SL8 Quality: Freq Split Tracks DO NOT NULL; nowhere near it

As per subject, I have freq split a track to 2 tracks with a band split at 100hz

  1. I set a freq selection at 100hz and saved the selection
  2. I copied the selection to a new layer and put that on tack A
  3. I recalled the selection and inverted frequencies (with a 10hz face)
  4. Copied to a new layer and put that on track B
  5. The original track was phase inverted
  6. Original Track, Track A & B were all sent to a summing buss

The output is nowhere near a null. no other processing was applied, a simple crossover in essence and it fails pretty badly

This raises serious concerns about frequency recombinationā€¦and this is without any processing being applied

Am I doing something wrong? Is there a transition beside freq fade that can razor cut and produce true null?

Unless Im missing something hereā€¦its a showstopper for me Im afraidā€¦as far as using it for this type of application; linear phase filters might be the real deal except the ones in Cubase are far from linkwitz/linear phase

Here is what happens graphically when I cut/invert/cut. The line of residue is clearly visibleā€¦how is this resolved?

Really? thats your problem?

Crossovers are no linear processes, they introduce distortionā€¦

linkwitz is not linear phaseā€¦ very far from linear in phase and frequency

Linear phase filters will introduce latency, low frequencies introduce very high latencyā€¦ at 100Hz I would estimate 40ms at least

@Aurasphere Trying to invert a selection with fading wonā€™t really get you what youā€™re looking for.
But hereā€™s the procedure for what you want to do:

  1. Duplicate original to layer A and layer B
  2. Cut some frequencies from layer A, with fading if you want
  3. Duplicate layer A to layer A (duplicate), phase invert it (next to the volume control) and put it below layer B
  4. Merge layer A (duplicate) into layer B using Layer > Merge Up.

Youā€™ll get layer A and layer B, a perfect decomposition of the Original layer into 2 layers.

1 Like

Im not treating it as realtimeā€¦Im copy/pasting the results of a static process; in the context of using an ara process ie dragging back into the timeline, latency is not relevant.

Thats my problem? Yep, need to know I can trust processesā€¦I want to know how to do things best practice in an abstract so I understand the process and can execute it a component level for propogation into composite processā€¦this is called learning.

Specifically Im trying to get the equivalent of using eg a linkwitz/linear phase eg
I use to just apply plugin process to split bands in cubase 5 ie using QRange which has linear phase/linkwitz. This produces a perfect nullā€¦and my question relates to this simple example but as a newb with much siltation, Im having trouble achieving the analog of thisā€¦

Even without fade, it doesnt seem to work

Thanks for your exampleā€¦thats exactly what I needed to know ie decomposition and reassembly is a trusted outcome, fidelity wise. Ill try it out.

Now to try and dissassemble a composite track eg cajon to extract not only a kick/snare freq wise but hopefully ā€˜choose similiarā€™ will be the solution

Thanks Robin