Research ... capture EQ curve

[Spogg]

For what it's worth (not a lot probably) the impression I have of the requirement is similar in some ways to a vocoder in function.

The envelope followers connected to each of the bank filters would need to have an average-hold behaviour for the duration of the track or representative track section played. This held or frozen bank of followers would then control an identical bank of filters to process the incoming new track.

A normal vocoder works dynamically in real time so what we'd need is a vocoder with a "memory"; the stored average from the reference source.

This could be set to process the new track at a low amplitude then the resulting recording could simply be normalised offline in any wave editor.

How far off the mark am I?

Am I seeing the birth of another community product?

Cheers

Spogg

[tulamide]

Just to make sure, RJ, are you talking of something like this and this ?

[RJHollins]

Hi tulamide,

Yes, Match EQ.

I've been using FabFilter-Q2 'match' feature for my experiments ... but not in the tradition way.
I had the Voxengo on my old 32-bit computer ... just haven't up'd to the latest 64-bit. [probably won't].

But yeah ... 'match eq' is the named used. I don't know what the internal [programming] technique is called.

[Spogg]

So I kind of think I might be on the right track (!)

Martin already has the filter core worked out. I used it in my Quilcom Bluesky vocoder and I was very satisfied with the filter bank accuracy etc.

What would be needed is to create the new averaging envelope followers with a hold-on-demand function... Martin?

Cheers

Spogg

[RJHollins]

I've been experimenting with the 10 band splitter that was posted, and looking at some of the 'band-pass' filters.

Cobbling ideas from various areas just to test out ideas. Not knowing a way to do this, can only experiment.

I'm taking level measurements off the multi-splits. Can probably feed a MAX SAMPLE HOLD to trigger a higher value. This idea seems destine to fail as I need to consider the gains in both the +/- range ref to 0. Maybe a re-scale could correct that.

With the captured gain outputs from each band, these values could feed a band-pass filter that has a wide enough Q to cover the zone, with minimal overlap or gap. Maybe a Reinkist filter a better choice ?!?

So this initial experiment would attempt to set 10 eq bands that would represent the basic coarse structure of the overall envelope.

This seems such a 'first-grade' approach

[RJHollins]

Continuing with my initial idea [right or wrong]

I now have an audio source feeding a 10 band splitter [martin i believe].

Then into a PEAK HOLD routine [RUBY], followed by an M2F primitive.

For viewing, I take M2F into 10 SLIDERS.

The sliders respond to the level of each band. Again, some of this is just for visual and is not really important or needed [but it is fun to see them respond].

Currently, I'm collapsing everything to MONO only. This may have to change. So I'm having to re-scale the M2F output with a .73 constant for the VCA. This keeps levels just below 0dB.

The next step is to use these slider values to feed a series of FILTERs [I'm first thinking BANDPASS ... but that is just a guess].

From there I need to look at re-scaling these filters around a new reference point. Right now, everything is positive gain. I'm thinking that I need to re-scale to a proper +/- scale. There probably is a known logic to do all this, but with this total experimentation, I hope to figure/learn what this might be.

To anyone thinking about FLOWSTONE. It is exactly projects like I'm attempting that make FLOWSTONE such a fantastic environment to develop an idea into a working model, regardless of the elementary approach with minimal knowledge of the subject or techniques. It's like an experimenter's 'breadboard'.

As I hope to learn more, I may thow this initial concept completely out, and look to pursue a more elegant or sophisticated design. FLOWSTONE is THE application for me.

I'm am still very open to any comments, ideas from the forum GANG to slap some sense into me

[RJHollins]

update:

Well, continuing the experiment.

Instead of taking the output of the faders into a series of EQ's ... a little twist ...

Since I have the source audio already split into 10 bands, I now have hooked in 10 VCA controls [one for each split band], and use the analysis GAIN as the level for the VCA.

Freezing this GAIN value, I think I have something much like an equalizer. There are no adjustable frequency or Q, but the tonal changes are there. In fact, since I have 'sliders' still in the circuit, I'm able to change the spectral envelope from any of these bands.

With them locked in, and a different audio source, the audio is quite affected.

I'm experimenting ... so some of these observation and test, are just that.

I did add a BYPASS and FREEZE/HOLD option to switch from analysis mode to modified playback mode.

Next step:

Re-scaling these 10 captured GAIN values. I need to have the bands centered around a middle 0dB line [ref], with the bands using both +/- vertical [not just +].

Anybody have a technique to re-scale 10 floating ARRAYs that also maintain the relative relationship to each other Off the top of my head ... I don't

[RJHollins]

progress.

I've 're-scaled' the slider gain level. Using the 'Float Re-scale' module, taking original range 0-1 and transposing to -1, 1.

At this point, we have both visual and values that reflect the source spectral waveform. The sliders position much the same as in the GRAPH display. I don't know if 10-bands are enough for reliable accuracy ... to be tested.

An idea I'm looking at is a type of high-speed rip of the audio waveform [like Martin used in his excellent Spectral-Envelop example]. Thinking further out, I could possible generate a TXT file of the [10-band] output, and load that into a mastering project. This writing could maybe done with a 'batch' process in mind ... just an idea.

2. Right now I'm using a collapsed to MONO source. It would be trivial to duplicate for discreet stereo ... but I'd like to look at combining the 2 audio signals to better optimize. Not sure if that would have any benefits.

I've read combining 4 audio channels .... is there any benefit if only combining 2 channels ?

Thanks for any insights !

[RJHollins]

question please ...

IMPULSE primitive - Generates an impulse signal. This has a value of one as the first sample and zero for all others. Use this to test the frequency response of a filter.

OK, I understand this ... and then I don't.

So ... can anyone expand on how this primitive can be used ?

It's specifically mentioned to test 'filters' ... that is of interest.

What is the bandwidth response [output] ? What output level?

Is it synchronized ? Is it a single pulse? can they be 'stacked' ?

OK ... bunch of questions ... but the other is, would IMPULSE replace PINK or WHITE noise as a test signal ?

any examples ... always, Thank-you.

[Spogg]

Below is a schematic that Martin made for me on which I based my Bluesky vocoder.
You'll see the impulse prim in use.
For a vocoder 16 filters is sufficient but I would say for mastering and eq matching it may be best to go for 24, i.e. 3 of these banks.
The pro eq matching apps that I've looked at create an initial curve but it's then up to the user to fine tune frequencies, q factor and amplitudes to get the best match to the reference track.

Cheers

Spogg