System overview

Most people understand why a stereo system with a center channel and a surround channel can produce more interesting audio effects than a system with just the left and right channels alone. Unfortunately, audio media has traditionally been recorded with only two channels, meaning that there simply isn't a convenient place to stick the center and surround channel information on a typical videocassette. This is where center and surround channel encoding comes into play. The encoding process stores the center and surround channel information along with the left and right channel information, effectively squashing 4 signals (left, right, center, and surround) into 2 signals (left and right). Eventually, the decoder extracts the left, right, center and surround channels from these two encoded signals.

Encoding the Surround and Center Channels

Even though the encoding mechanism is not a part of consumer electronics (encoding is performed at the recording studio), a brief discussion of the process is included in our report because the decoding circuitry is dependent on it.

Figure 1 illustrates how the center and surround channels are encoded into the left and right channels to produce the "Lt" and "Rt" outputs. Encoding the center first involves a 3 decibel signal reduction, effectively halving its magnitude. This halved center signal is then added to both the left and right signals. Encoding the surround channel also starts with a 3 decibel reduction in signal amplitude. The reduced surround signal then passes through a band pass filter, eliminating high and low frequencies. Presumably, this filtering makes the decoding process easier on hardware that has limited bandwidth, as it serves little other functionality. After being shifted 90 degrees in positive and negative directions, this filtered surround signal is then added to the left and right signals, respectively.

This may seem like a bunch of random signal arithmetic, but there are three important effects of this encoding mechanism that should be mentioned. First, the center channel information encoded in the left signal is in phase with the center channel information encoded in the right channel. Second, the surround channel information encoded in the left signal is exactly 180 degrees out of phase with the surround channel information encoded in the right channel. Third, the center and surround signals are only added to the left and right signals at half volume, meaning that the left and right channels still sound pretty close to what they did before the encoding process. These observations play an important role in the decoding mechanism, as described in the following section.

Decoding the Surround and Center Channels

After understanding how the encoding scheme works, we designed a simple method of decoding the left, right, center, and surround channels from the encoded left and right channels. Our method takes advantage of the general properties of the encoding process, without implementing some of the more complicated (and more expensive) operations of commercial decoders.

As Figure 1 illustrates, our "decoded" left and right signals are obtained by simply passing through the encoded left and right signals. In other words, there is really no decoding taking place at all! Recall that the center and surround channel information is added onto the left and right signals at half amplitude. Sound coming from the left and right speakers actually contains the center and surround channel information as well, but this is not necessarily a bad thing. What the listener can hear of the center and surround information will effectively create a subtle blending effect between adjacent channels.

To recover the encoded center information, the decoder performs a summing operation between the left and right input signals. Signals that are in phase on both input channels are added, effectively doubling the amplitude of this in phase information. Because of the -3db reduction in amplitude during the encode process, the center output from the decoder will be restored to the 0db reference level. Out of phase information is summed to equal zero by this information, meaning that the out of phase surround information is not present at all in the center channel. An artifact of this process is that some left and right channel information is added to the center channel, but is down 3db in magnitude. See Figure 2 for an illustration of how the center signal appears originally, encoded onto the Lt and Rt tracks, and after decoding. Note the waveform has additive noise as a result of decoding. This noise is actually sound originally recorded as left or right channel information.

To obtain the encoded surround information, the decoder performs a difference operation, subtracting the right input signal from the left input signal. This effectively adds signals that are exactly 180 degrees out of phase, precisely the phase difference of the encoded surround information. Again, the amplitude of this out of phase information is doubled, and the surround output from the decoder will be restored to the 0db reference level. The in phase center channel information will be subtracted to equal zero. Again, some left and right channel information is present in the surround center channel, but at half magnitude. Figure 3 illustrates the conversion of the original surround signal to the encoded signal and the decoded result.

Figure 4 shows the schematic of our circuit. The upper right op-amp performs the summing operation and the lower right op-amp performs the difference operation. The four op-amps on the left are simply voltage followers used to buffer the inputs to the sum and difference op-amps. Each of these buffers contains a 10 Ohm resistor used to offset the input voltage. This prevents a clipping effect on the buffers' outputs. For the difference op-amp, this offset is naturally eliminated when the two signals are subtracted. For the summer, the offsets will be added to each other, resulting in an higher input voltage for the summing op-amp. This effect is alleviated by the potentiometer connected to the positive terminal of the summing op-amp; it controls an offset voltage that compensates for this higher input voltage. The potentiometer on the input of the difference op-amp controls the relative balance between the left and right input signals. By tweaking the input resistance of this input, we could match its input signal very closely with the other signal and minimize any noise at the output of the difference op-amp. The remaining two potentiometers control the volume of the summing signal and differentiated signal, respectively.

Dolby Surround and Dolby Pro-Logic

Our testing also involved examining the Dolby Surround and Dolby Pro-Logic decoders. The model used was the Pioneer VSX-502 which is an integrated amplifier/decoder. We were unable to examine the specific circuit schematic of the Dolby decoder because a single IC was used to perform decoding. We could, however, examine the block diagrams of the Dolby decoding processes to determine how the Dolby circuit likely differs.

Dolby Surround is a three channel surround decoder that ouputs a left, right and surround channel. The surround channel undergoes a delay and is lowpass filtered before playback.

Dolby Pro-Logic is the modified form of Dolby Surround decoding that adds a center channel. Pro-Logic also adds a steering logic to "guess" which channel sounds should be coming from and cancel the signal at all other speakers besides that channel or to move into 'passive' operation when no dominant signal is detected. The surround channel is also filtered and delayed as in the Dolby Surround system. Dolby claims this logic results in greater channel separation over the 3db theoretical limit.

Both Dolby decoding technologies can be implemented in either analog or digital ICs. These ICs were found to range in price from under $100 up to $800 for versions that incorporated multiple surround decoding algorithms.