Also known in USA as "Hafler Stereo" and in some other countries, like in Sweden, as "Ambiofoni"



This is the definition of ambiophony we have used for the last 35 years or so:

Reproduction of sound using rear speakers connected so they will radiate the difference signal between right and left channel. Normal stereo recordings are used (with two channels recorded.)



The requirement for a successful ambiophonic reproduction is that the difference signal contains mainly room ambience information, which is also does, if the recording is made with either the X/Y or the M/S microphone placement method. It is assumed that the instruments are not placed too far apart since the difference signal in such a case will contain a large amount of directly radiated sound.

Pic. 1: Here it is seen how an X/Y or M/S microphone looks at it's surroundings and encodes it to right and left channels. Note that neither X/Y nor M/S (with 8-mics) can discern between font and back! What is encoded as "rear" is what is coming from the sides. Further, the sound from the rear is reversed right-left! If X/Y is used the right=X and left=Y, if M/S is used the right=M+S and left=M-S.



The direct sound is found in the sum signal on the other hand, if the recording is made with either the X/Y or the M/S microphone placement method.



In a matrix system the channel separation is limited, because it is not a "full" discrete four channel system. The channel separation may be spread differently depending on how the system is oriented.



Depth or front/back stereo.

If we, e.g. imagine that we reproduce a sum signal from two loudspeakers standing in front of us (or with a single mono loudspeaker) while the difference signal is reproduced by two loudspeakers behind us, the channel separation between font and rear - that is between direct and ambience - will be total (see pic. 2).

Between right and left the separation will be zero however! Not a perfect reproduction in other words.

Pic. 2: Here is a Depth Stereo setup. The two front channels may be replaced with a single speaker op front. Many people consider depth stereo to give a truer reproduction than usual two channel stereo although left-right info is missing.


Width or right-left stereo.

If we on the other hand start from an ordinary stereo set-up, (with two front speakers) and add a couple of rear speakers that we feed with the same information as the front speakers, that is the normal right and left channel, we will get a total right-left separation. Front-rear separation will then be zero however. (See pic. 3).

Pic. 3: Width stereo with four speakers may be like this. A characteristic of this setup is that front and rear sound stage is almost eliminated, instead a more or less irritating "inside the head experience" is experienced. Width stereo with four speakers adds nothing valuable to a two-speaker stereo setup.



The normal compromise is to use neither "sum-front/difference-rear" or "stereo-front/stereo-rear", but use "stereo-front/difference-rear" (se pic. 4).

The front channels may, if you prefer, be assisted by a speaker placed centrally, where a mono signal is played (R+L). The both rear speakers are phase inverted to each other, which is necessary for symmetry.

This gives a maximum channel separation of 6 dB. Not a very impressive figure, but a figure we should not reject until we have found out if it is sufficient.

To spare the readers the trouble to find out yourselves if 6 dB is enough so here is my experience on the matter: 6 dB in channel separation is enough to get a proper directional information, but is too little to not to impair the resolution, mainly the front channels.

Pic. 4: Hafler stereo may be considered a usual stereo connection with two speakers-width stereo supplemented by two additional speakers behind the listener, for the ambient information (room acoustics). Thus you combine the advantages of two speakers "Width Stereo" (normal stereo) with "Depth Stereo" without compromising the sound stage too much.



The first that springs to mind is to decrease the level of the rear channels until the front image is not impaired. Unfortunately, this method does not work very well, since the level where the rear channels are not heard, and the level where they do not impair the sound stage is very much the same.



The ear can prioritize between sounds both depending on strength, timbre and on arrival time. This time parameter is very important. If the difference is 10 ms between two identical signals with transient content, the later signal arriving may be up to 10 dB louder and we still will not recognise it as the main source. With 6 dB channel separation, we thus get 16 dB margin if we delay the rear channels! A small delay for the rear channels will not only remedy the problem of the rear channels stealing the interest from the front channels, it will also make the perceived resolution from the front channels remain close to unimpaired however unbelievable it may seem.



Also the timbre will add to how we prioritize directions, if not so much as the arrival time. A little simplified it may be said that we consider a sound source with a reduced treble content to be reflected sound if there are other sources with full frequency response reproducing the same signal. To cut off the rear channels above 5-10 kHz may therefore be suitable.

Pic. 5: With delayed and filtered rear channels we achieve the best of two worlds. A very distinctive sound stage ( that even expands outside the front speakers with good precision), an even transfer to the appropriately diffuse rear sound stage and it is very insensitive to the placement of the listeners if the centre speaker is used.


You may comment that a delay and filtering as the suggested may save the stereo image from bad resolution but will then reduce the resolution from the rear channels instead! This is in a way correct, but from experience we know that the rear sound stage does not contain such a lot of detailed information as the front channels do, and if it does, it is probably because it has "leaked" transient and detail information from the front to the rear channels, and then it will be a good thing to remedy the error.

The reason for using Ambiophony is to retrieve increased information from music recordings. Therefore we may accept that mainly the ambience is recreated from behind. The 6 dB margin that the channel separation gives will be enough since the ambience does not contain any distinctive transients. For pure effects, like barking dogs running around or explosions and such, ambiophony is of only limited value.



A simple analysis of an ambiophonic system with delayed rear channels will maybe give the impression that bad time errors would appear, but appearances are deceptive! The truth is (which will be shown later in the article) that the rear channel information is recorded too "early" because of the rearward placement of the ambience recording microphone compared to the stereo microphones recording the front sound stage. Further, the possible time errors we are talking about are small compared with the very large interval between the direct recorded sound and the information coming from the rear of the room where the recording is made, especially if it is a concert hall. Then we talk about roughly 100 ms, which is 5-10 times the "offset time" of the 10-20 ms that is suitable for ambiophonic recordings.



The desired delay may be achieved in many ways.

* Either by placing the rear loudspeakers further from the listener than the front speakers are. (10 ms represents a distance of 3.4m).

* Or the delay may be done by preventing the direct sound to reach the listeners until it has reflected on a few walls.

* A third possibility is to have an electronic delay. Delaying the line level is the only possibility with the delay times here, preferably digital. All pass filter delays will not give enough delay time.



The rear speakers should be more than 3 metres further away than the front speakers to get a good result. In practice few have a room large enough for this. A slightly inferior way is to have the sound reflected by the walls without radiating directly towards the listener.



Electronic delay is an alternative to acoustic delay. This is an alternative that can give an equally good illusion.

It should be noted that "time delay" is different from "decay time" that may be adjusted on units where reverb is added. These units create artificial ambience. That, to listen to an artificial room is not the idea with ambiophonic listening at all. The point with ambiophony is that you may experience the acoustics of the room where the recording was made!

A drawback with electronic delay, compared with acoustical, is that in practice you have the rear speakers physically closer and are not able to move around your listening seat without upsetting the intensity between right and left rear. The content of this is that you should try to get the distance to the rear speakers as large as possible although you use electronic delay.



In a stereophonic system predicting what tonal balance you "should have" is not possible in the listening seat, since only the front half sphere is encoded in the recording, that is the front half - the sound stage. The listening room has a marked influence on the frequency response when you measure from the listening seat, but does not influence the perception of the sound from individual instruments in the sound stage.

In an ambiophonic system however having the ambition to have a linear frequency response in the listening seat is fully realistic, at least from the system as a whole. It is a reasonable ambition since you want to reproduce the full sound sphere that was present at the recording session, with unchanged tonal balance.

In practice the bass range is increased more the more loudspeakers there are in the listening room. To some degree this is compensated by the fact that in an ambiophonic system the rear channels are working in opposite phase with each other (R-L and L-R). When you are placed sufficiently in the midst of the room, you often notice that an additional bass cut may be necessary. Speaking of individual directions, the best way to achieve a correct tonal balance it is to cut off some of the bass output to the rear channels.

© Text and illustrations, Ingvar Öhman.
A fusion of a series of articles published in Musik & Ljudteknik no. 3-93, no. 4-94 and no.1-97.
Translated by Per Arne Almeflo, Sonic Design, with permission from the author.