Ambiophonics

[Rocko]

Hey...

I've just read about 'Ambiophonics' technique. It is a method of expanding the stereo field of two short distant speakers, so stereo angle is 'too' small.

In a nutshell:
In an ideal setup, left ear will recieve only left channel of a stereo system in a room and right ear will receive only right channel.
However, in a real situation, left ear receives right channel information and vice versa.
This 'cross' information reaches the 'wrong' ear with some EQ (due to the skull filtering the signal), short latency and a lower SPL.
So, using this knowledge, one can 'inject' left side information into right channel, phase reverse it, attenuate, filter and delay for cancelling the 'cross information' (?!) and vice versa (right into left).
One can then create a feedback loop so that this cancelling process is endless.

Before I jump into experiments, let me ask if someone in this forum is experienced with this topic? Sounds an interesting topic to play around with.

http://www.ambiophonics.org/

[Spogg]

Hi Rocko

This is indeed a fascinating subject. Some time ago I looked into stereo enhancement and made the Quilcom STAGE which uses various techniques to experiment with:

viewtopic.php?f=3&t=5801&start=0&hilit=quilcom+stage

If you find a tehnique I missed I'd love to know, or if something could be done better too!

Cheers

Spogg

[Rocko]

The Quilcom STAGE is a great solution indeed.
There are many solutions for stereo widening out there all of which are content dependant. So the STAGES's ability to simply jump from one technique to another, while content is running, is very useful. It's like a few VST's in a single box. Impressive.

Testing:
In my setup, I'm aimed at stereo content as input, not 'mono to stereo effects'.
The listening angle is small, so distance between speakers is much smaller than distance from speakers to listener.

Using the STAGE In this set up I found the Mid/Side widener and Bedini FIRST B.A.S.E. to provide the best results, best being wide stage but keeping a natural sound (as far as possible), on different content. Content includes stereophonic sounds (reverbs, guitar effects, pads, etc.) with monophonic singing voices simulatenously.

Theory:
From theory side, ambiophonics takes a different approach. It is kind of 'HAAS effect' and 'MS widener' combined, but there is also some different EQ (to simulate the head) and "R.A.C.E" which is a feedback loop. This is what makes 'Ambiophonics' technically different.
Feel free to read about Ambiophonics in the link I had sent. Sounds interesting.

I'll start comparisons to Ambiophonics and my own experiments soon. Keep you posted.


Thanks Spogg ! Appreciated.

[Spogg]

Thanks Rocko

I really do appreciate the feedback and comments.

Here's another one of mine, the Quilcom M2S, which deals in turning a mono source into a stereo field, again with various methods.

viewtopic.php?f=3&t=5794&p=27835&hilit=m2s#p27831

I know it's a bit off-topic, but it's kinda related.

One of the issues, as I'm sure you know, is that if you introduce unequal eq (left Vs right) you can get phase shifts which muddy the stereo image, if you're not careful. But if it's mono to start with then it's no problem

I'll be VERY interested to see and hear what you come up with.

Good luck!

Spogg

[DigiTonix]

Hi!
The topic is very interesting, I'd like to read more...
But why removed the filter?

[Rocko]

Hi,

I removed the filter question because I had found a solution. Once I finlaize it, the complete model will be more understandable - so no need for that long question.


Rocko

[DigiTonix]

Oh!, then waiting for your further discussi!, I didn't understand in the past the filter, but I think all will become clear later...

I watched the video with the helicopter, which on the website Ambiphonics, in the headphones I get the effect of a multichannel system , it is possible to do this on FS?... This is a very cool idea!!!

[Spogg]

Hey Rocko

I just spent a couple of hours reading about this and auditioning example sounds and…
This really works and sounds great. As they say, speaker angle is important but when it’s right I can look at the speakers and my brain tells me they are silent, just sound coming from “out there”!

I wasn’t aware of this technique when I made the STAGE, otherwise I would’ve had a go. From the info on the site you linked, especially the block diagram, it shouldn’t be too tough to reproduce. So now I’m really looking forward to hearing what you can do.

Thanks for this!

Spogg

[Rocko]

I'm going through this topic step-by-step as this is new to me...

I'll need to pass the "original signal" in parallel to a "filtered copy" of it. Use case can be the original signal which is added to an HPF'd copy of it.
To do that (with minimum phase IIR with zero latency) I'll first need to go over Martin Vicanek's tutorial of phase and magnitude response in use with cross-overs (thanks again Martin):
http://flowstone.guru/downloads/multi-band-splitter-tutorial/#changelog
http://www.dsprobotics.com/support/viewtopic.php?f=4&t=8081&p=32980&hilit=crossover#p32980

I've let myself build a minimized copy of this tutorial which is more focused on our 'Ambiophonics' example. This copy is attached.
It shows that one needs to add AP (All Pass) filters to the 'original' signal in order of receiving the correct (and desired) magnitude response.

I've noticed that
For a 6db/octave filter slope, one does not need to add an AP (same as Spogg's Quilcom STAGE / Antiphase bleed).
For a 12dB/octave filter slope one needs a 1st order AP.
For a 18dB/octave filter slope one needs a 2nd order AP.
However for a 24dB/octave (two 12db/octave cascaded) one still uses a 2nd order AP.

Any explanation to this? I was expecting a 3rd order AP to fit into the last example ??

[Rocko]

So, here is my first version of Ambiophonics Basic setup.
Feel free to test run it and please share any findings.
Notice that it DOESN'T RUN A HARD LIMITER so the signal may 'clip'. I recommend to run it at low input volume or add a hard limiter.

The idea is to split the signal into three bands, low and high get no processing and mid-band gets the RACE feedback loop.
This feedback loop gets attenuated, channel swapped (L-R), delayed and phase inverted before being injected again to the system.
The theory is to cancel the left speakers content arriving at the right ear (and vice versa) [acoustic cross-talk] by injecting the cancelation signal into the right channel [electronic cross-talk], thus the electronic cross-talk cancels the acoustic one.
One should receive a wider stereo stage with much more "air" and "space" between the recorded instruments and make the "speakers" acoustically disappear, in a natural way.
As a reminder, for this to work, the speakers need to be facing listener with a small distance between the speakers.
Stereophonic recorded material is a must.

I have tested it with speaker distance = ~35 cm and listening distance ~1.5 meters on a pair of studio monitors.
I prefer not to share my conclusions at this stage to receive clear results from other users. I'm not objective...

Waiting for your comments.