PERCEPTIVE / ORGANIC PROCESSING
Observations :
Group 1 - Digital vs Analog - High Fi vs Low Fi - Grain
Group 2 - Linked Evolution Parameters
Group 3 - Human Hearing
PRINCIPLE & DEFINITION
We'll call "perceptive" or "organic" processing a way to process the audio signal which would be, presumably, more suited to the human ear than standard processing.
A complete "perceptive processing unit" would include dynamic processing, EQs, and reverberation.
( Note : from a certain point of view, reverberation and EQs are two different aspects of the same phenomenon - both can be mathematically expressed using convolution, a reverb is a sum of EQed delays, and the spectral representation of a delay is a comb filter - but this is not taken into account here, since it's not really true from a psychoacoustic point of view).
The notion of perceptive processing is based on three groups of observation.
Group 1 deals with pros and cons of existing signal processing techniques.
Group 2 give examples of "natural", "normal" trimbre evolutions - as opposed to studio / laboratory timbre evolutions.
Group 3 studies the ear behaviour.
OBSERVATIONS : GROUP 1
DIGITAL VS ANALOG, HIGH VS LOW FIDELITY, GRAIN
1. digital audio is often considered as sounding "cold"
=> this is probably a matter of grain - analog audio bring microscopic irregularities that digital audio usually doesn't.
2. digital EQs, especially inside a same host software, sound almost the same
=> grain, brought by the host audio engine, is even likely to override the EQ's intrisical characteristics
3. the best EQs (like the Manley Massive Passive or the GML8200) are analog
=> in processing, situations that are not entirely deterministic can bring better results
4. if the ear can be considered as an ADC, the analog pre processing before the actual conversion is extremely complex
=> this shows that highly non linear signal processing is not necessarily a problem, on the contrary.
5. putting an ADC in a mike (thus minimizing the signal's analog path) gives poor results
=> reducing the unpredictable to zero is a bad idea
6. electrodynamic speakers give more comfortable results than electrostatic speakers, though they are less accurate
=> the main difference between electrodynamic and electrostatic speakers in terms of behaviour is probably the dynamic aspect :
transients are more accurately rendered on electrostatic speakers, which have a very low integration time
this can bring the hypothesis that an audio system may be more suited to the ear if its integration time is close to the ear's integration time (~50ms)
7. non linear processing (like processing the signal with a tape) often give comfortable, "signal balancing" results
=> it's better if processing includes a physical aspect - purely mathematical signal processing (like digital) may bring unbalanced results
8. complex reverberations (like outdoors natural reverberations) are often very nice - meaning a lot of details, a "liquid" aspect etc.
=> another hint that detail, complexity, grain, is important.
OBSERVATIONS : GROUP 2
LINKED EVOLUTION PARAMETERS
Every "traditional" or "natural" or "realistic" timbre evolution is a result of the synchronised evolution of different parameters, parameters that would be considered as having nothing to do with each other in a studio environment.
Example 1
Moving away from a sound source is :
- level goes down
- high freqs go down, low freqs go down, medium freqs get more prominent
- direct sound / diffuse sound ratio gets down
- transients get less accurate, overall dynamics is reduced
Example 2
Doppler effect, same phenomenon as example 1 +
- pitch gets lower when the object passes
Example 3
A violin player plays "louder"
- levels gets higher
- vibrato gets faster
- harmonic / noise ratio gets higher
- level of the harmonics / level of fundamental gets higher
Example 4
An electric guitar player (+ amp with overdrive) does vibrato
- vibrato gets indeed more important
- the level of a particular harmonic gets drastically higher (fake feedback effect)
- the overall level thus gets higher
- sustain gets more important
OBSERVATIONS : GROUP 3
HUMAN HEARING
The combination of the human ear and the brain is known to exhibit extraordinarily good performances, like source discrimination, pitch and chord detection etc.
In parallel, it can be noticed that the ear as a transducer is a very bad system in terms of audio fidelity.That brings an hypothesis : audio fidelity, in certain situations, is a mistake, it's the engineering of the ear which is right - and organic processing would be well inspired to look for engineering tips there. This is linked to group 1, observations 6 and 7.
Let's have a closer look at certain aspects of the human ear.
Two main characteristics :
1. highly non linear processing
2. inter dependancy of all parameters (level, frequency, position, transitional behaviours)
-- More details about "highly non linear processing"
1. extensive use of complex comb filters, due to the morphology of the external ear
2. eardrum muscles act as a built in compressor
3. the signal gets through complex, resonating systems in the middle ear, thus bringing more spectrum distortion
4. those resonating, physical systems take time to react to stimulation, as all physical systems do : thus the existence of an integration time (~50ms)
-- More details about "inter dependancy of all parameters"
1. the perceived spectrum of a sound source depends of its position (due to HRTFs)
2. there is no such thing as a static spectrum, since we always make micro movements with the head, thus constantly modifying the HRTFs.
3. perceptive level is frequency dependant (see perceptive level vs physical level)
4. integration time, or ear latency (50ms) varies with the transient level, and probably with the transient's spectrum
One can notice that, in each detail that we highlighted, there are aspects of observations from group 1 and 2.
SUMMARY OF THOSE OBSERVATIONS
Group 1
Grain / detail complexity, physical (realistic) behaviour, non linear aspects --------- are important aspects.Group 2
Simultaneous evolution of radically different parameters is a good idea .Group 3
Looking at the human's ear behaviour, we find the same aspects : grain / detail / non linear + simultaneous parameter evolutions
CREATING A PERCEPTIVE / ORGANIC PROCESS
This is now straightforward : to experience organic processing, just play two notes on the violin or walk away from a sound source.
To experience it the studio, just make successful emulation of "walking away" or "making a violin note louder".Those would be realistic processings. Other examples can easily be found.
Realism in those process are a garantee that they will be coherent.Now the interesting points :
1. to create non realistic organic yet coherent evolutions
2. to adapt plain processing (EQ, dynamics, reverbs) towards something more organic.
Point 1
This is really difficult : the very notion of it seems out of the question by all standards, at least for the next 10 years.
A non realistic organic machine would be considered as a real UFO like : what's the point ?
However, several aspects of what it could resemble is shown here.
Point 2
EQ : The easy part is the "perceptiv-isation" or "organic-isation" of the EQ - that is shown in this chapter.
Dyn : Dynamics are, by essence, quite organic, since they are, of all three classical processes, the one that is closest to the ear. This is shown here.
Rev : Good studio reverbs are often already a bit organic : the "distance" parameter in the TCElectronics VSS3 algorithms is obviously organic.