Syntrillium Backup Forums :: View topic

Syntrillium M.D. · Location: USA Posts: 5124

Did you know that you can do multiband compression and limiting in Cool Edit Pro 2.0? Did you know that you can effectively apply any effect in a multiband scenario? You can. So whether you're mastering your tracks and need to apply some careful limiting or perhaps you want to add reverb to the high-end portion of a track without applying reverb to the bass, the key is in CEP2's FREQUENCY BAND SPLITTER.

The Frequency Band Splitter is located in the Multitrack view under the Effects Menu. To begin, select and highlight your file. Go to Effects>Frequency Band Splitter. What this enables you to do is to separate your audio into user-definable bands...so, whether you want to do simple low-Mid-hi, or more complex lo-loMid-Mid-HiMid-hi, you have up to 8 individual bands that you can set.

Let's start with something simple, say, three bands. The crossover field is where you determine the cut-offs. So start with a figure of 120. You'll notice that the first field now states a range of 0Hz-->120Hz. Click on the "3" radio button, and then in the crossover field, type "4500". Now, you'll see that you've selected to use 3 bands, with ranges of 0-120, 120-4500, and 4500-22k (max range determined by the sample rate you're using)

When you're done, click OK. Cool Edit will now extract these bands as individual waves in the multitrack, and insert them in a consecutive manner, below the original wave. Each file will also have the name of the original and the frequency band that it represents.

It's really that simple. Once the files are 'split' multitrack, you can add compression, limiting, reverb, EQ or any other effect and only apply it to the band that you desire. This also makes auditioning of multiband real-time processes (ie, adding a limiter to the low end only) very simple, seeing as you can just 'SOLO' the appropriate band and really hear what's happening, without having to mute or disable other bands in the process.

Now, this is one simple way to use the Frequency Band Splitter. Obviously, there are many uses. Of course, if you simply want to do something like multiband compression, but would rather work off one file in the MT view, simply click on the FX tab. Go into the effects rack and add three instances of the Dynamics Processor. Click Apply and OK. Inside the DP window, there is a Band Limit tab. This tab allows you to do essentially the same thing as the FBS's crossover field. Designate which bands you want to effect, type in the cut-offs, and you're ready to roll.

Multiband Processing for Mixing and Mastering has never been easier...and you've got complete control at your fingertips.

---Syntrillium, M.D.

MusicConductor · Location: USA Posts: 1524

I am looking forward to my next vinyl restoration so I can try the Frequency Band Splitter on it--it's a great way to sum the noise below 150 or 200 Hz to mono and leave the rest alone. Also for de-essing and other vocal tricks... ever see anybody run a vocal through a variable (4-band) Dolby A processor? This can be emulated by the multiband compressor.

If anyone has concern about how accurate it is, try this (make sure everything's 32-bit): split a source file into whatever bands you desire. Don't change or add anything, but simply mix the split files back down. Invert paste the mixdown over the source file. Of course, the main signal will cancel out leaving the noise, distortion, phase shifts, and other inaccuracies behind. What do you get? Well, try it, it's pretty impressive!

Nice going, Synt!

Syntrillium M.D. · Location: USA Posts: 5124

Glad you like it, MusicConductor! We think it's pretty cool too.

Happy recording!

---Syntrillium, M.D.

Heavens to Betsy · Location: USA Posts: 508

LEE TYLER · Location: USA Posts: 47

Bigggggggg Badda-bump!!

I have just spent an hour exploring splitting frequencies with the Frequency Band Splitter, and I noticed in the very low freek regions (20-100hz approx), a little crackle from each speaker on reviewing ONE track at a time, soloed. When I solo two tracks TOGETHER, it disappears. Like....uh....what is this crakling about? Doc?? Thanks in advance, ---Lee

bonnder · Posts: 215

SteveG · Location: United Kingdom Posts: 6695

ROBSCIX · Posts: 254

Ok, cool, I have just started to use the FBS.."VERY COOOL"....Ok now If I am reading this right then, what your actually doing is "Pulling" out the frequencies you like/need..using FSB, I balance the way you like then you Invert Paste over the original?...that I don't get. This could be a good NR process but, how much signal are you going to lose that you need? I am assuming that matching the bands up that close could be difficult. I have been restoring analog cassettes of old 80's Rock. I seem to be having difficult time getting a half decent punch without the WOOF. Could this be an alternate way of restroring old recordings cassettes/Vinyl. Maybe synt could help on this. What I would want to do is try and using the FSB to get resonable "envelopes" across the frequency response of the tape and adjust to taste..so to speak. Any help would be appreciated. BTW the FSB is COOOOOLLLLL;)

kylen · Posts: 290

OK, I'm interested ! This sounds really fun.

The art of dynamics processing is on my mind these days so besides doing some reasonable things with this I can see that I'll also try the 8 frequency split and hook up 8 Ozones giving me a 32 band compressor.

Doubting that 8 Ozones will run at the same time on my 2.66GHz Pent 4 then I'll try putting 8 Vintage Warmers up there (8 of those will run) which would give me 24 bands to go cross-eyed or cross-eared with. Then I'll try all my others and see you in a month. :D

Then I'll probably do something more reasonable like the MD suggested in the first place. Subbass (10-40Hz), Bass (40-120Hz), LoMid (120-500Hz), Mid (500-1K), Hi Mid (1K-4K), Hi (4k-10K), Treble (10K-20K), Dogs (20K-??) for doing dynamics balance and a little repair at the same time.

Plus the fact that I can run the dynamics processors in parallel in the FX rack (I think they refer to that as the New York sound in 'The Mixing Engineers' guide) should make for some precision control.

Thanks for the cool tip, I'm glad this one got a bump:) !
kylen

LEE TYLER · Location: USA Posts: 47

kylen · Posts: 290

Hi Lee,

I need to try this tonight (I'm at work now) using dynamics to smooth out and repair a very ruff mix that can't be remixed. I've tried to use the 4 band Ozone which is very transparant on the mids, and the 3 band Vintage Warmer which gives me some great ooomph (technical term Big Grin

) to the bottom end down around 40-60Hz, and splashy cymbols on the highs (still working on my splash).

Long story short, now I can 'glue' 2 Vintage Warmers and 1 Ozone together which I'm most excited about. I'll keep my ear peeled for the crackling you mentioned down low since that's where I'll be working a lot.

kylen

LEE TYLER · Location: USA Posts: 47

Thanks!

kylen · Posts: 290

Hi Lee,

I forgot to mention - cool palm trees. Mine are on the West coast...

LEE TYLER · Location: USA Posts: 47

Thanks for responding, Kylen. I'll be watching this thread for any answers. Thanks ----Lee

kylen · Posts: 290

Syntrillium MD & Syntrillium/Adobe Support:

I'm giving this a friendly bump until you folks have time to look at the problem of crackle and clicks using CEP 2.1 Frequency band splitter. :)

It is too good of a feature to ignore.

FBS allows me to split a song up, based on various crossover frequencies, into seperate wave tracks. I can then insert a compressor on each resulting track (up to Cool

. Each compressor can be different, with a different character for a very tailored sound.

I can easily glue a Vintage Warmer, PSP MixSaturator, Timeworks single band and an Ozone together to make one huge custom compressor. Tongue

Please help me out here.

Thanks,
kylen

dobro · Posts: 342

Ookayyyy... so explain me this...

I split a mixdown into a number of bands. Then I muted the two lowest bands (0-50 Hz and 50-100 Hz) and made a new mixdown out of what was left. I thought I was being really clever and dumping everything under 100 Hz. But when I then ran the splitter on the new mixdown, there was still stuff going on under 100 Hz!

So, here's my $17 question: does the splitter actually split the frequencies, or does it just analyze? If it splits, then how can the above scenario be explained?

See, I *really* want the splitter to work the way I was trying to make it work. You have no idea. I want to be able to absolutely AXE everything south of a particular frequency. The scientific filters have very steep filters, I know, but I want something even more radical. I want something that dumps EVERYTHING below the lowest note in the mix. Can the splitter do that?

kylen · Posts: 290

Hi dobro,

This is not intended to take the place of an answer from Support.:)

I think you and I have the same question pertaining to the slope amount per octave, 3db, 6 db, hopefully it is supposed to be 20 db or more. If you look at your wave in the spectrum analyzer you'll see the same stuff you're hearing. There is a hump or peak in the new band you have created but its' slope is such that a bunch of other frequncies outside of the cutoff are present also. So there is noise and slope issues concerning the lower freq splits. I've only had time to look at one band so far so I don't know what the other higher ones look like.

For what you want to do I used a band pass filter from one of the CEP native effects that I described above, I selected a patch called 'Subbass only' then adjusted the frequency to 50 Hz for my own experiment, and adjusted the slope to pretty much vertical (something you don't necessarily want in a crossover slope).

I can tell you which native effect I was using tonight after work if you can't find it, it's under effects-->filters maybe its the FFT filter.

kylen

bonnder · Posts: 215

dobro - have you tried the FFT filter? That should dump everything below a target frequency, provided you set it up correctly.

For those who might not know how - Use "Passive" (upper left)and "Log Scale" for low notes (mid-center). Drag both ends of the yellow line to the bottom of the window (you should have a flat line at "0 %"). Left-click on the yellow line and you will get a small square white box. Right-click on that box and type in your target frequency. "Amplification" should equal zero. Just to the right of that box, left-click again and get another box. Right-click on that box and type in your target frequency. "Amplification" should equal 100. Drag the white box at the bottom right of the window up to the top right. That should do it. Name and save your preset if you want. You must highlight something in the edit window before the preset will work.

If you don't know how to change a note to it's corresponding frequency, download this frequency converter:

http://www.analogx.com/contents/download/audio/freq.htm

If ya'll know all of this already, then nevermind.

Craig Jackman · Location: Canada Posts: 909

While I give Syntrillium full credit for FBS, and figuring out that you can do multiband dynamics with it ... it's still not as clean and elegent a solution as using somtihing like Ozone, Waves C4, or the db Audio Multiband limiter to achieve the same results. Perhaps in CEP 2.2?

dobro · Posts: 342

Craig, care to elaborate?

I'm still trying to figure out what the FBS can do, and what it can't do.

What makes it less than clean and elegant in your view?

MusicConductor · Location: USA Posts: 1524

Talk about dragging up an old thread! And a bunch of these questions are still open for discussion. I'll add my two cents.

First of all, Heavens to Betsy and Dobro were trying to get an idea how steep the FBS filter is. Let me illustrate with a rather odd set of parameters, which I used when testing the FBS before doing an organ project. I wanted to control the bass an octave apart, so I made the split points 40 and 80 Hz. In an effort to create the steepest filter possible, I used a FIR setting of 10,240--not that it makes much difference by going higher than 1,280 or so (at most a couple of dB). Then I used my perfectly flat-response white noise file, split it, and used the frequency analysis window to see how it all ended up. This is what I got:

This is good evidence that the FBS functions as a linear, not logarithmic, filter. Thus, "dB per octave" doesn't apply, and the separation between bands becomes quite steep in anything but the bass. This is consistent with dobro's misplaced expectations -- if you want to dump everything below a certain point, use the FFT or scientific filters as suggested, not the FBS.

The experiment that I did to create this graph resulted in my splitting at only one point (60Hz) instead of two because of the relatively low separation at low frequencies. This served my purposes very well, as it turns out.

CLICKING: it seems curious to me that I did an entire CD with bass splits as described above and never heard a single click. For both Lee Tyler and Kylen, please do this: when you hear a click, note the spot in the file, switch to Edit View/Spectral View, and determine if the click is actually in the file data or not. Subtle sound card sync problems are sometimes only obvious with pure bass tones and sounds; perhaps your card is dropping samples and you're not noticing it when the full frequency spectrum is present to mask it. Another possibility is perhaps from using headphones, which Kylen had mentioned (how about Lee Tyler, what are you monitoring on?). Do the clicks disappear altogether if you reduce the volume? Are the headphones' drivers perhaps unable to handle the large, pure bass excursions without a little whining?

INVERT PASTING AND MY OLD COMMENT FROM 4/24/02:
The point of invert pasting was simply to reveal artifacts introduced by the FBS alone. I hope I do a better job explaining myself on average than I did that day! Robscix, I wouldn't invert-paste only one part of the audio band as part of a normal processing job. Bonnder, my claims have a big loophole in them, which I'll try to explain in the next paragraph. SteveG's response include a very good, basic technique that is made possible by the FBS, but was not my point in mentioning invert-pasting. I'll try again.

My initial use of the FBS was to see just how good it was. I split a white noise file into a whole bunch of bands. Then I muted the original file, and without changing any levels or inserting any other changes to the audio, mixed down the split files. In theory, this would yield the data of the original file, plus any possible inconsistencies resulting from the splitting process (distortion, noise, and phase errors). An easy way to test for those inconsistencies is to invert-paste the mixdown file over the original (both of which, by the way, should be 32-bit). My little experiment was gratifying in that those side-effects were unbelievably low in amplitude, easily greater than 24-bit accuracy. When I posted those glowing comments over a year ago, I didn't account for the theoretical possibility that a phase error or other audio inconsistencies that are split symmetrically between two adjacent bands could conceivably cancel out upon mixdown but be present in a band by itself--and thus get added back in to the mix if processing a band negates the possibility of cancellation. That's the "loophole." That is also groundless speculation and paranoia. I've no reason to believe that the FBS is anything other than what it appears to be: another excellent tool of surgical accuracy and precision.

LEE TYLER · Location: USA Posts: 47

kylen · Posts: 290

MusicConductor,

Very cool of you to offer such a detailed explaination complete with pictures.:P

Long story short - It's effectively fixed for my use. I changed my fir setting from the default of 320 to 10240 like you mentioned and the whitenoise test is clickless at the 0-40Hz band now. When I move the mixer slider for that channel up so the peak meter reads about -16db or so I hear 3 or 4 'settle-in clicks' and no other noise or artifacts garbage, the same thing happens when I decrease the volume. This means no volume automation at this point which I don't plan to do anyway. I'll need to look at that further.

If you were to split out some white noise at 320 I'd be willing to bet you'll hear what we're hearing. How did you choose 10240 ? :???:

Just to clear up a couple of misc details you brought up for monitoring I use: Sony MDR7509 cans, Alesis M1 active nears, Realistic Optimus50 mains. The clicks and crackle don't sound like the kind of thing when a cone travels too far or otherwise distorts. They're quieter than that, I think you would hear them if you ran the 320 fir. I just tested mine at 320 again and I can make the problem come back using that bad setting.

Also the spectrum clears up quite nicely at 10240 max fir and the spectrum slope appears a lot more reasonable finally going into oblivion at 120Hz or so. The 0-40Hz band starts rolling off very nicely (-18db) by around 85Hz and it's down -12db at 60Hz.

This brings us back to slope which, to be honest with you I don't have any requirements or expectations at this point since I'm just beginning to play with this and maybe compare it to how Ozone rolls off or crosses over at the splits, ah - in my free time...

The big thing for me is how it sounds and can I control what I need to.

Now I can continue with my experiment I detailed earlier - a super compressor with multiple personalities. Well, after thanking you kind folks for the great gift I'm back to my anniversary evening Wink

kylen · Posts: 290

...one more thing, I found this about accuracy in the CEP manual:

kylen · Posts: 290

I have additional information from a reliable source who is looking into the matter that the default value for low frequencies is in question.

For the clicking and crackling issue the recommendation on this thread is a max FIR setting of anywhere between 4500 and 10240 works.

I have personal experience with the 10240 setting and can state that it is clean in my configuration on this side of the world doing what I'm doing at the low end. I have yet to listen to the other bands but from what I gather so far the expectation is this is currently just a low end issue.

I would expect to hear from the folks responsible for determining those settings in the future when they have the chance to get to it.:)

kylen

ps Don't forget if this is a production problem for your business and you need help you can contact Adobe/Syntrillium support also they're pretty responsive at the support level. Approve

MusicConductor · Location: USA Posts: 1524

My primary HDD failed last evening, very thankfully after a successful backup. Thus I can't try anything out for a while yet, and when I can, I'll try the 320 FIR on my test noise file. My original tests--with no clicks at any FIR setting--were done in 2.0, and perhaps a little glitch was introduced in the new version. I'll try to repro it soon.

How did I come up with 10,240? I kept using increasing multiples of 320 because the bass resolution wasn't good enough. Finally I just tried what seemed like a ridiculously high number and got about 3 dB improvement over the lower ones.

Thanks for all the details, guys. You really do sound like you know what you're doing, and my purpose wasn't to question that; rather, just to leave no stone unturned, including the obvious. Now on to the subtle!

kylen · Posts: 290

Thanks again for the help MusicConductor.

I was searching the web a little trying to see what the length of a fir filter even means and now I see why it's described as "the length of the fir filter". Argh !@#$! Holy smokes the things goin on underneath the U.I. Shock

I'll have a bad dreams over that for sure! I would certainly tell the boss some of those hairy formulas equal 320, yep every time !:D

Anyway I was just trying to see if I can shoot my self by setting the max fir filter value too high and I really can't say. I see things like ripple and distortion mentioned but it depends on CEP filter implementation I guess and I don't even know if the numbers we input at the U.I. level have any real meaning at the filter design level.

So I'll just listen (in headphones) to what I'm splitting out and make sure I save the original till we get more information. I'm worried that I'll split a low band out and won't notice the clicks till later when I compress it and they start sticking their heads out. But I'll just have to do that sooner.

kylen

LEE TYLER · Location: USA Posts: 47

I tried to investigate the infamous FIR filter at different levels whilst setting freek ranges 40 HZ apart starting at the 40HZ level. Tried an FIR of 500, no change. Then tried 1280, problem gone. Soloing all of the split tracks combined sounded exactly like the original unsplit track via A/B comparison. I guess that means we are all in business???? Does it? I am wondering. ---Lee

<~~~ME

dobro · Posts: 342

I found that a setting of FIR 312 doesn't find/get rid of all the frequencies below 90 Hz. I did almost the same thing MC did. I ran the splitter on a mixdown with the lowest band 0-90 Hz. Then I muted the original track and the 0-90 band, and made a new mixdown of what was left. I then ran the splitter on the new mixdown to see if everything below 90 Hz was gone. It was reduced, but not gone.

Okay. So after reading this thread I did exactly the same thing with a FIR of 5000. Again, there was still stuff going on below 90 Hz, but less than before.

My third step was to contact Synt. I'll let you know if I learn anything more useful than what's already in this thread.

2Bdecided · Posts: 340

I _still_ haven't bought CEP2, though I might now that I've read through this thread.

I have been building some multi-band processing within MATLAB recently, and as soon as someone mentioned that invert-mix-paste over the original was perfect, but the lowest band had clicks in it, I thought I knew where this was going!

I don't want to second-guess what synt have programmed here, because I'm sure they're a lot smarter than me. But I'm going to explain the way I was trying it, because it might explain the clicks to people. It may also suggest how we can confidently avoid them - if synt will explain what the parameters in their version mean!

Before we start: FFT means "Fast Fourier Transform". The Fourier transform converts signals from the time (waveform) domain into the frequency (pitch, or spectrum) domain, and/or back again.

OK, here we go.

Basically, this process involves filtering in the frequency domain rather than the time domain, and this causes the problems. You may think of a "filter" as a frequency domain thing, but in truth, all signals are time-domain, and any possible filter has a time domain response. This means you have two options:

1. Use the actual time-domain input signal (what you see in waveform view), and filter this. To do this, you need to get the filter into the time domain too. Let's say you draw the frequency response, and then FFT it to get a time-domain version of that filter. The result is a short click. (the frequency response of the click is equal to that of your filter). You convolve this click with the signal, and, as if by magic, you get the filtered result.

or

2. Use the actual frequency response of the filter to, er, filter! This means that you have to get the time-domain input signal (what you see in waveform view) into the frequency domain. To do this, you FFT it, usually in blocks. Then you multiply the result by the frequency response you want, and inverse FFT this to get back into the time domain. Magic - filtered signal.

The advantage of the first method is you only have one FFT calculation - to get the time-domain version of the filter. There are other ways of finding out what this should be, so you might have to do no FFT calculations at all. Cool!

However, the advantage of the second method is that it's much quicker. FFT is a relatively quick calculation - especially compared to convolution. So, lots of FFT processing is much better than lots of convolution processing. The two are mathematically equivalent (perfectly), so the output will be identical. It's just a question of speed. Very Cool! Well, almost...

Now here's the problem: A filter which does something to low frequencies will have a long time-domain response. Think about it: if the filter has to do something to a 20Hz tone, then it has to "see" a few cycles at 20Hz. This means what you convolve must be very long, and any single sample of the input signal will have an effect on several thousand samples of the filtered signal. Think about filtering a sharp click at 50Hz: you'll get a slow dull thud.

When convolving, this means that the time-domain version of the filter is very long, but we will get the result that we expect. If we can't be bothered to use the entire length of the time-domain filter, we can shorten it. This will change the frequency response (it won't be quite what we want anymore), but it may be close enough.

Using method 2, we have a problem. We're working in blocks. We can't FFT the whole file because this would be slow - it's much faster to FFT a thousand samples at a time, do the filtering, then move on to the next thousand.

But what if the filter is so long that it should have an effect beyond the length of the block that we've chosen? The answer is: it won't work. The value of a sample in one block cannot have any effect on the output of the next block, because the FFT and filter of block n will never see the samples from block n-1.

Worse still, the FFT and filter that does see the sample will still try to filter it. It can't put the result into the next block, so it just puts it into the current one. So, if the time-domain version of the filter (which we didn't bother calculating) is longer than the block length we've chosen, then the extra part of the result which doesn't fit properly into the current block gets added to the start of the current block - NOT the start of the next block where it belongs. You could say that it wraps around.

Result: a click at the start of the block!

You can only hear it when there's something in the input audio signal - if you filter silence, you'll still get silence. Even more bizare: if you design a set of filters to be complimentary (like in the multi-band filter), then these clicks will cancel out when you add the other frequency bands back in. But if you listen to one band on it's own (or process it on it's own) then you'll hear them.

It's obvious what we must do to prevent this: check the time-domain version of the filter, see how long it is, and make damn sure that the block length is longer.

(Smart readers may also realise that, no matter how long the block is, the last few samples of it would always have an effect on the next block - but instead will wreck the start of the current block because of the wrap around effect. For this reason, we always have to dump the first samples of the result because they're always wrong. The process is called overlap and add, and it's explained properly on lots of web pages. The "few samples" are actually as many as are needed for the time-domain representation of the filter).

The time-domain representation of a filter is longer if
a) lower frequencies are adjusted, or
b) greater slopes are enforced.

So, to make a gentle adjustment to only high frequencies, you could use a very short filter indeed. A steep cut at low frequencies will need a very long filter.

The MBS is hopefully designed such that (b) is not a problem. To prevent (a) being a problem, they seem to have left it to the user to set the correct FFT size.

You can (roughly) find out the length of the time-domain response of a filter. Generate 1 second of silence. Zoom in, and drag the centre sample up to full scale. Zoom out, and filter the wave with your filter. Now, zoom in vertically and see how long the "wiggles" in the impulse response last for. That's how many samples you need to allow for the FFT length. In practice, allow 2-10 times as many to let the overlap-and-add process work. Also, in practice, the "wiggles" last forever, but you have to pick a point beyond which they don't matter. How loud a click will you be able to ignore during the music? -90dB? -20dB? Your choice - see where the time-domain filter response )i.e. the wigles) have dropped off to this level, and use that length (or 2-10 times that) for the FFT length.

If synt explain the parameters, you can use this to set the FFT in the MBS. However, it would be nice if this were set to an optimum value automatically. So you say "I want it accurate down to -90dB" and the prog selects the FFT size to do it.

Enough guessing. They probably don't do it this way at all!

If you read all this post, congratulations! Even more so if you understood it! ;)

Cheers,
David.

SteveG · Location: United Kingdom Posts: 6695

Yes, I understood it... :D

Just as a matter of note, the Boss has spoken about this before - so there is an alternative explanation 'Parametric EQ Low/high Shelf' that you might like to read as well. It also mentions the other potential issue with filters - the difference between FIR and IIR.

kylen · Posts: 290

Thanks very much 2Bdecided and SteveG for my continued education concerning fir design which is rather limited at the present. It is a very interesting topic.

But for now I'm just a filter user and customer so I reported my findings to Syntrillium also for them to chew on via their email support mechanism.

What I'm curious about is the "clicks are present but we can't hear them cause they cancel out" comments. I'm thinking if I compress a low band a lot (like 12db or more of gain reduction) then the click as well as the program content won't have the same characteristics anymore as it once did with the clicks in the other bands and now they won't cancel. I haven't finished testing my paranoia here yet.

The other funny thing is that when max fir value is set so that you don't hear any clicks while a solo'd low band is playing (max fir 10240) then adjusting the track gain up or down causes about 3 or 4 fast clicks then the clicks stop for ever again. I set max fir up to 20000 last night (the split took about 3-4 minutes in that case, time of that scale is not an issue for me, the quality of the audio is though) and everything sounded great. I don't remeber if the 3-4 fast clicks occured though...

The next problem I'm running into is tuning Cool Edit Pro up to be able to allow 8 mastering compressors working at the same time without choking but thats another search & thread...

kylen

Craig Jackman · Location: Canada Posts: 909

MusicConductor · Location: USA Posts: 1524

Craig, it surprises me that such an astute user as yourself woulud be using Edit View to run dynamics. Add dynamics processing as a real-time effect in the multitrack window and aren't we closer to what those plugins do? Am I missing something?

"BUG" CONFIRMED

I'm sure it's no great surprise that I was able to duplicate the clicks. And they occur at perfectly regular intervals just as you'd expect from David's (2Bdecided) wonderful explanation (thanks for that, BTW). They're also visible in Edit View/Spectral.

The workaround, at least for my White Noise sample, was to raise the FIR to 1280. Less than that, and there were still clicks. It may take a bit longer to process, but the more accurate filter is a nice fringe benefit to make up for it. And what a great "trick" it is to use the FBS.

And Kylen, your paranoia is confirmed: compressing a "clicky" file will indeed likely result in the clicks not cancelling out when the mixdown is done. But you knew that.

LEE TYLER · Location: USA Posts: 47

kylen · Posts: 290

Hi Lee,

I'm still looking (listening mostly) to this to see if I have a problem. While I'm waiting for my 'chainer' registration to appear I can give you and any curiousity seekers an idea of what this crazy guy thinks he needs an 8 band split for.

The story begins:
As I progressed into the netherworld of home studios and pre-mastering I began to have a dream. This may have been brought on by reading about Avalon outboard gear and $15,000 tube gear and non-linearity and saturation and stuff. But just the same I thought what I really need is an EQ that shows me a spectrum that allows me to tweak the levels of bands while at the same time adjust compression, limiter, and saturation settings on top of that (or under it) for each band. Well, I've just described Ozone2 haven't I ?

Now what if in the event of trying to master, restore or further unwind some horrible mixes or mixes that were simply made too far in the past to have any decent sound to them I needed more power. What if I needed more bands ? That's easy use CEP and put 2 Ozones in parallel (if your setup will run them power wise - I don't know), or use some other DX effects rack if there are any.

But now what if some of the bands needed a certain kind of compression, while other bands needed a certain kind of saturation, and other bands needed a special limiter ? Well that's what the frequency band splitter will allow me to try, so that's why I nearly jumped thru the roof a few days ago.

Now that I have my 8 bands of wave files I just now plugged them in to Sonar where it has vu meters on each track. Now when I play the song each of the 8 tracks looks exactly like a band on an outboard 8 band EQ'd spectrum analyzer - that's what my dream had in it! Track 1 controls the subbass, I can move the slider up and down and see the vu bar bounce up and down according to the dynamics of the music. I can also insert any flavor of compressor or saturator on there that I want.

Anyone who knows the difference in sound of PSP, Timeworks, Ozone, Voxengo, and T-RackS may have at one time or another thought - Oh yes that band needs some saturation and it sounds fine on the next one but this Ozone saturation is a little too rough here - why can't I put a PSP saturator on there, or digitalfishphones ?

Well, in a few minutes I'm goin to. As soon as my chainer registration gets here...strap me in, this could be it :P

Oh yes and adjusting and controlling this big mess, I think Craig mentioned it, in Sonar (maybe CEP too) I can line up all the EQs and compressors and overthrusters (lampthruster voxengo.com) in the same window and tweak away so that the affect of an effect is easily discernable in real time.

OK, I'm goin in now Cool

kylen

MusicConductor · Location: USA Posts: 1524

Lee, you summed it up well. Just run that FIR number up and have a party. Which Kylen is already doing! And I'm not sure it's fair to call the clicking a bug, based on David's explanation, although it might be good if low FIR values weren't allowed if a split point is below a certain frequency.

Have fun, guys.

Craig Jackman · Location: Canada Posts: 909

Craig Jackman · Location: Canada Posts: 909

kylen · Posts: 290

Wowie, I just returned from the inner rings of Saturn on the wierdest auditory trip I've had yet.:P

I'm the guys who was going to split 8 different bands out and add various dynamics processors, compressors, limiters, expanders to push each band into a threshold, push it down with an expander of any make, etc. Well I have to say it's a fun learning exercise, everythink is very clean - no pops, clicks, or crackle from any of the bands I split out at 20000 fir each (a little overkill ?) no matter how I squish, pump, bloat, twist the waves together...

But for a more useful activitiy I think I'll go back to my Ozone approach which is. Listen to the wave, add compression only when necessary, and try to add the compression only to the band that needs it at a bandwidth that sounds good for that frequency range or dynamic event I'm trying to adjust. In other words all the changes have to be in context like Craig and MusicConductor said earlier.

So today I'll only make 3 splits, apply Vintage Warmer to the Bass 0-120 to smooth the lower bass guitar but bring out the kick thump, Ozone2 to the mids and some highs 120-8K where I want super transparancy , Vintage Warmer to the treble 8K-22.5K to make the drum machine hi-hat and cymbols sound like fine silk.

Since those guys are multiband compressors themselves I can further subdivide to repair a vocal that sticks out too much in 2 different zones at 400 and 600 Hz or there abouts. Cool

I like it !

Now another thing I'll be listening for are any artifacts produced when 2 multiband compressors by different vendors are sitting next to each other.

Note: For this experiment I'm doing the frequency split in CEP and the mixing and effects inserts over in Sonar because I have a VST CurveEQ plug inserted also for some saturation and stuff.

I've talked to support about this and when they come up with anything new I'll ask them to tell you guys too. I think the settings question is still open from a user perspective.

kylen

VoodooRadio · Location: USA Posts: 3971

jonrose · Location: USA Posts: 2901

Hi kylen,

I may have missed this somewhere, and if so, my apologies... but have you actually tried stacking multiple band-limited compressors in an FX rack yet? Just curious.

Best... -Jon

kylen · Posts: 290

jonrose · Location: USA Posts: 2901

Hi kylen,

Actually, the idea I was trying to get at, here, was using BAND-LIMITED compressors, instead of using the Frequency Band Splitter function and then compressing the resultant (separated) band-limited tracks (as you've already done). Points taken on your above discussion, though.

Best... -Jon

kylen · Posts: 290

OK, I have it now. The name of the dream I've been having that I mentioned earlier in the thread:

Cal · Location: USA Posts: 577

Hey Kylen.... since I particularly like the mastering process your references to Ozone, CurveEQ, etc. caught my attention. This, and Jonrose's comments, seem to be somewhat in contrast to one another... Unless I'm misunderstanding his intent, he, in referring to the tools already in CEP, appears to be steering you to a simpler way of accomplishing some of the things you hear in your brain.

It can be awfully intriguing to get down to the multi-levels of frequencies and compression and EQing and saturation and all, power-hungry artists that we are... But i find that after using Ozone and learning its behavior, AND, not to mention being very grateful for its optimal EQ overlay graph (great learning tool), well.... I really do find that CEP can do those things, with some knowledgeable and effective use of (Jonrose's ref) its own band-limited compression (may need to do several passes), Parametric EQ, effects, a smattering of plugins, and Analyzer. You'll find that if you get a really great mix mastered just right, the frequency slope in the CEP Analyzer will look just like the one in Ozone.

Yes, these other programs have the advantage of layering and chaining those processes in one pass. And i haven't tried CurveEQ but did go to the site for a read... One thing about Ozone, and comments in the past about this excellent package seem to corroborate: it's easy to overdo it, making the processing noticeable. It's easier for me to get those results much more transparently in CEP using its tools.

I guess I'm not yet a believer that h-u-g-e amounts of micro-multilevel-processing gains enough of an advantage that will make another listener's ears melt and eyes bug out in sheer joy for the dazzling difference it might make.

Audio can be a tweakers paradise/playground.

Well, I apologize for the lecture.... I might be a bit simplistic. And thanks for your contributions to the forum. I've enjoyed your enthusiasm and the creative thinking you get into. You're the kind of guy that, if you were working on cars, you'd try to put a chrome drive shaft in some vintage car, and keep it polished up even though no one could see it... and smile, because you know how cool it looks.

jonrose · Location: USA Posts: 2901

Hello again,

Well, between this and the other, related thread, I was only trying to point out that there is an alternative to the Frequency Band Splitter - Stacking band-limited compressors in an FX rack.

Since this forum is frequented by lots of folks who don't necessarily use (or wish to purchase) the kinds of plugins referred to here, I just thought it would be appropriate to point out that the multi-band compression component, at least, could be achieved in CEP2's Multitrack View, alone - in a couple of different ways, and without the necessity of doing multiple passes.

Happy experimenting! :)

Best... -Jon