Transient detector...

[Blue Cat Audio]

You can try our digital peak meter(free) or its pro version for your enveloppe following task. It'll enable you to have more control on the envelope settings.

[JonnySun 2.0]

The stereomatrix can also be implemented as:
Left=mono+stereo; // left=L+R+L-R => left = 2L;
right=mono-stereo; // right=L+R-L+R= right = 2R;

or:
Left=right+stereo; // left=R+L-R= Left= L;
right=left-stereo; // right=L-L+R= left = R;

left=left+stereo;
right=right-stereo;
is incorrect..

grt.


Evert
gtek

If it sounds much better than orban's thingie ..

It's a shame i never heard processed audio files from the 222a. I haven't found what i'm looking for till yet. If it sounds good you'll make the ultimate new imager for the vst-generation

[JonnySun 2.0]

Here are two links of bob katz' k-system:

http://www.proaudioreview.com/par/augus ... _dd2.shtml

http://www.digido.com/portal/pmodule_id ... age_id=48/

[citizenchunk]

this is a very interesting discussion. not having worked in radio myself, i had only heard of Orban products, but no actual experience with them. now i'm Googling my ass off.

gtek is right: this would be a very simple plugin. basically, this is just an upward compressor with M/S processing; but it only processes the difference (L - R) information. that's cake, man.

i already have a stockpile of envelope/dynamics classes that i've written for other projects (i love reusability!), so i might throw something together sometime.

== chunk

[citizenchunk]

... basically, this is just an upward compressor with M/S processing...

correction: it's a upward expander, with differential envelope transient tracking and M/S processing. still relatively simple to create, once you have a grip on the concepts.

== chunk

[gtek]

Well,

You could subscribe the orban 222a effect as a very glossy sound, without overexaggerated sideeffect. Because the function of the 222a is mainly to preserve the stereofield (l-r) loudness compared to the mono channel.Normally you will lose "stereo" loudness due to transients like bassdrums or snaredrums.
With linear stereo expansion ( non-mono envellope triggered ) you would hear a clear massive increment of reverberation effects placed in the stereofield. This sideeffect is very very unpleasant.
The 222A does not have this, due to the L+R envellope follower.

I made a piece of code on my analog devices dsp board to simulate the 222a. The results are very good.

>If it sounds much better than orban's thingie ..
That's indeed the biggest issue of homemade tools, they rely on the perception of 1 person, which is in this case a technical programmer and NOT an audiophile.


grt.

Gtek

Evert Verduin

[gtek]

Just to sketch the structure of the program, for other platform devellopers.

- Subtract stereo information from your samples (L-R).
- Make an envellope follower with attack and release timing and follow the L+R signal.
- COMPRESS the L+R signal based on the L+R envellope.
- use the compressed L+R signal to derrive a control signal.
- Expand the L-R signal based on the gain of the control signal of the L+R compressor.

Don't use expansion,but the inverted control signal of a compressor.
This sounds silly, but this is not an expander.
An expander will attack (expand ) after the attack time is passed. We want to Expand before the attack time has passed!
In this case we can expand the stereo signal during a transient.
The attack release time of the envellope follower are in fact time domain filtering parameters in order to "pick out" the transients.
Attack_time now represents the maximum transient time before expansion of the stereo signal will sease. i.e. if a L+R transient will come, the stereofield expands until the attack_time has passed.
Release_time now represents the minimum time allowed between L+R transients. Expansion after an attack will occur only if the release time has passed. In this way you could suppress reverbs of hihats etc.

Grt.

Evert

[gtek]

My way to derrive the compressor control signal is as follow:
( i'll put in C )
monosave=L+R;
mono=Compress(monoEnvellope,mono,thresshold,ratio);
stereo=stereo/(mono/monosave); // =stereo* 1/(mono/monosave); mono/monosave = control signal of compressor.

that's actually it. Be aware of logarithmic and linear algorithms. The above example is a linear algorithm.

Evert

[JonnySun 2.0]

I don't understand what you say but if it sounds right ... Looking forward to hear this what you describe

[gtek]

Well,

To check out how my routine sounds check the following links:

http://www.hockeyclub-alkemade.nl/evert ... iginal.mp3 for the oringinal track
http://www.hockeyclub-alkemade.nl/evert ... medium.mp3 medium enhanced.
and
http://www.hockeyclub-alkemade.nl/evert ... _heavy.mp3

grt.

Evert

[citizenchunk]

gtek: i think we need to stop thinking about this in terms of a traditional compressor/limiter; i.e. with threshold, ratio parameters. instead, we should think of this more as a transient designer (enhancer) for the decorrellated (side) information, triggered by the correllated (mono) information. as such, we need to track transients, which is best accomplished by using multiple envelopes--a tecnique that i believe S.P.L. pioneered, which they called, "Differential Envelope Technology (DET)". (for more info, check out their site, www.spl-usa.com, and look for "tech talk".)

i've already developed routines for detecting transients and using the differential between 2 envelopes to modulate a signal. this is trivial; it just takes some experimentation to find pleasant-sounding time constants for "dummy-proof" operation.

now, to get back to the original theory of how Orban's technique enhances the apparent stereo field. personally, i am a little dubious of the idea that modulating the decorrellated signal only during transients will somehow sound more realistic than a constant gain boost. i understand how the dynamic nature might be preferrable, but i think that the attack time itself is far too short. think about a concert hall: when someone hits a kick drum, the reverb isn't just a short spike; it has a long decay.

perhaps, if one were to track the transients (using envelope differentials), then apply a release (decay) to the gain control signal, it might be more realistic.

i've been messing around with this a little, so i'll post some results later on.

== chunk

[JonnySun 2.0]

Hi chunk,

it's nice to hear that devs show interest in these extraordinary and expensive machines

[citizenchunk]

...Don't use expansion,but the inverted control signal of a compressor.
This sounds silly, but this is not an expander.
An expander will attack (expand ) after the attack time is passed. We want to Expand before the attack time has passed!
In this case we can expand the stereo signal during a transient.
The attack release time of the envellope follower are in fact time domain filtering parameters in order to "pick out" the transients.
Attack_time now represents the maximum transient time before expansion of the stereo signal will sease. i.e. if a L+R transient will come, the stereofield expands until the attack_time has passed.
Release_time now represents the minimum time allowed between L+R transients. Expansion after an attack will occur only if the release time has passed. In this way you could suppress reverbs of hihats etc.

i think i'm beginning to understand this effect better, and my original suspicions were confirmed: this is essentially an upwards compressor that only affects the decorrellated information.

i wrote a couple variations on a transient-tracking-based scheme, but they all sound weird. definitely not natural sounding. i was mislead by the original claim that it needed to track the transients. the gain of the side is not modulated ONLY during transients. that just makes the side "jumpy".

i'll put together something that i think is more appropriate later on.

== chunk

[JonnySun 2.0]

Well,

To check out how my routine sounds check the following links:

http://www.hockeyclub-alkemade.nl/evert ... iginal.mp3 for the oringinal track
http://www.hockeyclub-alkemade.nl/evert ... medium.mp3 medium enhanced.
and
http://www.hockeyclub-alkemade.nl/evert ... _heavy.mp3

grt.

Evert

Hi gtek,

i have heard these audio_examples, and ....

yeeeeaaah ... very nice!
I don't know how the orban 222A sounds but your audio examples sound good for themselve!

Now we need to give the bass-power back and set it into the phantom-mid. There's a hole between right and left box.
I run two media players (Original and heavy processed)the same time to hear if the power comes back. (Sorry for tiny english .. )
Okay, besides the fact the two files don't run synchronized, i found it sounds better now.

So we need something like a low pass filter or so to save a specific range (e.g. 0 - 200 Hz) to keep the bassrange in the middle and unprocessed/dry.

How's that?!

Or trying to build a signal chain which adds the original signal to the processed one but placing it between the boxes to fill the "hole".

There are different possibilities for sure

[JonnySun 2.0]

Adding an idea:

Dear gtek,

is it possible to process the first part of the following audio-example with your processor-emulation?
It's an audiofile from Jeez. He makes 8 Parts with piano and a mix with different settings.

I'm interested in processing the raw-parts, so take the first part from

http://audioshots.com/auditorium/downlo ... e0385f217e

Here you can read additional comments to this file:

http://audioshots.com/auditorium/viewtopic.php?t=1124