DSP Robotics Support

Understood Martin. Always a welcome when you do have time.

As to first part of last message ... I'll re-state my, .... 'Uhh Huh, ...... ok'
Way over my head

I think I'm needing to try to simplify the concept/technique in order to have any chance with this. That being ...

I've got a 10-band splitter that feed individual PEAK-HOLDs. I'm then using each band w/Peak-magnitude to drive a associated VCA's.

This both measures the overall response and locks those band levels as a static snapshot.

I don't see an author name for this 10-band splitter [I'll track that down - got it here]. But I'm thinking I should bump this to 20 bands. Easier said then done for mu-waa.

I don't know how accurate or true this splitter is. I've not been able to analyze the shape of the band, the possible overlap or gap issue. But it does test the general concept.

The interest in Martin's 'SpectralEnvelope' example is the high-speed SCANNING through the audio source. That would be a welcomed feature. The 40-band aspect might just be overkill.

Bottom line .... this 'Matched EQ' will NOT be used to actually alter the audio source. I know this can sound confusing. I only want to get this snapshot to provide a good representation of the sonic shape.

The calibration PINK NOISE will be set to pass through this static filter representation in a hoped way that will bias the PINK-NOISE to appears as if it's the actual audio.

If we step back, it will be clear that we are using things like noise, and EQ, in a non-traditional manner.

I want to substitute the actual real music audio, whatever its' sonic envelope/energy, with a shaped PINK-NOISE ... like that of the real, actual music audio.

The DAW is stopped, and we can then activate the PINK and the snapshot into the following effects chain.

The LEVELS going into the chain would be just like playing the real music track ... but we are not

After adjusting the FX Chain to UNITY IN/OUT for each device ...

The PINK and the shaped Filter are By-Pass ... back to the music.

As fx changes are made ... rinse & repeat. Using familiar tools in a non-conventional way.



thanks Guys !

I think I finally found a way to clone the spectral envelope of an entire track onto another source - white noise in the example, however it is possible to "whiten" an arbitrary track by dividing through its own spectral envelope. Both the analysis and the filtering is FFT based, with an option to smear the (unwanted) high frequency resolution to obtain constant Q. The Q is freely adjustable as a post process once the spectral envelope is captured.

The filter is an FFT linear phase FIR which is a desirable property for mastering (so I am told). I haven't implemented it as a zero latency partitioned convolution here, although it could be done for real time scenarios.

off to the Lab for testing ...

Martin ... this is really something, but I do need to ask/check on one thing ...

I'm looking inside the 'FFT FIR' module.

In the 'FastUnpack (unmasked)' module, the L/R output links ... are these in the correct position [1 & 4] ?

I would have thought connection 1 & 3 ... but not clear why.

Thank-you for any clarification !

It's a bit tricky.. I'll upload a stereo version later.

I'm glad I asked first.

I've been doing several testing with very positive results. Very solid.

It takes [maybe 10 secs] to scan an audio track and post a 4096 data point output/display.

I've added some extra features as I've scan 7 tracks to be used as separate presets. Easily can switch between the spectral presets. very solid.

side note: It is quite interesting to listen to the white noise filtered through these spectral envelopes. You can actually sense the balance, energy, attitude, and the overall sound-stage of the original track.

This is exactly what I had hope to have ... a single snapshot that holds the original audios energy AND presentation.
This looks to be a specially 'focused', calibration tool. Granted, this is but one piece of the overall strategy/goal ... but it's a critical starting point for the process.

It's almost like we now have the laser-beam coherently focused, and now the test/experiments can begin.

Martin ... this is very special what you have designed. There are a few 'internal' settings that are hidden inside that I'm going to be looking into, but I may need to ask for additional information/explanation to realize their impact.

Meantime, I look to test this concept in real-world.

Thank-you ... Thank-you

Another Patent Application for MV?

There are no words! Cool!

Thanks for the kudos!
Here you go stereo cloner.

Thank you!