posted 14 September 2003 21:58            

DolbySpec.kym

and here
http://www.dolby.com/products/Model740/740_3.pdf

(Highly recommended viewing!)

(The Sound requires KymaX, but it might work on older versions -- who knows?)

This sound splits the incoming signal into 3 broad passbands (L+R treated separately, so modifications might be desired for true stereo processing...). Each passband goes through a 2-stage compressor to produce a compression characteristic with unity gain at -10 dBFS and some amount of boost at the threshold level. The threshold is the same for all 3 bands, but the amount of boost is controllable for each one. The 3 bands are controlled by two crossover frequency controls (FLow and FHi).

The gains specify how much boost should happen at the threshold level. The compression curves are unity gain from 0 dB down to -10 dB, then provide compression dependent on the requested threshold and band boost levels (GLow, GMid, and GHi).

An input gain block provides input attenuation or gain as needed (Dolby), and two toggles provide for enhancement (cmpr) and sidechain listening (side).

The idea here is that high level signals need no enhancing, while signals between -10 dB and threshold get progressively more boost until the maximum boost is achieved at threshold and below.

With toggles Cmpr and Side both unchecked the original signal is heard. With Cmpr checked the sidechain processing is added to the original signal. And with both Cmpr and Side checked, you get to hear only the sidechain signal that will be added to the original signal.

Read the Dolby user's guide for suggestions on settings. As a quick example, you can increase the appearent ambience by dropping the boost on the low and high channels, increase it in the mid channel, and set the crossover frequencies to around 300 Hz and 2 KHz. Threshold should be around -50 dB. No enhancement happens when there are spectral components at loud levels, but as the signal in the midband falls away, the ambience in the recording is enhanced by gentle compression to produce a more rich sound.

Another technique uses the high band to add more of the natural harmonic energy to the mix, enriching it. Too much also enhances noise. But this is a technique for boosting the harmonic richness without going through something like a harmonic exciter that adds artifical harmonics to the mix.

I have my version feeding off an input audio mix on channels 1&2, sending the result out to channels 3&4, and then routing them back in on channels 3&4 from my mixer for additional processing in another parallel chain of Sound blocks, before going back out on Channels 1&2 to the amplifier and speakers. This makes Kyma a really neat effects processing chain, with up to 4 stereo processing chains. Pretty neat!

[The filters are set up as 1-pole filters cascaded in an interesting way to keep the overall transfer function unit gain, without phase shifts, at unity gain settings. Bothersome phase shifts between the bands are eliminated at neutral gain settings. You can verify this for yourself by listening to both Cmpr and Side checked while dropping the gain (GLow, GMid, and GHi) to zero. What you hear is the subtraction of the compressed signal and the original signal -- essentially nothing.]