Various frequency splitters - hardware & software

I was reading an article in "Behind the Glass" (Howard Massey) where Mick Glossop (recordings by Frank Zappa & Van Mossison) used a frequency splitter then did a multi-band compression thing with a A&DR Scamp rack. The scamp rack appears to have about a dozen slots that you plug devices into like compressors, EQs, delays, etc.

Does anyone know about the Scamp or various professional grade frequency splitters (I guess a normal crossover unit might work but could be noisey) . I'd just like to find out a little more about these classic pieces of equipment. Or if there are any software frequency splitters that get used on commercial pieces.

All of this comes to mind while I'm reading of course because of the CEP Frequency Band Splitter that I'm currently using.

I just picked my favorite compressor (today it's Elephant Mastering Limiter - Voxengo) that I wish was multi-band and presto, now it is!

Anybody got any good frequency band splitter stories, hardware, software or Scamp ? Who is A&DR anyway, the book calls it AudioDesign but a search on the internet shows a lot of engineers with the A&DR Scamp.

kylen

Good question!  I am guessing they used bandpass filters like used in graphic equalizers for example.  

I am finding CEP eight band dynamics processing great for 'flattening out' a mixdown, with wide spectrum, and making it clear and sharp at the same time.  I have 8 band-limited dynamics-processor in series in the rack, so no need to split the track into frequencies.  I am presuming the band limiting crossovers sum smoothly, as per a graphic eq for analogy.    

The result is dramatic if the mixdown is not very flat to begin with.  A good mix is not affected much at all.  Either way the instruments seem eisier to hear, and each has better tone.  

Hi DeluXMan -

You're using 8 bands of band limiting.  Did you just select them by just thinking, I'll set one every octave and 1/2 or so. Or have you set them to 'fix' specific frequency zones that have bumps ?

From what you've said I'm guessing you have them spread somewhat systematically and are tying to lightly compress a mix.

Could you tell me how much gain reduction you set per band ? Oops sorry - trick question  you can't ! So it seems you would simply listen and drop the tresholds and make other adjustments by ear on the 8 different bands.

You mentioned one thing...

...the band limiting crossovers sum smoothly, as per a graphic eq for analogy.

I've used CEP band-limiting and Voxengo's 32-band limiting compressors and from talking to the developer over there crossover isn't the best analogy to make but I don't know what is better. Either you or some other posters can straighten me out on this if you would. Regardless of the internals CEP band-limiting is very smooth and transparant when used for light to light-medium compression of full mixes - I agree !

kylen

ps Someday I will be able to better characterize my use of compression on a full mix so I can discuss apples and apples. The term light compression - what is that ? Full mix - what is that ? Mix - what is that ? In other words if I had a full mix composed of instruments close mic'd and ambient mic'd I'd use very little compression if any (gain reduction 0db). If my mix was a pop tune full of close mic'd stuff I'd need lot's of compression (gain reduction 6db or more). So on, and so on... someday... thanks for your help!

There's another one gain reduction - of what headroom, rms min/max, peaks. Oh my the room is spinning again...

I worked out how I have my 5 bands split according to an article I found over here ( http://www.canimi.com/webnew/audiospectruma.html ) where a guy has 8 bands laid out by octave, I think it's more for display purposes as it's not a compression page.

Anyway I'm trying 5 bands for now because if I make the bands any tighter and I step on one too hard with a compressor I can hear it - so I'm going for about 2 octave bandwidth whenever possible.


Frequency Splits:
=================
BASS    20-120   = ~ 2 octaves (not counting low shelving EQ at 40Hz)
LOMID  120-480   = 2 octaves
  MID  480-1920  = 2 octaves
HIMID 1920-7680  = 2 octaves
HIGH  7680-22.5K = ~ 1.5 octaves

Here's how I figured it so if someone sees a mistake please bite my ankle!
(20-40=1 octave, 40-80=1 octave, 80-120=1/2 octave)
(120-240=1 octave, 240-480=1 octave)
(480-960=1 octave, 960-1920=1 octave)
(1920-3840=1 octave, 3840-7680=1 octave)
(7680-15360=1 octave, 15360-22.5K=.45 octaves)


Izotope seems to know a lot about this stuff so maybe I'll see how Ozone default bands are set and how that relates to octaves. I like the octave approach a little better for compression. I was trying to repair a vocal setting up a wide band that included the whole vocal - there were bumps in the lomid range around 300Hz and bumps in the mid range around 650Hz. Trying to compress both of those in the same band kind of defeats the purpose of using a multi-band compressor !

kylen

Right now i have a rack preset with nine bands actually in octaves up to 10kHz. plus the octave from 10kHz. to 20kHz.+ is split in two to get a bit more contour of the high frequencies.  They are a series of CEP dynamics processors set to band-limited.

Also part of the rack, at the front is the channel mixer default at 66%.  This allows you to set how hard you hit the compressers and acts like a master threshold.

The bands were calibrated against pink noise such that from low to high frequencies, the threshold gets lower to match the lower average levels at increasing fequencies.

I have the the attack and release setting to allow transients.  They are the same for all frequencies atm, but there may be a logical spread for attack/release values across the spectrum, to produce the sharpest tighest mix.  logic would suggest that longer attack times go hand in hand with lower frequency bands, but this needs to be verified with various material or by other engineers, before i tweek the compressers again.

I can only hear compression modulated frequency bands if i use way too much compression, way more than needed to get excellent results, so i can't say i've had this problem yet.

It is a pc hog so i do a mixdown to empty track and then remove it from the effects rack, and then use ozone from there.  

Thanks DeluXman !

Very Cool rack and setup ! Thanks for the explaination too ! I also like your use of MIX - very cool.

Right now my PC wouldn't run 9 dynamics processors, maybe when I tune it up.

A couple of questions whenever you get back to this...

The bands were calibrated against pink noise such that from low to high frequencies...

Pink noise has a -3db/octave slope. Did you use a spectrum analyzer, your ears (ouch), or some other type of band/octave pass filter to get the thresholds set using CEP Peak meters?

logic would suggest that longer attack times go hand in hand with lower frequency bands

I've noticed this myself which makes sense as you say. Adjustment of the attack/release (fast, slow, all the in between settings) depends what octave you're setting and if it's an individual track or full mix. Here we're talking about a mastering rack for anyone following this. Generally by the time you get to mastering you're only trying to adjust the dynamics a bit and smooth them a little better if they need it.

Also part of the rack, at the front is the channel mixer default at 66%. This allows you to set how hard you hit the compressers and acts like a master threshold.

That's interesting, it sounds like a send into the compressors then. The PSP compressors have a mix knob but it mixes in from the output and the threshold is always effectively the same. With the Cool Edit rack then you could simply hit the mix and gently push into the compressors till you hear something and back off a bit depending on how smooth you want it.

One other minor thing out of curiousity - are your detectors set to Peak or RMS ?

Thanks for yackin compressors !

kylen

Ya the channel mixer is used as a volume control.  The problem with a rack full of compressers, either virtual or real is that you have to set all the thresholds individually, but you can decrease or increase the input level to change the effective threshold equally on all channels that way.

As long as you arn't looking for instant attack the rms setting is more stable and closer matches our sense of loudness.  another feature of rms is that response-time is tied to the frequency, so in a multi-band setup rms has a head start on frequency dependent attack/release times.  It takes longer to average lower frequencies.  

I set the thresholds by reading off the spectrum analyzer graph of some generated pink noise that was captured on one of the four capture channels.    This is just a starting point too, the perfect curve will vary for each mix, but the pink noise overlaps nicely on most mixes.