Subject:multi-band dynamics

Posted by: tcrabb

Date:1/16/2003 12:05:57 PM

I am an audio hobbiest, not a professional. I am reading, learning, and trying to put it all together. I wanted to come up with a multi-band compressor that was close to the ear's critical bands. The following are the settings I came up with using the sound forge multi-band dynamics. I use the plug-in chainer with the multi-band dynamics 7 times. When I save the chain as a preset, it will save all four bands in each plug-in allowing me to have up to 28 bands of which I use 27. After I get all the bands set up I usually only adjust the threshold and the amount sliders on whatever band might need adjusting. I tend to keep the threshold pretty low like around -30 or so and the amount at around 1.5:1 with the idea of applying less compression to more of the signal. It appears to be working the way I think it should and it sounds OK to me on the equipment that I have. Here are my questions: 1-What kind of artifacts should I be listening for with this kind of set up? 2-Any advice on improving my attack, release times and frequency band settings. 3-I know it's overkill at least for what I am doing, but it's an exercise in learning so am I at least on the right track as far as compression goes? Thanks in advance for your time and advice.
Tim

My settings are:
band 01..Low-Shelf, below 50hz, attack 20.0, release 1,000
band 02..Band-Notch, cent. freq. 57hz, width 0.4, attack 20.0, release 1,000
band 03..Band-Notch, cent. freq. 71hz, width 0.4, attack 20.0, release 1,000
band 04..Band-Notch, cent. freq. 90hz, width 0.4, attack 20.0, release 1,000
band 05..Band-Notch, cent. freq. 113hz, width 0.4, attack 20.0, release 1,000
band 06..Band-Notch, cent. freq. 142hz, width 0.4, attack 20.0, release 1,000
band 07..Band-Notch, cent. freq. 180hz, width 0.4, attack 20.0, release 1,000
band 08..Band-Notch, cent. freq. 226hz, width 0.4, attack 20.0, release 1,000
band 09..Band-Notch, cent. freq. 285hz, width 0.4, attack 20.0, release 1,000
band 10..Band-Notch, cent. freq. 359hz, width 0.4, attack 10.0, release 700
band 11..Band-Notch, cent. freq. 426hz, width 0.4, attack 10.0, release 700
band 12..Band-Notch, cent. freq. 570hz, width 0.4, attack 10.0, release 700
band 13..Band-Notch, cent. freq. 718hz, width 0.4, attack 10.0, release 700
band 14..Band-Notch, cent. freq. 905hz, width 0.4, attack 10.0, release 700
band 15..Band-Notch, cent. freq. 1,140hz, width 0.4, attack 10.0, release 700
band 16..Band-Notch, cent. freq. 1,437hz, width 0.4, attack 10.0, release 700
band 17..Band-Notch, cent. freq. 1,810hz, width 0.4, attack 10.0, release 700
band 18..Band-Notch, cent. freq. 2,281hz, width 0.4, attack 10.0, release 700
band 19..Band-Notch, cent. freq. 2,847hz, width 0.4, attack 10.0, release 700
band 20..Band-Notch, cent. freq. 3,620hz, width 0.4, attack 5.0, release 500
band 21..Band-Notch, cent. freq. 4,562hz, width 0.4, attack 5.0, release 500
band 22..Band-Notch, cent. freq. 5,747hz, width 0.4, attack 5.0, release 500
band 23..Band-Notch, cent. freq. 7,241hz, width 0.4, attack 5.0, release 500
band 24..Band-Notch, cent. freq. 9,123hz, width 0.4, attack 5.0, release 500
band 25..Band-Notch, cent. freq. 11,494hz, width 0.4, attack 5.0, release 500
band 26..Band-Notch, cent. freq. 14,483hz, width 0.4, attack 5.0, release 500
band 27..High-Shelf, above 15,000hz, attack 5.0, release 500

Subject:RE: multi-band dynamics

Reply by: tcrabb

Date:1/30/2003 11:09:15 AM

Has anyone had a chance to try this out and see what they think? I would still like to get some kind of feed back.
Thanks for your time/Tim

Subject:RE: multi-band dynamics

Reply by: tcrabb

Date:2/2/2003 10:20:37 PM

I have lowered the attack times on the lower frequencies (bands 1 to 9) to 10.0 because I have not come accross any fast transients in the lower frequencies.

Tim

Subject:RE: multi-band dynamics

Reply by: tcrabb

Date:2/3/2003 10:38:33 PM

I think my release time on the low freqs are a bit long and not giving the compressor enough time to reset between beats. I know it will be different from one project to the next but, in general, how short could I set my release time for in the lower frequencies without having to worry about pumping?

Subject:RE: multi-band dynamics

Reply by: Geoff_Wood

Date:2/4/2003 2:26:35 AM

I've always understood the main use of multiband compressor was with more like 4 bands. Your example may be valid for a particular piece of (extremely problematical) music, but looks waaaaay OTT to me. The huge scale is such that I can't image learning anything from it.

I'd suggest sticking with 4 bands (more readily identifiable !) and experiment with different values in those.

geoff

Subject:RE: multi-band dynamics

Reply by: kilroy

Date:2/4/2003 10:41:41 AM

And there are far more ominous implications to using that many bands, or indeed any multiband dynamics processing period. That has to do with phase shift artifacts. Basically, when you are dealing with any multiple band splitting you have to implement some sort of crossover in order to accomplish your goal. And thats the tricky bit. It's just as tricky when designing speakers as well. Can you imagine a set of speakers with as many crossed over frequencies as what you have set up for your example?

In our mastering service we have never heard an analog multiband dynamics box we liked enough to use, not one, for the reason mentioned above. We could always notice the phase shift messing with the audio...always. On poorer quality playback systems this was alot less apparent. The lesson here is obvious, and shouldn't need explanation.

In the digital realm things are slightly different because a good mastering quality phase linear crossover can be implemented, depending of course on the skill of the programmer and his judgement. These dynamics modules don't have the same caveats. Theoretically the audio is "dismantled" and reassembled while maintaining the original phase relationship. Not all digital versions are created equal though, and you cannot assume that because a certain multiband dynamics module is digital it is also of the phase linear or phase coherant design.

If the mix is well balanced in the frequency domain then you should very rarely need any sort of multiband dynamics manipulation. If we are tempted to use this approach it is a sign that we should probably be asking for a different mix to work with. That is if there is one to be had. We make more frequent use of mutiband processing on restoration work, where what you are working on is all there is to work on. They can save the day for this work because you sometimes have to make dramatic spectral adjustments but you do not necessarily want them to be static in nature, something that you are stuck with using traditional EQing approaches.

Subject:RE: multi-band dynamics

Reply by: doctorfish

Date:2/4/2003 6:13:07 PM

I agree. Stick with four bands. 27 is overkill. Perhaps you should work
more with EQ to fine tune and use compression across a broader range
of frequencies.

Dave

Subject:RE: multi-band dynamics

Reply by: tcrabb

Date:2/4/2003 10:18:44 PM

Thanks everyone for taking the time. Sometimes I make things more complecated than they need to be. I will take your advice and stick to three or four bands.
Tim