
> In particular what we feel is missing is some description of where and
> why you
> might choose to deploy the filers in this effect - and some use cases
> or examples
> that can be included to aid the non-expert reader.
>
> Perhaps Martyn, if Norm cannot be persuaded to engage, we could
> persuade you
> to assist with the improvement of this page?
>
> None of us on the Manual team afaik has the necessary expertise to
> provide this,
> we've already taken it as far as we can with our limited knowledge.
>
> Cheers,
> Peter.
>
> Peter Sampson
> Tel: +44 (0)1625 524 780
>

Re: [Audacity-devel] Classic Filters status

[Federico M. <fm...@fc...>]

Dear all,

I think this page is enough stuff for a start:

> http://manual.audacityteam.org/man/Classic_Filters 

especially since it has links to wikipedia. It is true that these links 
do not provide insight to applications, but if the effect is working 
successfully and the documentation explains all the available 
parameters, users can start to play about with it.

There are a few things that I note and could be improved.

First, the input parameters differ between Butterworth and both 
Chebyshev's, but only the Butterworth case is depicted.

Moreover, when the user gets to

"-For Butterworth filters no value can be entered and any value 
displayed is ignored."

no value seems to be displayed so there seems to be no need for the 
comment and it might result confusing instead. But it would be useful to 
make clear this: "At the cutoff frequency this filter has an attenuation 
of 3 dB"

As regards

"-For Chebyshev Type I filters type in the acceptable amount of passband 
ripple. Higher values of passband ripple will also increase the cutoff 
slope"

it may be useful to include the trade-off remark by Martyn.

As regards

"-For Chebyshev Type II filters no value can be entered and any value 
displayed is ignored"

again, the "any value displayed is ignored" comment shouldn't be there.

The same is true for the case of the Minimun Stopband Attenuation for 
Butterworth and Chebyshev Type I.

I think Martyn has also pointed this out.

I believe an "application hints" section will be convenient indeed, but 
it should be posponed without preventing the inclusion of the effect.

Regards,

Federico

Re: [Audacity-devel] Classic Filters status

[Steve t. F. <ste...@gm...>]

On 2 January 2015 at 00:50, Martyn Shaw <mar...@gm...> wrote:
> Hi
>
> Sorry I have not got to this sooner.
>
> I have read all the rest of this thread and see that there is
> confusion over when to use the different filter types in an audio
> production context, and a desire to document that.  I cannot offer any
> advice on when to use which for audio production I'm afraid, other
> than listening to the result.  I think Roger said the same.
>
> Obviously the different 'classic' types were originally optimised for
> different things in the world of analogue (and then transferred into
> the digital world) as others have said.  I won't repeat what the
> optimisations were here, but to address the 'editor comments' on
> http://manual.audacityteam.org/man/Classic_Filters

> the ripple that you allow in the passband for Type 1 is traded off
> against the attenuation that you get in the stopband.

That makes sense where component count, physical size and
manufacturing costs of analog filters are relevant, but surely it is a
spurious trade off for digital filters. Yes a 2nd order Chebyshev
filter may have a slightly steeper roll-off compared with a second
order Butterworth, (at the expense of ripple in either the pass band
or stop band), but for digital filters there is absolutely no need for
that compromise - if you want a steeper roll-off, use a higher order
Butterworth filter and then you get a steeper roll-off with no ripple.


> For Type 2 the
> 'Minimum Stopband Attenuation' that you specify is traded off against
> the attenuation at the end of the passband.  You can see that in the
> graph if you try different values.  Some of the other comments are no
> longer valid as the boxes are invisible.
>
> In terms of the filter 'order' (or 'Rolloff (dB per octave)'/6) in the
> case of the Nyquist ones), it may be worth noting that the Nyquist
> ones go to 8 (1, 2, 4, 6, 8), 'Classic Filters' go 1 to 10, whereas
> the Equalization effect defaults to 4001 and can be pushed to 8191;
> OK, not quite comparing like with like but you get the idea of how
> powerful the Eq is compared to the others.

Yes, indeed.

>
> Use case:
> Say a user wanted to recover speech from a degraded source that had
> lots of interference, particularly at high frequencies (and they
> recognised this).  The easiest option would be the 'Low Pass
> Filter...' Nyquist effect (fewest parameters), then the 'Classic
> Filters...' where they could play with the different options to reduce
> the unwanted HF noise, then the 'Equalization...' where they have
> near-unlimited stuff to play with.  That is a situation that users may
> be faced with, and 'Classic' gives an extra option that could be
> educational.

The "educational value" of this example would appear to be the
opportunity to discover that the Chebyshev filters are the worst
choice of tool for the job.


>
> The Classic's GUI has a similar level of complexity as the EQ (in
> 'Draw Curves' mode) in terms of interpreting what they see in the
> graph, but they need some knowledge to use that.  They need to know
> how to use the 'Plot Spectrum...' tool and/or the Spectrogram view to
> make best use of it, but that is surely the purpose of a tutorial,
> rather than the manual.

Yes, I agree that would be "tutorial material".

>
> I do not see any explanation in
> http://manual.audacityteam.org/man/Low_Pass_Filter
> or
> http://manual.audacityteam.org/man/Equalization
> of which we should use when and so do not really understand the
> insistence that
> http://manual.audacityteam.org/man/Classic_Filters
> should have the same. :-)

For Low Pass and High Pass filters, I think that the documentation
(http://manual.audacityteam.org/man/Low_Pass_Filter) is quite clear
that:
"Low Pass Filter passes frequencies below its cutoff frequency and
attenuates frequencies above its cutoff frequency. This effect can
therefore be used to reduce high pitched noise." (and the
complimentary equivalent for High Pass).

I don't see the need to go into more detail than that, until we add in
the complication of Chebyshev vs. Butterworth.

My impression with this effect is that it has been created as a "good
idea" (yes I agree that it is a good idea, but with some reservations)
that Audacity should have these filters, rather than to meet a real
demand. In other words, a solution looking for a problem to solve.

The question that I would pose is; what would this effect be like if
it were designed from the view point of fulfilling audio production
needs?

If that were the design criteria I suspect that we would abandon
"classic" analog modelling and produce a variable slope high pass/ low
pass filter with zero ripple and probably with linear phase response.
Such an effect could be more powerful than the current high/low pass
filters, yet simpler (and less intimidating) than the full featured
Equalization effect. The graphical interface could easily be extended
to provide additional filter types, such as band pass/ band stop
filters, shelf filters, and yes could even include "classic analog"
modelled filters (Butterworth, Chebyshev type I, Chebyshev type II).

As Martyn said, just a few thoughts here. I'm not arguing against this
effect per se, but I am questioning whether its current form is in the
best interests of Audacity users. My view on that at present is that
the Chebyshev filters introduce unnecessary complication that is not
in the best interests of the vast majority of user, for very limited
benefit for a very small minority of users, and due to bug 660 is
inferior to what we already ship.

I would very much like to see this effect progress - it's been
standing on the side-lines too long. A suggestion for moving this
forward would be to initially introduce it with just low pass / high
pass options (Butterworth, but no need for the manual to go into
detail on that point), preferably with bug 660 resolved (though that
could be raised to P3 and release noted).


Steve

>
> So, just a few thoughts here.  I'd like Classic to be in this release
> for further comment, as you know.
>
> TTFN
> Martyn
>
>
> On 30/12/2014 09:54, Peter Sampson wrote:
>> Martyn responded to Gale:
>>  >> This is the page in the Manual as it stands:
>>  >> http://manual.audacityteam.org/man/Classic_Filters
>> <http://manual.audacityteam.org/man/Classic_Filters>.
>>  >>
>>  >> Not at all ready.
>>  >
>>  >I disagree.  That page isn't too bad for a new effect.  Somebody has
>>  >put a lot of useful effort into that.
>>
>> Your feedback as it being "not too bad" Martyn as someone with good
>> experience
>> in this area of expertise is useful to hear.
>>
>> I think that what we on the Manual team have always felt lacking on that
>> page was some editorial input by the effect's author - which is why
>> both I and Gale
>> have recently written, inviting Norm to participate.
>>
>> In particular what we feel is missing is some description of where and
>> why you
>> might choose to deploy the filers in this effect - and some use cases
>> or examples
>> that can be included to aid the non-expert reader.
>>
>> Perhaps Martyn, if Norm cannot be persuaded to engage, we could
>> persuade you
>> to assist with the improvement of this page?
>>
>> None of us on the Manual team afaik has the necessary expertise to
>> provide this,
>> we've already taken it as far as we can with our limited knowledge.
>>
>> Cheers,
>> Peter.
>>
>> Peter Sampson
>> Tel: +44 (0)1625 524 780
>>
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming! The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel

Re: [Audacity-devel] Classic Filters status

[Gary N. <por...@gm...>]

All
I've been kibitzing but perhaps it would be helpful to consider the FIR filter as a reference. We get linear phase => constant Group Delay. That is as good as it gets for filters. 

Analog Devices has an online analog filter designer that shows group delay of any design. Butterworths  can be crafted to have nearly constant group delay in pass band. Chebychevs have ripple in the group delay. Given a no cost choice, Butterworth is preferred because it is closer to FIR ideal. 
Gary Nelson


Sent from my iPhone

> On Jan 2, 2015, at 6:09 AM, Steve the Fiddle <ste...@gm...> wrote:
> 
>> On 2 January 2015 at 00:50, Martyn Shaw <mar...@gm...> wrote:
>> Hi
>> 
>> Sorry I have not got to this sooner.
>> 
>> I have read all the rest of this thread and see that there is
>> confusion over when to use the different filter types in an audio
>> production context, and a desire to document that.  I cannot offer any
>> advice on when to use which for audio production I'm afraid, other
>> than listening to the result.  I think Roger said the same.
>> 
>> Obviously the different 'classic' types were originally optimised for
>> different things in the world of analogue (and then transferred into
>> the digital world) as others have said.  I won't repeat what the
>> optimisations were here, but to address the 'editor comments' on
>> http://manual.audacityteam.org/man/Classic_Filters
> 
>> the ripple that you allow in the passband for Type 1 is traded off
>> against the attenuation that you get in the stopband.
> 
> That makes sense where component count, physical size and
> manufacturing costs of analog filters are relevant, but surely it is a
> spurious trade off for digital filters. Yes a 2nd order Chebyshev
> filter may have a slightly steeper roll-off compared with a second
> order Butterworth, (at the expense of ripple in either the pass band
> or stop band), but for digital filters there is absolutely no need for
> that compromise - if you want a steeper roll-off, use a higher order
> Butterworth filter and then you get a steeper roll-off with no ripple.
> 
> 
>> For Type 2 the
>> 'Minimum Stopband Attenuation' that you specify is traded off against
>> the attenuation at the end of the passband.  You can see that in the
>> graph if you try different values.  Some of the other comments are no
>> longer valid as the boxes are invisible.
>> 
>> In terms of the filter 'order' (or 'Rolloff (dB per octave)'/6) in the
>> case of the Nyquist ones), it may be worth noting that the Nyquist
>> ones go to 8 (1, 2, 4, 6, 8), 'Classic Filters' go 1 to 10, whereas
>> the Equalization effect defaults to 4001 and can be pushed to 8191;
>> OK, not quite comparing like with like but you get the idea of how
>> powerful the Eq is compared to the others.
> 
> Yes, indeed.
> 
>> 
>> Use case:
>> Say a user wanted to recover speech from a degraded source that had
>> lots of interference, particularly at high frequencies (and they
>> recognised this).  The easiest option would be the 'Low Pass
>> Filter...' Nyquist effect (fewest parameters), then the 'Classic
>> Filters...' where they could play with the different options to reduce
>> the unwanted HF noise, then the 'Equalization...' where they have
>> near-unlimited stuff to play with.  That is a situation that users may
>> be faced with, and 'Classic' gives an extra option that could be
>> educational.
> 
> The "educational value" of this example would appear to be the
> opportunity to discover that the Chebyshev filters are the worst
> choice of tool for the job.
> 
> 
>> 
>> The Classic's GUI has a similar level of complexity as the EQ (in
>> 'Draw Curves' mode) in terms of interpreting what they see in the
>> graph, but they need some knowledge to use that.  They need to know
>> how to use the 'Plot Spectrum...' tool and/or the Spectrogram view to
>> make best use of it, but that is surely the purpose of a tutorial,
>> rather than the manual.
> 
> Yes, I agree that would be "tutorial material".
> 
>> 
>> I do not see any explanation in
>> http://manual.audacityteam.org/man/Low_Pass_Filter
>> or
>> http://manual.audacityteam.org/man/Equalization
>> of which we should use when and so do not really understand the
>> insistence that
>> http://manual.audacityteam.org/man/Classic_Filters
>> should have the same. :-)
> 
> For Low Pass and High Pass filters, I think that the documentation
> (http://manual.audacityteam.org/man/Low_Pass_Filter) is quite clear
> that:
> "Low Pass Filter passes frequencies below its cutoff frequency and
> attenuates frequencies above its cutoff frequency. This effect can
> therefore be used to reduce high pitched noise." (and the
> complimentary equivalent for High Pass).
> 
> I don't see the need to go into more detail than that, until we add in
> the complication of Chebyshev vs. Butterworth.
> 
> My impression with this effect is that it has been created as a "good
> idea" (yes I agree that it is a good idea, but with some reservations)
> that Audacity should have these filters, rather than to meet a real
> demand. In other words, a solution looking for a problem to solve.
> 
> The question that I would pose is; what would this effect be like if
> it were designed from the view point of fulfilling audio production
> needs?
> 
> If that were the design criteria I suspect that we would abandon
> "classic" analog modelling and produce a variable slope high pass/ low
> pass filter with zero ripple and probably with linear phase response.
> Such an effect could be more powerful than the current high/low pass
> filters, yet simpler (and less intimidating) than the full featured
> Equalization effect. The graphical interface could easily be extended
> to provide additional filter types, such as band pass/ band stop
> filters, shelf filters, and yes could even include "classic analog"
> modelled filters (Butterworth, Chebyshev type I, Chebyshev type II).
> 
> As Martyn said, just a few thoughts here. I'm not arguing against this
> effect per se, but I am questioning whether its current form is in the
> best interests of Audacity users. My view on that at present is that
> the Chebyshev filters introduce unnecessary complication that is not
> in the best interests of the vast majority of user, for very limited
> benefit for a very small minority of users, and due to bug 660 is
> inferior to what we already ship.
> 
> I would very much like to see this effect progress - it's been
> standing on the side-lines too long. A suggestion for moving this
> forward would be to initially introduce it with just low pass / high
> pass options (Butterworth, but no need for the manual to go into
> detail on that point), preferably with bug 660 resolved (though that
> could be raised to P3 and release noted).
> 
> 
> Steve
> 
>> 
>> So, just a few thoughts here.  I'd like Classic to be in this release
>> for further comment, as you know.
>> 
>> TTFN
>> Martyn
>> 
>> 
>>> On 30/12/2014 09:54, Peter Sampson wrote:
>>> Martyn responded to Gale:
>>>>> This is the page in the Manual as it stands:
>>>>> http://manual.audacityteam.org/man/Classic_Filters
>>> <http://manual.audacityteam.org/man/Classic_Filters>.
>>>>> 
>>>>> Not at all ready.
>>>> 
>>>> I disagree.  That page isn't too bad for a new effect.  Somebody has
>>>> put a lot of useful effort into that.
>>> 
>>> Your feedback as it being "not too bad" Martyn as someone with good
>>> experience
>>> in this area of expertise is useful to hear.
>>> 
>>> I think that what we on the Manual team have always felt lacking on that
>>> page was some editorial input by the effect's author - which is why
>>> both I and Gale
>>> have recently written, inviting Norm to participate.
>>> 
>>> In particular what we feel is missing is some description of where and
>>> why you
>>> might choose to deploy the filers in this effect - and some use cases
>>> or examples
>>> that can be included to aid the non-expert reader.
>>> 
>>> Perhaps Martyn, if Norm cannot be persuaded to engage, we could
>>> persuade you
>>> to assist with the improvement of this page?
>>> 
>>> None of us on the Manual team afaik has the necessary expertise to
>>> provide this,
>>> we've already taken it as far as we can with our limited knowledge.
>>> 
>>> Cheers,
>>> Peter.
>>> 
>>> Peter Sampson
>>> Tel: +44 (0)1625 524 780
>> 
>> ------------------------------------------------------------------------------
>> Dive into the World of Parallel Programming! The Go Parallel Website,
>> sponsored by Intel and developed in partnership with Slashdot Media, is your
>> hub for all things parallel software development, from weekly thought
>> leadership blogs to news, videos, case studies, tutorials and more. Take a
>> look and join the conversation now. http://goparallel.sourceforge.net
>> _______________________________________________
>> audacity-devel mailing list
>> aud...@li...
>> https://lists.sourceforge.net/lists/listinfo/audacity-devel
> 
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming! The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel

Re: [Audacity-devel] Classic Filters status

[Martyn S. <mar...@gm...>]

Hi

This discussion has gone on a long time, just one more post from me, 
mostly agreeing with Steve...

On 02/01/2015 14:09, Steve the Fiddle wrote:
> On 2 January 2015 at 00:50, Martyn Shaw <mar...@gm...> wrote:
>> Hi
>>
>> Sorry I have not got to this sooner.
>>
>> I have read all the rest of this thread and see that there is
>> confusion over when to use the different filter types in an audio
>> production context, and a desire to document that.  I cannot offer any
>> advice on when to use which for audio production I'm afraid, other
>> than listening to the result.  I think Roger said the same.
>>
>> Obviously the different 'classic' types were originally optimised for
>> different things in the world of analogue (and then transferred into
>> the digital world) as others have said.  I won't repeat what the
>> optimisations were here, but to address the 'editor comments' on
>> http://manual.audacityteam.org/man/Classic_Filters
>
>> the ripple that you allow in the passband for Type 1 is traded off
>> against the attenuation that you get in the stopband.
>
> That makes sense where component count, physical size and
> manufacturing costs of analog filters are relevant, but surely it is a
> spurious trade off for digital filters.

Indeed.

  Yes a 2nd order Chebyshev
> filter may have a slightly steeper roll-off compared with a second
> order Butterworth, (at the expense of ripple in either the pass band
> or stop band), but for digital filters there is absolutely no need for
> that compromise - if you want a steeper roll-off, use a higher order
> Butterworth filter and then you get a steeper roll-off with no ripple.

Butterworth is also a trade off, only an 'ideal' for people wanting a 
particular thing.  And you have to put up with an (audible) -3dB at 
the cutoff you specify vs an (inaudible) 0.1dB (say) of ripple and a 
well-defined cutoff.  All trade-offs from a pre-digital age.  There 
are better ones now.

>> For Type 2 the
>> 'Minimum Stopband Attenuation' that you specify is traded off against
>> the attenuation at the end of the passband.  You can see that in the
>> graph if you try different values.  Some of the other comments are no
>> longer valid as the boxes are invisible.
>>
>> In terms of the filter 'order' (or 'Rolloff (dB per octave)'/6) in the
>> case of the Nyquist ones), it may be worth noting that the Nyquist
>> ones go to 8 (1, 2, 4, 6, 8), 'Classic Filters' go 1 to 10, whereas
>> the Equalization effect defaults to 4001 and can be pushed to 8191;
>> OK, not quite comparing like with like but you get the idea of how
>> powerful the Eq is compared to the others.
>
> Yes, indeed.

I think we all agree that long FIRs are 'better' than short IIRs.

>>
>> Use case:
>> Say a user wanted to recover speech from a degraded source that had
>> lots of interference, particularly at high frequencies (and they
>> recognised this).  The easiest option would be the 'Low Pass
>> Filter...' Nyquist effect (fewest parameters), then the 'Classic
>> Filters...' where they could play with the different options to reduce
>> the unwanted HF noise, then the 'Equalization...' where they have
>> near-unlimited stuff to play with.  That is a situation that users may
>> be faced with, and 'Classic' gives an extra option that could be
>> educational.
>
> The "educational value" of this example would appear to be the
> opportunity to discover that the Chebyshev filters are the worst
> choice of tool for the job.

Not always.  And even if it was, still an education, but I'm not 
wedded to it...

>>
>> The Classic's GUI has a similar level of complexity as the EQ (in
>> 'Draw Curves' mode) in terms of interpreting what they see in the
>> graph, but they need some knowledge to use that.  They need to know
>> how to use the 'Plot Spectrum...' tool and/or the Spectrogram view to
>> make best use of it, but that is surely the purpose of a tutorial,
>> rather than the manual.
>
> Yes, I agree that would be "tutorial material".

And appears to acknowledge that EQ-type effects should (if possible) 
have a graph of what they will do (???).

>>
>> I do not see any explanation in
>> http://manual.audacityteam.org/man/Low_Pass_Filter
>> or
>> http://manual.audacityteam.org/man/Equalization
>> of which we should use when and so do not really understand the
>> insistence that
>> http://manual.audacityteam.org/man/Classic_Filters
>> should have the same. :-)
>
> For Low Pass and High Pass filters, I think that the documentation
> (http://manual.audacityteam.org/man/Low_Pass_Filter) is quite clear
> that:
> "Low Pass Filter passes frequencies below its cutoff frequency and
> attenuates frequencies above its cutoff frequency. This effect can
> therefore be used to reduce high pitched noise." (and the
> complimentary equivalent for High Pass).

Yes, that's basic and clear.

> I don't see the need to go into more detail than that, until we add in
> the complication of Chebyshev vs. Butterworth.

Which would be a learning experience for the user.

> My impression with this effect is that it has been created as a "good
> idea" (yes I agree that it is a good idea, but with some reservations)
> that Audacity should have these filters, rather than to meet a real
> demand. In other words, a solution looking for a problem to solve.

Hmm, you may well be right...

> The question that I would pose is; what would this effect be like if
> it were designed from the view point of fulfilling audio production
> needs?

A very good question, but perhaps that should be 'from the point of 
view of a user', since not all users are doing 'audio production'.  We 
need to consider all users.

> If that were the design criteria I suspect that we would abandon
> "classic" analog modelling and produce a variable slope high pass/ low
> pass filter with zero ripple and probably with linear phase response.
> Such an effect could be more powerful than the current high/low pass
> filters, yet simpler (and less intimidating) than the full featured
> Equalization effect. The graphical interface could easily be extended
> to provide additional filter types, such as band pass/ band stop
> filters, shelf filters, and yes could even include "classic analog"
> modelled filters (Butterworth, Chebyshev type I, Chebyshev type II).

That's certainly a good thought, and presumably mixing FIR and IIR 
types into one interface.  The code is probably there for both, the 
interface you suggest is clearly the tricky bit :-).

> As Martyn said, just a few thoughts here. I'm not arguing against this
> effect per se, but I am questioning whether its current form is in the
> best interests of Audacity users. My view on that at present is that
> the Chebyshev filters introduce unnecessary complication that is not
> in the best interests of the vast majority of user, for very limited
> benefit for a very small minority of users, and due to bug 660 is
> inferior to what we already ship.

I think that back last Jan the 660 problem was fixed, or at least 
'worked around'.

> I would very much like to see this effect progress - it's been
> standing on the side-lines too long. A suggestion for moving this
> forward would be to initially introduce it with just low pass / high
> pass options (Butterworth, but no need for the manual to go into
> detail on that point), preferably with bug 660 resolved (though that
> could be raised to P3 and release noted).

So what do others think about that idea?

TTFN
Martyn

> Steve
>
>>
>> So, just a few thoughts here.  I'd like Classic to be in this release
>> for further comment, as you know.
>>
>> TTFN
>> Martyn
>>
>>
>> On 30/12/2014 09:54, Peter Sampson wrote:
>>> Martyn responded to Gale:
>>>   >> This is the page in the Manual as it stands:
>>>   >> http://manual.audacityteam.org/man/Classic_Filters
>>> <http://manual.audacityteam.org/man/Classic_Filters>.
>>>   >>
>>>   >> Not at all ready.
>>>   >
>>>   >I disagree.  That page isn't too bad for a new effect.  Somebody has
>>>   >put a lot of useful effort into that.
>>>
>>> Your feedback as it being "not too bad" Martyn as someone with good
>>> experience
>>> in this area of expertise is useful to hear.
>>>
>>> I think that what we on the Manual team have always felt lacking on that
>>> page was some editorial input by the effect's author - which is why
>>> both I and Gale
>>> have recently written, inviting Norm to participate.
>>>
>>> In particular what we feel is missing is some description of where and
>>> why you
>>> might choose to deploy the filers in this effect - and some use cases
>>> or examples
>>> that can be included to aid the non-expert reader.
>>>
>>> Perhaps Martyn, if Norm cannot be persuaded to engage, we could
>>> persuade you
>>> to assist with the improvement of this page?
>>>
>>> None of us on the Manual team afaik has the necessary expertise to
>>> provide this,
>>> we've already taken it as far as we can with our limited knowledge.
>>>
>>> Cheers,
>>> Peter.
>>>
>>> Peter Sampson
>>> Tel: +44 (0)1625 524 780
>>>
>>
>> ------------------------------------------------------------------------------
>> Dive into the World of Parallel Programming! The Go Parallel Website,
>> sponsored by Intel and developed in partnership with Slashdot Media, is your
>> hub for all things parallel software development, from weekly thought
>> leadership blogs to news, videos, case studies, tutorials and more. Take a
>> look and join the conversation now. http://goparallel.sourceforge.net
>> _______________________________________________
>> audacity-devel mailing list
>> aud...@li...
>> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming! The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>

Re: [Audacity-devel] Classic Filters status

[Martyn S. <mar...@gm...>]

Incidentally, can you hear the differences in these (440Hz and 880Hz 
tones), I can't.

https://dl.dropboxusercontent.com/u/1327769/phases.wav

TTFN
Martyn

Re: [Audacity-devel] Classic Filters status

[Federico M. <fm...@fc...>]

Martyn,

> Incidentally, can you hear the differences in these (440Hz and 880Hz 
> tones), I can't. https://dl.dropboxusercontent.com/u/1327769/phases.wav

Except for the last one, nobody can, according to Ohm's law for mono 
signals (you cannot hear any phase difference):

http://en.wikipedia.org/wiki/Ohm%27s_acoustic_law

In the last one, if you listen attentively, you will notice a second 
order effect that is explained in the book by Roederer, "The Physics and 
Psychophyisics of Music". It is a case where the second harmonic is not 
exactly in tune. The phase modulation is similar either to frequency 
modulation or a mistuned second harmonic, for instance 440 Hz and 880.44 Hz.

As regards phase differences that depart from linear phase with 
frequency, it is generally accepted that they can be noted in stereo 
signals since they represent different delays for different frequencies. 
This affects the sharpness of the stereo image. Probably most people 
cannot tell the difference, but experienced audiophiles can.

But the subtle difference probably cannot be perceived at all in the 
presence of fan noise from the computer if not acoustically isolated.

In the case of the antialiasing filters used in legacy A/D converters, 
I'm not sure if the audibility of this phase problem is not really due 
to the channel mismatch.

A 1 kHz order 10 lowpass Chebyshev has at 1 kHz a phase delay of 720º, 
about 2 ms, while at 100 Hz it has a delay of 0.13 ms. That means that 
there is a difference of about 1.9 ms, which corresponds to a distance 
of about 2 ft, that's why the source may, in theory, seem blurred.

Regards,

Federico

[Audacity-devel] was: 'Re: Classic Filters status', now a little OT

[Martyn S. <mar...@gm...>]

Hiya

On 03/01/2015 08:13, Federico Miyara wrote:
>
> Martyn,
>
>> Incidentally, can you hear the differences in these (440Hz and 880Hz
>> tones), I can't. https://dl.dropboxusercontent.com/u/1327769/phases.wav
>
> Except for the last one, nobody can, according to Ohm's law for mono
> signals (you cannot hear any phase difference):
>
> http://en.wikipedia.org/wiki/Ohm%27s_acoustic_law

I believe nobody can, but it seems contested.  If the phase changes 
over a second or less, certainly the difference can be heard.

> In the last one, if you listen attentively, you will notice a second
> order effect that is explained in the book by Roederer, "The Physics and
> Psychophyisics of Music". It is a case where the second harmonic is not
> exactly in tune. The phase modulation is similar either to frequency
> modulation or a mistuned second harmonic, for instance 440 Hz and 880.44 Hz.

I can't hear it, can you?

How did you get the 880.44 Hz?  I'm interested to hear your analysis.

> As regards phase differences that depart from linear phase with
> frequency, it is generally accepted that they can be noted in stereo
> signals since they represent different delays for different frequencies.
> This affects the sharpness of the stereo image. Probably most people
> cannot tell the difference, but experienced audiophiles can.

Hmm.  Can they really?  I'm a bit sceptical, since as I've got older 
and better at listening, my hearing has got worse.  And are these 
'experienced audiophiles' lucky or not? is something I've often wondered.

'generally accepted' references would be nice.

> But the subtle difference probably cannot be perceived at all in the
> presence of fan noise from the computer if not acoustically isolated.

Can you hear it?

> In the case of the antialiasing filters used in legacy A/D converters,
> I'm not sure if the audibility of this phase problem is not really due
> to the channel mismatch.
>
> A 1 kHz order 10 lowpass Chebyshev has at 1 kHz a phase delay of 720º,
> about 2 ms, while at 100 Hz it has a delay of 0.13 ms. That means that
> there is a difference of about 1.9 ms, which corresponds to a distance
> of about 2 ft, that's why the source may, in theory, seem blurred.

OK, it's a theory.  Do you (or anybody else) have a procedure that I 
can follow to demonstrate the difference in that delay and so (maybe) 
be able to hear the problem?  Something that we can do in Audacity by 
generating waveforms, applying filters and so on? I'd love to try that 
out on my young-eared students.

Thanks
Martyn

> Regards,
>
> Federico
>
>
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming! The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>

Re: [Audacity-devel] was: 'Re: Classic Filters status', now a little OT

[Federico M. <fm...@fc...>]

Martin,

> I believe nobody can, but it seems contested. If the phase changes 
> over a second or less, certainly the difference can be heard.

This is the sort of case discussed by Roederer.

> I can't hear it, can you?

Yes, I can hear some modulation.

> How did you get the 880.44 Hz? I'm interested to hear your analysis.

Sorry to disappoint you, it's very basic. I just looked at the waveform 
o rather the envelope and observed that it took about 2.5 s to complete 
sort of a cycle, then I generated a mistuned harmonic multiplyng by 
2.001 (which made a beat about that long with 2.000*f). The envelope is 
similar, not identical, but enough to test the idea. The same as in the 
case of frequency modulation, the spectrum of a very slight FM is very 
similar as the spectrum of an amplitude modulation (one tone and two 
side bands, with the only phase differences)

> Hmm. Can they really? 

I can't tell... Haven't run a double blind test. But more than once I 
have been surprised by people systematically hearing things I did not.

> I'm a bit sceptical, since as I've got older and better at listening, 
> my hearing has got worse. And are these 'experienced audiophiles' 
> lucky or not? is something I've often wondered. 'generally accepted' 
> references would be nice.

I don't have any at hand right now, but now and then I come across some 
comment of the like. I'll pay attention next time and send it to you.

>> But the subtle difference probably cannot be perceived at all in the 
>> presence of fan noise from the computer if not acoustically isolated. 
> Can you hear it?

I don't have an adequate environment. I don't believe even that I am the 
best subject to test these subtleties, since I have some tinnitus

>> A 1 kHz order 10 lowpass Chebyshev has at 1 kHz a phase delay of 
>> 720º, about 2 ms, while at 100 Hz it has a delay of 0.13 ms. That 
>> means that there is a difference of about 1.9 ms, which corresponds 
>> to a distance of about 2 ft, that's why the source may, in theory, 
>> seem blurred. 
> OK, it's a theory. Do you (or anybody else) have a procedure that I 
> can follow to demonstrate the difference in that delay and so (maybe) 
> be able to hear the problem? Something that we can do in Audacity by 
> generating waveforms, applying filters and so on? I'd love to try that 
> out on my young-eared students.

OK, I'll try to think of some test.

Regards,

Federico

Re: [Audacity-devel] Classic Filters status

[Steve t. F. <ste...@gm...>]

On 30 December 2014 at 03:40, Federico Miyara <fm...@fc...> wrote:
>
> Dear all,
>
> As regards classic filters, there is a method to get a closer
> approximation to the real continuous-time filters than the bilinear
> transformation. This is to find the analytic version of the impulse
> response, compute it at the sampling intervals and convolve it with the
> signal. The interesting thing is that this method is much more general
> than just classic filters. It is possible to simulate any system with
> known poles and zeros. So the full range of classic, "esoteric" and
> vintage filters can be easily implemented.
>
> One can compute the impulse response up to a point where it falls under
> -60 dB of the maximum which for typical filters amounts to less than one
> second. Not only us the frequency response but also the transient
> response is accurately simulated.
>
> I'm currently testing this approach with Matlab.

Sounds interesting. I look forward to hearing more about this.


>
> Regards,
>
> Federico
>
> On 29/12/2014 22:21, Martyn Shaw wrote:
>> Hi. I guess Gale is where this is aimed primarily...
>>
>> On 28/12/2014 15:28, Gale wrote:
>>> Hi Norm,
>>>
>>> I'm not strongly opposed but do see docs and use cases as an outstanding
>>> issue.
>> 'Classic Filters' (that is, traditional analog IIR filters implemented
>> in the digital domain) are just as valid filters, to a user, as the
>> rather crude pre-sets available in the current Equalization effect.
>> But read on.
>>
>> I do not agree that documentation is an issue for this new effect.

There are a number of issues in the editor notes on that page for
which we currently don't have answers. As far as possible we like to
resolve editor note issues before release. Unless the editor note
issues are "P1" (or "P2" at release manager's discretion) we may
release without resolving the issues, but the presence of lots of
editor notes in an article is a warning sign that the documentation is
not up to scratch.


>>
>>> Norm C wrote
>>>> I've been out of touch for a few months but I gather there's been some
>>>> rather polarized discussions regarding the Classic Filters and they're now
>>>> out of the build. In my view that's unfortunate. I'll summarize what I see
>>>> as the issues:
>>>>
>>>> - badly named: I agree, that's why I called it Scientific Filters when I
>>>> submitted it. There's ample justification for this name:
>>>> CoolEditPro/Audition calls the same effect Scientific Filters, and Matlab
>>>> (certainly a major scientific application) uses these same filters
>>>> (butter, cheby1 and cheby2) as its workhorse filters
>> Agreed, workhorse filters.  Should be in the build.
>>
>>> I always felt to justify "Scientific" moniker there ought to be more
>>> esoteric
>>> filters e.g. Elliptic, Gaussian and any others that could be useful for
>>> modelling
>>> or educational/research interests (which you give as a use case).
>> Well, Gale, I don't think you have the background to discuss Elliptic
>> filters, or any others.  The ones selected here are the most common by
>> a long way, I think.

but other than Butterworth, they are not commonly used in audio
production software.

I think that we all agree that the primary purpose of Audacity
software is for audio production, so it would seem strange to me if we
shipped "obscure" (in audio production) filters (such as the Chebyshev
type II) and not include much more common (in audio production) ones
(such as band pass, band stop, sinc, Linkwitz-Riley...)

As I wrote previously, I can completely see why we would want to
include the Chebyshev filters in SPICE software, but I still don't
know the answer to the audio production question "when and why should
I use a Chebyshev filter?" For most effects the manual says why/when
an effect would be used - a typical job - we still don't have that for
Chebyshev filters.

Steve

>>
>>> Do you or anyone buy into that? Even if you do, I guess new filters would
>>> have to wait until release after next.
>> Why would it have to wait?  More esoteric filters could, obviously, be
>> added.  Maybe you could find yet another word on wikipedia and hold
>> that up longer, making the effect more complex and documentation
>> demands higher.  You say that you are "not strongly opposed", so still
>> opposed to this being released, and making the barrier still higher.
>>
>>> Norm C wrote
>>>> - the filters are "unorthodox": Certainly not - as I pointed out,
>>>> CEP/Audition and Matlab use these same filters (with the same behaviour
>>>> near the Nyquist frequency), as do many other audio applications (eg.
>>>> Logic, Goldwave and Wavosaur). Yes, the high frequency response is
>>>> different from the other filters in Audacity, but it's good to have
>>>> choices.
>>> Yes but the greater HF attenuation in your first order Low Pass compared
>>> to docs that most users will find or understand is a drawback if we're
>>> presenting these as "Classic".
>> I disagree, as I have before.  Just because you, Gale, don't have a
>> thorough understanding of DSP, it does not mean that users won't find
>> these filters useful.
>>
>>> Alternatives aren't usually "classic" though I understand the sense in
>>> which these are "classic".
>> No, I don't think you do understand Gale.  Please read, for example,
>> http://web.mit.edu/~cjoseph/Public/Discrete-Time_Signal_Processing_-_Oppenheim_-_2nd_Edition.pdf
>> especially chapter 7 on 'Filter Design Techniques', or trust that
>> people who do understand that stuff know what they are doing.
>>
>>> Norm C wrote
>>>> - documentation is inadequate: OK. CEP's Scientific Filter documentation
>>>> would be a good place to start. And I provided a fair amount of commentary
>>>> on the filter when I first submitted it. I'd be glad to provide more -
>>>> just let me know exactly what needs to be addressed.
>>> Yes this is the main problem (and the first order Low Pass question above
>>> would I think need a sentence in the docs).
>>>
>>> This is the page in the Manual as it stands:
>>> http://manual.audacityteam.org/man/Classic_Filters .
>>>
>>> Not at all ready.
>> I disagree.  That page isn't too bad for a new effect.  Somebody has
>> put a lot of useful effort into that.
>>
>> Being quite close to 2.1.0, I feel you would need to work
>>> directly on that page yourself (I can sign you up there). You would need to
>>> get it passed by someone on the Manual team who only had average
>>> competence in filtering, to verify that it was written in basic enough
>>> fashion
>>> and said why one would want to use these.
>> That sounds fair enough.  Make sure that the manual page is
>> understandable by the people who usually write those pages.  It looks
>> OK to me.
>>
>> TTFN
>> Martyn
>>
>>> I know quite a bit of information has been given already about "why" but
>>> it's quite scattered around in different threads.
>>>
>>>
>>> Norm C wrote
>>>> - doesn't work properly on tracks with multiple sample rates (same problem
>>>> as Equalization has, I think): True, and I don't think there's a simple
>>>> consistent way to make it work, as the filter response varies with sample
>>>> rate. At a minimum we'd need to display two different frequency responses
>>>> on the same graph and I think that would be both complex and confusing to
>>>> the user. In my view the solution is to prevent the user from applying the
>>>> same filter simultaneously to tracks with multiple sample rates and
>>>> explain what he needs to do.
>>> That is already addressed so isn't an obstacle. There is an error message
>>> that "all selected tracks must be the same sample rate."
>>>
>>>
>>> Gale
>>>
>>>
>>> Norm C wrote
>>>> - Chebyshev filters are too uncommon to be used here: Certainly they're
>>>> less common than Butterworth but it's far from unheard of to find them in
>>>> audio gear, both analog and digital. I've got lots of examples if anyone
>>>> cares.
>>>>
>>>> Steve, Gale: you may still have concerns that I've missed here. Let me
>>>> know if so, I'll try to address them.
>>>>
>>>> Norm
>>>
>>>
>>>
>>>
>>> --
>>> View this message in context: http://audacity.238276.n2.nabble.com/Classic-Filters-status-tp7565152p7566434.html
>>> Sent from the audacity-devel mailing list archive at Nabble.com.
>>>

Re: [Audacity-devel] Classic Filters status

[Steve t. F. <ste...@gm...>]

On 30 December 2014 at 13:45, Federico Miyara <fm...@fc...> wrote:
>
> Steve,
>
>
> but other than Butterworth, they are not commonly used in audio production
> software.
>
>
> Probably not even Butterworth is actually used, since it doesn't ensure
> linear phase response as FIR (finite impulse response) does.

Nevertheless, digital implementations of Butterworth filters are still
very widely used in audio production (for example, for rolling off low
bass in microphone recordings).

>
> I think that we all agree that the primary purpose of Audacity software is
> for audio production,
>
>
> When any software reaches certain level of sophistication its usage range
> widens a lot. I think it is the case of Audacity. Many people around the
> world use it for teaching rather than just audio production. Goldwave, for
> instance, more than a decade ago, included means to implement any kind of
> digital filter that can be specified in terms of its coefficients.
> Flexibility is a very appreciated feature by the one-percent people that
> will use that kind of feature.
>
> so it would seem strange to me if we shipped "obscure" (in audio production)
> filters (such as the Chebyshev type II) and not include much more common (in
> audio production) ones (such as band pass, band stop, sinc,
> Linkwitz-Riley...)
>
>
> Band pass may also be Butterworth or even Chebyshev (they are used in
> spectral analyzers). Sinc is used for signal reconstruction and for signal
> oscillogram rendering, mainly in digital oscilloscopes (audio software uses
> some kind of spline, since the image doesn't need to be that accurate).
> Linkwitz-Riley is used in speaker crossovers, not in audio processing. They
> cpould be used to simulate the result of crossover.

Sinc filters are useful in audio production to avoid aliasing
distortion (for example, when speeding up a recording).
Digital implementations of Linkwitz-Riley filters may be used for
splitting out an LF channel for multi-channel audio.

It seems a simple and obvious question, but so far no-one has provided
an answer - why would I choose to use a Chebyshev filter rather than
any other type of filter? (clearly, component count for hardware
implementation is irrelevant in Audacity). It bothers me that we are
considering adding a feature for which there is apparently no
practical use.


>
> As I wrote previously, I can completely see why we would want to include the
> Chebyshev filters in SPICE software, but I still don't know the answer to
> the audio production question "when and why should I use a Chebyshev
> filter?" For most effects the manual says why/when an effect would be used -
> a typical job - we still don't have that for Chebyshev filters.
>
>
> Chebyshev has been used in vintage and low-cost antialiasing filters. I
> think it is not frequently used nowadays since sigma-delta converters are
> much mor common. They could be used for demonstrating the effect of analog
> Chebyshev implementations, i.e., for filter simulation.

Surely there is much better software for analog circuit simulation, for example:
http://en.wikipedia.org/wiki/List_of_free_electronics_circuit_simulators

Steve


>
> Regards,
>
> Federico
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming! The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>

Re: [Audacity-devel] Classic Filters status

[Federico M. <fm...@fc...>]

Steve,

Your examples are good, but I just would like to add some comments.

> Nevertheless, digital implementations of Butterworth filters are still 
> very widely used in audio production (for example, for rolling off low 
> bass in microphone recordings).

This is an example of a "vintage" application. With a digital filter it 
is possible to perfectly compensate, at a given distance (which could be 
an input parameter), the proximity effect. Butterworth cannot do that; 
moreover, microphone hardware roll-off usually features a single cutoff 
frequency. As regards to just removing rumble or low frequency hum, any 
linear phase FIR filter can do a better job over the pass band.

What I mean is that Butterworth is a valid choice but not the only one, 
and not even the best.

> Sinc filters are useful in audio production to avoid aliasing 
> distortion (for example, when speeding up a recording).

Correct, but note that it is impossible to get a true sinc filter, they 
are always windowed sinc, so some amount of aliasing is to be present. 
Chebyshev can be used with the same purpose since it attains a very fast 
roll-off close to the Nyquist limit with a smaller order.

Maybe the order seems unimportant but it could matter in real time 
applications. In analog breadboard circuit, order is equivalent to 
component count. In the digital counterpart, high order means inefficiency.

Matlab uses, for the purpose of resampling, a filter based on the Kaiser 
window. Probably this is more computationally efficient than the sinc 
filter (though, arguably, any nearly brick-wall filter may be called a 
sinc filter...)

> Digital implementations of Linkwitz-Riley filters may be used for 
> splitting out an LF channel for multi-channel audio.

Linkwitz-Riley is frequently two Butterworth LP in cascade and two 
Butterworth HP in cascade. The interesting thing is that time alignment 
could be easilly accomplished by means of a simple delay (which could be 
required as a parameter). What cannot be accomplished is perfect 
equalization of loudspeaker frequency response, so often L-R is just a 
theoretical ilusion.

> It seems a simple and obvious question, but so far no-one has provided 
> an answer - why would I choose to use a Chebyshev filter rather than 
> any other type of filter? (clearly, component count for hardware 
> implementation is irrelevant in Audacity). It bothers me that we are 
> considering adding a feature for which there is apparently no 
> practical use.

I think we shouldn't pose the quest in terms of "rather than any other", 
since the perfect filter for a given application doesn't exist. It 
suffices that there be some situations where the choice is a valid option.

The most obvious one is to simulate, not a circuit (for which, as you 
say, there are many circuit simulators), but the /behavior/ of the 
circuit or system, for instance an antialiasing filter or a highly 
selective bandpass filter as some used in spectrum analyzers. Now that 
you have mentioned Linkwitz-Riley, Chebyshev is one of the possible 
alignments in Thiele-Small theory of loudspeaker vented boxes. See, for 
instance, this paper by Richard Small: 
http://diyaudioprojects.com/Technical/Papers/Vented-Box-Loudspeaker-Systems-Part-IV.pdf

I hope you finally consider that this may be a start for an answer to 
your question.

>> As I wrote previously, I can completely see why we would want to 
>> include the Chebyshev filters in SPICE software, but I still don't 
>> know the answer to the audio production question "when and why should 
>> I use a Chebyshev filter?" For most effects the manual says why/when 
>> an effect would be used - a typical job - we still don't have that 
>> for Chebyshev filters.

Audio production may be the main intended application of Audacity, but 
it isn't the only one.

> Surely there is much better software for analog circuit simulation, 
> for example: 
> http://en.wikipedia.org/wiki/List_of_free_electronics_circuit_simulators

Circuit simulators are not intended for simulating a mathematical 
formula, but to include in the analysis the full range of non-idealities 
of real-life components, including nonlinear behavior, thermal effects, 
electrical noise, Montecarlo analysis of tolerance and so on. On the 
other hand, this horse power is very time-consuming, so it may take 
quite long to simulate the effect on several minutes of audio. Circuit 
simulators excel in computing the frequency response, the noise 
behavior, distortion and so on.

Regards,

Federico

Re: [Audacity-devel] Classic Filters status

[Federico M. <fm...@fc...>]

Steve,

> That makes sense where component count, physical size and
> manufacturing costs of analog filters are relevant, but surely it is a
> spurious trade off for digital filters. Yes a 2nd order Chebyshev
> filter may have a slightly steeper roll-off compared with a second
> order Butterworth, (at the expense of ripple in either the pass band
> or stop band), but for digital filters there is absolutely no need for
> that compromise - if you want a steeper roll-off, use a higher order
> Butterworth filter and then you get a steeper roll-off with no ripple.

Higher order in Butterworth means more phase distortion (as in any 
analog filter), what in turn means worse transient response (for 
instance, more ringing. It ios true that for the same order Chebyshev is 
worse than Butterworth, but a Butterworth of larger order has worse 
transient behavior,

Regards,

Federico

Re: [Audacity-devel] Classic Filters status

[Steve t. F. <ste...@gm...>]

On 2 January 2015 at 19:24, Federico Miyara <fm...@fc...> wrote:
>
> Steve,
>
> That makes sense where component count, physical size and
> manufacturing costs of analog filters are relevant, but surely it is a
> spurious trade off for digital filters. Yes a 2nd order Chebyshev
> filter may have a slightly steeper roll-off compared with a second
> order Butterworth, (at the expense of ripple in either the pass band
> or stop band), but for digital filters there is absolutely no need for
> that compromise - if you want a steeper roll-off, use a higher order
> Butterworth filter and then you get a steeper roll-off with no ripple.
>
>
> Higher order in Butterworth means more phase distortion (as in any analog
> filter), what in turn means worse transient response (for instance, more
> ringing. It ios true that for the same order Chebyshev is worse than
> Butterworth, but a Butterworth of larger order has worse transient behavior,

Sorry but that is just not true, but it does demonstrate the point
that including Chebyshev filters makes the effect too complicated and
confusing for the vast majority of users (including many very
experienced users).

Buterworth filters have a more linear phase response in the pass band
than either type I or type II Chebyshev filters. A type 1 Chebyshev
low pass filter will ring more at frequencies a little below the
corner frequency than a Butterworth filter, and a type 1 Chebyshev
high pass will ring more at frequencies a little above the corner
frequency than a Butterworth filter.

Type 2 Chebyshev filters ring less near the corner frequency, but at
the expense of significantly worse roll-off characteristics and
stop-band performance.

Stephen Butterworth owes his immortal fame to finding the sweet spot
between type I and type II, where an IIR filter of a given order most
closely resembles the ideal high/low pass.
http://en.wikipedia.org/wiki/Butterworth_filter

Attached is a screen shot demonstrating the ringing effect of a 4th
order Butterworth high pass filter compared with 3rd order Chebyshev
type I and type II filters on a 2 ms noise pulse (Waveform (dB) view).

Steve

>
> Regards,
>
> Federico
>
>
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming! The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net
> _______________________________________________
> audacity-devel mailing list
> aud...@li...
> https://lists.sourceforge.net/lists/listinfo/audacity-devel
>

Re: [Audacity-devel] Classic Filters status

[Steve t. F. <ste...@gm...>]

On 2 January 2015 at 21:52, Steve the Fiddle <ste...@gm...> wrote:
> On 2 January 2015 at 19:24, Federico Miyara <fm...@fc...> wrote:
>>
>> Steve,
>>
>> That makes sense where component count, physical size and
>> manufacturing costs of analog filters are relevant, but surely it is a
>> spurious trade off for digital filters. Yes a 2nd order Chebyshev
>> filter may have a slightly steeper roll-off compared with a second
>> order Butterworth, (at the expense of ripple in either the pass band
>> or stop band), but for digital filters there is absolutely no need for
>> that compromise - if you want a steeper roll-off, use a higher order
>> Butterworth filter and then you get a steeper roll-off with no ripple.
>>
>>
>> Higher order in Butterworth means more phase distortion (as in any analog
>> filter), what in turn means worse transient response (for instance, more
>> ringing. It ios true that for the same order Chebyshev is worse than
>> Butterworth, but a Butterworth of larger order has worse transient behavior,
>
> Sorry but that is just not true, but it does demonstrate the point
> that including Chebyshev filters makes the effect too complicated and
> confusing for the vast majority of users (including many very
> experienced users).
>
> Buterworth filters have a more linear phase response in the pass band
> than either type I or type II Chebyshev filters. A type 1 Chebyshev
> low pass filter will ring more at frequencies a little below the
> corner frequency than a Butterworth filter, and a type 1 Chebyshev
> high pass will ring more at frequencies a little above the corner
> frequency than a Butterworth filter.
>
> Type 2 Chebyshev filters ring less near the corner frequency, but at
> the expense of significantly worse roll-off characteristics and
> stop-band performance.
>
> Stephen Butterworth owes his immortal fame to finding the sweet spot
> between type I and type II, where an IIR filter of a given order most
> closely resembles the ideal high/low pass.
> http://en.wikipedia.org/wiki/Butterworth_filter
>
> Attached is a screen shot demonstrating the ringing effect of a 4th
> order Butterworth high pass filter compared with 3rd order Chebyshev
> type I and type II filters on a 2 ms noise pulse (Waveform (dB) view).

For completeness, I've attached a "Spectrum (log f)" view of the same
impulses to illustrate the low frequency roll-off of the different
filter types.

Steve


>
> Steve
>
>>
>> Regards,
>>
>> Federico
>>
>>