frequency splitting

  • Thread starter Thread starter dobro
  • Start date Start date
Sure, you're talking about how accurate Cool's splitter is, and I've been wondering about that myself. I mean, the thing I don't like about highpass filters is that they're *approximate*.

But what you're ignoring here is this: the splitter produces eight tracks out of one track. And then you can delete one or more of those tracks and make a mixdown of what's left, completely eliminating everything south of the lowest note in the mix.

You say you could do the same with a steep highpass filter. Well, how steep? I put a Q of 1.6 on the track I was working on, and there was still loads of stuff going on below the cutoff point. But using the splitter and then dumping the range below the cutoff point, there's NOTHING going on below the cutoff point.

Look, try it yourself. Before using the splitter and dumping the range you don't want, the track's got bottom end you don't want, EVEN IF YOU'VE USED A HIGHPASS FILTER. After using the splitter and dumping the range you don't want, the track's empty of all that range.

In the first case, there are traces of what you don't want in the bottom of the barrel. In the second case, the bottom of the barrel is scrupulously empty of what you don't want. Make your choice.
 
Just so Blue Bear doesn't jump my shit - I mentioned fast attack times on the low end...if you don't know what I'm talking about, then it's probably best that you ignore that until you do, lol.

If you DO know what I'm talking about, understand that a fast attack time on low frequencies DOES have the effect of distorting the signal, b/c lower frequencies take longer to develop. To me, the overall effect is worth it, as long as the distortion is kept in check in a "musical" way. You can obviously check this a lot more easily if you have a track split into different bands. I usually don't have a lot of big peaks in the low end, b/c the bass is already compressed, and my kick drums usually have 90% of their energy between 50Hz & 120Hz...If you have a real "punchy" kick drum, you'll probably want higher attack times on the low end.

After all, I'm a freaking amateur, lol.

Dobro - lmao @ "stardust." ...hehehee...more like Angel Dust.

Change Of Poets - What Dobro's talking about is the very very low shit, like below 40Hz. Technically, the human ear CAN hear all the way down to 20Hz, but this is well within the realm of stuff that's mostly felt instead of heard. It's also the stuff that creates problems in a few ways. For example, if you get a "plosive" sound on a mic, this is what you're hearing...you can eliminate it by running a High pass filter (low cut) on the track to kill everything below 40Hz. Cleans it right up. Did you bump the mic stand? Did somebody slam a door while you were tracking? All cleaned up by cutting stuff below 40Hz.

I'll add that another eye opening experience was looking at commercial mixes on some kind of a spectrum analyzer. Do this, and you'll usually see a common pattern in terms of EQ on the top and bottom. The low end usually peaks somewhere around 63Hz, then drops off DRAMATICALLY below that. I'm not saying they shelf the mix, b/c there's still energy needed at 50Hz (kick, for example), but there's definitely a big cut that happens in mastering usually below 40Hz or 30Hz. There's also a similar cut that seems to happen around 16K...(above it, I should say). Now, don't try this with an mp3, b/c mp3's cut off everything above 16K as part of the compression process...it's like a brick wall if you look at an mp3 on an analyzer...but it's similar in commercial mixes on CD.

Here's my theory. Commercial mixes tend to sound good on a bunch of different systems, right? Well, some systems have GREAT low end, and some (desktop computer speakers) have almost none, so the goal is to get the mixes to sound similar on all systems, and you can't do that if you've got shat below 40Hz in your mix...on some systems, it'll be pushing a TON of air, and it'll have the effect of masking the top end. So on a great system, the mix may sound super bassy. The same tune, on desktop speakers, would probably sound really tinny, b/c it's not playing the low stuff, so there's no "masking" of the upper frequencies, so you hear the top end too prominently.

By containing the mix within frequencies that are easily played back on most systems, you can get a much better idea of how your mix is gonna' sound wherever it's played...it'll sound like it does on your monitors. :D

That's my theory, anyway.
 
I'm not ignoring anything. Yes, you end up with 8 separate files. But that does not mean that the cutoffs are any more precise. It is not like teh computer goes bit-by-bit through the file and says "ok, this bit goes to the 400-600hz file, and that one goes to the 800-1000hz file." It applies filters.

As a repeatable test, I created a new wav file at 44.1khz, and generated about 10 seconds of a 440Hz tone at -6dB, using the "Generate....Tone" function.

I then used the frequency splitter to split this file into 6 bands: 0-200,200-400,400-600,600-800, and 800-22050.

Guess what? I can hear the 440hz tone in every one of the files. Here are the levels in each of the files:

0-200: -36 dB
200-400: -14 dB
400-600: -18 dB
600-800: -24 dB
800-1000: -31 dB
1000-up: -23 dB


The original level was -6db, and by the time the target range is raised by an octave to be centered roughly at 880, the signal has dropped by roughly 25dB. Seems awful close to a 24/db octave filter to me.
 
You're talking precision. I'm talking 'getting rid of'.

See, what I'm talking about is this: you take that lowest range you got after you generated and split, and then DUMP IT. Make a new mix. It won't have that range no more. There won't be anything 1-200, cuz you axed that frequency range from the mix.

As I see it, the only way I'm wrong is if there's this Catch 22 built into the equation that goes something like this: 'you can't trust the frequency splitter to dump frequencies you don't want cuz you can't trust your frequency splitter to recognize them properly'.

I still say it dumps more frequencies than a highpass filter.
 
dobro said:
You're talking precision. I'm talking 'getting rid of'.
Click to expand...

Ok, so which of those ranges should I delete if i want to get rid of that 440hz tone? 400-600? Whichever you get rid of, you'd still hear it.

I don't doubt it has a steeper slope than whichever filter you were using before. But it is still far from "scrupulously empty of what you don't want."
 
if you want real control over the slope of your high or low pass filter, use a high-order Butterworth, which you can find in the "scientific filter" menu.

not to suggest that it would necessarily be any better than the Frequency Splitter's "magic." just to say it wouldn't be any worse.
 
LOL LOL

I believe you. But could I afford one? LOL A scientific filter, I mean?
 
cooledit offers them. Effects .... Filters .... Scientific Filters
 
Okay, I found them. Tell you what. I'll do a test. I'll run a Butterworth filter on a session. Then I'll do a frequency split on the session. Results posted here the next time I'm sober.
 
jrosenstein said:
Ok, so which of those ranges should I delete if i want to get rid of that 440hz tone? 400-600? Whichever you get rid of, you'd still hear it.

I don't doubt it has a steeper slope than whichever filter you were using before. But it is still far from "scrupulously empty of what you don't want."
Click to expand...

Its easy ;) Just generate a profile for a 440hz sine wave in the noise reduction filter section and apply it. The precision of each of the filters differ because the programmers pick a common slope, cutoff, Q or any other parameter they deem as being the end to their means.

SoMm
 
dobro said:
Okay, I found them. Tell you what. I'll do a test. I'll run a Butterworth filter on a session. Then I'll do a frequency split on the session. Results posted here the next time I'm sober.
Click to expand...

I'll be interested to hear about that. 6th Order or so should make for a damn clear cutoff.
 
Son of Mixerman said:
Its easy ;) Just generate a profile for a 440hz sine wave in the noise reduction filter section and apply it.
Click to expand...

Cheater!!
 
Okay, jrosenstein, I tried the Butterworth scientific filter on the same track with an Order of 10 - nice and steep. I didn't hear any of the phasing problems that the Help said might crop up. I liked it better than the highpass I usually use on low end stuff I don't want.

Then I ran the frequency splitter on the resultant track with three bands:

0-50 Hz
50-100 Hz
100- the top

The outcome was similar to the result on the un-EQed track. There was stuff in both ranges which was easy to hear when each track was soloed - not loud, but it was there. The filter gets rid of lots of the lower end stuff beyond the cutoff point, but not all of it, not even down below 50 Hz. So, again, I muted the two lower ranges (dumping everything below 100 Hz) and made a mixdown of what was left, and the result was a less dark track.

I don't know how accurate the splitter is in terms of nailing the frequencies it says it's nailing, but it does ISOLATE frequency bands in a way that lets you dump them completely.
 
Fair enough.

I just wanted to make sure it was clear that the frequency splitter isn't magic, and it doesn't actually "split" one file into many -- it is just an automated series of filters, though it may or may not have documented the exact process it uses.
 
i just re-read your post --- did you mean that you "split" the track after you had applied the butterworth to it?
 
jrosenstein said:
i just re-read your post --- did you mean that you "split" the track after you had applied the butterworth to it?
Click to expand...

That's the way I read it.

He said he split the rusultant track... meaning the result of the Butterworth...
 
Change of POETS said:
That's the way I read it.

He said he split the rusultant track... meaning the result of the Butterworth...
Click to expand...

If that's the case, then the track was filtered twice .... of COURSE more lows freqs were gone afterwards!

"There was stuff in both ranges which was easy to hear" .... this was my point with the 440hz tone demonstration. When you listen to these ranges, you are hearing predominantly the frequencies you want to, but others are there too. So in listening to the band after the high-pass, you are hearing "frequency band leakage", so to speak, from BOTH the butterworth filter and from the frequency splitter.
 
Dunno, man, you know lots more stuff than me, and you're saying things I don't understand.

For me, what it comes down to is this: do you trust the Cool Edit frequency splitter to actually tell you what's going on in those different frequency ranges?

If yes, then my approach is a sound one for getting rid of low end noise.

If no, then that calls into question all of Cool Edit's filters, including the Butterworth, the parametric, the...

I find, however, that usually when there's this mutual incomprehension thing going on between me and someone else that it's my incomprehension that's the bigger problem. LOL
 
Sure, I'd trust the splitter. I don't mean to come across as saying it's useless, or anything close to that. But the first few things people were saying about it were hailing this feature as a miracle, leaps and bounds beyond any high or low pass filters' wildest dreams.

In fact, the two are one and the same. The frequency splitter automates the process of applying a whole bunch of filters on different tracks, which makes it very cool and useful. But all in all, it's not doing anything that you couldn't do manually if you felt like taking 10 minutes more.

The End.
 
You must log in or register to reply here.
Back
Top