Applying Nyquist effects to regions containing envelope points may corrupt track

classic Classic list List threaded Threaded
13 messages Options
Reply | Threaded
Open this post in threaded view
|

Applying Nyquist effects to regions containing envelope points may corrupt track

Stevethefiddle
Repeatable example on Linux:

1) Generate a default Chirp
2) Add an envelope point at about 0.8 / 15 seconds
3) Select from about 10 to 20 seconds
4) Apply Low Pass filter.

Track becomes flat line from start of selection to end of the track.

I assume that this is caused by the changes to envelope point handling
that Paul has been working on, as is the debug crash that I reported
on Windows.

Paul, are you on to this or does it need logging on bugzilla? I'd say
it's a P1 regression issue if we're logging it.

Steve

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Gale
Administrator
On Windows the selection suffers a large fade in to distortion
levels from the start of the selection to the end of the track
(also seen applying a flat EQ curve).



Gale


On 7 June 2017 at 10:56, Steve the Fiddle <[hidden email]> wrote:

> Repeatable example on Linux:
>
> 1) Generate a default Chirp
> 2) Add an envelope point at about 0.8 / 15 seconds
> 3) Select from about 10 to 20 seconds
> 4) Apply Low Pass filter.
>
> Track becomes flat line from start of selection to end of the track.
>
> I assume that this is caused by the changes to envelope point handling
> that Paul has been working on, as is the debug crash that I reported
> on Windows.
>
> Paul, are you on to this or does it need logging on bugzilla? I'd say
> it's a P1 regression issue if we're logging it.
>
> Steve
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Audacity-quality mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/audacity-quality

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Paul Licameli


On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews <[hidden email]> wrote:
On Windows the selection suffers a large fade in to distortion
levels from the start of the selection to the end of the track
(also seen applying a flat EQ curve).

Which commit, and what steps to reproduce, does this describe?
PRL

 



Gale


On 7 June 2017 at 10:56, Steve the Fiddle <[hidden email]> wrote:
> Repeatable example on Linux:
>
> 1) Generate a default Chirp
> 2) Add an envelope point at about 0.8 / 15 seconds
> 3) Select from about 10 to 20 seconds
> 4) Apply Low Pass filter.
>
> Track becomes flat line from start of selection to end of the track.
>
> I assume that this is caused by the changes to envelope point handling
> that Paul has been working on, as is the debug crash that I reported
> on Windows.
>
> Paul, are you on to this or does it need logging on bugzilla? I'd say
> it's a P1 regression issue if we're logging it.
>
> Steve
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Audacity-quality mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/audacity-quality

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Gale
Administrator
On 8 June 2017 at 01:39, Paul Licameli <[hidden email]> wrote:

>
>
> On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews <[hidden email]> wrote:
>>
>> On Windows the selection suffers a large fade in to distortion
>> levels from the start of the selection to the end of the track
>> (also seen applying a flat EQ curve).
>
>
> Which commit

Before your fix "01450e8".


>  and what steps to reproduce, does this describe?

Same as Steve's steps below, with am optional variant of Flat EQ
instead of Low Pass.

After your fix, on Linux and Windows, I see after Step 4 below that
the blue envelope has risen to 1.0 amplitude between 0 seconds
and 10 seconds.

As you know I am not an expert in envelopes, but why as a user would
I want the audio before the selection to be boosted when I apply Low
Pass or EQ to the selection?  2.1.3 release does not do that.


Gale


>> On 7 June 2017 at 10:56, Steve the Fiddle <[hidden email]>
>> wrote:
>> > Repeatable example on Linux:
>> >
>> > 1) Generate a default Chirp
>> > 2) Add an envelope point at about 0.8 / 15 seconds
>> > 3) Select from about 10 to 20 seconds
>> > 4) Apply Low Pass filter.
>> >
>> > Track becomes flat line from start of selection to end of the track.
>> >
>> > I assume that this is caused by the changes to envelope point handling
>> > that Paul has been working on, as is the debug crash that I reported
>> > on Windows.
>> >
>> > Paul, are you on to this or does it need logging on bugzilla? I'd say
>> > it's a P1 regression issue if we're logging it.
>> >
>> > Steve
>> >
>> >
>> > ------------------------------------------------------------------------------
>> > Check out the vibrant tech community on one of the world's most
>> > engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> > _______________________________________________
>> > Audacity-quality mailing list
>> > [hidden email]
>> > https://lists.sourceforge.net/lists/listinfo/audacity-quality
>>
>>
>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> _______________________________________________
>> Audacity-quality mailing list
>> [hidden email]
>> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>
>
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Audacity-quality mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Paul Licameli


On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews <[hidden email]> wrote:
On 8 June 2017 at 01:39, Paul Licameli <[hidden email]> wrote:
>
>
> On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews <[hidden email]> wrote:
>>
>> On Windows the selection suffers a large fade in to distortion
>> levels from the start of the selection to the end of the track
>> (also seen applying a flat EQ curve).
>
>
> Which commit

Before your fix "01450e8".


>  and what steps to reproduce, does this describe?

Same as Steve's steps below, with am optional variant of Flat EQ
instead of Low Pass.

After your fix, on Linux and Windows, I see after Step 4 below that
the blue envelope has risen to 1.0 amplitude between 0 seconds
and 10 seconds.

As you know I am not an expert in envelopes, but why as a user would
I want the audio before the selection to be boosted when I apply Low
Pass or EQ to the selection?  2.1.3 release does not do that.

I see all is not yet working as intended.  I think I got everything right when the exiting envelope has more than one point, but the case of only one point  needs work.

In 2.1.3 the effect applies to the waveform as it is without the envelope, and loses envelope information for that selection, and replaces it with a default level portion of envelope.

Right or wrong, that should be so for many effects besides Nyquist, and that is not meant to change.

What is meant to change is the treatment of discontinuities.  If you zoom very close, you should see a true vertical in 2.2.0 and not the short steep slope of 2.1.3.

PRL

 


Gale


>> On 7 June 2017 at 10:56, Steve the Fiddle <[hidden email]>
>> wrote:
>> > Repeatable example on Linux:
>> >
>> > 1) Generate a default Chirp
>> > 2) Add an envelope point at about 0.8 / 15 seconds
>> > 3) Select from about 10 to 20 seconds
>> > 4) Apply Low Pass filter.
>> >
>> > Track becomes flat line from start of selection to end of the track.
>> >
>> > I assume that this is caused by the changes to envelope point handling
>> > that Paul has been working on, as is the debug crash that I reported
>> > on Windows.
>> >
>> > Paul, are you on to this or does it need logging on bugzilla? I'd say
>> > it's a P1 regression issue if we're logging it.
>> >
>> > Steve
>> >
>> >
>> > ------------------------------------------------------------------------------
>> > Check out the vibrant tech community on one of the world's most
>> > engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> > _______________________________________________
>> > Audacity-quality mailing list
>> > [hidden email]
>> > https://lists.sourceforge.net/lists/listinfo/audacity-quality
>>
>>
>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> _______________________________________________
>> Audacity-quality mailing list
>> [hidden email]
>> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>
>
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Audacity-quality mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Paul Licameli


On Thu, Jun 8, 2017 at 2:02 PM, Paul Licameli <[hidden email]> wrote:


On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews <[hidden email]> wrote:
On 8 June 2017 at 01:39, Paul Licameli <[hidden email]> wrote:
>
>
> On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews <[hidden email]> wrote:
>>
>> On Windows the selection suffers a large fade in to distortion
>> levels from the start of the selection to the end of the track
>> (also seen applying a flat EQ curve).
>
>
> Which commit

Before your fix "01450e8".


>  and what steps to reproduce, does this describe?

Same as Steve's steps below, with am optional variant of Flat EQ
instead of Low Pass.

After your fix, on Linux and Windows, I see after Step 4 below that
the blue envelope has risen to 1.0 amplitude between 0 seconds
and 10 seconds.

As you know I am not an expert in envelopes, but why as a user would
I want the audio before the selection to be boosted when I apply Low
Pass or EQ to the selection?  2.1.3 release does not do that.

I see all is not yet working as intended.  I think I got everything right when the exiting envelope has more than one point, but the case of only one point  needs work.

In 2.1.3 the effect applies to the waveform as it is without the envelope, and loses envelope information for that selection, and replaces it with a default level portion of envelope.

Right or wrong, that should be so for many effects besides Nyquist, and that is not meant to change.

What is meant to change is the treatment of discontinuities.  If you zoom very close, you should see a true vertical in 2.2.0 and not the short steep slope of 2.1.3.

PRL

Gale, Steve, please try it again at 33c9aa9
PRL
 

 


Gale


>> On 7 June 2017 at 10:56, Steve the Fiddle <[hidden email]>
>> wrote:
>> > Repeatable example on Linux:
>> >
>> > 1) Generate a default Chirp
>> > 2) Add an envelope point at about 0.8 / 15 seconds
>> > 3) Select from about 10 to 20 seconds
>> > 4) Apply Low Pass filter.
>> >
>> > Track becomes flat line from start of selection to end of the track.
>> >
>> > I assume that this is caused by the changes to envelope point handling
>> > that Paul has been working on, as is the debug crash that I reported
>> > on Windows.
>> >
>> > Paul, are you on to this or does it need logging on bugzilla? I'd say
>> > it's a P1 regression issue if we're logging it.
>> >
>> > Steve
>> >
>> >
>> > ------------------------------------------------------------------------------
>> > Check out the vibrant tech community on one of the world's most
>> > engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> > _______________________________________________
>> > Audacity-quality mailing list
>> > [hidden email]
>> > https://lists.sourceforge.net/lists/listinfo/audacity-quality
>>
>>
>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> _______________________________________________
>> Audacity-quality mailing list
>> [hidden email]
>> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>
>
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Audacity-quality mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality



------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Paul Licameli


On Thu, Jun 8, 2017 at 2:41 PM, Paul Licameli <[hidden email]> wrote:


On Thu, Jun 8, 2017 at 2:02 PM, Paul Licameli <[hidden email]> wrote:


On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews <[hidden email]> wrote:
On 8 June 2017 at 01:39, Paul Licameli <[hidden email]> wrote:
>
>
> On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews <[hidden email]> wrote:
>>
>> On Windows the selection suffers a large fade in to distortion
>> levels from the start of the selection to the end of the track
>> (also seen applying a flat EQ curve).
>
>
> Which commit

Before your fix "01450e8".


>  and what steps to reproduce, does this describe?

Same as Steve's steps below, with am optional variant of Flat EQ
instead of Low Pass.

After your fix, on Linux and Windows, I see after Step 4 below that
the blue envelope has risen to 1.0 amplitude between 0 seconds
and 10 seconds.

As you know I am not an expert in envelopes, but why as a user would
I want the audio before the selection to be boosted when I apply Low
Pass or EQ to the selection?  2.1.3 release does not do that.

I see all is not yet working as intended.  I think I got everything right when the exiting envelope has more than one point, but the case of only one point  needs work.

In 2.1.3 the effect applies to the waveform as it is without the envelope, and loses envelope information for that selection, and replaces it with a default level portion of envelope.

Right or wrong, that should be so for many effects besides Nyquist, and that is not meant to change.

What is meant to change is the treatment of discontinuities.  If you zoom very close, you should see a true vertical in 2.2.0 and not the short steep slope of 2.1.3.

PRL

Gale, Steve, please try it again at 33c9aa9
PRL

Please note too that one purpose of this project with envelopes is that you you cut then immediately paste again in a place where the envelope is more complicated than a single level, then the net result should be no change; this was not so in 2.1.3.  Try it to confirm that remains so with the recent fix.

If you do so with an envelope that has only one point, then the envelope is also unchanged in case the selection does not include that point.

Ah, but now I see that if it does include the point, then the cut-and-pase results in one more, unnecessary point.

I may try to do something about this minor case too.

PRL

 
 

 


Gale


>> On 7 June 2017 at 10:56, Steve the Fiddle <[hidden email]>
>> wrote:
>> > Repeatable example on Linux:
>> >
>> > 1) Generate a default Chirp
>> > 2) Add an envelope point at about 0.8 / 15 seconds
>> > 3) Select from about 10 to 20 seconds
>> > 4) Apply Low Pass filter.
>> >
>> > Track becomes flat line from start of selection to end of the track.
>> >
>> > I assume that this is caused by the changes to envelope point handling
>> > that Paul has been working on, as is the debug crash that I reported
>> > on Windows.
>> >
>> > Paul, are you on to this or does it need logging on bugzilla? I'd say
>> > it's a P1 regression issue if we're logging it.
>> >
>> > Steve
>> >
>> >
>> > ------------------------------------------------------------------------------
>> > Check out the vibrant tech community on one of the world's most
>> > engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> > _______________________________________________
>> > Audacity-quality mailing list
>> > [hidden email]
>> > https://lists.sourceforge.net/lists/listinfo/audacity-quality
>>
>>
>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> _______________________________________________
>> Audacity-quality mailing list
>> [hidden email]
>> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>
>
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Audacity-quality mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality




------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Gale
Administrator
On 8 June 2017 at 19:49, Paul Licameli <[hidden email]> wrote:

>
>
> On Thu, Jun 8, 2017 at 2:41 PM, Paul Licameli <[hidden email]>
> wrote:
>>
>>
>>
>> On Thu, Jun 8, 2017 at 2:02 PM, Paul Licameli <[hidden email]>
>> wrote:
>>>
>>>
>>>
>>> On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews <[hidden email]>
>>> wrote:
>>>>
>>>> On 8 June 2017 at 01:39, Paul Licameli <[hidden email]> wrote:
>>>> >
>>>> >
>>>> > On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews <[hidden email]>
>>>> > wrote:
>>>> >>
>>>> >> On Windows the selection suffers a large fade in to distortion
>>>> >> levels from the start of the selection to the end of the track
>>>> >> (also seen applying a flat EQ curve).
>>>> >
>>>> >
>>>> > Which commit
>>>>
>>>> Before your fix "01450e8".
>>>>
>>>>
>>>> >  and what steps to reproduce, does this describe?
>>>>
>>>> Same as Steve's steps below, with am optional variant of Flat EQ
>>>> instead of Low Pass.
>>>>
>>>> After your fix, on Linux and Windows, I see after Step 4 below that
>>>> the blue envelope has risen to 1.0 amplitude between 0 seconds
>>>> and 10 seconds.
>>>>
>>>> As you know I am not an expert in envelopes, but why as a user would
>>>> I want the audio before the selection to be boosted when I apply Low
>>>> Pass or EQ to the selection?  2.1.3 release does not do that.
>>>
>>>
>>> I see all is not yet working as intended.  I think I got everything right
>>> when the exiting envelope has more than one point, but the case of only one
>>> point  needs work.
>>>
>>> In 2.1.3 the effect applies to the waveform as it is without the
>>> envelope, and loses envelope information for that selection, and replaces it
>>> with a default level portion of envelope.
>>>
>>> Right or wrong, that should be so for many effects besides Nyquist, and
>>> that is not meant to change.
>>>
>>> What is meant to change is the treatment of discontinuities.  If you zoom
>>> very close, you should see a true vertical in 2.2.0 and not the short steep
>>> slope of 2.1.3.
>>>
>>> PRL
>>
>>
>> Gale, Steve, please try it again at 33c9aa9
The behaviour seems OK now with the caveat that I don't understand
why Flat EQ changes the envelope when "flat" Bass & Treble" rightly
(?) does not. But that is the same in 2.1.3 and not of direct relevance
here.


> Please note too that one purpose of this project with envelopes is that you
> you cut then immediately paste again in a place where the envelope is more
> complicated than a single level, then the net result should be no change;
> this was not so in 2.1.3.  Try it to confirm that remains so with the recent
> fix.

That mostly seems the case, but I found if I cut the audio as seen
in "before_cut.png" , then immediately paste, I see "after paste.png"
which is not identical because the envelope at the start of the
selection has been changed. See attached images.


Gale

> If you do so with an envelope that has only one point, then the envelope is
> also unchanged in case the selection does not include that point.
>
> Ah, but now I see that if it does include the point, then the cut-and-pase
> results in one more, unnecessary point.
>
> I may try to do something about this minor case too.
>
> PRL
>
>
>>
>>
>>>
>>>
>>>
>>>>
>>>>
>>>>
>>>> Gale
>>>>
>>>>
>>>> >> On 7 June 2017 at 10:56, Steve the Fiddle <[hidden email]>
>>>> >> wrote:
>>>> >> > Repeatable example on Linux:
>>>> >> >
>>>> >> > 1) Generate a default Chirp
>>>> >> > 2) Add an envelope point at about 0.8 / 15 seconds
>>>> >> > 3) Select from about 10 to 20 seconds
>>>> >> > 4) Apply Low Pass filter.
>>>> >> >
>>>> >> > Track becomes flat line from start of selection to end of the
>>>> >> > track.
>>>> >> >
>>>> >> > I assume that this is caused by the changes to envelope point
>>>> >> > handling
>>>> >> > that Paul has been working on, as is the debug crash that I
>>>> >> > reported
>>>> >> > on Windows.
>>>> >> >
>>>> >> > Paul, are you on to this or does it need logging on bugzilla? I'd
>>>> >> > say
>>>> >> > it's a P1 regression issue if we're logging it.
>>>> >> >
>>>> >> > Steve

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality

before_cut.png (8K) Download Attachment
after_paste.png (8K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Paul Licameli


On Fri, Jun 9, 2017 at 2:34 PM, Gale Andrews <[hidden email]> wrote:
On 8 June 2017 at 19:49, Paul Licameli <[hidden email]> wrote:
>
>
> On Thu, Jun 8, 2017 at 2:41 PM, Paul Licameli <[hidden email]>
> wrote:
>>
>>
>>
>> On Thu, Jun 8, 2017 at 2:02 PM, Paul Licameli <[hidden email]>
>> wrote:
>>>
>>>
>>>
>>> On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews <[hidden email]>
>>> wrote:
>>>>
>>>> On 8 June 2017 at 01:39, Paul Licameli <[hidden email]> wrote:
>>>> >
>>>> >
>>>> > On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews <[hidden email]>
>>>> > wrote:
>>>> >>
>>>> >> On Windows the selection suffers a large fade in to distortion
>>>> >> levels from the start of the selection to the end of the track
>>>> >> (also seen applying a flat EQ curve).
>>>> >
>>>> >
>>>> > Which commit
>>>>
>>>> Before your fix "01450e8".
>>>>
>>>>
>>>> >  and what steps to reproduce, does this describe?
>>>>
>>>> Same as Steve's steps below, with am optional variant of Flat EQ
>>>> instead of Low Pass.
>>>>
>>>> After your fix, on Linux and Windows, I see after Step 4 below that
>>>> the blue envelope has risen to 1.0 amplitude between 0 seconds
>>>> and 10 seconds.
>>>>
>>>> As you know I am not an expert in envelopes, but why as a user would
>>>> I want the audio before the selection to be boosted when I apply Low
>>>> Pass or EQ to the selection?  2.1.3 release does not do that.
>>>
>>>
>>> I see all is not yet working as intended.  I think I got everything right
>>> when the exiting envelope has more than one point, but the case of only one
>>> point  needs work.
>>>
>>> In 2.1.3 the effect applies to the waveform as it is without the
>>> envelope, and loses envelope information for that selection, and replaces it
>>> with a default level portion of envelope.
>>>
>>> Right or wrong, that should be so for many effects besides Nyquist, and
>>> that is not meant to change.
>>>
>>> What is meant to change is the treatment of discontinuities.  If you zoom
>>> very close, you should see a true vertical in 2.2.0 and not the short steep
>>> slope of 2.1.3.
>>>
>>> PRL
>>
>>
>> Gale, Steve, please try it again at 33c9aa9

The behaviour seems OK now with the caveat that I don't understand
why Flat EQ changes the envelope when "flat" Bass & Treble" rightly
(?) does not. But that is the same in 2.1.3 and not of direct relevance
here.

I now see there is inconsistency among the effects.

Those that do override ProcessBlock should not display change the envelope.

Some that do not override ProcessBlock will change the envelope.  This includes Nyquist effects, also noise reduction.

Fixing that was not part of the scope of my project.  But I will keep it in mind.
 


> Please note too that one purpose of this project with envelopes is that you
> you cut then immediately paste again in a place where the envelope is more
> complicated than a single level, then the net result should be no change;
> this was not so in 2.1.3.  Try it to confirm that remains so with the recent
> fix.

That mostly seems the case, but I found if I cut the audio as seen
in "before_cut.png" , then immediately paste, I see "after paste.png"
which is not identical because the envelope at the start of the
selection has been changed. See attached images.


Gale

Is there an exact procedure for reproducing this?

I have some suspicions about this, and I think behavior might vary with slight changes of the left-hand edge of the selection.

PRL

 

> If you do so with an envelope that has only one point, then the envelope is
> also unchanged in case the selection does not include that point.
>
> Ah, but now I see that if it does include the point, then the cut-and-pase
> results in one more, unnecessary point.
>
> I may try to do something about this minor case too.
>
> PRL
>
>
>>
>>
>>>
>>>
>>>
>>>>
>>>>
>>>>
>>>> Gale
>>>>
>>>>
>>>> >> On 7 June 2017 at 10:56, Steve the Fiddle <[hidden email]>
>>>> >> wrote:
>>>> >> > Repeatable example on Linux:
>>>> >> >
>>>> >> > 1) Generate a default Chirp
>>>> >> > 2) Add an envelope point at about 0.8 / 15 seconds
>>>> >> > 3) Select from about 10 to 20 seconds
>>>> >> > 4) Apply Low Pass filter.
>>>> >> >
>>>> >> > Track becomes flat line from start of selection to end of the
>>>> >> > track.
>>>> >> >
>>>> >> > I assume that this is caused by the changes to envelope point
>>>> >> > handling
>>>> >> > that Paul has been working on, as is the debug crash that I
>>>> >> > reported
>>>> >> > on Windows.
>>>> >> >
>>>> >> > Paul, are you on to this or does it need logging on bugzilla? I'd
>>>> >> > say
>>>> >> > it's a P1 regression issue if we're logging it.
>>>> >> >
>>>> >> > Steve

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality



------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Gale
Administrator
On 10 June 2017 at 02:19, Paul Licameli <[hidden email]> wrote:

>
>
> On Fri, Jun 9, 2017 at 2:34 PM, Gale Andrews <[hidden email]> wrote:
>>
>> On 8 June 2017 at 19:49, Paul Licameli <[hidden email]> wrote:
>> >
>> >
>> > On Thu, Jun 8, 2017 at 2:41 PM, Paul Licameli <[hidden email]>
>> > wrote:
>> >>
>> >>
>> >>
>> >> On Thu, Jun 8, 2017 at 2:02 PM, Paul Licameli <[hidden email]>
>> >> wrote:
>> >>>
>> >>>
>> >>>
>> >>> On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews <[hidden email]>
>> >>> wrote:
>> >>>>
>> >>>> On 8 June 2017 at 01:39, Paul Licameli <[hidden email]>
>> >>>> wrote:
>> >>>> >
>> >>>> >
>> >>>> > On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews
>> >>>> > <[hidden email]>
>> >>>> > wrote:
>> >>>> >>
>> >>>> >> On Windows the selection suffers a large fade in to distortion
>> >>>> >> levels from the start of the selection to the end of the track
>> >>>> >> (also seen applying a flat EQ curve).
>> >>>> >
>> >>>> >
>> >>>> > Which commit
>> >>>>
>> >>>> Before your fix "01450e8".
>> >>>>
>> >>>>
>> >>>> >  and what steps to reproduce, does this describe?
>> >>>>
>> >>>> Same as Steve's steps below, with am optional variant of Flat EQ
>> >>>> instead of Low Pass.
>> >>>>
>> >>>> After your fix, on Linux and Windows, I see after Step 4 below that
>> >>>> the blue envelope has risen to 1.0 amplitude between 0 seconds
>> >>>> and 10 seconds.
>> >>>>
>> >>>> As you know I am not an expert in envelopes, but why as a user would
>> >>>> I want the audio before the selection to be boosted when I apply Low
>> >>>> Pass or EQ to the selection?  2.1.3 release does not do that.
>> >>>
>> >>>
>> >>> I see all is not yet working as intended.  I think I got everything
>> >>> right
>> >>> when the exiting envelope has more than one point, but the case of
>> >>> only one
>> >>> point  needs work.
>> >>>
>> >>> In 2.1.3 the effect applies to the waveform as it is without the
>> >>> envelope, and loses envelope information for that selection, and
>> >>> replaces it
>> >>> with a default level portion of envelope.
>> >>>
>> >>> Right or wrong, that should be so for many effects besides Nyquist,
>> >>> and
>> >>> that is not meant to change.
>> >>>
>> >>> What is meant to change is the treatment of discontinuities.  If you
>> >>> zoom
>> >>> very close, you should see a true vertical in 2.2.0 and not the short
>> >>> steep
>> >>> slope of 2.1.3.
>> >>>
>> >>> PRL
>> >>
>> >>
>> >> Gale, Steve, please try it again at 33c9aa9
>>
>> The behaviour seems OK now with the caveat that I don't understand
>> why Flat EQ changes the envelope when "flat" Bass & Treble" rightly
>> (?) does not. But that is the same in 2.1.3 and not of direct relevance
>> here.
>
>
> I now see there is inconsistency among the effects.
>
> Those that do override ProcessBlock should not display change the envelope.
>
> Some that do not override ProcessBlock will change the envelope.  This
> includes Nyquist effects, also noise reduction.
>
> Fixing that was not part of the scope of my project.  But I will keep it in
> mind.
>
>>
>>
>>
>> > Please note too that one purpose of this project with envelopes is that
>> > you
>> > you cut then immediately paste again in a place where the envelope is
>> > more
>> > complicated than a single level, then the net result should be no
>> > change;
>> > this was not so in 2.1.3.  Try it to confirm that remains so with the
>> > recent
>> > fix.
>>
>> That mostly seems the case, but I found if I cut the audio as seen
>> in "before_cut.png" , then immediately paste, I see "after paste.png"
>> which is not identical because the envelope at the start of the
>> selection has been changed. See attached images.
>>
>>
>> Gale
>
>
> Is there an exact procedure for reproducing this?
>
> I have some suspicions about this, and I think behavior might vary with
> slight changes of the left-hand edge of the selection.
>
> PRL

I generate 30s default Chirp (no effects applied) then add
envelope points as shown then drag a selection as shown
and Copy then Paste.

Usually one end of the selection or the other gets a changed
envelope.

Open env40.aup from http://gaclrecords.org.uk/bugs/env40.zip .

Copy the selection already there then paste. The paste has its
envelope changed at right of selection.


Gale

>> > If you do so with an envelope that has only one point, then the envelope
>> > is
>> > also unchanged in case the selection does not include that point.
>> >
>> > Ah, but now I see that if it does include the point, then the
>> > cut-and-pase
>> > results in one more, unnecessary point.
>> >
>> > I may try to do something about this minor case too.
>> >
>> > PRL
>> >
>> >
>> >>
>> >>
>> >>>
>> >>>
>> >>>
>> >>>>
>> >>>>
>> >>>>
>> >>>> Gale
>> >>>>
>> >>>>
>> >>>> >> On 7 June 2017 at 10:56, Steve the Fiddle
>> >>>> >> <[hidden email]>
>> >>>> >> wrote:
>> >>>> >> > Repeatable example on Linux:
>> >>>> >> >
>> >>>> >> > 1) Generate a default Chirp
>> >>>> >> > 2) Add an envelope point at about 0.8 / 15 seconds
>> >>>> >> > 3) Select from about 10 to 20 seconds
>> >>>> >> > 4) Apply Low Pass filter.
>> >>>> >> >
>> >>>> >> > Track becomes flat line from start of selection to end of the
>> >>>> >> > track.
>> >>>> >> >
>> >>>> >> > I assume that this is caused by the changes to envelope point
>> >>>> >> > handling
>> >>>> >> > that Paul has been working on, as is the debug crash that I
>> >>>> >> > reported
>> >>>> >> > on Windows.
>> >>>> >> >
>> >>>> >> > Paul, are you on to this or does it need logging on bugzilla?
>> >>>> >> > I'd
>> >>>> >> > say
>> >>>> >> > it's a P1 regression issue if we're logging it.
>> >>>> >> >
>> >>>> >> > Steve
>>
>>
>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> _______________________________________________
>> Audacity-quality mailing list
>> [hidden email]
>> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>>
>
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Audacity-quality mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Paul Licameli


On Sat, Jun 10, 2017 at 11:51 AM, Gale Andrews <[hidden email]> wrote:
On 10 June 2017 at 02:19, Paul Licameli <[hidden email]> wrote:
>
>
> On Fri, Jun 9, 2017 at 2:34 PM, Gale Andrews <[hidden email]> wrote:
>>
>> On 8 June 2017 at 19:49, Paul Licameli <[hidden email]> wrote:
>> >
>> >
>> > On Thu, Jun 8, 2017 at 2:41 PM, Paul Licameli <[hidden email]>
>> > wrote:
>> >>
>> >>
>> >>
>> >> On Thu, Jun 8, 2017 at 2:02 PM, Paul Licameli <[hidden email]>
>> >> wrote:
>> >>>
>> >>>
>> >>>
>> >>> On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews <[hidden email]>
>> >>> wrote:
>> >>>>
>> >>>> On 8 June 2017 at 01:39, Paul Licameli <[hidden email]>
>> >>>> wrote:
>> >>>> >
>> >>>> >
>> >>>> > On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews
>> >>>> > <[hidden email]>
>> >>>> > wrote:
>> >>>> >>
>> >>>> >> On Windows the selection suffers a large fade in to distortion
>> >>>> >> levels from the start of the selection to the end of the track
>> >>>> >> (also seen applying a flat EQ curve).
>> >>>> >
>> >>>> >
>> >>>> > Which commit
>> >>>>
>> >>>> Before your fix "01450e8".
>> >>>>
>> >>>>
>> >>>> >  and what steps to reproduce, does this describe?
>> >>>>
>> >>>> Same as Steve's steps below, with am optional variant of Flat EQ
>> >>>> instead of Low Pass.
>> >>>>
>> >>>> After your fix, on Linux and Windows, I see after Step 4 below that
>> >>>> the blue envelope has risen to 1.0 amplitude between 0 seconds
>> >>>> and 10 seconds.
>> >>>>
>> >>>> As you know I am not an expert in envelopes, but why as a user would
>> >>>> I want the audio before the selection to be boosted when I apply Low
>> >>>> Pass or EQ to the selection?  2.1.3 release does not do that.
>> >>>
>> >>>
>> >>> I see all is not yet working as intended.  I think I got everything
>> >>> right
>> >>> when the exiting envelope has more than one point, but the case of
>> >>> only one
>> >>> point  needs work.
>> >>>
>> >>> In 2.1.3 the effect applies to the waveform as it is without the
>> >>> envelope, and loses envelope information for that selection, and
>> >>> replaces it
>> >>> with a default level portion of envelope.
>> >>>
>> >>> Right or wrong, that should be so for many effects besides Nyquist,
>> >>> and
>> >>> that is not meant to change.
>> >>>
>> >>> What is meant to change is the treatment of discontinuities.  If you
>> >>> zoom
>> >>> very close, you should see a true vertical in 2.2.0 and not the short
>> >>> steep
>> >>> slope of 2.1.3.
>> >>>
>> >>> PRL
>> >>
>> >>
>> >> Gale, Steve, please try it again at 33c9aa9
>>
>> The behaviour seems OK now with the caveat that I don't understand
>> why Flat EQ changes the envelope when "flat" Bass & Treble" rightly
>> (?) does not. But that is the same in 2.1.3 and not of direct relevance
>> here.
>
>
> I now see there is inconsistency among the effects.
>
> Those that do override ProcessBlock should not display change the envelope.
>
> Some that do not override ProcessBlock will change the envelope.  This
> includes Nyquist effects, also noise reduction.
>
> Fixing that was not part of the scope of my project.  But I will keep it in
> mind.
>
>>
>>
>>
>> > Please note too that one purpose of this project with envelopes is that
>> > you
>> > you cut then immediately paste again in a place where the envelope is
>> > more
>> > complicated than a single level, then the net result should be no
>> > change;
>> > this was not so in 2.1.3.  Try it to confirm that remains so with the
>> > recent
>> > fix.
>>
>> That mostly seems the case, but I found if I cut the audio as seen
>> in "before_cut.png" , then immediately paste, I see "after paste.png"
>> which is not identical because the envelope at the start of the
>> selection has been changed. See attached images.
>>
>>
>> Gale
>
>
> Is there an exact procedure for reproducing this?
>
> I have some suspicions about this, and I think behavior might vary with
> slight changes of the left-hand edge of the selection.
>
> PRL

I generate 30s default Chirp (no effects applied) then add
envelope points as shown then drag a selection as shown
and Copy then Paste.

Usually one end of the selection or the other gets a changed
envelope.

Open env40.aup from http://gaclrecords.org.uk/bugs/env40.zip .

Copy the selection already there then paste. The paste has its
envelope changed at right of selection.


In the bad examples, is it always so, that the selection encompasses all of the control points?

PRL

 

Gale

>> > If you do so with an envelope that has only one point, then the envelope
>> > is
>> > also unchanged in case the selection does not include that point.
>> >
>> > Ah, but now I see that if it does include the point, then the
>> > cut-and-pase
>> > results in one more, unnecessary point.
>> >
>> > I may try to do something about this minor case too.
>> >
>> > PRL
>> >
>> >
>> >>
>> >>
>> >>>
>> >>>
>> >>>
>> >>>>
>> >>>>
>> >>>>
>> >>>> Gale
>> >>>>
>> >>>>
>> >>>> >> On 7 June 2017 at 10:56, Steve the Fiddle
>> >>>> >> <[hidden email]>
>> >>>> >> wrote:
>> >>>> >> > Repeatable example on Linux:
>> >>>> >> >
>> >>>> >> > 1) Generate a default Chirp
>> >>>> >> > 2) Add an envelope point at about 0.8 / 15 seconds
>> >>>> >> > 3) Select from about 10 to 20 seconds
>> >>>> >> > 4) Apply Low Pass filter.
>> >>>> >> >
>> >>>> >> > Track becomes flat line from start of selection to end of the
>> >>>> >> > track.
>> >>>> >> >
>> >>>> >> > I assume that this is caused by the changes to envelope point
>> >>>> >> > handling
>> >>>> >> > that Paul has been working on, as is the debug crash that I
>> >>>> >> > reported
>> >>>> >> > on Windows.
>> >>>> >> >
>> >>>> >> > Paul, are you on to this or does it need logging on bugzilla?
>> >>>> >> > I'd
>> >>>> >> > say
>> >>>> >> > it's a P1 regression issue if we're logging it.
>> >>>> >> >
>> >>>> >> > Steve
>>
>>
>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> _______________________________________________
>> Audacity-quality mailing list
>> [hidden email]
>> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>>
>
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Audacity-quality mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/audacity-quality
>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Gale
Administrator
On 10 June 2017 at 16:55, Paul Licameli <[hidden email]> wrote:

>
>
> On Sat, Jun 10, 2017 at 11:51 AM, Gale Andrews <[hidden email]>
> wrote:
>>
>> On 10 June 2017 at 02:19, Paul Licameli <[hidden email]> wrote:
>> >
>> >
>> > On Fri, Jun 9, 2017 at 2:34 PM, Gale Andrews <[hidden email]>
>> > wrote:
>> >>
>> >> On 8 June 2017 at 19:49, Paul Licameli <[hidden email]> wrote:
>> >> >
>> >> >
>> >> > On Thu, Jun 8, 2017 at 2:41 PM, Paul Licameli
>> >> > <[hidden email]>
>> >> > wrote:
>> >> >>
>> >> >>
>> >> >>
>> >> >> On Thu, Jun 8, 2017 at 2:02 PM, Paul Licameli
>> >> >> <[hidden email]>
>> >> >> wrote:
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>> On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews
>> >> >>> <[hidden email]>
>> >> >>> wrote:
>> >> >>>>
>> >> >>>> On 8 June 2017 at 01:39, Paul Licameli <[hidden email]>
>> >> >>>> wrote:
>> >> >>>> >
>> >> >>>> >
>> >> >>>> > On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews
>> >> >>>> > <[hidden email]>
>> >> >>>> > wrote:
>> >> >>>> >>
>> >> >>>> >> On Windows the selection suffers a large fade in to distortion
>> >> >>>> >> levels from the start of the selection to the end of the track
>> >> >>>> >> (also seen applying a flat EQ curve).
>> >> >>>> >
>> >> >>>> >
>> >> >>>> > Which commit
>> >> >>>>
>> >> >>>> Before your fix "01450e8".
>> >> >>>>
>> >> >>>>
>> >> >>>> >  and what steps to reproduce, does this describe?
>> >> >>>>
>> >> >>>> Same as Steve's steps below, with am optional variant of Flat EQ
>> >> >>>> instead of Low Pass.
>> >> >>>>
>> >> >>>> After your fix, on Linux and Windows, I see after Step 4 below
>> >> >>>> that
>> >> >>>> the blue envelope has risen to 1.0 amplitude between 0 seconds
>> >> >>>> and 10 seconds.
>> >> >>>>
>> >> >>>> As you know I am not an expert in envelopes, but why as a user
>> >> >>>> would
>> >> >>>> I want the audio before the selection to be boosted when I apply
>> >> >>>> Low
>> >> >>>> Pass or EQ to the selection?  2.1.3 release does not do that.
>> >> >>>
>> >> >>>
>> >> >>> I see all is not yet working as intended.  I think I got everything
>> >> >>> right
>> >> >>> when the exiting envelope has more than one point, but the case of
>> >> >>> only one
>> >> >>> point  needs work.
>> >> >>>
>> >> >>> In 2.1.3 the effect applies to the waveform as it is without the
>> >> >>> envelope, and loses envelope information for that selection, and
>> >> >>> replaces it
>> >> >>> with a default level portion of envelope.
>> >> >>>
>> >> >>> Right or wrong, that should be so for many effects besides Nyquist,
>> >> >>> and
>> >> >>> that is not meant to change.
>> >> >>>
>> >> >>> What is meant to change is the treatment of discontinuities.  If
>> >> >>> you
>> >> >>> zoom
>> >> >>> very close, you should see a true vertical in 2.2.0 and not the
>> >> >>> short
>> >> >>> steep
>> >> >>> slope of 2.1.3.
>> >> >>>
>> >> >>> PRL
>> >> >>
>> >> >>
>> >> >> Gale, Steve, please try it again at 33c9aa9
>> >>
>> >> The behaviour seems OK now with the caveat that I don't understand
>> >> why Flat EQ changes the envelope when "flat" Bass & Treble" rightly
>> >> (?) does not. But that is the same in 2.1.3 and not of direct relevance
>> >> here.
>> >
>> >
>> > I now see there is inconsistency among the effects.
>> >
>> > Those that do override ProcessBlock should not display change the
>> > envelope.
>> >
>> > Some that do not override ProcessBlock will change the envelope.  This
>> > includes Nyquist effects, also noise reduction.
>> >
>> > Fixing that was not part of the scope of my project.  But I will keep it
>> > in
>> > mind.
>> >
>> >>
>> >>
>> >>
>> >> > Please note too that one purpose of this project with envelopes is
>> >> > that
>> >> > you
>> >> > you cut then immediately paste again in a place where the envelope is
>> >> > more
>> >> > complicated than a single level, then the net result should be no
>> >> > change;
>> >> > this was not so in 2.1.3.  Try it to confirm that remains so with the
>> >> > recent
>> >> > fix.
>> >>
>> >> That mostly seems the case, but I found if I cut the audio as seen
>> >> in "before_cut.png" , then immediately paste, I see "after paste.png"
>> >> which is not identical because the envelope at the start of the
>> >> selection has been changed. See attached images.
>> >>
>> >>
>> >> Gale
>> >
>> >
>> > Is there an exact procedure for reproducing this?
>> >
>> > I have some suspicions about this, and I think behavior might vary with
>> > slight changes of the left-hand edge of the selection.
>> >
>> > PRL
>>
>> I generate 30s default Chirp (no effects applied) then add
>> envelope points as shown then drag a selection as shown
>> and Copy then Paste.
>>
>> Usually one end of the selection or the other gets a changed
>> envelope.
>>
>> Open env40.aup from http://gaclrecords.org.uk/bugs/env40.zip .
>>
>> Copy the selection already there then paste. The paste has its
>> envelope changed at right of selection.
>>
>
> In the bad examples, is it always so, that the selection encompasses all of
> the control points?
>
> PRL

Not always. If I add envelope points after the selection shown in env40.aup
then copy and paste the unchanged selection, the envelope problem still
occurs.

But selecting over just three or four closely bunched envelope points
wherever they are in the track (start, middle, end) seems to paste fine.

If the three or four points are a long way apart, there is often still
a problem.


Gale



>> >> > If you do so with an envelope that has only one point, then the
>> >> > envelope
>> >> > is
>> >> > also unchanged in case the selection does not include that point.
>> >> >
>> >> > Ah, but now I see that if it does include the point, then the
>> >> > cut-and-pase
>> >> > results in one more, unnecessary point.
>> >> >
>> >> > I may try to do something about this minor case too.
>> >> >
>> >> > PRL
>> >> >
>> >> >
>> >> >>
>> >> >>
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>>>
>> >> >>>>
>> >> >>>>
>> >> >>>> Gale
>> >> >>>>
>> >> >>>>
>> >> >>>> >> On 7 June 2017 at 10:56, Steve the Fiddle
>> >> >>>> >> <[hidden email]>
>> >> >>>> >> wrote:
>> >> >>>> >> > Repeatable example on Linux:
>> >> >>>> >> >
>> >> >>>> >> > 1) Generate a default Chirp
>> >> >>>> >> > 2) Add an envelope point at about 0.8 / 15 seconds
>> >> >>>> >> > 3) Select from about 10 to 20 seconds
>> >> >>>> >> > 4) Apply Low Pass filter.
>> >> >>>> >> >
>> >> >>>> >> > Track becomes flat line from start of selection to end of the
>> >> >>>> >> > track.
>> >> >>>> >> >
>> >> >>>> >> > I assume that this is caused by the changes to envelope point
>> >> >>>> >> > handling
>> >> >>>> >> > that Paul has been working on, as is the debug crash that I
>> >> >>>> >> > reported
>> >> >>>> >> > on Windows.
>> >> >>>> >> >
>> >> >>>> >> > Paul, are you on to this or does it need logging on bugzilla?
>> >> >>>> >> > I'd
>> >> >>>> >> > say
>> >> >>>> >> > it's a P1 regression issue if we're logging it.
>> >> >>>> >> >
>> >> >>>> >> > Steve

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality
Reply | Threaded
Open this post in threaded view
|

Re: Applying Nyquist effects to regions containing envelope points may corrupt track

Paul Licameli


On Sat, Jun 10, 2017 at 12:38 PM, Gale Andrews <[hidden email]> wrote:
On 10 June 2017 at 16:55, Paul Licameli <[hidden email]> wrote:
>
>
> On Sat, Jun 10, 2017 at 11:51 AM, Gale Andrews <[hidden email]>
> wrote:
>>
>> On 10 June 2017 at 02:19, Paul Licameli <[hidden email]> wrote:
>> >
>> >
>> > On Fri, Jun 9, 2017 at 2:34 PM, Gale Andrews <[hidden email]>
>> > wrote:
>> >>
>> >> On 8 June 2017 at 19:49, Paul Licameli <[hidden email]> wrote:
>> >> >
>> >> >
>> >> > On Thu, Jun 8, 2017 at 2:41 PM, Paul Licameli
>> >> > <[hidden email]>
>> >> > wrote:
>> >> >>
>> >> >>
>> >> >>
>> >> >> On Thu, Jun 8, 2017 at 2:02 PM, Paul Licameli
>> >> >> <[hidden email]>
>> >> >> wrote:
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>> On Thu, Jun 8, 2017 at 1:28 PM, Gale Andrews
>> >> >>> <[hidden email]>
>> >> >>> wrote:
>> >> >>>>
>> >> >>>> On 8 June 2017 at 01:39, Paul Licameli <[hidden email]>
>> >> >>>> wrote:
>> >> >>>> >
>> >> >>>> >
>> >> >>>> > On Wed, Jun 7, 2017 at 6:24 PM, Gale Andrews
>> >> >>>> > <[hidden email]>
>> >> >>>> > wrote:
>> >> >>>> >>
>> >> >>>> >> On Windows the selection suffers a large fade in to distortion
>> >> >>>> >> levels from the start of the selection to the end of the track
>> >> >>>> >> (also seen applying a flat EQ curve).
>> >> >>>> >
>> >> >>>> >
>> >> >>>> > Which commit
>> >> >>>>
>> >> >>>> Before your fix "01450e8".
>> >> >>>>
>> >> >>>>
>> >> >>>> >  and what steps to reproduce, does this describe?
>> >> >>>>
>> >> >>>> Same as Steve's steps below, with am optional variant of Flat EQ
>> >> >>>> instead of Low Pass.
>> >> >>>>
>> >> >>>> After your fix, on Linux and Windows, I see after Step 4 below
>> >> >>>> that
>> >> >>>> the blue envelope has risen to 1.0 amplitude between 0 seconds
>> >> >>>> and 10 seconds.
>> >> >>>>
>> >> >>>> As you know I am not an expert in envelopes, but why as a user
>> >> >>>> would
>> >> >>>> I want the audio before the selection to be boosted when I apply
>> >> >>>> Low
>> >> >>>> Pass or EQ to the selection?  2.1.3 release does not do that.
>> >> >>>
>> >> >>>
>> >> >>> I see all is not yet working as intended.  I think I got everything
>> >> >>> right
>> >> >>> when the exiting envelope has more than one point, but the case of
>> >> >>> only one
>> >> >>> point  needs work.
>> >> >>>
>> >> >>> In 2.1.3 the effect applies to the waveform as it is without the
>> >> >>> envelope, and loses envelope information for that selection, and
>> >> >>> replaces it
>> >> >>> with a default level portion of envelope.
>> >> >>>
>> >> >>> Right or wrong, that should be so for many effects besides Nyquist,
>> >> >>> and
>> >> >>> that is not meant to change.
>> >> >>>
>> >> >>> What is meant to change is the treatment of discontinuities.  If
>> >> >>> you
>> >> >>> zoom
>> >> >>> very close, you should see a true vertical in 2.2.0 and not the
>> >> >>> short
>> >> >>> steep
>> >> >>> slope of 2.1.3.
>> >> >>>
>> >> >>> PRL
>> >> >>
>> >> >>
>> >> >> Gale, Steve, please try it again at 33c9aa9
>> >>
>> >> The behaviour seems OK now with the caveat that I don't understand
>> >> why Flat EQ changes the envelope when "flat" Bass & Treble" rightly
>> >> (?) does not. But that is the same in 2.1.3 and not of direct relevance
>> >> here.
>> >
>> >
>> > I now see there is inconsistency among the effects.
>> >
>> > Those that do override ProcessBlock should not display change the
>> > envelope.
>> >
>> > Some that do not override ProcessBlock will change the envelope.  This
>> > includes Nyquist effects, also noise reduction.
>> >
>> > Fixing that was not part of the scope of my project.  But I will keep it
>> > in
>> > mind.
>> >
>> >>
>> >>
>> >>
>> >> > Please note too that one purpose of this project with envelopes is
>> >> > that
>> >> > you
>> >> > you cut then immediately paste again in a place where the envelope is
>> >> > more
>> >> > complicated than a single level, then the net result should be no
>> >> > change;
>> >> > this was not so in 2.1.3.  Try it to confirm that remains so with the
>> >> > recent
>> >> > fix.
>> >>
>> >> That mostly seems the case, but I found if I cut the audio as seen
>> >> in "before_cut.png" , then immediately paste, I see "after paste.png"
>> >> which is not identical because the envelope at the start of the
>> >> selection has been changed. See attached images.
>> >>
>> >>
>> >> Gale
>> >
>> >
>> > Is there an exact procedure for reproducing this?
>> >
>> > I have some suspicions about this, and I think behavior might vary with
>> > slight changes of the left-hand edge of the selection.
>> >
>> > PRL
>>
>> I generate 30s default Chirp (no effects applied) then add
>> envelope points as shown then drag a selection as shown
>> and Copy then Paste.
>>
>> Usually one end of the selection or the other gets a changed
>> envelope.
>>
>> Open env40.aup from http://gaclrecords.org.uk/bugs/env40.zip .
>>
>> Copy the selection already there then paste. The paste has its
>> envelope changed at right of selection.
>>
>
> In the bad examples, is it always so, that the selection encompasses all of
> the control points?
>
> PRL

Not always. If I add envelope points after the selection shown in env40.aup
then copy and paste the unchanged selection, the envelope problem still
occurs.

But selecting over just three or four closely bunched envelope points
wherever they are in the track (start, middle, end) seems to paste fine.

If the three or four points are a long way apart, there is often still
a problem.


Gale

I gave envelope some more attention today.

 
PRL




>> >> > If you do so with an envelope that has only one point, then the
>> >> > envelope
>> >> > is
>> >> > also unchanged in case the selection does not include that point.
>> >> >
>> >> > Ah, but now I see that if it does include the point, then the
>> >> > cut-and-pase
>> >> > results in one more, unnecessary point.
>> >> >
>> >> > I may try to do something about this minor case too.
>> >> >
>> >> > PRL
>> >> >
>> >> >
>> >> >>
>> >> >>
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>>>
>> >> >>>>
>> >> >>>>
>> >> >>>> Gale
>> >> >>>>
>> >> >>>>
>> >> >>>> >> On 7 June 2017 at 10:56, Steve the Fiddle
>> >> >>>> >> <[hidden email]>
>> >> >>>> >> wrote:
>> >> >>>> >> > Repeatable example on Linux:
>> >> >>>> >> >
>> >> >>>> >> > 1) Generate a default Chirp
>> >> >>>> >> > 2) Add an envelope point at about 0.8 / 15 seconds
>> >> >>>> >> > 3) Select from about 10 to 20 seconds
>> >> >>>> >> > 4) Apply Low Pass filter.
>> >> >>>> >> >
>> >> >>>> >> > Track becomes flat line from start of selection to end of the
>> >> >>>> >> > track.
>> >> >>>> >> >
>> >> >>>> >> > I assume that this is caused by the changes to envelope point
>> >> >>>> >> > handling
>> >> >>>> >> > that Paul has been working on, as is the debug crash that I
>> >> >>>> >> > reported
>> >> >>>> >> > on Windows.
>> >> >>>> >> >
>> >> >>>> >> > Paul, are you on to this or does it need logging on bugzilla?
>> >> >>>> >> > I'd
>> >> >>>> >> > say
>> >> >>>> >> > it's a P1 regression issue if we're logging it.
>> >> >>>> >> >
>> >> >>>> >> > Steve

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Audacity-quality mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/audacity-quality