"Show Clipping" Options

[kozikowski]

I'm not necessarily in love with "three samples over." I want a graceful supposition of probable clipping damage and that it matches as default on both tools. I strongly disagree they should measure the same thing differently.

I don't care where the setup panels is, either, but something that simple is warranted in my opinion. A case can be made for splitting them up, but the two tools really are two views of the same thing. If a meter flashes (briefly), I expect to see a red bar present. I expect to be able to count the rapidly dying flashes in past history by counting the timeline red bars. My panic at needing to adjust levels (or restart the job) is the presence of repeated flashes, not the presence of one stuck flash.

Being able to easily reset the trigger points to the production needs is the treacle on top.

Koz

[Gale Andrews]

If the only way we can agree to make Show Clipping and metering trigger on the same number of clipped samples is to make the meter trigger on 1 clipped sample, I suggest we do that. I don't think it will make many recordings seem clipped when you can't hear the clipping. It may have that effect for playback of some files.

I guess we could then have a meter preference that defaults to 1 clipped sample but if changed, only affects the meter. I do see use cases for having Show Clipping and metering trigger on different number of samples, but I really don't think that should be default behaviour.

Would a new View Menu item "Configure Show Clipping..." or some such leading to a dialogue be regarded as important enough to add to the length of that menu?


Gale

[steve]

I think we've moved this from "bug" to "enhancement", and I support the idea of enhancing clip detection/notification.

I think that providing preference settings is only justified in cases where the default settings work well for one job and badly for another. If the default setting work well for all jobs, then there is little need to complicate the user experience by requiring that the user chooses appropriate settings for their task in hand.

What level of clipping (how many samples in a row) would indicate "stop the show now and start again" ?

[Edgar]

What level of clipping (how many samples in a row) would indicate "stop the show now and start again" ?

What if we get data which looks like: clip sample sample clip sample sample sample clip sample sample clip sample …etc. (where "clip" is a clipped sample and "sample" is not)? In other words, clipping is occurring frequently but we have set our level greater than one. The non-real-time Effect takes this into consideration but in live recording it would be imperative to be notified of this condition whereas if it were just a rare "clip sample sample clip" which might have no audible impact on the recording it could be ignored.

Another thought - there are at least three situations in which clipping is of interest: 1) a live (no second chances) recording; 2) a studio ("Let's try that again." – anything from recording a live band to transferring vinyl) recording; 3) editing existing material. I contend that Steve's "level" might well be different for each of these situations. Personally, for live recording, if even one sample clips, I want "something" to slap me in the face HARD (e.g. turn the clipping indicator red and keep it that way until I make a physical intervention – adjust some gain down and/or reset the indicator). For a studio recording I MIGHT be slightly more lenient about the occasional/rare couple-of-samples clipping based on the understanding that, immediately after the take, I listen to some or all of the regions containing clipping VERY carefully trusting my ears to decide if another take is warranted. When editing existing material over which I have no control of the origin I might be considerably more lenient situationally; and, "clipping" becomes "clipped" irretrievably unless the material is in floating-point data.

So, my position would be that the show clipping algorithm for the meter & the wave track need to be in lockstep by default – every time one triggers, the other should also trigger. The Effect's controls (since it is not automatic/real-time and, given that it may be rerun and is nondestructive) seem to be adequate but should have no relationship with the meter and the wave track settings. Power to the people – give the user the ability to affect the defaults for the meter & wave track independently giving the user the ability to take the two out of lockstep.

[steve]

I think that the red line indicators should always show (when enabled) at 0 dB, even if it is just one sample. This is nothing to do with "real time", so the user can examine the parts indicated at their leisure.

Koz want clip indicators to go off after a few seconds, but Edgar wants to be slapped hard in the face (I presume not literally ).
So how about, for the meters:

The meter clip indicator always comes on as soon as one sample touches or exceeds 0 dB.
If only one sample does that, then the clip indicator will automatically reset after a couple of seconds and go off again.
In more than "N" sequential samples do that, then the meter clip indicator latches ON.

The only thing that then requires configuring is the size of "N", which should be easy to get at because, as Edgar describes, the user may want to change it according to the job.
As this only affect the meters, I think it should be in the meter dropdown > preferences.

What if we get data which looks like: clip sample sample clip sample sample sample clip sample sample clip sample …etc.

The meter clip indicator remains on until a couple of seconds after the it stops doing that.
The waveform red lines will show a solid block of red.

As a further enhancement, being able to reduce the opacity of the red lines on the waveform might be nice (control how hard the slap is), but that is a separate issue which could be discussed later.

[waxcylinder]

As a further enhancement, being able to reduce the opacity of the red lines on the waveform might be nice (control how hard the slap is).

+1

Peter

[Edgar]

reduce the opacity of the red lines on the waveform

+1

Since y'all seem to be concerned about slapping me around too hard…

Support for Transparency / Alpha Channel
In general wxDC methods don't support alpha transparency and the alpha component of wxColour is simply ignored and you need to use wxGraphicsContext for full transparency support.

While there is a passing mention in a comment of wxGraphicsContext:

// As of 2.8.9 (possibly earlier), wxDC::DrawLine() on the Mac draws the
// last point since it is now based on the new wxGraphicsContext system.
// Make the other platforms do the same thing since the other platforms
// "may" follow they get wxGraphicsContext going.

wxWidgets 2.8 makes it very difficult for us to do transparency. When we switch to 3 we will need to also start using wxGraphicsContext if we wish to use transparency. I may give that a try with SVN HEAD. In the meantime, one possible solution would be to draw a dotted line instead of a solid line (the dot/skip ratio could be user controlled). Giving the user control over the color of the line might also be nice.

[Edgar]

Since transparency is off the table I decided to try a dotted line. In srcTrackArtist.cpp at or near line number 1205:

Code: Select all

   // Draw the clipping lines
   if (clipcnt) {
      dc.SetPen(muted ? muteClippedPen : clippedPen);
      while (--clipcnt >= 0) {
         int xx = clipped[clipcnt];
#define CLIPPINGGAP 4//Add this line so you can test various values but they must be positive whole numbers – 2, 4, 6 etc.
         for (int i = 0; i < r.height; i += CLIPPINGGAP) AColor::Line(dc, xx, r.y + i, xx, r.y + i);//Add this line
            //AColor::Line(dc, xx, r.y, xx, r.y + r.height);//Comment out this line
      }
   }

A CLIPPINGGAP of 2 is barely discernible with the default colors; 4 seems to work pretty well until you get zoomed down to where individual samples are depicted at which point the math fails and you get a solid vertical line. If this looks interesting I could play around with other algorithms and make the math work for all zoom levels.

[kozikowski]

Since y'all seem to be concerned about slapping me around too hard…

It's the only excitement we get. Just go with it.

Collecting graphically restrained red markers on a timeline is lovely, but there's nothing like repeated flashing flashers to get your attention. And yes, I'm not at all concerned with transferring my Andrews Sisters vinyl. I'm live recording a voice track for a television commercial. And no, I have never stopped a recording for one single overload which, as above, makes the current flasher useless to me.

To be clear, every time a damage condition is sensed, the flasher comes on full and starts a fade cycle. The flasher either continues to fade to completion, or can be interrupted by another overload and start the fade cycle from full brightness again. For a serious overload condition, the flashers will pulse at you trying really hard to get your attention which, correct me, is their job description.

I suppose some provision must be made for a greatly distorted show that never comes down from overload. A stuck-on flasher because of continuous overload is not useful and is counterproductive in that case. The nightmare job is indicated by flashers that flash at you not because they occasionally activate, but because they occasionally release.

And lest we get too far afield, the "bug" is the timeline red bars that appear with no associated sound damage.

Koz

[Edgar]

To be clear, every time a damage condition is sensed, the flasher comes on full and starts a fade cycle.

And lest we get too far afield, the "bug" is the timeline red bars that appear with no associated sound damage.

For clarity, from a programmer's point of view, the timeline is that tiny little strip of numbers above the waveform. I do not see any clipping indication drawn in the timeline, only in the waveform. Since each recordist is going to define "no associated sound damage" differently – 1, 2, 4 consecutive clips samples; 5 clipped samples scattered randomly within 25 samples; something which is perceptible to "the ear" given the delivery format etc. – this seems to imply giving the user full control.

Personally, I really like the idea of the flasher which comes on and fades away as you describe. We had an ancient board in the stable which did that and when I was using it for onstage monitor mixing (think of sitting in the pitch black darkness of the wings) it always caught my eye.