dual band splitter

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

erikb1971

Well-known member
Joined
Jun 13, 2010
Messages
398
Location
Holland
Hi all

Is there a relatively easy way to split a signal in two eq bands (preferably choosing that split frequency with a pot) so the highs and lows of that signal can be processed seperatlely? And after that procession, could those two signals be summed together with a standard summing circuit?

Cheers

Erik
 
While any amount of processing will interfere with ability to recombine, I would consider a derived xover
http://sound.westhost.com/articles/derived-xovers.htm
since these in principle will recombine to unity if you don't mess with the intermediate products too much.

It is worth note that when dealing with higher than one pole transition, you will get an asymmetrical slope (i.e. 2 pole in one direction but only one pole in the other). You can flip around which bandpass gets the steep and which gets the shallow roll-off as serves your needs, by subtracting either a HPF or LPF from unity to derive the other passband.

Have fun...

JR
 
JohnRoberts said:
While any amount of processing will interfere with ability to recombine, I would consider a derived xover
http://sound.westhost.com/articles/derived-xovers.htm
since these in principle will recombine to unity if you don't mess with the intermediate products too much.

It is worth note that when dealing with higher than one pole transition, you will get an asymmetrical slope (i.e. 2 pole in one direction but only one pole in the other). You can flip around which bandpass gets the steep and which gets the shallow roll-off as serves your needs, by subtracting either a HPF or LPF from unity to derive the other passband.

Have fun...

JR

I had read the article and I was sure it said NOT to use a derived crossover...
 
erikb1971 said:
I had read the article and I was sure it said NOT to use a derived crossover...
True for loudspeaker crossovers; but for split-band processing, it has a definite advantage. Now it depends on what you expect in terms of "naturalness".
Higher order xovers such as LR24 offer better band separation at the cost of some impredictability. For a two-band split, it is acceptable, but if you wanted more than that, the mid bands tend to disappear.
 
erikb1971 said:
I had read the article and I was sure it said NOT to use a derived crossover...

Are you making a loudspeaker??? If yes I agree the derived has limited utility for several reasons. However if you are doing some random processing, then trying to put humpty dumpty back together, the subtractive approach has merit (IMO).

Loudspeaker crossovers are like a 3 dimensional problem as we try to acoustically combine the output from two physical drivers... I am guessing that you dealing with more of a 2 dimensional problem. Where you are electrically recombining your two processed stems.

The subtractive approach IMO is more promising than conventional loudspeaker crossovers for possibly preserving some signal integrity in the recombined result. 

But what would I know... ?

JR
 
And if I understand you guys correct, I would need a 4th order subtractable crossover network as in attached schematic?

Where does the adjustable pot for the freq come in?
 

Attachments

  • 4th order subtr crossover network.png
    4th order subtr crossover network.png
    64.1 KB · Views: 53
erikb1971 said:
And if I understand you guys correct, I would need a 4th order subtractable crossover network as in attached schematic?
You would still have one of the slopes at 6dB/octave (the lowpass in that case).
Where does the adjustable pot for the freq come in?
You would need to replace R1, 2 3 &4 with pots. Quite difficult because 4-gang reverse-log pots are not particularly common, but even more with different values (2x10k + 2x20k here).
 
abbey road d enfer said:
True for loudspeaker crossovers; but for split-band processing, it has a definite advantage. Now it depends on what you expect in terms of "naturalness".
Higher order xovers such as LR24 offer better band separation at the cost of some impredictability. For a two-band split, it is acceptable, but if you wanted more than that, the mid bands tend to disappear.

I should have asked first what you meant with LR24.... because I just thought: hey, I should use a Linkwitz-Riley crossover! And then.. aaahhh... that is what he meant with LR!. Thank you. Would attached schematic work easier with a pot for adjustable crossover frequency?
 

Attachments

  • 4th order LR filter.png
    4th order LR filter.png
    62.2 KB · Views: 33
We are getting way past relatively easy...

Any multipole filters will require one gang of pot per pole, unless you use voltage controlled filters. (not easy either).

How much slope do you need ?

A one or two pole derived is "relatively easy".

JR
 
Surely this is something better done in the digital domain.?  You did not mention the setup you're using, maybe you're completely analog.
 
note my new signature :)

How much slope do I need... good question. What I want to do is split a signal (my guess is quite often a drumhit or drumloop) into a low and high freq signal, and process them with a different overdrive / distortion. Part of my distortion / mangling unit GOD (Golden Over Drive) based on the rocktron austin gold overdrive. Attaches a first scetch of part of the front panel... maybe a bit clearer...

So how much slope woulld you need for that... I would say more is better, but on the other hand... it is distortion / drum processing...

Would a limited number of crossover frequency options make things easier?
 

Attachments

  • 2ru-austin-gold.jpg
    2ru-austin-gold.jpg
    27.7 KB · Views: 31
What I of course also could do is split the signal and put a steep high pass filter on one channel and a steep low pass on the other... or would that cause horribel problems when putting the signals back together?
 
One thing I would suggest is to do some experiments and try to figure out what sounds good.

Get a good sounding effect with fixed filters on the bench first, then figure out how to make it adjustable, or if you can cover it with few switch positions.

Since this is an effect,,, disregard all my comments about signal integrity...

JR
 
Yeah that is a good idea... I have a tendency of outrunning myself sometimes. Iam gonna set up a test environment with Cubase and a small analogue mixer and the distortion effects to see what I like. Will report on progress!

Thanks all so far

Erik
 
Ok one morning and half afternoon of testing later... Quick conclusions:

1. dual band overdrive on drums is great
2. dual band overdrive on drums needs to be done parallel, so mixing the clean signal with the two distorted bands
3. the split frequency for overdrive on drums is somewhere between 350hz and 500hz
4. the Austin Gold is perfect for overdriving the lower freq band
5. all distortions I tried for the high band clutter the mid highs /highs
6. drumloops need a fairly clean overdrive on high band, individual kick and snare can do with some more dirt
7. overdrives and distortions act as compressors, the only nice added compressor sound for me was a parallel compressor on setting "are you out of your mind?".
8. blending all those clean and processed bands does funny shit to your audio

Of course, all these conclusion are my personal ones, and me being:
1. almost tonedeff
2. convinced that the first Stone Roses album is great but sounds terrible
3. out of ADD medication
I would advice all of you  not to start building commercial products based on these findings. Unless you got shitloads of cash to burn of course.

For me however, these experiments have proven that Iam right on one or two subjects, and have to rethink my concept completely on all others. So that is a good thing.

Iam pretty sure about:
1. parallel dual band overdrive
2. Austin Gold on lower freq

Serious thought needs to go into:
1. the overdrive for the higher band. It needs something with a wide range of overdrive / distortion and a "open" sound on the (mid) highs. 
2. filtering of the processed bands. The parallel overdrive makes it a pretty precise tool for colouring loops. I have the idea though that is could be even more precise if I was able to add smaller bandwidths to the original.
3. the crossover filter. Especially when blending in both bands again with the original, all sorts of audio things happen. Most of them not particularly usefull.
4. EQ. Espcially with loops, I found that I quickly went for eq-ing the clean signal (pre compression) or the end signal (post compression) for more sound sculpting. Lost of fun and great results. I havbe a pair of boards and front-plates for bluzzi's EZ1084 EQ on the shelf. It could be interesting to use these on insert points, preferably switchable between clean (just the original) pre compressor and post compressor.

First things first though: The crossover filter and the high band overdrive.

In the mean time, I would like to hear your comments on everything, but especially on the filtering of the two distortion bands and on the eq (placement and is the EZ1084 the right eq for this job?)

Cheers

Erik

 

Attachments

  • Gold concept after test.jpg
    Gold concept after test.jpg
    46.9 KB · Views: 21
Maybe you don't need to distort the high pass.

More isn't always more...

I know a little about drums and your fundamental energy is all pretty low frequency. The higher frequency content is more diffuse so will not respond the same to a distortion efx as the lower band.

JR
 
JohnRoberts said:
Maybe you don't need to distort the high pass.

More isn't always more...

JR

You might be right... although I really love the little overdrive on the mid freq.... studying on that...
 
About the filter...

Let's say I want to build a 4th order LR filter. From Wikipedia:
Fourth-order Linkwitz–Riley crossovers (LR4) are probably today's most commonly used type of audio crossover. They are constructed by cascading two 2nd-order Butterworth filters. Their steepness is 24 dB/octave (80 dB/decade). The phase difference amounts to 360°, i.e. the two drives appear in phase, albeit with a full period time delay for the low-pass section.


So first a 2nd order butterscotch filter. At http://www.daycounter.com/Filters/Sallen-Key-LP-Calculator.phtml I choose:
450 hz as the crossover frequency
resistor: 10k (why? now I just took same value as article in previous post)
2 poles (pole equals order right?)
type: butterworth
pass band riple 0,1 (irrelevant for this kind of filter)

I press calculate and:

C1: 5.0043902601185E-8 F
C2: 2.5021471379566E-8 F

What's with the E in there?
 
E is an exponent , -8 means its a small number  1/10^8

uF is already E^-6  so E^-8 is uF x 1/100, or .00x uF

======
Another thought about drums and distortion EFX, typical distortion EFX add harmonic distortion components. Drums naturally exhibit resonances that are not harmonically related (nominal for two head drum with both heads tuned similarly is 1.7x between fundamental and first resonance. 

So,, narrowly band-passing the fundamental and distorting that will add a bunch of harmonic content to an otherwise not very harmonic instrument. Distorting the upper resonances will add harmonic content to them that is unrelated to the lower fundamental harmonics.  I speculate distorting the HP will add chaos, distorting the lower band pass might actually add musicality or more of a tonal voice.

or not....

JR

ps:  or 10x nF since nF= E^-9
 
Back
Top