Questions about Frequency Selective (Multiband) Limiting

[ptfigg]

As always, thanks for answering my questions.

I understand the possible occurrence of Spectral Gain Intermodulation in wide band dynamics processing. For example: voice frequency range sucking up the spectral energy and dictating gain reduction. In this case the loudness of additional and less dominant elements may be modulated by the dominant voice frequencies.

My question is - How is proper multiband limiting set up in a Broadcast Processor? Is this a manual process, where you would estimate the frequency range(s) of the passing audio, set crossovers, and implement custom limiting from band to band? Or - do modern Broadcast Processors make frequency range (and crossover) adjustments in real time based on the signal passing through it?

The reason I ask? I designed a Broadcast Processor preset for a plugin that I use in post production and spoken word processing. It's essentially a Loudness Maximizer using Dynamics and Limiting modules. This particular plug does not support multiband limiting. However I have access to one that does. I need to grasp multiband limiting in a broadcast environment and how to determine proper setup in order to enhance what I have already done.

thanks, again.

-paul.

 

[GeneSavage]

As I understand, crossover points are fixed, & limiting per band-pass is independent... even if it is pre-set at the factory & you're only given a "less / more" knob, chances are they handle each band a little differently.

In playing around with Sound Solution & Adobe Audition I've noticed that it seems you can use a faster attack & release for the higher octaves than for the lower... seems like I may have read something from the godfathers of multiband processing suggesting the same. Chances are the faster the vibration (the higher the frequency) the faster you need the processor to react... or something. (I'm no expert!)

To get fancy with it, you can put in a delayed-release circuit of your own design, allowing it to sit still if there's about to be another peak, but fall quickly if that 2nd peak never comes. There's also all kinds of options to cancel out distortion from the limiting / clipping but using bandpass as well, but I'm still learning those details, too.

The challenge with multiband processing is that if it's compression, you can end up push various instruments "forward" and "back" in the mix, even as the texture of the sound doesn't change. You also can have issues if you're not feeding a consistent density to the limiters. Feed a very open, dynamic recording through only a very slow leveler before hitting the limiters, your sound is very open and clear. Run a very DENSE recording (like many of today's pop records) through that same gentle leveler and even though the peaks are hitting the limiter the same, the greater density of the recording is passed to the limiters and you ended up with a very thick, smashed, crammed-up-against-the-speaker sound... so there's a real trick to getting your broadband (or dual-band, or tri-band, or even 4 band) compressor to feed the right levels and density to your limiters.

I don't know if I'm helping, or telling you things you already know; hopefully some of this is useful, and if I've gotten anything wrong maybe this will get people to come out of hiding and give you a REAL answer.

Good luck with your efforts!

 

[ptfigg]

-A,

Thanks for your response.

if you intend to record spoken word for broadcast, keep it as clean and neutral as possible and let the audio chain of the broadcast station handle subsequent processing

Of course, and totally understood. In fact in the past I have designed presets for clients that emulate the functionality and results of hardware voice processors. So equalization including band passing, de-essing, dynamics, etc. They would submit a dry recording the lacks any sort of loudness maximization and I would set things up based on what they were looking for. Most of the time these clients were distributing audio on the web as podcasts. Depending on their production methods and workflows, the preset would be used in real time or in post production after recording. In general this pre-processed audio was not suitable for distribution due to low output levels, insufficient average loudness, etc.

The next challenge is advising clients on how to maximize loudness in preparation for distribution. I do this as well.

The preset that I just built can be used in various ways to bump up levels of pre-processed audio in a typical workflow. To keep things simple I implemented a Macro slider/control that is essentially a fader that controls the cumulative effect of multiple processing modules in the plugin. A higher setting results in an increase of average loudness (RMS). For example - as you increase the slider the single band limiter Threshold is decreased. Compression becomes more aggressive and the makeup gain is increased accordingly. What I want to do next is apply what I've already done to a plugin that supports multi band limiting. I want to emulate a broadcast processor that can be used with applications like Skype or Nicecast in RT ... or used in post as a processed (bounced) effect. The technical aspects are not a problem. What I seek is a better understanding of how a modern broadcast processor handles things like AGC and Multi band limiting.

As a side note - I've come across a few software "broadcast processor" options. (Breakaway Broadcast processor, AudioProc, etc). The problem here is no Mac versions.

I hope this makes sense, and thanks again.

-paul.

 

[Goat Rodeo Cowboy]

Paul, don't allow yourself to get caught up in the whirlpool that seems to have enslaved parts of the broadcast industry. I suggest that you not define your goal as getting the LOUDEST possible signal at all times.

Break up your definition at bit: The overall signal has to be loud enough to overcome the ambient noise of the delivery system. Modern digital devices such as portable mp3 players, etc. do not have the same nigh noise floor that we used to have with tube devices and cassette tapes. And depending on the subject and content, you may not actually want obsessive amounts of sameness. (Ever been in a factory where stamping presses are pumping out automobile parts like door, trunk lids, hoods, fenders? Monotonous as hell! The people you record/process may sound much better with just a little bit of breathing room on the dynamic levels.

One more suggestion: Find out what the "delivery vehicle" will be commonly used by the people who consume the product you are processing. Don't always judge the impact of your work on what you hear on your studio monitors. Strap on an mp3 player and go for a walk and listen to some of your product. Sometimes I load up a collection of mp3 cuts on my player and go for a walk. Sometimes I shut the blasted thing off after about 34 minutes. I get agitated at the factory stamping press gait.

I guess I am a slow learner. My only multi-band processor is the version of Izotope licensed for use by Adobe Audition. It has taken me close to three years to finally get to where I am comfortable using it. It lets me reach up and drag the boundaries between the four bands back and forth. I find that with female voices I have to raise the dividing boundary between the lowest band and the one above it, or it becomes a wasted band. Moving the boundaries becomes a bit of an equalization process which may overemphasize what you just did in an equalization process, or it may nullify the improvement in sound you just accomplished.

I process the same small handful of voices over and over. Each time I find myself having to tweak both the EQ and the Multi-band processor because on different dates the person speaking reinvents his/her voice slightly.

 

[ptfigg]

Thanks, Cowboy.

Paul, don't allow yourself to get caught up in the whirlpool that seems to have enslaved parts of the broadcast industry. I suggest that you not define your goal as getting the LOUDEST possible signal at all times.

Indeed, and I appreciate that . However it's not my goal. I'm not looking to build something that simply spits out the loudest possible signal. Personally I have no problem delivering produced masters to clients that exhibit sufficient loudness for web-based distribution. The problem is there are many so called "producers" out there that lack the tools and knowledge to get this done properly. My preset, if properly designed - would allow user defined results depending on what and how they plan to distribute. Also, I realize there is no real reference standard, and it's totally subjective.

By the way, the preset I built is for iZotope Alloy. I want to move this over to iZotope Ozone (the same basic processor bundled in Audition, I think) due to it's support for multi band limiting. Needless to say your experience with it is of interest to me. I use this plugin all the time in single (wide-band) modes.

I also appreciate your advise on judgment including what I hear on studio monitors as opposed to MP3 players, stepping away, etc. I'm with you on this.

-paul.