Circle Surround

Circle Surround® White Paper

INTRODUCTION

Circle Surround was originally developed to offer the benefits of surround sound for music applications that had been available for cinematic use for years. There are side effects in the cinematic surround matrix systems that do not provide a high quality format for music. After first addressing the requirements of matrix surround for music, a video mode was developed which offers an improved stereo image for the front, as well as stereo surround channels.

All Circle Surround designs to date have provided multi-band left/right steering in the surround channels. In general, this has been accomplished by splitting the L-R signal into multiple bands and steering each band to the left or right surround channel based on the dominant left or right input signal energy in that specific band. These systems are also capable of producing simultaneous left and right stereo imaging (i.e. if the input signal contained dominant mid band signal energy in the left input and dominant high band signal energy in the right input, the resulting surround channels would provide mid band information steered to the left, and high band information steered to the right.). However, there has never been a matrix system available which provided the ability to localize a specific broadband audio signal in the left or right surround channel as an individual, independent sound source-at least, not until now!

The Circle Surround® 5.2.5™ surround matrixing system is the latest improvement in RSP Technologies' patented Circle Surround surround sound processing system. With the capabilities of the 5.2.5 system, it is now possible to encode 5 independent channels down to an LT/RT (stereo) signal, with the ability to then decode those five signals as independent sound sources which can be fed to any predetermined speaker location. Prior to the introduction of the Circle Surround 5.2.5 matrix system, 4-2-4 matrix technology allowed only for the decoding of a left, center, right or single surround signal as a dominant channel. 5.2.5 now provides the capability to encode a left, center, right, leftsurround or right surround signal as a dominant channel. Whether producing sound for music, film or multimedia, this major breakthrough in matrix technology greatly enhances the listening experience. In addition, the 5.2.5 system allows for backward compatibility with all existing matrix formats, as well as normal stereo material.

With the Circle Surround 5.2.5 system, it is now possible to produce audio where sounds originate from the left or right surround channel, with 30dB of separation to all other channels. Sounds can be panned from left or right front, to the left surround or right surround position-independent of all other channels. What this means is that the sonic impact of the discreet 5.1 digital systems is achieved with full backward compatibility to all other matrix systems and stereo.

Audio mixed on one of the 5.1 systems can only be played back on that exact system, requiring a costly digital decoder in order to decode any audio. A Circle Surround 5.2.5 mix can be decoded in Circle Surround 5.2.5 for full five-channel impact. When decoded with other matrix systems (such as Dolby ProLogic*), full compatibility is achieved. The left and right surround information will appear in the mono surround channel, and the front channels (left, center and right) play back as encoded. And, if 5.2.5 audio is played back without any surround decoder, the mix allows for full stereo reproduction without any sonic degradation.

With the 5.2.5 breakthrough, the optical sound track of motion pictures can now carry five channels of audio. The cinema digital systems that use the film as the recording medium suffer from a common problem of drop-out, where excessive data errors cause the system to immediately switch to the optical sound track. This problem increases with the number of times the film is shown. With a Circle Surround 5.2.5 soundtrack, the difference between the digital soundtrack and that of 5.2.5 would be considerably less noticeable than that of the digital sound and current film matrix soundtrack.

VHS video tape (which is the most commonly used storage medium for consumer video) can also provide five channels of decodable audio, as well as radio, television, and cable. Audio encoded in one of the other matrix formats can be decoded with a 5.2.5 decoder and still provide stereo imaging in the surround channels. However, this does not allow for precise placement of audio in one of the independent surround channels.

Analog Devices will be producing Circle Surround 5.2.5 decoder ICs for OEM applications in both analog and digital form.

Utilization of the complete Circle Surround system (both encode and decode) allows audio engineers to place any voice, instrument or sound effect at any predetermined location within the 360° radius surrounding the listener.

This white paper provides a brief background of surround sound matrix systems, including discussion of some of the more common surround systems, followed by an explanation of Circle Surround - its general operation, its advantages over other systems, and recent improvements that have been made.

BACKGROUND

Methods of providing multidimensional sound for both film and music applications have been in existence since the early 1960s. One of the original matrix systems was developed by David Hafler, founder of Dynaco. Hafler's system was strictly a passive decoder which decoded a standard stereo recording into four channels. In the late 1960s, Peter Scheiber filed U.S. Patent No. 3,632,886, in which he disclosed an encode/decode matrix system which was the basis of one of the major competing formats for quadraphonic sound in the early 1970s. Of the many attempts that have been made to introduce a multidimensional sound system, some of the most notable are the several rival quadraphonic systems introduced in the early 1970s. In hindsight, it is easy to see why the quadraphonic era was short-lived. All of the systems introduced had considerable technical problems, and were incompatible with one another. None of the systems had a good variety of software available and, in addition, the public had to be persuaded to buy extra speakers and power amplifiers. They did not work particularly well with non-encoded material, as they suffered from adverse image wandering effects due to the broadband gain riding implemented. Even so, the aforementioned Scheiber patent, as well as his subsequent patent numbers 3,746,792 and 3,959,590, are the patents cited by Dolby Laboratories for the Dolby Surround* system. In light of this fact, it becomes evident that the original concepts of what has come to be known as "Dolby Surround", as well as all other matrix systems, were originally founded by Peter Scheiber in his efforts to develop a system for quadraphonic audio use with phonograph records in the late 1960s and early 1970s.

As mentioned, the original passive decoding system accepted a standard stereo signal and, from it, produced four signals which could each be fed to a speaker located at specific position around the listener (as shown in Fig. 1).

The systems disclosed in the Peter Scheiber patents are known as 4-2-4 matrixes - wherein four discrete signals are encoded into a two channel stereo signal. The encoded signal can then be played back through a decoder to extract the original four signals from the encoded two-channel stereo signal - allowing each signal to be fed to its intended speaker location (see Fig. 2).

Although passive matrix systems are capable of providing infinite separation between the left and right channels, as well as the front and rear channels, a primary disadvantage lies in that such systems are only capable of approximately 3dB of separation between adjacent channels (i.e. left/center, center/right, right/surround and surround/left). Due to this drawback, it was desirable to develop a steered system, incorporating gain control and steering logic, to enhance the perceived separation between channels. The inventions disclosed in the U.S. Patents issued to Peter Scheiber also incorporate early implementations of such technology, which is further discussed later in this paper.

CINEMATIC CONSIDERATIONS

Following the demise of quadraphonic sound, companies such as Dolby Laboratories adapted the matrix scheme to cinematic applications in an attempt to provide additional realism to feature films. However, the nature of the cinematic listening environment requires that different criteria be met than would be needed for 4-channel matrixed music applications.

In cinematic applications, it is desirable that character dialogue be localized to any characters shown on the motion picture screen, therefore a center channel was incorporated to exclusively provide dialogue which emanates from a speaker located behind the screen. The need for a hard center channel with its own respective speaker (rather than utilizing a phantom center within a stereo system) results from the need to stabilize the center image for listening positions which are off to either side of the theater.

The left and right front channels provide stereo information and sound effects, and the surround channel is used primarily for ambience and effects.

Standard Surround Configuration

The standard Dolby ProLogic system in use for home theater applications utilizes five speakers - configured as left front, center, right front and left and right surround (see Fig. 3). (However, the left and right surround speakers are both fed from the same mono surround channel.) The steering logic of the system was designed specifically for cinematic applications, where the main criteria was keeping dialogue focused at the picture screen (i.e., in the center channel) and out of the surround channel.

TYPICAL SURROUND OPERATION

Surround Channel

The surround channel of the ProLogic decoding system is derived by subtracting L-R and feeding this mono difference signal to each of the rear surround speakers. A limited 100Hz-7kHz bandwidth and a simplified implementation of Dolby B noise reduction is applied to reduce any perception of sibilance splattering in the surround channel. This is primarily due to the inherent grain structure of the medium used for cinematic reproduction (i.e. optical soundtrack on 35mm film). Today, high fidelity VHS tape is capable of far better balance and frequency response than could be provided from an optical soundtrack on motion picture film. In addition, the limited bandwidth is applied to the rear channel due to the fact that the surround speakers often used in the typical home theater system are very small, and not capable of reproducing any bass information below 100Hz.

Delay is also added to the surround channels to contribute to the Haas effect (also commonly referred to as the "precedence" effect). This slight time delay ensures that any leakage of dialogue that may unintentionally emanate from the surrounds will still be perceived by the listener to originate from the front channels. The actual delay time applied for this purpose is based on the distance between the front and rear speaker locations, and is typically calculated at 1 millisecond per foot.

Front Channels

The center channel of the ProLogic decoder is derived by adding L+R information (i.e. all elements of the left and right channels are fed to the center channel). The left channel is the pure left signal from the left stereo input, while the right channel provides the pure right signal fed from the right stereo input. However, the operation of the ProLogic decoder is such that, at any given point in time, one channel is considered to be the "dominant" channel. To accomplish this, the system monitors the level between front and surround, and also between left and right. When a defined threshold has been exceeded, anti-phase information is added to both the left and right channels from their respective opposite channel to cancel out any center channel (L+R) information that is present in the front left and right channels. This acts to increase the separation between the left, center and right channels.

While this technique succeeds in removing the center channel information from the front left and right channels, it also removes the bass components. Therefore, a 6dB/octave low pass network is employed which adds low band information back into the left and right front channels.

Conversely, the center channel is also configured such that hard left or hard right detected input will result in an attenuation of the center channel.

Implementing a Surround Format for Music

With the added dimension and notable improvements that the surround environment provides the film-viewing community, it would seem only natural to apply such principles to music applications to greatly enhance the experience of listening to music.

However, a number of significant drawbacks become apparent when attempting to utilize a Dolby-style surround decoding system for exclusive music applications. One such drawback is that the surround channel of the Dolby-style surround decoder is mono and, as a result, the surround lacks any of the directional realism of a common stereo recording. Automotive sound systems incorporating four-speaker stereo have provided stereo operation in the rear channels for many years. Therefore, attempting to implement a system providing mono surround channel operation, such as Dolby, would be less than desirable. Add to this that the surround channel of such systems also consists of primarily ambient information - thereby not providing the required bass response through the rear speakers. Based on the fact that automotive sound systems derive the bulk of the system's bass through the rear speakers, any attempt to adapt an audio surround system to automotive applications requires that the bass emanates from the rear speakers. It becomes obvious that such a system could not be applied to automotive applications.

In addition to the shortcomings caused by the operation of the surround channel, additional drawbacks are present due to the operating nature of the front channels. A pronounced monophonic emphasis is produced across the front three channels when music is played exclusively through a common decoder designed for cinematic applications. Though this effect is not apparent when monitoring exclusive dialogue (as would be found in a feature film), it is unacceptable for the stereo imaging required for music related applications.

The operation of the Dolby-style adaptive matrix dictates that the system produce a slight cancellation of signals in the left and right channels when input signals are not steered hard left or hard right. This condition is always present unless a hard left or hard right input is detected. Therefore, the system is most often steered between a center-steered signal and the pure left and right that is input to the system - thus clouding the stereo imaging of the front three channels and producing a decidedly more mono sounding front soundfield.

In addition, the detection of a strong center channel input results in the left and right channels suddenly converting to difference signals - thereby producing undesirable image wandering effects across the front three channels, as well as a mono left and right signal. This audible side effect is very objectionable when listening to high fidelity music.

These numerous drawbacks clearly illuminate the fact that a different type of surround system is required for music applications than has been commonly known and used for cinematic applications.

THE ADVANTAGES OF CIRCLE SURROUND®

Other surround system developers have come up with different designs to essentially duplicate the end result of the Dolby system, striving for better performance, better speed and channel separation. Circle Surround, however, was developed under a different approach than any of these other systems. It was developed to provide a multidimensional surround sound system specifically for high fidelity audio applications. It was critical that the Circle Surround system operate effectively with both encoded and non-encoded material. The development of Circle Surround's "Cinema" mode for use with video applications followed only after the system was perfected for use in audio-related applications.

All of the inherent disadvantages of implementing a Dolby-style surround system for exclusively audio-based applications were addressed in the development of the Circle Surround Music mode.

Surround Channel Operation - Music Mode

The surround channels of the Circle Surround system provide full bandwidth, stereo operation. The 100Hz-7kHz bandwidth limitation is not applied, as Circle Surround is most commonly used in music and home theater environments. The time delay applied to the surround channels of typical surround systems can also act to "smear" sounds between the front and surround channels. It is for this reason that a time delay is not applied to the surround channels in the Music mode of Circle Surround. The only condition in which it would be desirable to apply a time delay to the rear channels in a music application would be when a music system is installed in a very large venue (such as a dance club), where the distance between front to rear speaker locations would require a time delay to compensate for the time required for the sound to arrive to the listener from the front vs. the surrounds.

When developing Circle Surround, it was critical that the system provide surround channel directional steering without the necessity of adding any artificial information (such as delays, reverb, phase correction or harmonics regeneration) that was not already present in the original source material. Some manufacturers of home theater equipment have recognized the need for a music reproduction mode. Generally, artificial reverberation and/or delays are applied to the surround channels to simulate the effects of rear reflections in a performance hall. At best, such systems only add an artificial element to the music which is simply not present in the recording - and not intended by the artist.

Other systems have attempted to provide directional surround channel steering capabilities, but have done so using broadband steering designs. Under many conditions, broadband steering is objectionable due to the unnatural pumping effects inherently produced from such steering schemes. To help compensate for this, these systems typically limit steering to something on the order of 10-15dB.

Other methods have also been applied to enhance the performance of the surround channels of surround systems. Lucasfilm's THX® system, for example, applies a "decorrelation" technique to the single monophonic surround channel. This is accomplished by splitting the mono surround channel into two signals and applying pitch shifting techniques to one or both of the channels. Methods such as this produce unacceptable results for music applications, as they merely detune information in the surround channels and still do not place instruments in the left or right surrounds based on their location in the original panoramic soundfield.

Therefore, it was critical that the system provide rear directional steering without encountering the objectionable pumping effects perceived with a single band system. This is accomplished by initially deriving a composite rear signal by subtracting L-R. This L-R signal is divided into at least two bands - mid and high. The crossover for the surround channels of a two band implementation of Circle Surround is a Linkwitz-Riley design with 24dB/octave response and a crossover frequency of 2kHz. This ensures both good separation between the bands as well as correct phase response at the crossover point. This also allows the portion of the audio spectrum which contains most of the high frequency transient and directional information to be processed with proper speed and accuracy. The audible side effect of pumping is greatly diminished as a result of steering the highs separate from the mids and lows. With an instantaneous left band signal in the input, the high band portion of Circle Surround will steer the rear high band to the left in approximately 500µs. It has been documented that the human ear acts as an integrator to signals in the first millisecond, therefore was imperative that the system respond faster than the ear to transients in order to provide the proper definition and transparency. In complex music, the mids do not necessarily follow the high frequency transients - this means that the mids will steer based on the mid band directional bias in the input audio. When a broadband system steers complex audio to the left or right, the opposite channel will produce an absence of audio across the entire spectrum - creating a "gated" effect and increasing the perception of pumping. The multiband scheme of Circle Surround eliminates this problem, as a high frequency left or right bias in the input audio will not necessarily provide a mid band bias in the same direction. This means that the mids will properly track the mid band bias and provide a correct surround sound field without the objectionable pumping or "gated" effect.

The typical method of producing the control signals to steer the surround matrix uses a single capacitor charged positive for one directional dominance (such as left) and will alternately charge the capacitor negative for the opposite directional dominance (i.e. right). The Dolby matrix operates this way, using two RC networks charging both positive and negative - with one network providing a fast time constant , and the other providing a slow (or long) time constant to help avoid the side effects of pumping (see Figure 4).Threshold detectors determine when there is a dominance signal present and will switch in the fast time constant so as to improve steering speed. This signal is fed to a polarity splitter which will produce a left output when the voltage is positive on the capacitor, and will produce a right output when the voltage is negative on the capacitor. Thus, the polarity splitter basically functions as a half-wave rectifier to produce the left and right control voltage from the single time constant. Another high end decoder design which utilizes broadband rear steering implements what is referred to as the Servo Logic™ system. In this design, the broadband left/rear steering also uses a single capacitor charged positive for one directional dominance and negative for the opposite directional dominance. The design switches an analog switch on and off to short out the resistor in the RC time constant circuit. The analog switch is controlled by a PWM (Pulse Width Modulation) circuit which is modulated based on the presence of a directional dominance signal. (The idea is to increase the charging time constant so as to provide a faster response time, which is claimed to be 3.5 milliseconds.) There is an inherent flaw in these designs which greatly reduce their ability to steer fast enough to satisfy the requirements for music. If, for example, a left dominance signal is detected (causing the timing capacitor to charge positive corresponding to a 20dB increase in the left input) and this was instantaneously followed by a right dominance signal, the design flaw becomes apparent. At this instant, the system not only has to charge the capacitor negative corresponding to the right dominance signal, but it also has to overcome the positive charge associated with the original left dominance signal. Thus, it becomes obvious that the actual time period may be several time constants (upwards of 20 milliseconds) before the system can actually provide the proper directional control for the matrix. By this time, the initial transient and directional information is either incorrectly decoded or has been smeared across the channels. This condition is not nearly as common in cinematic productions as it is in music applications. Because Circle Surround was designed for music applications first and cinema second, it was an initial design requirement to eliminate this problem. This problem is solved by dividing the rear channel steering into at least two steered bands, and processing each band with a different and optimum time constant. In a broadband system, a fast time constant (which is desirable for fast steering) can cause distortion due to VCA control ripple-especially at low frequencies. Therefore a system with broadband steering is limited to an attack that is slow enough to avoid control ripple or, by design, has an inherent problem of intermodulation distortion. The Circle Surround design solves this problem by steering in multiple bands. Thus, the high band (typically above 2kHz) can provide an extremely fast attack time without the concern for low frequency ripple causing distortion in the audio signal, since the low frequency audio is steered separately. The low frequency portion of the spectrum does not require an extremely fast time constant, since the upper portion of the spectrum contains most of the initial transient information.

The absence of dominant signal energy in the left or right input for a specific band will result in the rear channel remaining mono in that band.

Virtually all other surround processors generate steering control signals by monitoring the levels between left/right and between front/surround. These control signals are then used to control the entire matrix for all four channels (left, right, front and surround). However, the Circle Surround system incorporates a circuit dedicated to generating the control voltages for the surround steering independent of the front channels for improved performance. The method of deriving directional information utilized by the Circle Surround system also allows for localizing simultaneous images in the surround speakers, such that predominant right mid band information will cause the mid band to steer to the right while the high band can be simultaneously steered to the left. The advantages derived due to the enhanced operation of the multiband scheme provided by Circle Surround provides the perception that there are two virtually discrete channels in the rear.ain

All Circle Surround systems currently available provide a three band system in the rear, with at least two bands (mid and high) steered. Future systems may provide greater rear channel resolution (i.e., a larger number of bands that are steered in the rear channels) to further enhance the performance of Circle Surround. A simplified block diagram of a typical 3-band Circle Surround Decoder is shown in Fig. 5.

It is recommended that speakers of the same type be used for all channels, with equal power applied to all channels. It is also recommended that all speakers be located at an equal distance from the listener for this mode.

The 5.2.5 Decoding System

5.2.5 decoding is the latest improvement in Circle Surround technology. As previously stated, the patent-pending 5.2.5 system allows audio engineers to encode five discrete channels down to a 2-channel signal, then extract those five channels during the decode process and place specific sounds at any one of five or more predetermined locations as individual, independent sound sources.

Unlike a fully discrete digital system, signals can not be fed simultaneously into all five channels with full separation. However, if encoded as a dominant signal, a signal can be placed in any one of the five channels. The 5.2.5 system does have a distinct advantage over fully discrete digital systems in that it is backwardly compatible with all material produced with other matrix surround systems, as well as normal stereo material.

To see the full diagram please click on thumbnail.

Figure 6 discloses a simplified block diagram of the steering control generator for an implementation of the 5.2.5 decoding system which incorporates two steered bands for the surround channels.

Steering Control Generato

The Steering Control Generator monitors the audio in the input and, based on the encoded dominance signals, produces the control signals to steer the VCAs in the audio path so as to correctly position the audio signals in the soundfield. The Steering Control Generator shown in Figure 6 can be viewed as three basic sections: the upper section comprises the left/right high band surround steering control generator, the middle section comprises the left/right low band surround steering control generator, and the lower section comprises the front/back steering control generator. We will begin by examining the operation of the lower section (front/back steering) first.

Steering Control Operation

An L-R signal is fed to the input of filter F3, which provides a single-pole high pass response with a corner frequency of 480Hz. This removes the bass and very low mid band audio from the input to level detector L3 so that it can more accurately track the dominance signal. In the analog implementation, all level detectors produce the log of the absolute value of the input. Once filtered, this provides a signal which is linear in volts per decibel. All of the voltage controlled amplifiers also provide gain control which is linear in volts per decibel. This provides easy-to-implement ratios for the control of gain. Thus, the output of level detector L3 provides the log of the absolute value of the high pass filtered input signal, which is then fed to the negative input of differential amplifier A9.

An L+R signal is fed to filter F4, which is also a single-pole high pass with a corner frequency of 480hz. Filter F4's output is fed to level detector L4, which functions as described above. The output of detector L4 is fed to the positive input of difference amp A9. With no audio at the input of the decoder, the level detector outputs will be at the same potential (typically 0 volts). Therefore, the output of A9 will also be at 0 volts. The output of differential amplifier A9 will be positive-going when the input signal to the decoder contains front dominance (center information), and negative-going when the input signal contains rear dominance (surround information).

The output of differential amplifier A9 is fed to a single pole filter to effectively filter off the ripple from the level detectors and provide a quasi DC voltage. This DC voltage is then buffered and fed to fullwave rectifier FR3 and variable resistor VR3, which dynamically determines the time constant for attack and release characteristics for the steering circuit. The buffered outputs of amplifiers A9, A21 and A25 all feed full wave rectifiers to form a composite DC signal representative of any present dominant signal in either the left/right input or front/surround input. This composite DC signal controls the variable resistor blocks VR1, VR2 and VR3 in the steering circuitry.

The Variable Timing Circuit

As stated, the filtered outputs of the differential amplifiers (A9, A21 and A25) are each fed to fullwave rectifiers, and the quasi DC voltage at the output of each fullwave rectifier controls the variable resistance which drives each timing circuit. Effectively, the voltage applied to each variable resistor block is determined by deriving the peak of the absolute value of the three filtered outputs of difference amplifiers A9, A21 and A25. The output of each variable resistance block feeds a capacitor tied directly to ground, which sets the initial time constant for attack and release. This is followed by a second capacitor tied to ground, but fed through a series-connected resistor. When there is a strong dominance signal present at the input, the resistance of Variable Resistor Blocks VR1, VR2 and VR3 is equal to or less than the resistance of the series resistor which feeds the second capacitor.

The value of the single capacitor is such that the resistance driving it provides an extremely fast time constant. In the high band portion of Circle Surround, this time constant can be fast enough to accurately position sound based on transient information. Since this portion of the rear band steering operates only in the high band, the potential impact of distortion is avoided. In the low band steering and front/surround steering portions of Circle Surround, the time constant is much larger so as to avoid ripple in the control signal, which could result in distortion. However, it is obvious that the low band portion does not require as fast of a time constant to provide proper transient response as is required for the high band.

As the resistance of the variable resistance block increases, the second capacitor and resistor become a greater factor in determining the time constant.

This provides a continuously variable time constant circuit that varies over an extremely large range. Another advantage of this design is that, in IC form, this circuit will only require a single pin for adding the time constant capacitors.

The Center Voltage (C(V))

The output of timing amplifier A16 feeds diode D21, which provides a positive-going output when the output of amplifier A16 is positive (i.e. dominant center channel information is present at the input), to provide center steering voltage C(V.) The C(V ) voltage controls the dynamic operation of the center channel. When the output of amplifier A16 is negative, it is inverted and fed through diode D22, which provides the surround voltage S(V) when there is dominant surround information.

The Front Left and Front Right Voltages (F(LV) and F(RV) )

The C(V.)signal also feeds amplifier A17, which generates the front voltages with a gain of 1.5. Resistor R17 is connected between the negative input of A17 and the negative supply rail - thus producing a positive offset at the output of A17. With no front dominance signal present at the decoder input, this offset is present. Diodes D9 and D10 then provide the front steering voltages F(RV) and F(LV).

Pan Correction

In the cinema mode of operation, the output of the front steering signal is connected as a peak OR function with the output of pan correction amplifiers A21 and A20, which process the mid-band left and right steering signal. Amplifiers A21 and A20 provide the proper volt-per-decibel response for front channel VCAs to cancel audio panned from center to one of the front left or right channels. This improves channel separation for panned audio signals across the front three channels.

The Surround (Back) Voltage (B(V) )

The output of timing amplifier A16 is also inverted and fed through diode D22 to produce the S(V) steering voltage. The S(V) signal is then processed by exponential ratio circuit ER1 to produce the B(V) steering voltage. Exponential ratio circuit ER1 accepts the linear volts/dB response generated by dominant surround (or back) signals and produces an exponentially increasing output voltage. The B(V) signal is used to determine the amount of attenuation that occurs in the front channels when dominant surround information is present.

Generating 5.2.5 Steering Aspects

The 5.2.5 design uses a variable multiplier in the high band and low band left/right steering generators which provides the 5.2.5 steering aspects. A 4-2-4 matrix system encodes the surround information as an equal amplitude anti-phase signal that the decoder detects as dominant L-R information. By causing a slight imbalance or amplitude bias in the L-R encoded signal, a 4-2-4 matrix decoder will still decode this signal as surround channel audio. The Circle Surround 5.2.5 matrix will detect this left or right bias and decode the audio as left or right surround. If the encoded L-R signal is exactly equal (as with 4-2-4 encoding) the 5.2.5 matrix will detect this as dominant surround with no left or right bias and produce a decoder output of equal amplitude in both the left and right surround channels. This provides backward compatibility with 4-2-4 matrix encoded audio. If a 5.2.5 encoded signal contains a L-R signal with a left amplitude bias of 1-3dB, the 5.2.5 decoder will detect this as dominant surround with a left bias and decode this as left surround. The 5.2.5 decoder can reproduce smoothly panned audio from one surround channel to the other, or pan from one or both surround channels to the front channels with up to 60dB of channel separation and 30dB of separation between the surround channels. The variable multipliers in the left and right multiband steering circuits provide variable gain of the left and right steering signals that control the VCAs in the surround channels. The gain of the multipliers increase exponentially with increasingly dominant surround signals, therefore only an encoded signal with dominant surround information will cause the multipliers to increase the gain of the steering signals. A dominant surround (or L-R) signal can only be present in audio that has been encoded, therefore a stereo non-encoded signal can never cause the variable multipliers to change gain. This means that excessive steering or pumping is avoided. The quiescent gain of the multiband steering is sufficient to produce stereo imaging in the surround channels, since signals panned to the left or right in the stereo mix contain dominant left or right audio at various frequencies. The Circle Surround decoder will accurately place these signals in the correct surround channel based on their panoramic position. As previously stated, 4-2-4 matrixed audio with dominant surround will appear in both surround channels. However, there is an interesting result with some 4-2-4 encoded material. If the 4-2-4 mix contains a dominant surround signal with a small amount of left or right front audio (such as might be encoded when transferring a 5.1 mix to a 4-2-4 matrix mix), the 5.2.5 decoder will position this surround audio with a left or right bias. The end result is that some 4-2-4 encoded material will decode in 5.2.5 with strikingly close performance to that of the original 5.1 mix. The positional accuracy is not quite as good as it is when encoded in 5.2.5, however it does point out yet another advantage of the Circle Surround 5.2.5 decoding system. In many cases the sonic performance of Circle Surround 5.2.5 may actually be better than that of the 5.1 digital systems, especially those that operate at a bit rate of 384 kilobits per second. At this bit rate if all channels were to simultaneously have nominal signal level full bandwidth audio this allows approximately 76.8 kilobits per second per channel. This is certainly not enough bits to support high quality audio. Some of the 5.1 systems use a bit pool and allocate bits to the various channels based on the audio level and bandwidth in the channels. This means that when only a dominate front center channel is required most of the bits are available to support the center channel data. To avoid audible degradation of the signal most of these 5.1 mixes avoid high amplitude full bandwidth audio in all channels. In fact most of the cinema soundtracks use the surround channels for periodic left or right effect. With the capabilities of Circle Surround 5.2.5 in most cases similar stereo surround impact can be realized without any concern for sonic artifacts or audible degradation.

It should be noted that the DTS 5.1 system for laser disk and CD uses a considerably higher bit rate and does not suffer from the previously described problems and is therefore capable of simultaneous full bandwidth high level audio in all channels.

Referring back to Figure 6, when increasing dominant surround information is detected (i.e. when S(V) increases), the voltage S(V) is fed to Variable Multiplier Control VC1 which determines the gain provided by Variable Multipliers VM1 and VM2. When S(V) is at 0 volts, the Variable Multipliers VM1 and VM2 provide a gain of .55. As S(V) increases, Variable Multiplier Control VC1 increases the gain of variable multipliers VM1 and VM2 to a maximum gain of 10x when the S(V) signal reaches a predetermined voltage. The gain factor of the variable multipliers increases exponentially over a predetermined voltage range. However, Variable Multiplier Control VC1 clamps at a specified voltage, and therefore will not allow any higher gain factor than 10x.

When a signal appears at the decoder input with dominant surround information, the outputs of the difference amps for the left/right high band (A25) and left/right low band (A21) are monitored to determine whether the left or right signal is dominant. When a 1-3dB left or right surround dominance is detected, it will then be amplified by as much as 10x to provide the proper steering to the left or right surround channel. Therefore, the Variable Multipliers VM1 and VM2 simply provide an exponential function between the surround voltage S(V) and the multiplication factor over a given voltage range.

The 5.2.5 function can be defeated via switch SW2. When SW2 is closed, the S(V) signal from diode D22 is fed to Variable Multiplier Control VC1 and functioning 5.2.5 steering aspects are provided. When SW2 is open, an increasing surround voltage (SV) at the output of diode D22 does not produce any change in the variable multipliers VM1 and VM2, therefore the gain factor of both VM1 and VM2 remains constant at .55.

Split Band Left/Right Surround Steering

The left (L) and right (R) input signals are each divided into two steering generation paths - a left/right high band path (the upper path shown in Figure 6) and a left right low band path (the center path shown in Figure 6). In the upper path, the L and R inputs are each processed by high pass filters F9 and F10 which have a corner frequency of 16kHz, so that the upper path generates steering signals based on the high band spectrum information of the input audio. In the center path, the L and R inputs are each processed by band pass filters F7 and F8, which provide a center frequency of 480Hz, so that the center path generates steering signals based on the mid band information of the input audio.

The outputs of difference amplifiers A25 and A21 in the high and low band paths are each fed to low pass filters to filter off ripple components from the level detectors and provide a quasi-DC output. As stated previously, the outputs of fullwave rectifiers FR1, FR2 and FR3 are summed to generate the control voltage for the variable resistor blocks VR1, VR2 and VR3 for the three steering generation paths.

Generating the High Band Surround Steering Voltages

The high band and mid band steering signals are also each processed by a variable multiplier to provide 5.2.5 steering functions. In the high band, the output of the variable multiplier VM1 is fed to variable resistor block VR1, which generates dynamically-changing timing signals. These timing signals change both attack and release characteristics based on a dominance signal detected in the inputs to the Steering Control Generator.

After the timing signal is generated, the output of amplifier A14 feeds diode D17, which goes positive when the output of amplifier 14 is positive. Amplifier A14 also is also inverted and fed to diode D18, which goes positive when the output of amplifier 14 is negative. When dominant right high band information is present at the input, the output of amplifier A14 will be positive, and a positive voltage will be present at the output of D17 - thus providing a positive response at the L(HA) output. The L(HA) output feeds the control port of a VCA to attenuate the left high band when dominant right rear high band information is present in the input. The positive voltage at the output of diode D17 is also inverted and multiplied by a factor of .2 to provide a negative response at the R(HG) output. The R(HG) and L(HG) outputs clamp at .375 volts, which means that a maximum gain of 3dB will be provided from the surround VCAs. The R(HG) output feeds the control port of a VCA to increase the gain in the right surround high band to a maximum of 3dB when dominant right surround high band information is present in the input.

Conversely, when dominant left high band information is present at the input, the output of amplifier A14 will be negative, and a negative voltage will be present at the output of D17. However, the negative voltage at the output of timing amplifier A14 is inverted and fed through diode D18 - thus providing a positive response at the R(HA) output. The R(HA) output feeds the control port a VCA to attenuate the right high band when dominant left surround high band information is present in the input. The positive voltage at the output of diode D18 is also inverted and multiplied by a factor of .2 to provide a negative at the L HG output. The L HG output feeds the control port of a VCA to increase the gain in the left surround high band to a maximum of 3dB when dominant left surround high band information is present in the input.ain

Generating the Low Band Surround Steering Voltages

In the low band steering generator path, the filtered output from difference amplifier A21 is first fed to variable resistor block VR2 and the timing circuit prior to variable multiplier VM2. This is done so that pan correction and the R/L steering voltage can be generated prior to applying the variable gain block of variable multiplier VM2.

Note: As previously described, pan correction corrects for signals that are panned from center through to left or center through to right, and cancels that signal out of the opposite channel over the pan until the channel that is being panned to becomes the dominant channel.

Dominant right mid band information at the input will cause the output of difference amplifier A21 to be positive. Therefore, the output of timing amplifier A15 will also be positive. After being fed through variable multiplier VM2, the positive signal is fed to amplifier A10 to provide the proper V/dB response at the surround low band outputs. This positive voltage from amplifier A10 is then fed through diode D19, which then provides a positive response at the L(LA) output. The L(LA) output feeds the control port of a VCA to attenuate the left low band when dominant right rear low band information is present in the input. The positive voltage at the output of diode D19 is also inverted and multiplied by a factor of .2 to provide a negative response at the R(LG) output. This signal also clamps at .375V to provide a maximum of 3dB attenuation. The Rn6(LG) output feeds the control port of a VCA to increase the gain in the right surround low band when dominant right surround low band information is present in the input.

Conversely, when dominant left low band information is present at the input, the output of amplifier A10 will be negative, and a negative voltage will be present at the output of D19. However, the negative voltage at the output of gain amplifier A10 is inverted and fed through diode D20 - thus providing a positive response at the R(LA) output. The R(LA) output feeds the control port of a VCA to attenuate the right low band when dominant left surround low band information is present in the input. The positive voltage at the output of diode D20 is also inverted and multiplied by a factor of .2 to provide a negative response at the L(LG) output. The L(LG) output feeds the control port of a VCA to increase the gain in the left surround low band to a maximum of 3dB when dominant left surround low band information is present in the input.

The outputs of A20 and A21 are peak detected by diodes D13 and D14 to produce and R/L output signal. The R/L output will be positive-going when either a left or right dominance signal is present at the input. The R/L output will be 0V when a front or surround dominance is detected. The R/L signal is fed to a VCA which attenuates the center channel when dominant left or right information is detected.

The R/L signal is also applied to a VCA which attenuates the surround channels when a dominant left or right input signal is detected. This allows signals panned hard left or hard right to remain in the front channels.

Auto Balance Requirements

Due to the nature of the 5.2.5 system, greater accuracy for automatic balancing is required than for previous designs. 5.2.5 Decoders for consumer applications monitor the input for differences in encoded left and right surround signal levels of only 1-3dB. As a result of this requirement, the 5.2.5 system uses a feed-forward auto balance design. Previous designs for automatic balancing have been implemented as feed-back designs, which have limited range accuracy. The use of the feed-forward method allows for balancing within roughly .25dB over a +-5dB range.

Front Channel Operation - Music Mode

Due to the de-emphasis of the stereo image and the image wandering effects produced by the steering scheme of the common decoders, it was imperative that the left and right channels of the Circle Surround system remain unaltered. Like the other systems, the center channel signal consists of L+R information. However, Circle Surround incorporates a dynamic center channel - where strong, predominant center channel information results in the center channel level increasing to unity gain. If a strong center signal is not detected, the center channel level is reduced by as much as 10dB to avoid collapsing the stereo imaging of the left and right front channels. Input signals panned hard to the left or right will cause the center channel to steer down completely to eliminate any collapse toward center of signals panned hard left or right. When used in a four speaker configuration without a dedicated center channel, center channel information is divided equally between the left and right front channels. However, the center channel still operates dynamically in such a configuration. (Automotive applications may require a configuration such as this.)

A signal panned hard to surround will result in the attenuation of the left and right front channels to provide a dominant signal in the surround channels.

Note: There is no broadcast compatibility for signals that are located fully in the surround, since they are out-of-phase and cancel out of the L+R monophonic signal. They can be used to good effect on programs that will never be broadcast, such as trade shows. In music or cinema productions, hard surround signals should only be used for non-essential audio (such as sound effects).

This provides the producer additional potential directional impact (for effect only) for signals panned hard to the surround position in the absence of any other audio. Anti-phase information in the left and right channels does not appear in the center channel, therefore center channel steering is not required. No objectionable impact will be apparent due to these steering characteristics, as this steering condition will only occur under a hard surround pan - which can only be achieved when intentionally encoding material to take advantage of this particular feature.

The incorporation of the dynamic center channel, coupled with the pure, unaltered left and right channels, results in a very stable front sound stage where the stereo imaging is not adversely affected - even in the presence of diffuse, non-correlated audio. Thus, all the benefits of having a center speaker are gained without destroying the normal stereo image.

Circle Surround Video/Cinema Mode

Although Circle Surround was initially developed as a surround system for music applications, it also provides a Video mode for Cinematic use. The Circle Surround Video mode also provides additional improvements over the standard surround systems.

An 18dB per octave low pass network is applied to the front channels to maintain a stable low band when steering is taking place. Though other systems typically utilize a 6dB per octave low pass network for this purpose, an 18dB per octave network is implemented to attenuate mid band dialogue information to a greater level in the left and right front channels.

As previously mentioned, the Dolby system has been designed to provide a single dominant channel at any point in time. As a result, the front soundfield tends to collapse towards center instead of maintaining a wide stereo front image. Coupled with the fact that the rear channel is mono, this produces a very one-dimensional soundfield which goes almost directly from front to rear. As previously stated, this is due the fact that the Dolby matrix produces a slight cancellation of the signals in the left and right channels when it is not steered hard left or hard right. Therefore, even in cinema applications where program material contains stereo background music information, the system will collapse toward center and produce a notably narrower soundfield than would be derived with a normal stereo signal. Circle Surround has been designed to avoid this drawback and provide full high fidelity left and right stereo information under conditions where a dominant center signal, or a dominant left or right signal, is not present.

The center channel operates dynamically, as described in the Music mode, so as to avoid collapsing any stereo imaging that may be present toward the center channel. The center channel level rises to unity gain only under hard center conditions, and attenuates under conditions involving stereo music in the background of a cinematic production. This works to maintain a wide left/right soundfield in the front channels.

The Circle Surround 5.2.5 decoding system provides noticeable performance benefits over other decoding systems - even when used with conventional 4-2-4 surround encoded material. Even though the surround channel of conventionally encoded source material is mono, the Circle Surround Decoder maintains the ability to often extract independent left and right surround signals. When 4-2-4 encoded material is panned strong to the surround position, and a portion of the signal is also slightly panned to the left or right front channel, the 5.2.5 Decoder will respond by directing the surround signal to the left or right rear channel.

The Circle Surround decoder does not apply any bandwidth limitations to the rear channels, or the modified Dolby B noise reduction. The system is also compatible with any of the 2-channel enhancement formats, such as QSound* or the Roland RSS* system.

The 5.2.5 Encoding System

Although the 5.2.5 decoding system can be used with material produced with any other matrix system, as well as standard stereo material, it is most effective when used with material specifically encoded as 5.2.5 surround material.

Figure 7 discloses an encoder which accepts five discrete signals and encodes them down to a two-channel L(T)/R(T) signal.

The left audio input (L), at a gain of 1, is combined with the center audio input (C), at a gain of .707 (-3db), to provide a summed output fed to an allpass network (F1). The allpass network produces a constant phase shift with frequency from 20Hz to 20kHz. The output of all-pass network F1 feeds a summing amplifier (A3), with a gain of 1, to provide part of the final LT output. Likewise, the right audio input (R), at a gain of 1, is also combined with the center audio input (C), at a gain of .707, to provide a summed output to feed a second all-pass network (F4) with an identical phase vs. frequency response to that of F1. The surround left (S(L) ) and surround right (S(R) ) inputs each feed identical all-pass filters (F2 and F3), which have been modified to provide a 90° phase shift at all frequencies from F1 and F4. The surround left audio (S(L) ), after being processed by filter F2, is fed to summing amplifier A3. Summing amplifier A3 sums the surround information in-phase with both the left information from the output of F1 as well as with the output of VCA V1 to produce the composite LT output. The output of filter F2 also feeds VCA V2, which controls the gain of the surround left (S(L) ) audio that is fed to summing amplifier A4.

VCAs V1 and V2 dynamically change the gain to provide a variable level from -3dB to -6dB to amplifiers A3 and A4.

The surround left (S(L) ) and surround right (S(R) ) signals are also fed to a processing circuit (Surround Encode Positional Bias Generator), which generates the controls signal for VCAs V1 and V2. Depending on a surround left or surround right dominance, the Surround Encode Positional Bias Generator will determine the proper gain of the VCAs to provide proper encoding of the bias signal for the composite LT/RT output.

By providing the 90° phase shift for filters F2 and F3, a signal fed simultaneously to a surround input (S(L) or S(R) ) and the right input (R), cancellation of the input signal will not occur at summing amplifier A4. Without processing these signals through the all-pass networks, signals that appear in both a surround channel and the right channel would cancel.

This technique allows for the encoding of an L-R signal wherein the left or right surround information will be encoded with a positional bias based on the left or right input level difference. This allows the 5.2.5 decoder to differentiate between left surround and right surround in the encoded audio signal.

Thus, a 5.2.5 matrixing system can be achieved which allows any encoded signal to be fed exclusively to the left front, right front, center, left surround or right surround channels.

In Conclusion

Circle Surround 5.2.5 is the pinnacle of evolution in surround matrix technology, offering tremendous benefits to the high-end audio community. In addition, Circle Surround also furthers the performance previously available from surround matrix decoders for video applications.

The Circle Surround system was co-developed by Derek Bowers, who also devised the original concepts of the system, and worked through the development of the system to its current embodiment.

The Circle Surround® 5.2.5™ system is currently covered by U.S. Patents #5,319,713 and #5,333,201, with other patents pending and foreign patents pending on both the Circle Surround encode and decode processes.