Wow some of the replies to this seem very very technical, and not something even as an engineer myself (with reasonble golden ears), would want to bother learning about the real nitty gritty to understand the numbers of adjustment.
You should keep everything in terms of simple, ie: in dB not percent and numbers and divisions and stuff.
I noticed on page 1 or 2 someone mentioned inovonics. I think this is here John Burnill started out or a processor that had the MLB software or something in it? I could be mistaken.
I have slow PC's so I can't run stereotool, I still run Sonos and MBL4 (which John made) and would use stereotool if I could.
The AGC has an emergency attack/release setting - real simple to understand for the user - you set the emergency attack threshold in dB, say a sudden 8 to 10dB RMS jump, and an emergency release threshold of say -8 to -10dB RMS jump, and the AGC will act instantly based on those figures in an emergency, such as a one off loud spike of thumping a microphone or the table etc... Remembering, it'll only sound louder if you are acting on average or RMS changes in the audio, not so much transient peaks to the human ear.
If that fails, then it's even easier to implement emergency attack/release, go thump your audio engineer and tell him to learn about sound and levels and voice processors! AGC's should only be for lack of RMS or average level, not take the roll of a peak limiter or voice processor.
I still use VU metering and all audio is set on PC's to the -20dBFS SMPTE recommendations as found in most TV studios as well (a 1KHz sinewave at -20dBFS digital peak = 0VU or +4dBu on a balanced sound card and console). This means peaks can dance anywhere from about -13 to -6dBFS (so the gear peaks output from say about +11dBu to +18dBu with room to the full +24dBu out for live production environments [+24dBu take away 20dB headroom = +4dBu normal operating level or 0VU]. Because I follow this standard, audio stored on PC is amplified not normalised - to 0VU on average (not -0dBFS on a digital peak meter as some incorrectly call VU meters) and no two songs have the same peak level. The PC's full scale peak level (incorrectly called VU sometimes) is just there for clip protection, it shouldn't be what you follow, that's what ears and VU is for. To me, this is how a real broadcast studio library should be stored for consistant mixes and segues and especially, when using various/random hooks from songs in promo's etc.... so they all sound tidy to the ear before processing and fattening up (made louder) by the processor last. This way, heavily compressed audio is stored on average along with really older dynamic stuff... stored to the same average level not peak level. That way, the more dynamic song has higher peaks, and the processor chops them off, as for less dynamic hyper-compressed CD, the peaks are lower yet the RMS is still the same and the processor downstream can then back off it's limiters on an already loud track since it lacks the higher peak output. I think this is basic sound engineering, but most of this has been lost in the digital age of CD's limited to a dB of their life, and there's more domestic not commercial sound engineers around learning by what they buy at the CD store or iTunes, not what they produce and how they manage studio levels in a live broadcast environment. In the studio, production is very different from the final CD or broadcast. Less is more in the studio, much more attention should be paid here than on the final processor.
Yes there is always need for an AGC, but realistically, if you need emergency response in an AGC, you're lacking in studio design... voice processor/limiter, that's about all it takes to solve that live peak based problem. That's the job of limiters, not AGC's full stop.
Cheers,
Gavin.
PS: I have worked hundreds of hours on sound presets and dialing up that 'sweet spot' on broadcast processors, and there's only one way to sound loud and the best on the dial, less is more in the studio, processing is the last thing in the chain, and never should be treated as the first - ie: AGC's acting like limiters.
|