[quote author="TedF"]At the risk of frying in the pan :shock:
In the early days of pan control development, we tried different laws:
If you stick to a 'constant energy' type of control (6dB down in the middle for each side) then the centre image appears to dip slightly. It works best with a dip of about 2 to 3dB in the centre; this is a compromise so that a panned signal does not increase in level too much as it approaches the centre, and does not suffer from the 'hole in the middle' effect.
BUT..... All this is really tosh :roll: Pan controls actually don't work! Spatial positioning to the ear is a combination of variations in amplitude and timing of the signal that hits the ears. Amplitude difference without timing difference sort of works; well it's been done that way for about 50 years, but it's a horribly poor compromise compared to real spatial effects created naturally by 'sum and difference' recording.[/quote]
I'd like to take issue with Ted's reply as a generalisation, and for a couple of points of explanation.
Certainly, -3dB @ centre works best for stereo listening. -2dB @ center is never the right thing to do. It's too loud in stereo and it's 4dB too loud when you sum to mono.
The problem is that when you sum to mono, a fader with a constant-level signal which is slowly panned from left-to-right (or vice versa) will need to be 6dB down at the center compared to the edges, in order for there to be a constant level across the travel. This is easy. Two linear pots wired on opposition will give you the perfect law for mono.
In stereo, two equal powers sum to 3dB and not 6dB, so the same pan law will sound 3dB too quiet in the middle, when listened to in stereo.
So you have a simple choice if you want absolute equal level in one or other format: -3dB for stereo or -6dB for mono. Some consoles -like older Harrisons and Ameks- had pan law switches which allowed the user to choose on a channel-by-channel basis.
The majority of major manufacturers nowadays split the difference. SSL for example uses -4.5dB. This means that in stereo it sounds a little over a dB quiet in the middle, and in mono, it's a-dB-and-a-bit stronger in the middle. Nobody complains.
I've never ever heard or a -2dB pan law, and I shoudl imagine that the 4dB error in mono would be disasterous. -Ted, if you know of a manufacturer who's ever implemented that one, I'd be VERY interested to try their stuff... I should imagine I'd loathe it!
As for pan controls not working, again I disagree. The mechanics of this are pretty well known and I'm sure most of us are familiar with the influence of timing differences. Certainly for example, spaced mics offer timing cues, but sum to mono like crap. Coincident mics sum to mono beautifully, but offer notiming cues, only amplitude. To say that "Panning doesn't work" is to impute that X/Y micing doesn't work, since it produces similar amplitude distribution with no timing cues. Likewise the inference would be that Soundfield microphones don't work either, since they're fully-coincident, and offer amplitude-only decoded outputs.
I'm also a little confused by the use of the term "sum & difference" other than in the M/S, FM stereo or B-format applifations...
Sorry to put you through the wringer Ted, but I think the "Pan controls don't actually work" is an extremist statement. -Bearing in mind that for timing cues to arrive unpolluted at the listener, acoustic loudspeaker inter-aural crosstalk is a big no-no... so that leaves binaural only... While I LOVE binaural and have enjoyed making many such recordings over the years, it's just impractical for many applications.
I've also used the Studer 950 digital console with VSP, -and was at one point several years ago asked if I'd like to move to Switzerland and take over the product management for it. That console has many different pan modes: amplitude-only (which -as you very correctly point out has its failures) two different HRTF (head-related transfer function) modes, which add timing shifts to the secondary channel, and slightly shift the amplitude & phase responses to a greater or lesser degree, to recreate a cosine-related timing delay and 'shadow' effect (amongst other things... it really is VERY clever!) and an additional "ultra" 'VSP" panning which adds after-event reflections with timing, response and angles which vary with the source angle in a VERY realistic way.
-Of course, even with something as fabulous as the D950S, only amplitude-only panning sums down to mono with no ill effects. You have to consider the purpose.
Keith