Question on Channel Tools L and R Width Controls

I’m wondering if anyone knows the answer to how Channel Tools manages to pull off its L and R individual width adjustments.  I fully understand how M/S alterations can let one manipulate the overall width and placement of a stereo signal.  But what I can’t conceive of is how the individual width adjustments on the L and R controls (the little knobs beside the main one in the L and R sliders) can actually do anything.

Consider this thought experiment.  Assume the R channel is complete silence.  It doesn’t matter where we place it using Channel Tools, it should have zero effect on the output.  Now, put the L slider at dead center.  This is a mono signal.  How can adjusting the width of L do anything?  The information is not present.  No matter how “wide” we set the width control, there’s still just a single (mono) signal.

We give a mono signal fake width only by manipulating it and coming up with two similar but not identical mono signals.  We can use delay, verb, EQ or a variety of other tricks to do this.  But there’s no way to take a mono signal and widen it without creating two non-identical signals.

Yet, Channel Tools claims to be able to do this.  The documentation tells us nothing about what’s going on under the covers.  Is Channel Tools engaging in some signal manipulation trickery and not telling us what it is?

Before I get in and do some rigorous testing to solve this mystery, I thought I’d ask if anyone has previously done some investigation and figured it out.  I hope this question is making sense.

Here's an example of what I mean about S-1:
 
http://forum.cakewalk.com...-Imaging-m2850597.aspx
 

mike_mccue
You might want to ask Craig,
 
He uses the word "width" 6 times in one paragraph:
 
http://www.soundonsound.com/sos/jul09/articles/sonarworkshop_0709.htm
 
but I still don't quite know what "width" means in this context.

 
The context was in the paragraph's first sentence - "There's also an alternative panning control option." So width refers to the stereo imaging associated with panning.
 

Anderton

mike_mccue
You might want to ask Craig,
 
He uses the word "width" 6 times in one paragraph:
 
http://www.soundonsound.com/sos/jul09/articles/sonarworkshop_0709.htm
 
but I still don't quite know what "width" means in this context.

 
The context was in the paragraph's first sentence - "There's also an alternative panning control option." So width refers to the stereo imaging associated with panning.
 

 
 
I'm no closer to understanding what you mean than I was the first couple times I read it.
 

Hi David, FWIW I think you have posed an interesting question.
 
I just generated a 120 second long custom fabricated "stereo" test tone with Adobe Audition.
 
The tone had a 600Hz sine wave peaking at -6dBFS on the left and it had a 1200Hz sine wave peaking at -6dBFS on the right channel.
 
I imported the tone track into SONAR and placed an instantiation of Channel Tools on the track.
 
I placed an instance of NUGEN's Visualizer on the master bus.
 
I used Visualizer's Stereo Spectrum Analyzer and I used the Vector Scope.
 
When I brought either the right or the left halfway, or a 15 degree angle, with minimum "width" I could see the amplitude on the Stereo Spectrum Analyzer change appropriately and I could see the tilt on the vector scope.
 
When I left the center at the 15 degree angle and maximized the width so that the far side was at the limit I noted that the amplitude of the side being adjusted decreased by approximately 2dB, ( which was not substantially different then simply moving the L or R closer to the center a couple more degree) and there was a very slight increase in the spread on the vector scope which would indicate that there is some aspect of widening occurring with a very slight phase shift or very short delay.
 
Interestingly, if you adjusted the width for a spread further than 30 degrees the results seemed to mimic the effect of simply adjusting the L or R closer to its respective mate, because of course to spread the width further than 30 degree you had to move the L or R so as to get enough spread from the closer of the two side limitations.
 
I'd have to run some more tests before I thought that I, personally speaking, understood what was going on, but my gut feeling is that "not much" is close to an accurate description.
 
I would be curious to learn what others discovered, thought, knew etc.
 
Thank You.

Hi Craig,
 It's not like I haven't used Channel Tools before. ;-)
 
 I checked it out extensively back when it came out.

mike_mccue
Hi Craig,
 It's not like I haven't used Channel Tools before. ;-)
 
 I checked it out extensively back when it came out.

That doesn't matter, replies aren't just for you but for others who are reading this thread and want to know more about Channel Tools. You said you didn't understand what I was saying, yet included a link to the article. If someone follows that link and is not familiar with Channel Tools, it is unlikely they will understand what you didn't. So it's best to err on the side of completeness...especially because unlike print, here there are no column-inch or word count limitations.

If you're referring to post #12, conventional panning uses a single point source with a constant level where the level is distributed differently between the left and right channels. What I described is two point sources with equal levels capable of panning (i.e., distributing levels differently) between the left and right channels.
 
Think of the Channel Tools scenario I described as the equivalent of two concentric pots, each capable of panning an input signal independently.

mike_mccue
Just to be clear, I was hoping you knew something about what Channel Tools was doing when the width adjustment was made.

 
I think post #12 and #15 describe what it's doing. It's all level-based, there is nothing time-based about it. This is the "executive summary":
 
Think of the Channel Tools scenario I described as the equivalent of two concentric pots, each capable of panning an input signal independently.

Think I'll just stick with LCR.

If you're going to narrow a spread, to then move it around, don't you have do some cancelling in the side the sound is coming from that you want it to move away from ?  I've imagined that Channel Tools did it's thing by playing around with phase to get the cancellation required.  I don't see how else it could work.
 
Someone recently mentioned they didn't like what Channel Tools was doing to their signal phase-wise and I just figured that it was the price to pay for what it does.
 

Paul P
If you're going to narrow a spread, to then move it around, don't you have do some cancelling in the side the sound is coming from that you want it to move away from ?  I've imagined that Channel Tools did it's thing by playing around with phase to get the cancellation required.  I don't see how else it could work.

 
Please see post #12. You can also think of Channel Tools as a more flexible implementation of how Pro Tools uses two panpots, one for each channel, to be able to change the image width and move it around. I'm talking straight level-based stereo applications here, not M/S decoding or anything fancy.

dmbaer
For the moment, I'm going to be skeptical about Craig's claim of being able to place two identical sound sources in the stereo panorama and have anything but a result that sounds like they are at the phantom center between the two points.  I'll just have to play around with the math to prove it - and that may take a bit of thought.

 
If someone told you on a sunny day that the sky was blue and you didn't believe him, it's true one could use math to show how Rayleigh scattering affects shorter wavelengths, and therefore, the blue light (which has a shorter wavelength) is scattered more by the molecules in the atmosphere. Therefore, the sky appears blue. But a simpler option would be to step outside, then look up.
 
Here's an experiment to lift the veil of skepticism. Use headphones to take room acoustics out of the picture.
 

 

1. Set up something like the above with identical mono clips in each track.
2. Adjust the level for both tracks to -6.0.
3. Pan Track 1 full right.
4. Vary the panpot in Track 2 (circled in red) from full left to full right.

 
At full left, the sound will be centered. As you move it more toward the right, the image width will narrow and shift more and more toward the right until it eventually becomes an extremely narrow point source panned full right. The two Width controls in Channel Tools are doing essentially the same thing, with the added bonus that you can vary the center point from which they extend to the right or left. 
 

Take an edge case where the two mono points are at extreme left and extreme right.  That's two points in the panorama, right?  Now, do you expect to hear anything other than a single placement at the phantom center?  Now, move the two points to the center of the panorama.  Do you expect it to sound any different that the previous case?  Do you expect the content of L and R signals to differ between the two cases?

 
Please see post #12.
 

One other thing: the term "pan law" was mentioned earlier.  As far as I know, Channel Tools has no support for pan law, right?  I'm pretty certain it's never mentioned in the documentation.  I just want to confirm that pan law will not be a factor in any testing anyone does.

This article will tell you what you need to know about pan laws.

mike_mccue
oh my...
 
... and to think I was using a vectorscope when I shoulda been using my Studio Beats by Dr. Dre.

Personally, I find the sound quality of a vectorscope lacking. Then again, I find Beatz lacking too...except under 120 Hz, where they're quite tight.

OK, I did the following test.  I set up a synth (stereo output, no filtering no effects).  I put a square wave in a single oscillator and panned it far left.  Output was as expected.  R channel is totally silent and becomes irrelevant.
 
When I panned the Channel Tools (hereafter CT) to center with zero width, I got a phantom image dead center.  When I expanded the L width control, the output in both channels got about 3 dB louder as I approached the extremes and the phantom center remained dead center.  Metering confirmed both were L and R levels were equal at all width settings. 
 
I froze the track and looked at the waveform.  As expected both L and R were identical and perfectly in phase.  In fact, no phase alteration was ever noticed at any point in this testing (which I think we can all agree is a good thing).
 
When placing the L control about -15 (halfway left) and initially setting L width to zero, here's what happened.  As I expanded the width control, the left channel amplitude stayed nearly constant (increased but only very marginally).  The right channel level increased more.  Frozen waveforms remained identical in all cases except for amplitude (as expected based on what the meter output was indicating).
 
So, I'm going to stick to my guns here and say that in the second case (L at -15), increasing the width changed the stereo image, but it did so by moving the phantom center to the right, and the overall level increased.  One could produce exactly the same result by leaving width at zero, moving the CT L pan position a little to the right and increasing the level slightly.
 
In a situation like this, we've got two identical audio streams in L and R - identical except for amplitude that is.  As we increase/decrease either side, the perceived panorama position will naturally change.  But if the two waveforms are identical other than amplitude, that's all that changes.  That implies that I could produce two exports, one using CT width manipulation and one where I just altered pan and overall output level and I could achieve two exports that are identical.  And if this is true, then there's no width manipulation actually happening.  Changing CT width is just moving the panorama position and messing with amplitude.  Thus, CT width is not doing what it's claiming to do.
 
Anyway, that's my story and I'm stickin' to it.