Fades

From Audacity Development Manual
Revision as of 11:27, 13 January 2013 by PeterSampson (talk | contribs) (archived a now-redundant ednote on the talk page)
Jump to: navigation, search
Audacity ships with several types of fade effects: a pair of simple linear fades, a pair of fades to help you make a crossfade, a more musically pleasing fade out (similar to that used in many recording and mastering studios) and a powerful customizable fade.

Linear Fades

These basic fades apply a fade to the selected audio such that the amplitude of the selection goes from absolute silence to the original amplitude (Fade In), or from the original amplitude to absolute silence (Fade Out). The shape of the fade is linear, so it appears as a straight line from beginning to end (when viewed in the default linear Waveform View Mode). The speed of the fade in or out is therefore constant throughout its length and depends entirely on the length selected for the fade.

Fade In

Accessed by: Effect > Fade In
Linear Fade In applied to the start of a constant frequency and amplitude sine wave

A fade in is often applied over a very short audio selection (less than a second). You may get a more "musical" result by applying a linear fade in three times to the same audio selection. This approximates an exponentially shaped fade in.

Fade Out

Accessed by: Effect > Fade Out
Linear Fade Out applied to the end of a constant frequency and amplitude sine wave

A fade out is often applied to a longer selection than a fade in, typically to a selection up to about ten seconds long.

Cross Fades

Despite their names these effects do not perform an automatic "crossfade" between two tracks or adjoining audio, where the audio coming to an end would fade out at the same time the following audio fades in. Instead Cross Fades apply a curve that is different from the linear Fade in and Fade Out. The curve used is one that will result in equal RMS (average) volume throughout the faded section once the faded in and faded out sections are mixed.

As with linear fades and Studio Fade Out, the fades are between silence and original volume.

Cross Fade In

Accessed by: Effect > Cross Fade In
Cross Fade In applied to the start of constant amplitude white noise

Cross Fade Out

Accessed by: Effect > Cross Fade Out
Cross Fade Out applied to the end of constant amplitude white noise
Gale 11Jan13: ToDo-1 You may get your wish not to have an image of "after" because the image below is obviously incorrect. Since two tones are phase-coherent, the result will be a bulge of peak amplitude (though rms will be constant). It might actually be useful to show this with the recommendation to use an "equal gain" curve (S-Curve in Adjustable Fade?) for phase-coherent material. There would be an equivalent image in the Adjustable Fade section to show the peak level constant but the rms falling away mid fade.
  • Peter 11Jan13: The proper "after" image looks distictly weird with the bulge - it will look wrong to many users. Even Steve's "Cross Fade Classic" plug-in produces a power bulge, but that is biased towards the beginning of the crossfade rather than centred as with Cross Fade In/Out. I would vote strongly for removing this image altogether.
  • Gale 11Jan13: Yes the "proper" image does look weird and would have to show a lower amplitude tone. It can't stop as it is, so I think this is probably out, but I do think some text could remain about "after cross fade". The current almost complete lack of help with crossfading is a significant blemish in the Manual IMO. I'm going to sleep on it, I suggest you do too.
  • Bill 11Jan13: Even with noise you get the bulge, but not with music. Music may not show the shape as well, but does show the "equal volume" better.
    There's a problem with the existing 'before' images - the Cross Fade effects create clips when applied and this is not shown in the images.
    Would these images be better?
  • Gale 11Jan13: Thanks, Bill. Good catch, the Cross Fade "before" images must have split lines. I think quite strongly that the the "before" images should be tones in order to clarify the shape. The upper track in your before image could easily look linear to an inexperienced eye.

    We could show only the "after" music image to demonstrate that volume "looks equal" throughout the processed fade. Or we could reduce the height a little of the two "before" music images (or reduce height of the "after" music image as well). ATM I would tend to favour just showing the "after" music image, enhanced with arrows to show where the fade was, since render removes selectedness of the mixed track. Probably we should make that enhancement even if we show all three music images.

  • Peter 12Jan12: This just shows the conundrum here. Like Bill I began experimenting with fades on real music as neither tones nor noises gave a "sensible looking" after image. And I agree with Gale that the musical examples normally fail to properly show the fade shape. This is why I cheated with a faux-image in the first place :=)) Personally I still don't think we need to bother with an after-crossfade image on this page - it will be better dealt with in Steve's developing fades page - which increasingly, I think, should become a tutorial in this Manual and not the Wiki where it is now. For now it would leave a slight gap in the manual on the usage of crossfades - but that gap exists now in the current manual and the one for 2.0.0. I would just like to see it as a priority for us in the editorial team to help Steve get his work on the fades tutorial polished and ready for publication in the 2.0.4 Manual that we will start on soon.
  • Bill 12Jan13: New images using heavily compressed music. I think these show the "bullet nose" shape adequately. After image has fade region selected for clarity.
    On further reflection I think the current after image is likely to cause confusion even though it is technically correct. We say the RMS level is contant, but a reader who looks at the image and doesn't read the text (or doesn't click through to the definition of RMS) is going to question our assertion that these effects, when mixed, produce an "equal volume" cross-fade. However if we go back to the "faked" image, the technically minded are sure to call us out on it. So, my preference is either to have no 'after' image, or use music for the three images.
  • Steve 12Jan13: The original "faked" image was just wrong and stood out like a sore thumb. I like Bill's new image (below) but it does not show that peak levels will often be higher than the pre-mixed tracks (which is clearly shown with white noise. I like the white noise image a lot because it shows an increased peak level and the assertion that the "loudness" is constant. I think that the apparent contradiction is one of the clearest demonstrations that "loudness" relates more closely to rms level than to peak level. Could we include both a cross-faded white noise and a cross-faded music example? Can images be captioned?
  • Gale 13Jan13: Thanks, all. I removed the P1 (agree with Steve's reasoning). I vote for the current noise image. I dislike the advice div (it is not a second level warning) and prefer more explanation above the image, and an explanation of why we used noise rather than tones. I have tried to tweak in this way, but please check for sanity. I think we can do without music images if we use the text I am suggesting, but if someone wants realistic crossfaded music images (that I would expect to show some peak increase) please put them in an editornote.

    Yes we can have caption text but please not Wiki captions as they are unreasonably small. Chirp shows caption style (italic, normal size) that I like.

CrossFadeBefore.png

CrossFadeAfter.png

After a Cross Fade

If the two waveforms above (Cross Fade In and Out) were on separate tracks then rendered so as to mix the waveforms together, a section of audio with equal average volume (the inner light blue part of the waveform) would result as below. The peak volume of the crossfade (the outer dark blue part) does increase, but the volume still sounds constant because average volume level is a much stronger indicator than peak levels of "perceived" loudness.

Constant amplitude white noise after cross-fading
Advice We chose constant amplitude white noise rather than tones to illustrate the crossfade because two sections of noise are much less correlated (alike) than two sections of tone. As a result, the raising of peak level in the fade is much less extreme than if we had crossfaded tones.

Had we crossfaded music (which is usually a little more correlated than noise), average volume would still have remained constant. The peak level could possibly have risen too, though this would appear much less obvious because of the naturally more "spiky" (variable) peak levels.

Crossfading by hand should be performed by overlapping the parts to be faded using two separate tracks. A general rule that avoids a loss of perceived volume mid-fade is to use the two Cross Fade effects, or the "Cosine" presets in the Adjustable Fade effect.
  • Depending on the material, both these methods might cause a gain in peak amplitude (as illustrated above) with a risk of clipping. In that case, try using the linear fades or the "S" curves in Adjustable Fade.
Gale 01Jan13: Probably we should not have a "Musical Fades" (plural) section until we have Pro Fade Out released on Wiki (or in 2.0.x)
  • Peter 1Jan13: I added a temporary header "Additional Shaped Fades" until we can un-leash "Musical Fades". SFO and AF are not part of "Crossfades" so need their own umbrella header.

Musical Fades

These fades apply a more "musical" fade to the selected audio, giving a more pleasing sounding result.

Additional Shaped Fades

Studio Fade Out

Accessed by: Effect > Studio Fade Out
Gale 01Jan13: This is confusing. Which of the mirrored S or 1/2 period curves does it use, when? Or are they the same thing?
  • Peter 1Jan13: I changed the description to that on the Effect Menu page. Previous text was "Produces a smooth and musical sounding fade out. A mirrored S-curve or 1/2 period cosine is used for the volume and a linear ramp for the low pass filter (from 22050 Hz down to 100 Hz)."
  • Gale 02Jan13: It's up to Steve if he wants to explain about the curves, but I think it is un-necessary detail on this page. Removed "professional" in favour of "studio". "Professional" is too contentious, as Vaughan says.
    • Peter 4Jan13: I softened the description by moving the techy details into a tech-note rather than having them up-front first in the desciption - this should enable us to demote the P1 to >P2.
    • Peter 6Jan13: I moved the technote into an ednote to comment it out - and accordingly removed the P1 from here.

Produces a smooth and "musical" sounding fade out from original volume to silence by applying a doubly-curved (S-shape) fade and fading out the higher frequencies a little quicker than the lower frequencies. It gives a "going off into the distance" type of sound rather than sounding like a "mechanical" fade out. This technique has found favor in recording studios for CD and LP tracks.

Studio Fade Out applied to the end of a constant frequency and amplitude sine wave
Peter 6Jan13: I decided to comment-out this technote, at least for now. I don't see what it adds over the simpler textual description provided (except for the experts in digital signal processing). This enables us not just to be able to demote the above P1 but remove it.
  • Gale 11Jan13: Nonetheless there is some confusion invited now as it may appear that retaining the bass is giving this distinctive "S" shape however AFAIK the filter is not shaping the curve that much, the shape comes mainly from the choice of shape. But you should ask Steve.
    • Peter 11Jan13: I have added some text about the fade shape - my understanding from earlier long discussions with Steve is that this fade achieves its musically pleasing nature fro a combination of the S-shaped (sinusoidal) curve together with the use of the progressive filtering. But I think that we probably don't need that level of detail - we do beed to encourage users to try it however and I feel that too much techy detail may put them off.
    • Gale 11Jan13: I'm happy with your text. AFAICT the essential "S" shape persists even with a very high or very low frequency tone. There is now no implication the filter is creating the "S" which was my concern.
    • Steve 12Jan13: For clarification, Gale is correct. The filtering can affect the shape a little, but only a little. The shape is essentially sinusoidal and comes from the "fade" rather than the "filter".
Technical note: this effect works by applying a sinusoidal fade with a progressive low pass filter from full spectrum at the start of the selection to 100 Hz at the end.
You can make a simple customization of the shape of any of the above fades by repeating it using Repeat Last Effect ( CTRL + R ). If you repeat a fade in, the audio will remain quiet for longer then rise more rapidly to the original level. Repeating a fade out will cause the level to initially drop more rapidly then tail off at a lower level.
Peter 12Jan13: Looking at this page as a "civilian user" I would prefer us to move the content of this note into the intro div as it seems to get somewhat lost down here.
  • Gale 13Jan13: I really cannot see that as part of the intro - is it that fundamental? I don't think it is unreasonably lost if viewed as logged out. Even if in a note div under the intro, it looks too "early" there to me. You could try it after the linear fades then explain it applies to the other fades except AF?
Gale 01Jan13: ToDo Pro Fade Out has been committed for 2.0.3 as "Studio Fade Out", but as yet we have no Studio Fade In ready to release.

Studio Fade In

Accessed by: Effect > Studio Fade In provided you have installed it.
Example of a musical fade in applied to a constant frequency and amplitude sine wave.

Adjustable Fade

This has a dialog box where you can choose the shape of the fade in or fade out to be applied. You can also create "partial" fades to and from other than silence and original volume. An example of this might be a fade in from 20% of the original volume to 80% of the original volume. The "Handy Presets" at the bottom offer a choice of six pairs of fade in or fade out shapes, fading between silence and original volume.

Accessed by: Effect > Adjustable Fade...
Dialog box for Adjustable Fade
Advice Choosing a preset disables all the other controls irrespective of their setting. Therefore to make a partial fade you must choose the Start and End values then use the other controls above to determine the fade shape and direction.

Examples of fades that can be achieved with Adjustable Fade

Gale 13Jan13: ToDo I have strong reservations about reusing this image "as is". OK, showing multiple fades shows the power of AF, but we don't make clear this is done in multiple passes, and the mid-fade boost levels are unexplained. I think the caption for mid fade boost should be removed or some attempt made to explain it.

But why not show a simpler image (without the mid-fade complication) showing something else, such as a "stepped" fade up (two different partial fades). And IMO we should either make clear (in text as well as hover text) that these fades were done on a constant amplitude sine tone (if they were) or show unaffected audio abutting a non-100% fade so you can see the effect of the partial fade more clearly.

Partial Fade Down to 30% applied to a tone using Adjustable Fade

Example of a partial fade down to 30% applied to a constant amplitude sine tone using Adjustable Fade.

ToDo Gale 01Jan13: Adjustable Fade has been committed for 2.0.3. Review if this should go at the bottom of this page and the text here.
  • Gale 11Jan13: I'm not sure all of these tips below are in the correct place here. I removed the last tip about longer fade outs which was duplication.
    • Peter 11Jan13: I'm not sure now that I like any of these tips, not just here but anywhere on this page. For a start is the Envelope Tool advice still vaid now that we have Adjustable Fade? I've always been iffy about the so-called 2/3 rule in point 3 in the note - I seem to remeber some (largely unresolved) debate about this in the past; so is it valid, do we need it?
    • Gale 11Jan13: Re the "repeating" tip, I think it remains useful as a less scary (and quick) alternative to Adjustable Fade. I've tried moving it up so it's under Studio Fade Out.

      Yes, we must mention Envelope Tool I think. Some people will prefer the visual GUI and others will want multiple envelope points. I tweaked the note, leaving it where it is for now.

      The objection to the "two-thirds" rule is that it may increase peak amplitude even for linear fades. But since I hear more complaints about "mid-fade drop" than clipping, I think we want this hint at least until we actually have Steve's definitive Cross Fade Wiki page (and even then it provides somewhere to give the link). I moved this note underneath Cross Fades. I think these three changes are something of an improvement.

    • Steve 12Jan13: I have removed the tip about the "two-thirds" rule as that "rule" exists as a rough approximation to an "equal power fade" (from when Audacity did not have such a thing). As Audacity now has multiple ways to produce "equal power fades" (the "Cross Fade In / Out" and "Rounded" or "Cosine" presets in "Adjustable Fade", we don't need to resort to the 2/3 rule.
Envelope Tool on Tools Toolbar enables you to make completely customizable exponential fades having unlimited points along the fade, using a volume envelope in a separate track.

Links

>  Forward to: Invert

<  Back to: Equalization

|< Effect Menu