Btsc encoder

Field of the Invention

The present invention relates to an adaptive digital signal weighting arrangement, and is generally related to stereophonic audio encoders used for broadcasting. More particularly, the invention relates to an encoder. digitai to generate the audio signals used in the broadcast of stereo television signals in the United States and other countries.

Background of the invention

In the 1980s, the Federai Commission of

United States Federai Communications Commission (FCC) adopted new provisions regarding the audio frequency portion of the television signals that allowed television programs to be issued and received with audio on two channels, ie sound stereophonic. In these provisions, the FCC recognized and gave special protection to a method of broadcasting additional audio channels, backed by the Electronic Industries Association and the National Association of Broadcasters, and called Broadcast Television Systems Committe (BTSC). This well-known standard is sometimes referred to as Multichannel Television Sound (MTS) and is described in the FCC document entitled MUTICHANNEL TELEVISION SOUND TRANSMISSION AND AUDIO PROCESSING REQUIREMENTS FOR THE BTSC

<img file="AR024581A2_D0001.tif" />

SYSTEM (Requirements for Sound Transmission and

Audio Processing in Multichannel Television for the BTSC System) (OET Bulletin No. 60, Revision A, February 1986), as well as in the document published by the Electronic Industries Association entitled MULTICHANNEL TELEVISION SOUND BTSC SYSTEM RECOMMENDED PRACTICES (Recommended Practices for the BTSC Multichannel Television Sound System) (EIA Television Systems Bulletin No. 5, July 1985). The television signals generated in accordance with the BTSC standard will be referred to as BTSC signals'¹.

The original monaural television signals carried only a single audio channel. Due to the configuration of the monophonic television signal and the need to maintain compatibility with existing television sets, the stereo information was necessarily located in a higher frequency region of the BTSC signal, making the stereo channel much louder than the monophonic audio channel. This resulted in an inherently higher noise threshold for the stereo signal than for the monophonic signal. The BTSC standard overcame this problem by defining an encoding system that provided additional signal processing for the stereo audio signal. Prior to the broadcast of a BTSC signal through a broadcasting station, the portion

<img file="AR024581A2_D0002.tif" />

<img file="AR024581A2_D0003.tif" />

the manner prescribed by the BTSC standard and then a receiver (for example a television set), upon receiving a BTSC signal, decodes the audio frequency portion in a complementary manner. This complementary coding and decoding ensures that the signal-to-noise ratio of the entire stereo audio signal is maintained at acceptable levels.

Figure 1 is a block diagram of a prior art BTSC coding system, or more simply, a BTSC 100 encoder, as defined by the BTSC standard. The encoder 100 receives audio input signals from the left and right channel (indicated in Figure 1 as I "and" D respectively) and generates from them a conditional sum signal and an encoded difference signal. It should be noted that while the prior art system and that of the present invention is described as useful for encoding the left and right audio signals of a stereo signal that is subsequently transmitted as a television signal, the BTSC system also provides means to encode a separate audio signal, for example sound information in a different language, which is separated and selected by the final receiver. In addition, the noise reduction components of the BTSC coding system can be used for other purposes.

<img file="AR024581A2_D0004.tif" />

in addition to broadcasting, such as to improve audio recordings.

System 100 includes an input section 100, a suina channel processing section, 120, and a difference channel processing section, 130. Input section 110 receives the audio input signals of the left and right channels and it generates from them a serial sum (indicated in Figure 1 as an R&amp;D) and a difference signal (indicated in Figure 1 as an ID). It is well known that for the stereo signals the sum R + D signal can be used by itself to provide monophonic sound reproduction, and it is this signal that is decoded by the existing monophonic sound television sets to reproduce the sound. In stereo devices, the sum and difference signals can be added or subtracted from each other to retrieve the two original stereo signals (I) and (D). Input section 110 includes two signal additives 112, 114. Adder 112 adds the audio input signals of the left and right channel to generate the sum signal, and adder 114 subtracts the audio input signal from the right channel of the audio input signal from the left channel to generate the difference signal. As described above, the sum sum R&amp;D is transmitted through a transmission medium with the same signal / noise ratio as that obtained.

<img file="AR024581A2_D0005.tif" />

with the previous monophonic signals. However, the IR difference signal is transmitted through a very noisy channel, particularly in the upper frequency portion of the relevant spectrum, so that the decoded difference signal has a lower signal / noise ratio, because of the noisy medium and the reduced dynamic range of the environment. The dynamic range is defined as the range of signals between the level of the noise threshold and the maximum level where signal saturation occurs. In the signal channel, the dynamic range decreases at higher frequencies. Accordingly, the difference signal is subjected to further processing than the sum signal so that the dynamic range can be substantially preserved.

More particularly, the processing section of the sum channel, 120, receives the sum signal and generates the conditional sum signal from it. Section 120 includes a pre-emphasis filter 122 of 75ps and a band limiter 124. The sum signal is applied to the input of the filter 122 which generates from it an output signal that is applied to the input of the band limiter 124 The output serial generated by the latter is then the conditioned sum signal.

The processing section of the difference channel, 130, receives the difference signal and generates from it the

<img file="AR024581A2_D0006.tif" />

signal coded difference. Section 130 includes a fixed pre-emphasis filter 132 (shown implemented as a cascade of two filters 132a and 132b), a variable gain amplifier 134 preferably in the form of a voltage-controlled amplifier, a variable pre-emphasis / de-emphasis filter (in hereinafter referred to as a variable accent filter) 136, an overmodulation guard and band limiter 138, a fixed gain amplifier 140, a bandpass filter 142, an effective level detector 144 (RMS level detector), a fixed gain amplifier 146, a bandpass filter 148, an effective level detector 150, and a reciprocal generator 152.

The difference signal is applied to the input of the fixed pre-emphasis filter 132, which generates from it an output signal that is applied via line 132d to an amplifier input terminal 134. An output signal generated by it. Reciprocal generator 152 is applied via line 152a to a gain control terminal of amplifier 134. The amplifier 134 generates an output signal by amplifying the signal on line 132d, using a gain that is proportional to the value of the signal on line 152a. The output signal generated by the amplifier 134 is applied via line 134a to an input terminal of the variable accent filter 136, and an output signal.

<img file="AR024581A2_D0007.tif" />

sanili generated by the effective level detector is applied via line 144a to a control terminal of filter 136. Variable accent filter 136 generates an output signal by pre-emphasizing or de-emphasizing the high frequency portions of the signal on the line 134a under the control of the signal on line 144a. The output signal generated by the filter 136 is applied to the input of the overmodulation protector and band limiter 138, which generates the encoded difference signal therefrom.

The coded difference signal is applied through feedback path 138a to the outputs of the fixed gain amps 140, 146, which amplify the coded difference signal by Gain A and Gain B, respectively. The amplified signal generated by the amplifier 140 is applied to an input of the bandpass filter 142, which generates from that signal an output signal that is applied to the input of the effective level detector 144. The latter generates an output signal as a function of the effective value of the level of the input signal received from the filter 142. The amplified signal generated by the amplifier 146 is applied to the input of the bandpass filter 148, which generates from it a output signal that is applied to the input of the effective level detector 150. The latter generates an output signal as a function of the effective value of the signal level of

<img file="AR024581A2_D0008.tif" />

input received from filter 148. The output signal of detector 150 is applied via line 150a to the generator

S.

reciprocal 152, which generates a signal on line 150a.

As stated before, the output signals generated by the effective level detector 144 and the reciprocal generator 152 are applied to the filter 136 and the amplifier 134, ₍respectively.

As shown in Figure 1, the processing section of the difference channel, 130, is considerably more complex than the processing section of the channel sum, 120. The additional processing provided by the difference channel processing section, 130, in combination with the complementary processing provided by a decoder (not shown) that receives a BTSC signal, maintains the signal / noise ratio of the difference channel at acceptable levels even in the presence of the upper noise threshold associated with the transmission and reception of the difference channel. The difference channel processing section, 130, essentially generates the coded difference signal by dynamically compressing or reducing the dynamic range of the difference signal, so that the coded signal can be transmitted through the transmission path of the associated limited dynamic range. with a BTSC signal, and so that a decoder that receives the encoded signal

<img file="AR024581A2_D0009.tif" />

can recover the entire dynamic margin in the original difference signal by expanding the compressed difference signal in a complementary way. The difference channel processing section, 130, is a particular form of the adaptive signal weighting system described in US Patent No. 4,539,526, which is known to be advantageous for transmitting a signal having a relatively wide dynamic range to through a transmission path that has a relatively narrow and frequency dependent dynamic range.

In summary, it can be conceived that the difference channel processing section includes a broadband compression unit 180 and a spectral compression unit 190. The broadband compression unit 180 includes the variable gain amplifier 134, preferably in the form of a voltage controlled amplifier, and the components of the feedback path for generating the control signal to the amplifier 134, which comprise the amplifier 146, the bandpass filter 148, the effective level detector 150, and the reciprocal generator 152.

The bandpass filter 148 has a relatively wide through band, weighted towards lower audio frequencies, such that in operation the output signal generated by the filter 148 and applied to the effective level detector 150 is substantially representative of the difference signal v

<img file="AR024581A2_D0010.tif" />

<img file="AR024581A2_D0011.tif" />

coded Accordingly, the effective level detector 150 generates on the line 150a an output signal representative of a weighted average of the energy level of the coded difference signal, and the reciprocal generator 152 generates on the line 152a a representative signal of the reciprocal of Your weighted average. The signal on line 152a controls the gain of amplifier 134, and since this gain is inversely proportional to a weighted average (ie weighted to lower audio frequencies) of the energy level of the coded serial difference, The broadband compression unit 180 compresses or reduces the dynamic range of the signal on line 132a by amplifying signals that have relatively low amplitudes and attenuating signals that have relatively large amplitudes.

The spectral compression unit 190 includes the variable accentuation filter 136 and the components of the feedback path that generate a control signal to the filter 136 and which comprise the amplifier 140, bandpass filter 142 and effective level detector 144. Unlike the filter 148, the bandpass filter 142 has a relatively narrow through band that is weighted towards higher audio frequencies. As is well known, the transmission medium associated with the difference portion of the BTSC transmission system has a dynamic range dependent on the frequency and the pass band of the filter 142 is chosen to correspond to

<img file="AR024581A2_D0012.tif" />

-li-

<img file="AR024581A2_D0013.tif" />

<img file="AR024581A2_D0014.tif" />

the spectral portion of that transmission path that has the narrowest dynamic range (ie the higher frequency portion). In operation, the output signal generated by the filter 142 and applied to the effective level detector 144 primarily contains the high frequency portions of the coded difference signal. Accordingly, the effective level detector 144 generates an output signal on line 144a, representative of the energy level in the high frequency portions of the coded difference signal. This signal then controls the pre-emphasis / de-stress applied by the variable accent filter 136 so that, in effect, the spectral compression unit 190 dynamically compresses the high frequency portions of the signal in line 134a in an amount determined by the level of energy in the high frequency portions of the sehal coded difference, as determined by filter 142. The use of the spectral compression unit 190 thus provides additional compression of the signal to the higher frequency portions of the difference signal, which is combined with the broadband compression provided by the variable gain amplifier 134 to cause there is actually more global compression at high frequencies than compression that occurs at lower frequencies. This is done because the sehal difference tends to be louder in the part of

<img file="AR024581A2_D0015.tif" />

higher frequencies of the spectrum. When the coded serial difference is decoded with a broadband expander and a spectral expander in a decoder (not shown), in a manner complementary to the broadband compression unit 180 and the spectral compression unit 190 of the encoder respectively, the signal / noise ratio of the signal ID applied to the processing section of the difference channel, 130, will be substantially preserved.

The BTSC standard rigorously defines the desired operation of the pre-stress filter 122 of 75με, of the fixed pre-emphasis filter 132, of the variable accent filter 136, and of the bandpass filters 142, 148, in terms of idealized analog filters. Specifically, the BTSC standard provides a transfer function for each of these components and the transfer functions are described in terms of mathematical representations of idealized analog filters. The BTSC standard also defines the gain, Gain A and Gain B values ​​of amplifiers 140 and 146 respectively, and further defines the operation of amplifier 134, effective level detectors 144, 150, and reciprocal generator 152. The BTSC also provides suggested guidelines for the operation of the overmodulation guard and band limiter 138 and the band limiter 124. Specifically, the limiter of

<img file="AR024581A2_D0016.tif" />

band 124 and the band limiter portion of the overmodulation guard and band limiter 138 are described as low pass filters with cutoff frequencies of 15kHz, and the overmodulation protection portion of the overmodulation guard and band limiter 138 is described as a threshold device that limits the amplitude of. The difference signal coded 100% of the full modulation, where the full modulation is? the maximum permissible deviation level to modulate the audio frequency subcarrier on a television signal.

Since the encoder 100 is defined in terms of mathematical descriptions of idealized filters, it can be conceived as an idealized or theoretical encoder, and those skilled in the art will appreciate that it is virtually impossible to construct a physical embodiment of a BTSC encoder that exactly matches the performance of the 100 theoretical encoder. Therefore, it is expected that the performance of all BTSC encoders will deviate somewhat from the theoretical ideal, and the BTSC standard defines the maximum limits of acceptable amounts of deviation. For example, the BTSC standard states that a BTSC encoder must provide at least 30 db of separation from 100 Hz to 8,000 Hz where separation is a measure of how much a serial applied only to one of the left or right channel inputs appears erroneously at the other exit of the

<img file="AR024581A2_D0017.tif" />

composite stereo (hereinafter referred to as composite signal) that is used to generate the audio frequency portion of a BTSC signal. The composite signal is generated using the conditioned sum signal, the coded difference signal and a tone signal, commonly designated a pilot or simply pilot tone, which is a sine wave at a frequency f_Hwhere f_H it is equal to 15.734 Hz. The presence of the pilot in a received television signal indicates to the receiver that the television signal is a BTSC signal instead of a monophonic signal or another signal that is not BTSC. The composite signal is generated by multiplying the coded signal difference by a waveform that oscillates twice the frequency of the pilot, according to the cosine function cos (4tff „t) where t is time, to generate a signal with carrier suppression, double sideband and amplitude modulated, and then adding to this signal the conditioned sum and the pilot tone.

Figure 2 is a graph of the spectrum of the composite signal. In Figure 2 the spectral band of interest that contains the content of the conditioned sum signal (or the sum channel signal) is indicated with R + R, the two spectral side bands that contain the content of the frequency offset signal ( or channel signal

<img file="AR024581A2_D0018.tif" />

Pilot is indicated by the fleet in the frequency f „. As shown in Figure 2, in the composite signal the coded difference signal is used at 100% of the complete modulation, and the pilot tone is used at 10% of the complete modulation.

Stereo television has been widely successful and existing encoders have performed admirably; however, virtually all BTSC encoders that are now in use have been constructed using analog circuit technology. These analog BTSC encoders, and particularly the analog processing sections of the difference channel, due to their increased complexity, have been relatively difficult and expensive to build. Due to the variability of the analog components, a complex selection of components and extensive calibration have been necessary to produce acceptable analog processing sections of the difference channel. In addition, the tendency of analog components to deviate over time from their calibrated operating points has also made it difficult to produce an analogue section of processing of the difference channel that functions systematically and repeatedly within a given tolerance. A digital channel processing section differs, if it could be

<img file="AR024581A2_D0019.tif" />

built, it would not suffer from these problems of component selection, calibration, and deviation from operation, and s

could potentially provide increased performance.

In addition, the analog nature of existing BTSC encoders has made them not suitable for use with newly developed and increasingly popular digital equipment. For example, television programs can now be stored using digital storage media such as a hard disk or digital tape, instead of traditional analog storage media, and in the future there will be increasing use of digital storage media. Generate a signal! BTSC from a digitally stored program now requires converting the digital audio signals into analog signals, and then applying the analog signals to an analog BTSC encoder. A digital BTSC encoder, if one could be built, could directly accept the digital audio signals and could therefore be more easily integrated with other digital equipment.

Even if a digital BTSC encoder would potentially offer several advantages, there is no simple way to build, using digital technology, an encoder that is functionally equivalent to the idealized encoder 100 defined by the BTSC standard. One problem is that the BTSC standard

<img file="AR024581A2_D0020.tif" />

It defines all the critical components of the idealized encoder 100 in terms of transfer functions of the analog filter. As is well known, although it is generally possible to design a digital filter so that the magnitude or phase response of the digital filter equals that of an analog filter, it is extremely difficult to match the responses of both phase-like amplitude without they require large amounts of processing capacity to process sampled data at very high sampling rates, or without significantly increasing the complexity of the digital filter. Without increasing either the sampling frequency or the order of the filter, normally the amplitude response of a digital filter can only be made more closely matched to that of an analog filter at the cost of increasing the disparity between the phase responses of the Two filters, and vice versa. However, since small errors in either amplitude or phase decrease the amount of separation provided by BTSC encoders, it would be essential for a BTSC digital encoder to closely match both amplitude and phase responses of an idealized type encoder. shown as 100 in Figure 1.

For a digital BTSC encoder to provide acceptable performance, it is critical to preserve the characteristics of the analog filters of an idealized 100 encoder. There are

<img file="AR024581A2_D0021.tif" />

the performance of an analog filter; however, in general none of these techniques produces a digital filter (in the same order as the analog filter) that has amplitude and phase responses that exactly match the corresponding responses of the analog filter. The ideal encoder 100 is defined in terms of analog transfer functions specified in the frequency domain or piano s, and to design a digital BTSC encoder these transfer functions must be transformed to the z plane. Such a transformation can be executed as a multi-to-one transformation from the s-plane to the z-plane, which attempts to preserve the characteristics of temporal domain. However, in such a transformation the responses in the frequency domain are subject to deviations and can be significantly altered. Alternatively, the transformation can be executed as a one-to-one transformation from the plane s to the z plane, which compresses the entire plane s in the unit circle of the z plane. However, such compression undergoes the familiar frequency warping between the analog and digital frequencies. Pre-warping can be used to compensate for this frequency warping effect; however, pre-warping does not completely eliminate deviations from the desired frequency response. These problems would have to be overcome to

<img file="AR024581A2_D0022.tif" />

produce a BTSC encoder di ^ „____________

<img file="AR024581A2_D0023.tif" />

<img file="AR024581A2_D0024.tif" />

Do not be unduly complex and expensive.

There is therefore a need to overcome these difficulties and develop a BTSC digitai encoder. Objects of the Invention

It is an object of the present invention to substantially reduce or overcome the previously identified problems of the prior art.

Another object of the present invention is to provide an adaptive digital weighting system.

A further object of the present invention is to provide an adaptive digital weighting system for encoding an electrical information signal of a predetermined bandwidth, so that the information signal can be recorded or transmitted by a frequency dependent channel and dynamically limited that in a first spectral region has a dynamically limited portion narrower than in at least one other spectral region of the predetermined bandwidth.

And another object of the present invention is to provide a BTSC digitai encoder.

Another object of the present invention is also to provide a digital BTSC encoder that prevents ticking sound, a problem that may occur when the input signal has practically zero levels.

* go λΚα wi

<img file="AR024581A2_D0025.tif" />

Digital BTSC encoder that uses a sampling frequency that is multiple of a pilot tone signal frequency of 15,734 Hz, so as to prevent interferences between signal signal information! coded and pilot tone signal.

A further object of the invention is to provide a digital BTSC encoder for generating a serial conditioned sum and a coded difference signal that do not substantially include any signal energy in the pilot tone frequency of 15, 734 Hz.

Another object of the present invention is to provide. a digital BTSC encoder that includes a sum channel processing section to generate the conditioned sum signal, and a difference channel processing section to generate the coded difference signal, including the channel processing section sum devices to introduce compensatory phase errors in the sum conditioned signal to compensate for any phase error introduced in the difference signal encoded by the processing section of the difference channel.

And another object of the present invention is to provide a digital BTSC encoder that includes a digital type variable accent unit, said unit including a digital variable accent filter characterized by

<img file="AR024581A2_D0026.tif" />

The unit also includes a device for selecting the coefficients of the variable coefficient transfer function, as a function of the signal energy of the coded signal difference.

It is also another object of the present invention to provide a digital BTSC encoder that includes a composite modulator to generate a composite modulated signal from the conditioned sum signal and the coded difference signal.

Another object of the present invention is also to provide a digital BTSC encoder that can be implemented in a single integrated circuit.

Summary of the Invention

These and other objects are provided by an improved BTSC encoder that includes an input section, a summation channel processing section and a difference channel processing section, all of which are implemented using digital technology. In one aspect, the input section includes high pass filters to prevent the BTSC encoder from displaying ticking sound. In another aspect, the BTSC encoder uses a sampling frequency that is equal to an integer multiple of the pilot frequency.

In another aspect, the processing section of the

<img file="AR024581A2_D0027.tif" />

Sum channel generates a conditioned sum signal signal, and the difference channel processing section generates an encoded difference signal, and the sum channel processing section includes components to introduce a phase error in the conditioned sum signal to compensate for any phase error introduced. in the difference signal encoded by the processing section of the difference channel.

In accordance with yet another aspect, the invention provides an adaptive digital weighting system for encoding an electrical information signal of a predetermined bandwidth, so that the signal of information can be recorded or transmitted by a frequency-dependent and dynamically limited channel that has a dynamically narrower portion in a first spectral region than in at least one other spectral region of the predetermined bandwidth.

Other objects and advantages of the present invention will be readily apparent to those understood in the art from the following detailed description where various embodiments are shown and described by way of illustration of the best form of the invention. As will be appreciated, the invention is suitable for other and different embodiments, and its various details allow modifications in various respects, all without deviating from the invention. In accordance with this, the drawings and description shall

<img file="AR024581A2_D0028.tif" />

be considered as illustrative in nature and not in a restrictive or limiting sense, indicating the scope of the application in the claims.

Brief Description of the Drawings

For a more complete understanding of the nature and objects of the present invention, the following detailed description, considered in relation to the accompanying drawings, in which the same reference numerals are used to indicate the same parts or references should be taken as reference. similar parts, where:

Figure 1 shows a block diagram of an idealized BTSC encoder of the prior art;

Figure 2 shows a graph of the spectrum of the composite signal generated in accordance with the BTSC standards;

Figure 3 shows a block diagram of an embodiment of a BTSC digitai encoder constructed in accordance with the invention;

Figure 4A-C shows block diagrams of low pass filters used in the BTSC digitai encoder shown in Figure 3;

Figure 5 shows a detailed block diagram of the broadband compression unit used in the digital BTSC encoder shown in Figure 3;

Figure 6 shows a block diagram of the spectral compression unit used in the BTSC encoder

<img file="AR024581A2_D0029.tif" />

digital shown in Figure 3;

Figure 7 shows a flow chart used to calculate the filter coefficients of the variable accent filter used in the spectral compression unit shown in Figure 6;

Figure 8A-D shows block diagrams illustrating signal demultiplication that can be used to preserve resolution and decrease the possibility of saturation in embodiments, at fixed points, of digital BTSC encoders constructed in accordance with the invention;

Figure 9 shows a detailed block diagram of the composite modulator shown in Figures 8B-C; Y

Figure 10 shows a block diagram of a preferred embodiment of sum and difference channel processing sections that can be used in digital BTSC encoders constructed in accordance with the invention.

Detailed Description of the Drawings

Figure 3 is a block diagram of an embodiment of a digital BTSC encoder 200 constructed in accordance with the invention. The digital encoder 200 is constructed to provide a behavior that is functionally equivalent to the behavior of the idealized encoder 100 (shown in Figure 1). As with idealized encoder 100, digital encoder 200 receives

<img file="AR024581A2_D0030.tif" />

j / generates from them the conditional sum signal and the coded difference signal; however, in the digital encoder 200 these input and output signals are digitally sampled signals instead of continuous analog signals.

The choice of sampling frequency f. for the audio input signals of the left and right channel significantly affects the design of the digital encoder 200. In preferred embodiments, the sampling frequency f₃ is chosen to be an integer multiple of the pilot frequency f_H, so that f_s= Nf_H where N is an integer; and in the most preferred embodiments N is selected to be greater than or equal to three. It is important for the encoder 200 to ensure that the signal sum conditioned and the signal difference encoded do not contain enough energy in the pilot frequency f „corno to interfere with the pilot tone that is included in the composite signal. As will be discussed in greater detail below, it is then convenient, for at least some of the filters in the digital encoder 200, to provide an exceptionally high degree of attenuation at the pilot frequency f „, and this choice of the sampling frequency f_s Simplify the design of such filters.

The digital encoder 200 includes a section of

<img file="AR024581A2_D0031.tif" />

entry 210, a summation channel processing section,

220, and a difference channel processing section,

230. Instead of simply implementing the processing section of the difference channel, 230, using digital technology, the three sections 210, 220, 230 are implemented entirely using digital technology. Many of the individual components in the digital encoder 200 correspond respectively to individual components in the idealized encoder 100. In general, the components of the digital encoder 200 have been selected so that their amplitude responses closely equate to the respective amplitude responses of their corresponding components in the encoder 100. This often results in the existence of a relatively large difference between phase responses of the corresponding components.

In accordance with one aspect of the present invention, means for compensating or canceling these phase differences, or phase errors, are provided in the digitai encoder 200. As those skilled in the art will appreciate, the relatively small phase errors in the processing section of the difference channel, 230, can be compensated by introducing similar phase errors in the processing section of the sum channel, 220; and the implementation of the processing section of the sum channel through the use of digitai technology simplifies the introduction of taies

<img file="AR024581A2_D0032.tif" />

It includes two high pass filters 212, 214, and two signal additives 216, 218. The digital audio input serial I of the left channel is applied to the input of the high pass filter 212, generating the latter from it. an output signal that is applied to the positive input terminals of the additives 216, 218. The audio input signal D of the right channel is applied to the input of the high pass filter 214, which generates from it an output signal that is applied to a positive input terminal of the additive 216 and to a terminal of Negative input of additive 218. Additioner 216 generates a sum signal (indicated in Figure 3 as R&amp;D) by adding the output signals generated by filters 212 and 214. The additive 218 generates a difference signal (indicated in Figure 3 as ID) by subtracting the output signal generated by the filter 214 from the output signal generated by the filter 212. The input section 210 is therefore similar to the section of Ilo input (shown in Figure 1); however, section 210 additionally includes the two high pass filters 212, 214, and generates digital sum and difference signals.

High pass filters 212, 214 preferably have substantially identical responses and preferably

<img file="AR024581A2_D0033.tif" />

Eliminates the DC components of the audio input signals of the left and right channels. As discussed in more detail below, this DC removal prevents the encoder 200 from exhibiting a behavior called ticking sound. Since the audible information content of the audio input signals that are of interest, of the left and right channels, is considered to be within a frequency band between 50 Hz and 15,000 Hz, the removal of the components of CC does not interfere with the transmission of the information content of the audio signals. Therefore, filters 212, 214, preferably have a cut-off frequency below 50 Hz, and more preferably have a cut-off frequency below 10 Hz, so that they will not eliminate any audible information contained in the input signals of audio Filters 212, 214 also preferably have a flat magnitude response in their pass band. In a preferred embodiment, filters 212, 214, are implemented as first order infinity imprint (IIR) filters, each having a transfer function H (z) given by the formula shown in the following Equation ( one).

H (z) =

1-z '¹

1 + α, ζ '¹ (D

<img file="AR024581A2_D0034.tif" />

<img file="AR024581A2_D0035.tif" />

<img file="AR024581A2_D0036.tif" />

Referring again to Figure 3, the processing section of the sum channel, 220, receives the sum signal and generates the conditioned sum sum from it. In particular, the sum signal is applied to a pre-emphasis filter 222 of 75ps. In turn, filter 222 generates an output signal that is applied to a static phase equalization filter 228. Filter 228 generates an output signal that is applied to a low-pass filter 224 of section 220, which in turn generates the conditioned sum signal.

The pre-emphasis filter 222 of 75 ps provides a signal processing that is partially analogous to the filter 122 (shown in Figure 1) of the idealized encoder 100. The amplitude response of the filter 222 is preferably selected to closely match that of the filter 122. As will be discussed below, in the processing section of the difference channel, 230, means are preferably provided for compensation of all differences in the phase responses of filters 222 and 122. In a preferred embodiment, filter 222 is implemented as a first-order IIR filter that has a transfer function H (z) described by the formula shown in the following Equation (2).

H (z) =

1 + az '¹

<img file="AR024581A2_D0037.tif" />

(2)

<img file="AR024581A2_D0038.tif" />

The static phase equalization filter 228 executes the processing that is not directly analogous to any of the components in the idealized encoder 100 (shown in Figure 1). As will be analyzed in more detail below, the static phase equalization filter 228 is used to introduce phase errors that compensate for phase errors introduced by the processing section of the difference channel, 230. In summary, the static phase equalization filter 228 is preferably an all-pass filter having a relatively flat amplitude response and a selected phase response. In a preferred embodiment, filter 228 is implemented as a first-order IIR filter that has a transfer function H (z) described by the formula shown in the following Equation (3).

tf (z) =

Qp ^ '¹

1 + ûo²¹ (3)

The low pass filter 224 provides processing that is partially analogous to the band limiter 124 (shown in Figure 1) of the encoder 100. The low pass filter 224 preferably provides a flat amplitude response in a zero through band. 15 kHz and a relatively fast cut above 15 kHz. The filter

<img file="AR024581A2_D0039.tif" />

224 it also provides an exceptionally large degree of attenuation at the frequency f „of the pilot tone (ie 15,734 Hz). Providing this exceptionally large degree of attenuation, filter 224 ensures that the summed conditional signal does not include sufficient energy in the pilot frequency f_or as to interfere with the pilot tone used in the composite signal. As discussed above, select the sampling frequency f_s so that it is equal to an integer multiple of the pilot frequency f „simplifies the design of a filter that provides an exceptionally large degree of attenuation in the pilot frequency, and therefore simplifies the design of the filter 224. The filter 224 preferably has a value null in the pilot frequency f „and preferably provides at least 70 dB of attenuation for all frequencies, from the pilot frequency f„ and up to half of the sampling frequency.

Figure 4A is a block diagram illustrating a preferred embodiment of the low pass filter 224. As shown in Figure 4A, the filter 224 can be implemented by cascading five filter sections 310, 312,

314, 316, 318. In a preferred embodiment, each of the five filter sections 310, 312, 314, 316, 318, is implemented as a second-order IIR filter having transfer functions H (z) that are described by the

<img file="AR024581A2_D0040.tif" />

-3;

formula shown in the following Equation (4).

<img file="AR024581A2_D0041.tif" />

<img file="AR024581A2_D0042.tif" />

<img file="AR024581A2_D0043.tif" />

l + a₁z '^l+ fl₂z '² (4)

Then in the embodiment shown in Fig. 4A, filter 224 is a tenth order IIR filter.

Referring again to Figure 3, the processing section of the difference channel, 230, receives the difference signal and generates from it a coded difference signal. The difference signal is applied to a low pass filter 238a, which generates from it an output signal that is applied to a fixed preamp filter 232a. The latter generates an output signal that is applied via line 239 to an input terminal of a broadband compression unit 280, and the coded difference signal is applied via feedback line 240 to a detector terminal of the unit 280 broadband compression. The latter generates an output signal that is applied via line 281 to an input terminal of a spectral compression unit 290, and the coded difference signal is also applied via feedback line 240 to a detector terminal of unit 290. The latter generates an output signal that is applied to a fixed pre-emphasis filter 232b, which in turn generates an output signal that is applied to a trimmer 254. The trimmer

<img file="AR024581A2_D0044.tif" />

254 generates an output signal that is applied to a low pass filter 238b, which in turn generates the encoded difference signal.

The low pass filters 238a, 238b, together form a pass filter 238 that executes a processing that is partially analogous to the band limiter portion of the overmodulation protector and band limiter 138 (shown in Figure 1). of the idealized encoder 100. Preferably, the filter 238 is implemented so that it is substantially identical to the low pass filter 224, which is used in the processing section of the sum channel, 220. Therefore, all the phase errors introduced by the filter 238 in the coded difference signal are compensated by balancing the phase errors that are introduced in the sum conditioned by the filter 224. The filter 238 is preferably divided into two sections 238a, 238b , as shown, for reasons that will be analyzed in more detail below, and filter 238a preferably has a null value in the pilot frequency

Figures 4B-C are block diagrams illustrating a preferred embodiment of the respective filters 238a and 238b. As shown in Figure 4B, filter 238 a can be implemented by cascading three filter sections 310, 314, 318, which are identical to three of the

<img file="AR024581A2_D0045.tif" />

filter sections used in filter 224 (shown in Figure 4A), and as shown in Figure 4C, filter 238b can be implemented by cascading two filter sections 312, 316, which are identical to the two remaining sections used in filter 224.

The fixed pre-stress filters 232a, 232b (shown in Figure 3) together form a fixed pre-stress filter 232 that executes a processing that is partially analogous to the filter 132 (shown in Figure 1) of the idealized encoder 100 The amplitude response of the filter 232 is preferably selected to closely match the amplitude response of the filter 132. In one embodiment, the phase responses of filters 232 and 132 are significantly different and, as will be analyzed in more detail below, the resulting phase errors are compensated by filters 222 and 228 in the processing section of the sum channel, 220. Filter 232 is preferably divided into two sections 232a, 232b, as shown, for reasons that will be analyzed below. In a preferred embodiment, each of the filters 232a, 232b are implemented as first-order IIR filters that have transfer functions H (z) that are described by the formula shown in Equation (2). Then in this embodiment the filter 232 is a second order IIR filter.

<img file="AR024581A2_D0046.tif" />

-35 In a preferred embodiment, the difference between the phase responses of filters 232b and 132a closely equates to the difference between the phase responses of filters 222 and 122. Accordingly, the phase error introduced in the encoded difference signal by the fixed pre-emphasis filter 232b it is balanced by the phase error introduced in the sum signal conditioned by the pre-accent filter 222 of 75 με. In addition, in this embodiment, the phase response of the static phase equalization filter 228 is selected to closely match the difference between the phase responses of the fixed pre-emphasis filter 232a and 132b, so that any phase error introduced in the difference signal encoded by the filter 232a, it is balanced by a compensatory phase error in the conditioned sum signal that is introduced by the static phase equalization filter 228.

The trimmer 254 executes a processing that is partially analogous to the overmodulation protection portion of the overmodulation protector and band limiter 138 (shown in Figure 1) used in the idealized encoder 100. In summary, the trimmer 254 is implemented as a threshold formation device; however, the operation of trimmer 254 will be analyzed in more detail below.

The 280 broadband compression unit and the unit

<img file="AR024581A2_D0047.tif" />

Spectral compression 290 execute processing functions that are partially analogous to those of units 180 and 190, respectively, of the idealized encoder 100 (shown in Figure 1). In summary, the broadband compression unit 280 dynamically compresses the signal in line 239 as a function of the overall energy level in the coded difference signal, and the spectral compression unit 290 also compresses high frequency portions of the signal in the line 281 as a function of high frequency energy in the signal coded difference.

Figure 5 shows a block diagram of a preferred embodiment of a digital broadband compression unit 280. Unit 280 includes a digital signal multiplier 434, a digital s multiplier; efiales 446, a broadband digital bandpass filter 448, an effective level 450 digital detector, and a 458 digital reciprocal generator. These components perform processing functions partially analogous to those performed by amplifier 134, amplifier 146, bandpass filter 148, effective level detector 150, and reciprocal generator 152, respectively, of the idealized encoder 100 (shown in Figure 1). The coded signal difference is applied through feedback path 240 to an input of the digital bandpass filter

<img file="AR024581A2_D0048.tif" />

<img file="AR024581A2_D0049.tif" />

<img file="AR024581A2_D0050.tif" />

/ broadband 448, which generates from it an output signal that is applied to the effective level detector 450. The latter generates an input signal that is representative of the effective value of the output signal generated by the filter. 448, and apply this output signal via line 450a to reciprocal generator 458. The reciprocal generator 458 then generates a representative output signal from the reciprocal of the signal on line 450a and applies this output signal via line 458a to multiplier 446. The digital signal multiplier 446 multiplies the signal on lines 458a by the determined gain value, Gain D, thereby generating an output signal that is representative of D times the reciprocal of the effective value and that is applied via line 446a to a multiplier input terminal 434. The output signal generated by the fixed pre-emphasis filter 232a is applied via line 239 to another multiplier input terminal 434. The multiplier 434 multiplies the signal on line 239 by the signal on line 446a, thereby generating the output of the broadband compression unit 280, which is applied via line 281 to the input of the spectral compression unit. 290.

The 448 broadband digital bandpass filter is designed to have an amplitude response that closely matches the bandwidth filter amplitude response

<img file="AR024581A2_D0051.tif" />

148 (shown in Figure 1). A preferred choice is to select filter 448 so that the mean square difference between its amplitude response and that of filter 148 is minimized. In one embodiment, the phase responses of filters 448 and 148 are substantially different, but since the output signal of the effective level detector 450 is substantially insensitive to the phase of its input signal, these phase differences can be ignored. . In a preferred embodiment, the broadband bandpass filter 448 is implemented as a second-order IIR filter that has a transfer function H (z) that is described in the formula shown in Equation (4).

The effective level detector 450 is designed to approximate the performance of the detector 150, which is used in the idealized encoder 100 (shown in Figure

one). The detector 450 includes a device for raising signals to square 452, a device for averaging signals 454, and a device for extracting square roots 456. The device for raising to square 452 squares the signal generated by the bandpass filter 448 and apply this signal squared via line 452a to the averaging device 454. The latter calculates a weighted average time of the signal on line 452a and applies the average via line 454a to the square root extraction device 456. The device

<img file="AR024581A2_D0052.tif" />

output generated by the 448 broadband digitai bandpass filter.

The averaging device 454 includes a digital signal multiplier 460, a digital signal additive 462, a digital signal multiplier 464, and a delay recorder 465. The output signal generated by the device for squareing 452 is applied via line 452a at an output of multiplier 460, which generates an output serial by multiplying the signal on line 452a by a constant a. The demultiplicated output signal generated by the multiplier 460 is applied to one input of the additive 462 and an output serial generated by the delay recorder 465 is applied to the other input of the additive 462. The additer 462 generates an input serial by adding the signals present in its two inputs, and this summed signal is the output signal of the averaging device 454, and is applied to the square root extraction device via line 454a. This summed signal is also applied to an input of multiplier 464, which generates an input signal by multiplying the signal summed by the constant (1-a). The output signal generated by multiplier 464 is applied to an input of the recorder

<img file="AR024581A2_D0053.tif" />

of delays 465. Those skilled in the art will appreciate that the averaging 454 is a recursive filter and implements a digital averaging function that is described by the recursion formula shown in the following Equation (5). , y (n) = ax (n) + (la) and (nl) (5) in which y (n) represents the digital sample in use of the signal coming from the pro-mediator 454 to line 454a, and (nl ) represents the previous digital sample of the signal leaving the averaging 454 to line 454a, and x (n) represents the digital sample in use of the signal leaving the device to raise the square 452 to the line 452a. Those skilled in the art will appreciate that the averager 454 provides a digital approximation of the analog averaging function defined in the BTSC standard and supplemented by the effective level detector 150 (shown in Figure 1) of the idealized encoder 100. The constant to. It is preferably chosen so that the time constant of the effective level detector 450 closely approximates the corresponding time constant specified in the BTSC standard for the effective level detector 150.

The device for the extraction of square roots

456 and the 458 digital reciprocal generator are shown in the

Figure 5 as two separate components; nevertheless, fcüado I

<img file="AR024581A2_D0054.tif" />

Those skilled in the art will appreciate that these two components can be implemented using a single device that generates a representative output signal from the reciprocal of the square root of its input signal. Such a device can be implemented for example as a search table stored in memory (LUT), or alternatively implemented using processing components that calculate a polynomial approximation by Taylor series of the inverse square root function.

Figure 6 shows a block diagram of a preferred embodiment of the spectral compression unit 290. Unit 290 includes a variable pre-emphasis / de-emphasis unit (hereinafter referred to as variable accent unit) 536, a signal multiplier 540, a spectral bandpass filter 542, and an effective level detector 544, and these components provide a processing that is partially analogous to that of the variable accent filter 136, the amplifier 140, the bandpass filter 142, and the effective level detector 144, respectively, of the idealized encoder 100 (shown in Figure 1). The encoded difference signal is applied via feedback line 240 to an input of signal multiplier 540, which generates an output signal multiplying the signal encoded by the determined fixed gain value of Gain C. The signal

<img file="AR024581A2_D0055.tif" />

of amplified output generated by signal multiplier 540 is applied to spectral bandpass filter 542, which generates an output signal that is applied to the effective level detector 544. The latter generates an output signal that is applied via line 544a to a control terminal of the variable accent unit 536, and the output signal generated by the broadband compression unit 280 is applied via line 281 to an input terminal of the unit 536. The latter dynamically varies the frequency response applied to the signal on line 281 according to a function of the signal on line 544a, the latter being a function of the signal energy of the signal difference encoded within the last frequency band through the spectral bandpass filter 542. In this way, the output signal of the unit 290, which is generated by the unit 536 and is applied to the input of the fixed pre-emphasis filter 232b, is dynamically compressed in a larger amount in the high frequency portions of the signal than in the rest of the spectrum of interest.

The spectral bandpass filter 542 is designed to have an amplitude response that closely matches the amplitude response of the bandpass filter 142 (shown in Figure 1) of the idealized encoder 100. As with filter 448 (which is shown in Figure 5), a preferred choice is to select filter 542 so that

<img file="AR024581A2_D0056.tif" />

The difference between your effective amplitude response and 'the' of filter 142 is minimized. In one embodiment, the phase responses of filters 542 and 142 are substantially different, but since the effective output of the effective level detector 544 is substantially insensitive to the phase of the input to the detector, these phase differences can be ignored. In a preferred embodiment, the spectral bandpass filter 542 is implemented as a cascade of three second order IIR filter sections 542a, 542b, 542c (as shown in Figure 6) each having a transfer function H (z ) which is described in the formula shown in Equation (4).

The effective level detector 544 is designed to approximate the performance of the detector 144, which is used in the idealized encoder 100 (shown in Figure

one). The detector 544 includes a device for raising signals to square 552, a device for averaging signals 554, and a device for extracting square roots 556. The device for raising square 552 squares the signal generated by the spectral bandpass filter 542 and apply this signal squared to the averaging device 554 via line 552a. The latter works similarly to the averaging device 454 (shown in Figure 5), which is used in the broadband compression unit 280,

<img file="AR024581A2_D0057.tif" />

different from the constant a. The behavior of the averaging device 554 is also described of course by Equation (5) when ß replaces a. The constant β is preferably selected for the device 554 so that the time constant of the effective level detector 544 closely approximates the corresponding time constant specified by the BTSC standard for the effective level detector 144 (shown in Figure one). The averaging device 554 calculates a weighted average time of the signal on line 552a and applies the average to the square root extraction device 556 via line 554a. The square root extraction device 556 calculates the square root of the signal on line 554a and then generates a signal on line 544a as a function of the effective value of the output signal generated by the spectral bandpass filter 542.

The signal on line 544a is applied to the control terminal of the variable accentuation unit 536. The variable accentuation unit 536 performs a processing that is partially analogous to the filter 136 (shown in Figure 1) of the idealized encoder 100. As defined by the BTSC standard, filter 136 has amplitude and phase responses that vary depending on the output signal generated by the effective level detector 144. A preferred way of •

<img file="AR024581A2_D0058.tif" />

Similar variable is to use a digital filter that has variable coefficients that determine its transfer function, and select the value of the coefficient during any given sampling period or group of sampling periods, based on the value of the signal on line 544a.

Figure 6 shows an embodiment of the variable accentuation unit 536, which includes a logarithmic generator 558, a variable accentuation filter 560, and a search table LUT 562. The output signal generated by the effective level detector 544 is applied via line 544a to the logarithmic generator 558. The latter generates a serial in line 558a that is representative of the logarithm of the signal in line 544a, and applies this serial to LUT 562. The LUT 562 generates an output signal selected from the LUT and representative of the filter coefficients that will be used by the variable accent filter 560. The coefficients thus generated by the LUT 562 are applied via line 562a to a selection terminal of coefficients of the variable accent filter 560. The output signal generated by the broadband compression unit 280 is applied to an input terminal of the variable accent filter 560 via line 281. The variable accent filter 560 generates the output serial of the spectral compression unit 290, which is applied to the input of the filter of

<img file="AR024581A2_D0059.tif" />

fixed pre-emphasis 232b.

Variable accent filter 560 is designed to have a variable amplitude response that closely matches the variable amplitude response of filter 136 (shown in Figure 1) of the idealized encoder 100. The variable accent filter 560 provides a similar variable response using a variable coefficient transfer function (ie the coefficients of the transfer function H (z) of the filter 560 are variable), and allowing the LUT 562 to select the value of the coefficients during intervals based on the sampling period. As will be described in greater detail below, the LUT 562 stores the values ​​of the filter coefficients used by the filter 560, and during each sampling period or during each selected group of sampling periods, the LUT selects a set of coefficients of filter as a function of the output signal generated by the logarithmic generator 558 on line 558a. In a preferred embodiment, the variable accent filter 560 is implemented as a first-order IIR filter that has a transfer function H (z) that is described by the formula shown in the following Equation (6)

H (z) = b ^ by¹1 + a ^ z ¹ (6)

r. '

<img file="AR024581A2_D0060.tif" />

in which the filter coefficients b_or/ b_x already_x they are variables that are selected by LUT 562. The methods for selecting the values ​​for the filter coefficients used by the filter 560, as well as the other filters of the encoder 200, will be analyzed below.

In Figure 6, the logarithmic generator 558 and the square root extraction device 556 are shown, for convenience, as two separate components.

However, those skilled in the art will appreciate that these two components can be implemented using a single device, such as a LUT, or alternatively using processing components that calculate a Taylor polynomial approximation of the logarithm of the signal on line 554a , and then dividing this value by two. Similarly, in alternative embodiments, the functions performed by the logarithmic generator 558, the square root extraction device 556, and the LUT 562, can be incorporated into a single device.

As stated before, the high-pass filters 212, 214 (shown in Figure 3) are useful for blocking the DC components so as to prevent the encoder 200 from exhibiting a known behavior like ticking sound. In the context of a stereo encoder, ticking refers to a behavior

<img file="AR024581A2_D0061.tif" />

<img file="AR024581A2_D0062.tif" />

relatively low frequency oscillator that the encoder presents, caused when there is no signal present in the audio inputs of the left and right channel. The desired behavior of a stereo system when there is no signal present in the audio inputs is to remain silent; However, an encoder connected to loudspeakers through a decoder that exhibits ticking sound makes the loudspeakers emit an audible sound, called ticking, with a somewhat regular period that is partially dependent on the time constant of the effective level detector. in the broadband compressor. More particularly, in the encoder 200, when only very low level signals are present in the audio inputs, and when there is a DC component or a drop in the signal 239, the broadband compression unit 280 It tends to behave in an unstable way that causes ticking.

Consider the case where only a low level audio signal is present on line 239. In this case, the output of the effective level detector 450 on line 450a becomes very small, which in turn causes the gain of multiplier 434 to become very large. If one such low-level audio signal on line 239 is constant in its amplitude, the broadband compression unit 280 reaches a permanent regime after some time.

<img file="AR024581A2_D0063.tif" />

(determined by the time constant a applied to multiplier 460), because the coded difference signal is applied back on line 240 to the broadband compression unit 280. Because the feedback is arranged to be negative, when the Audio signal on line 239 increases in its amplitude, the signal on line 450a increases, which in turn causes the gain of multiplier 434 to decrease. When the audio signal on line 239 decreases in its amplitude, the signal on line 450a decreases, which in turn causes the gain of multiplier 434 to increase.

However, if there is a significant DC signal on line 239 in addition to a low-level audio signal, the DC signal is blocked from the feedback process by the action of the broadband bandpass filter 448, which has a response zero to CC signals. In particular, any CC present in the signal difference encoded in line 240 is blocked by filter 448, and is not detected by effective level detector 450. Any DC signal present on line 239 will be amplified by multiplier 434 together with any audio signal present on line 239, but the amplification factor or gain will be determined only by the amplitude of the audio signal as detected by the 450 effective level detector after filtering by the filter

<img file="AR024581A2_D0064.tif" />

As indicated above, as long as the amplitude of the audio signal on line 239 varies, the gain of multiplier 434 varies inversely. During such gain variations, any CC present on line 239 will also be subject to variable amplification, effectively modulating the CC signal and then producing an AC signal. In this way, such DC signals can be modulated so as to create significant audio frequency band signals that will not be rejected by the filter 448, and are therefore detected by the detector 450. When the audio signal on line 239 is small compared to the DC on line 239, small variations in the level of the audio signal, which cause changes in the gain of amplifier 434, can cause a large change in the level. of the CC (which amounts to an AC signal) on line 281 through this modulation process. The produced AC signal tends to increase the total signal that passes through filter 448, regardless of whether the variation of the audio signal that gave rise to the AC signal was an increase or decrease in the signal level. In particular, if the level of the audio signal on line 239 decreases, the negative feedback process usually increases the gain of multiplier 434. However, if a sufficient DC signal is present on line 239,

<img file="AR024581A2_D0065.tif" />

A decrease in the audio signal on line 239 may cause an increase in the signal detected by detector 450, forcing the gain of multiplier 434 to increase.

In this way, the negative feedback process is reversed, and the feedback becomes positive.

Such positive feedback persists only as long as the DC signal modulated on line 281 is large enough compared to any audio signal present on line 281, when measured by means of the response of all filters and signal modifiers between line 281 and filter output 448. Once the gain of multiplier 434 decreases sufficiently so that the DC signal modulated on line 281 no longer provides a significant input to detector 450, the feedback is reversed to its normal negative direction. In accordance with the time constant of the detector 450, the system will again acquire an appropriate gain level based on the level of the audio signal on line 239. But if there is enough CC left in the signal on line 239, the cycle will be repeated once the gain of multiplier 434 is increased sufficiently. During each such positive feedback period, there is an acute change in the DC level of line 281. This change is audible, and the sounds are somewhat similar to the ticking of a clock. Since such changes of CC will occur with some regularity, in

<img file="AR024581A2_D0066.tif" />

Based on the time constant of detector 450, the phenomenon is often referred to as ticking sound.

One method to prevent the ticking sound is to eliminate all DC components present in the input signal to encoder 200. This is done by high pass filters 212 and 214. In addition, high pass filters 212 and 214 help to Maximize the dynamic range of the encoder 200 by eliminating DC components that could otherwise consume valuable dynamic range. As said before and as shown in Figure 3, the low pass filter 238 is preferably implemented as two filters 238a and 238b. Splitting the filter 238 in this way offers several advantages. If filter 238a were removed, and the entire filter 238 were located after trimmer 254 (i.e. at the location of filter 238b), then all components above 15 kHz in the audio input signals may cause instability in the broadband compression unit 280, similar to the ticking behavior described above. This occurs because all serial components above 15 kHz on line 239 will be amplified by multiplier 434 (shown in Figure 5) and because such components will not be detected by the effective level detector 450 since such components they are filtered and removed by the low pass filter that follows trimmer 254 (shown in Figure 3). Since the detector

<img file="AR024581A2_D0067.tif" />

450 the multiplier gain 434 decreases when it detects the absence of a signal, the gain of the multiplier 434 can be made relatively large when the signal on line 239 is composed of little audio signal information (below 15 kHz), but significant information high frequency (above 15 kHz). The multiplier 434 then amplifies the high frequency information, which can generate large signals that are likely to be clipped by the components of the processing section 230. This cut can produce harmonics that can be diverted at low frequencies that will be detected by the detector. effective level 450 causing the system to emit ticking sound as previously described. Alternatively, if the filter 238b were removed and the entire filter 238 were located before the fixed pre-emphasis filter 232a (i.e. at the location of the filter 238a), then the high frequency artifacts generated by the trimmer 254 would be included in the signal. coded difference and could interfere with the pilot tone in the composite signal. Consequently, dividing the filter 238 as shown showed an optimal configuration whereby filter 238a prevents the ticking sound in the compression unit 280 and the filter 238b filters the high frequency artifacts that can be generated by the trimmer 254.

<img file="AR024581A2_D0068.tif" />

The fixed pre-emphasis filter 232 is preferably also divided into two filters 232a, 232b, as * x is shown in Figure 3. Filter 232 typically requires high gain at high frequencies, as specified in the BTSC standard, and use only one single section to implement filter 232 increases the probability that filter 232 causes clipping. It is advantageous to apply some of the gain of the filter 232 on the input side of the broadband compression unit 280 (with the filter 232a), and apply some of the gain of the filter 232 on the output side of the compression unit broadband 280 (with filter 232b). Since unit 280 normally compresses its input signal, distributing the gain of filter 232 throughout the compression provided by unit 280 decreases the probability that the gain of filter 232 causes an overflow regime.

To minimize size, energy consumption and cost, the encoder 200 is preferably implemented using a single digital signal processing chip. The encoder 200 has been successfully implemented using one of the well-known Motorola DSP 56002 digital signal processing chips (this embodiment will be referred to hereafter as "DSP realization"). The Motorola 56002 DSP is a twenty-four bit fixed point chip; however, other types of chips of course can of course be used

<img file="AR024581A2_D0069.tif" />

processing, such as fixed point chips or fixed point chips that have other word lengths. The DSP embodiment of the encoder 200 uses a sampling frequency f_s which is equal to three times the pilot frequency F „(ie f₃ = 47202 Hz). The following Table 1 lists all the filter coefficients used in the DSP Embodiment of the encoder 200, except those used in the variable accent filter 560.

TABLE 1

<td>Low Pass Filter (Section # 1) 310 (Equation 4)</td><td>Low Pass Filter (Section # 2) 312 (Equation 4)</td>

<td>b_or = 0,18783270</td><td>b_or = 0,44892888</td>

<td>b, = 0.36310206</td><td>b_x = 0,70268024</td>

<td>b₂ = 0,18783270</td><td>b₂ = 0,44892888</td>

<td>to_x = 0,388832539</td><td>a, = 0.12638618</td>

<td>to₂ = 0,12709286</td><td>to₂ = 0,47415181</td>

<td>Low Pass Filter (Section # 3) 314 (Equation 4)</td><td>Low Pass Filter (Section # 4) 316 (Equation 4)</td>

<td>b₀ = 0,70674027</td><td>b₀ = 0,85733126</td>

<td>b₂ = 0,87637648</td><td>b_x = 0,91505047</td>

<td>b₂ = 0,70674027</td><td>b₂ = 0,85733126</td>

<td>to₂ = 0,53702472</td><td>ai = 0.74320197</td>

<td>to₂ = 0,75298490</td><td>to₂ = 0,89832289</td>

<td>Low Pass Filter (Section # 5) 318 (Equation 4)</td><td>Broadband Bandpass Filter 448 (Equation 4)</td>

<td>b_or = 0,92737972</td><td>b₀ = 0,02854672</td>

<img file="AR024581A2_D0070.tif" />

<td>b, = 0.92729649</td><td>b_x = 0,18789051</td>

<td>φ₂ = 0,92737972</td><td>b₂ = 0,21643723</td>

<td>to_x = 0,82951974</td><td>a, = 1,75073141</td>

<td>to₂ = 0,97259237</td><td>to₂ = 0,75188028</td>

<td>Precentration Filter Fixed 238a (Equation 2)</td><td>Precentration Filter Fixed 238b (Equation 2)</td>

<td>b_or = 9,50682180</td><td>b_or = 4,357528</td>

<td>b_x = 9,00385663</td><td>b_x = 3,24843271</td>

<td>a, = 0.497064357</td><td>to_x = 0,10881833</td>

<td>Spectral Bandpass Filter (Section # 1) 542a (Equation 4)</td><td>Spectral Bandpass Filter (Section # 2) 542b (Equation 4)</td>

<td>b₀ = 0,646517841</td><td>b_or = 0,850281278</td>

<td>b_x = 0,649137616</td><td>b, = 0.850247036</td>

<td>b₂ = 0,0</td><td>b₂ = 0,0</td>

<td>to₁ = 0,557821757</td><td>a, = 0.602159890</td>

<td>to₂ = 0,0</td><td>to₂ = 0,0</td>

<td>Spectral Bandpass Filter (Section # 3) 542c (Equation 4)</td><td>Static Filter Phase 224 Match (Equation 3)</td>

<td>b₀ = 0,597678418</td><td>to₀ = 0,9029</td>

<td>b_x = 1,195357770</td><td></td>

<td>b₂ = 0,597679348</td><td></td>

<td>a, = 0.776566094</td><td></td>

<td>to₂ = 0,352824276</td><td></td>

<td>Precentration Filter 222 of 75ps (Equation 2)</td><td>High Pass Filters 212, 214 (Equation 1)</td>

<td>b_or = 4,57030583</td><td>to_x = 0,999</td>

<td>b_x = 3,43823487</td><td></td>

<td>a, = 0.131778883</td><td></td>

<img file="AR024581A2_D0071.tif" />

-57 In the DSP embodiment of the encoder 200, the constant 1 valor value that is used by the averager 454 (shown in Figure 5) in the broadband compression unit 280 is set equal to 0.0006093973517, and the The value of the constant ß that is used by the averaging 554 (shown in Figure 6) in the spectral compression unit 290 is set equal to 0.001825967. In addition, the Gain C and Gain D values ​​used by the amplifiers 540 and 446 respectively, in the spectral and broadband compression units, are set equal to 0.5011872 and 0.08984625 respectively, to ensure that the DSP Realization of the encoder 200 perform similarly to the encoder

100.

Figure 7 shows a flow chart 700 describing a preferred method for precalculating all sets of filter coefficients used by the variable accent filter 560 (shown in Figure 6) in the DSP Embodiment of encoder 200. Prior to the operation of the encoder 200, all sets of filter coefficients used by the filter 560 are pre-calculated (for example by means of a general purpose digital computer) and loaded into the LUT 562. In the DSP Embodiment of the encoder 200, filter 560 has a transfer function H (z) that is described by the

<img file="AR024581A2_D0072.tif" />

-58 Equation (6), so that flow chart 700 describes the calculation of the coefficients b_0/ b_x already_x. As specified in the BTSC standard, the transfer function of S (f, b) of analog filter 136 (shown in Figure 1) to which filter 560 partially corresponds, is described by the formula shown in the following Equation (7).

,&gt;*⁵¹⁾

S (f, bX —-ff⁺¹⁾ - (7) ® W51Z »in which F is equal to 20.1 kHz.

The first step in flow chart 700 is an initialization step 710 during which several variables are initialized. Specifically, the sampling frequency f_to is set equal to 47202 Hz, and the period T is set equal to / f_to. The variable W is a digital version of the variable F used in Equation (7) and is set equal to π (20.1 kHz) / f_s. The MARGEN dB variable represents the desired signal range of the effective level detectors in the spectral compression unit, and for the DSP MARGEN dB (dbRANGE) realization it is set equal to 72.25 dB. The variable RES dB (dbRES) refers to the sensitivity of filter 560 at

<img file="AR024581A2_D0073.tif" />

changes in the energy level of the code difference signal In the DSP Embodiment of encoder 200, RES dB is set equal to 0.094 dB so that filter 560 will use coefficients based on the signal value on line 558a quantified to the nearest 0.094 dB. The variable N equals the total number of sets of filter coefficients used in filter 560 and N is calculated by dividing the sensitivity (RES dB) by the margin (MARGIN dB) and rounding to the nearest integer. In the DSP Realization, N is equal to 768, although those skilled in the art will appreciate that this number can be changed, which will vary the sensitivity or the margin. In the DSP Embodiment, LUT 562 stores 769 coefficient values ​​for filter 560, and of course if N is increased, a larger LUT will be used to store surplus values ​​of filter coefficients. In addition, those understood in the art will appreciate that the logarithmic generator 558 demultiplies the signal on line 558a and then reduces the number of values ​​of filter coefficients stored by LUT 562, for a given minimum quantification of the value of the signal on the line 558a. However, in other embodiments, logarithmic generator 558 can be removed and LUT 562 can store a correspondingly larger number of sets of coefficients of

<img file="AR024581A2_D0074.tif" />

filter. Finally, the variables Scale and Direction are set equal to 32 and zero, respectively. The variable

s.

Scale, which is used only in fixed point embodiments, is selected so that all filter coefficients have a value greater than or equal to less than one and less than one (where filter coefficients are represented in complement two).

Following the initialization step 710, a coefficient generation step, 720 is executed. During the first execution of step 720, the variables b are calculated_or(0), b_x(0) already_x(0) corresponding to values ​​of the coefficients b₀, bj ya_x which will be stored in the location of the zero address of LUT 562. Following this execution of step 720, an incremental step 730 is executed during which the value of the address variable is increased. Following step 730, a comparison step is performed during which the values ​​of the Address and N variables are compared. If Address is less than or equal to N, then steps 720, 730 and 740 are again executed iteratively so that the values ​​of the coefficients b_0/ b_x already_x be calculated for each of the 769 addresses of LUT 562. When step 740 detects that the address value is greater than N, then the 769 coefficient values ​​have been calculated, and the execution of flow chart 700

<img file="AR024581A2_D0075.tif" />

proceed to a final step 750.

In the coefficient generation step, 720, the variable FS dB (dbFS) corresponds to the output of the logarithmic generator 558. As the value of the address variable varies from zero to 769, the value of FS dB varies from approximately 72.25 at zero dB, corresponding to the signal range of approximately 72.25 dB provided by the DSP realization of encoder 200 (where zero dB corresponds to full modulation). The RMSd variable corresponds to the output of the analogue effective level detector 144 (shown in Figure 1), and as the value of the Direction variable varies from zero to 769, the RMSd value varies from approximately -36 to 36 dB , corresponding to the 72 dB signal range provided by the analog BTSC encoders typical of the prior art. The RMSb variable is a line version of the RMSd variable, and RMSb corresponds to the variable b in the transfer function S (f, b) described in Equation (7). The variables Kl and K2 correspond respectively to the terms (b + 51) / (b + l) and (51b + l) / (b + l) in Equation (7). The coefficients b₀b_x already_x they are calculated as shown in step 720 using the variables Kl, K2, W and Scale.

Figure 8A shows a block diagram illustrating a method for using DSP Realization in a system

<img file="AR024581A2_D0076.tif" />

analog, and in Figure 8A all the components that are implemented in the integrated circuit 56002 are indicated in 200a. The analog system provides analog audio input signals of the left and right channel (shown in Figure 8A as I and D respectively) and these signals are applied to the inputs of the six-bit 810 and 812 analog-digital converters, respectively. Converters 810, 812 sample their analog input signals using a sampling frequency f₃ which is equal to 47.202 Hz (ie 3f_H) and converters 810, 812 thereby generate sequences of sixteen bit digital samples that are representative of the audio input signals of the left and right channel, respectively. The signals generated by converters 810 and 812 are applied to encoder 200a where they are received by modules 292 and 294, respectively. The modules 292, 294 are modules divided by sixteen (which divide the amplitude of their inputs by a factor 16) and therefore generate output signals that are equal to their input signals divided by sixteen. Since the division by any power of two is easily performed in a digital system using a shift register, modules 292, 294 are implemented as displacement records that displace

<img file="AR024581A2_D0077.tif" />

As stated before, the 56002 chip is a twenty-four bit fixed point processor, and the samples applied to the chip by the 810, 812 converters have a representation of the two complement type. Modules 292, 294 divide by sixteen the samples generated by converters 810, 812 and thereby place each of the samples in the middle of a twenty-four word word. Then in all the samples generated by modules 292, 294, the four most significant bits are sign bits, and the least significant four bits are zero, and the sixteen bits in the middle of the word correspond to a sample generated by one of 810,812 converters. Filling in this way each twenty-four word word with sign bits at the upper end and with leading zeros preserves the ëxactitude and allows the intermediate signals generated by the encoder 200a to exceed sixteen bits without causing a state of error such as an overflow.

In encoder 200a, each bit of the twenty-four word word corresponds to approximately 6 dB of signal range, and therefore modules 292, 294 correspond to attenuators of -24 dB (ie -6 times 4). If the analog input audio signals applied to the

<img file="AR024581A2_D0078.tif" />

810, 812 converters are considered as zero dB signals for reference purposes, then the signals generated by x modules 292, 294 are attenuated by 24 dB.

The input section 210 receives the twenty-four bit words generated by the modules 292, 294 and generates from them the sum signal that is applied to the processing section of the sum channel, 220. The output signal generated by the section The processing of the sum channel, 220, is applied to a 16 ”296 multiplication module (which can be considered as a 24 dB amplifier). The module 296 then compensates for the attenuators 292, 294 of 24 dB and carries the output of the processing section of the sum channel, 220, back to 100% modulation (ie back to full scale). The output signal generated by module 296 is applied to a six-bit digital analog converter 814, which in turn generates an analog sum sum signal.

The input section 210 also generates the difference signal that is applied to the processing section of the difference channel, 230. As stated above, as a result of the modules 292, 294, it can be considered that the difference signal is attenuated by 24 dB. In the DSP embodiment of the encoder 200a, the trimmer 254 (shown in Figure 3) of the channel processing section

<img file="AR024581A2_D0079.tif" />

difference, 230, includes an 18 dB amplifier (which is implemented as a multiplication by eight). That is, the * x trimmer 254 amplifies by 18 dB the signal generated by the fixed pre-emphasis filter 232b, and then cuts this amplified signal so that the output signal generated by the trimmer 254 does not exceed a number that is 6 dB per under full modulation. The signal applied from trimmer 254 to the low pass filter 238b therefore has a bit (or 6 dB) of free space, so filter 238b can generate an output serial that is 6 dB larger than its input signal without causing saturation. It is convenient to leave this bit of free space because the transient response of the filter 238b includes some transient oscillation that may cause it to temporarily generate an instant output serial that is larger than its instantaneous input signal, and then the free space prevents any transient oscillation in filter 238b cause a saturation state. Referring again to Figure 8A, the output signal generated by filter 238b is applied to a six-bit digital analog converter 816, which in turn generates an output signal that is applied to a 6 dB analog amplifier, 820. Both D / A converters 814 and 816 are intended to be complete converters, which include as part of their

<img file="AR024581A2_D0080.tif" />

functionality the well known analog filters _xantiimage The anti-image filters are analog filters applied to the analog signal that follows the digital-analog conversion, which serve to attenuate any image of the desired signal that is reflected in the entire sampling frequency and multiples thereof. It is assumed that converters 814 and 816 are substantially identical to each other, operating at the same sampling frequency and containing substantially the same anti-image filtering. Such converters can commonly be obtained in commercial embodiments such as the Crystal Semiconductor CS4328. The 820 amplifier amplifies its input serial by 6 dB, which brings the coded difference signal back to full scale. Although Figure 8A shows the encoder 200a coupled to the analog-digital converters 810, 812 to receive analog audio signals, in digital systems the converters 810, 812 can of course be eliminated so that the encoder 200a receives the digital audio signals directly.

Figure 8B shows a block diagram of a preferred embodiment of a BTSC 200b encoder constructed in accordance with the invention and configured as part of an analog system. The encoder 200b is similar to

<img file="AR024581A2_D0081.tif" />

module 296 amplifies its input signal by 18 dB * x (multiplying by 8) instead of 24 dB as in encoder 200a. The output signal generated by module 296 is a demultiplicated version of the conditional sum signal and is shown in Figure 8B as S. Also, encoder 200b includes a module 298 to amplify the signal by 6 dB (multiplying by two) output generated by the processing section of the difference channel, 230. The output signal generated by module 298 is a demultiplicated version of the coded difference signal and is shown in Figure 8B as D. In addition, encoder 200b includes a composite modulator 822 to receive signals S and D and to generate a from them a digital version of the composite signal. The digital composite signal generated by the modulator 822 is applied to a digital analog converter 818 whose output is an analogue version of the composite signal. The D / A 818 converter is designed to be a complete converter that includes as part of its functionality the analogue anti-image filter mentioned above. Such converters can be obtained commonly in commercial embodiments such as the Burr-Brown PCM1710. In preferred embodiments, modules 292, 294, input section 210, the channel processing section

<img file="AR024581A2_D0082.tif" />

In addition, 220, the difference channel processing section, 230, modules 296, 298, and composite modulator 822, are all implemented in a single digital signal processing chip.

Since the composite signal is generated as a digital signal in encoder 200b, module 298 is included to scale up the output signal generated by the processing section of the difference channel, 230, instead of waiting until after digital analog conversion, and using an analog amplifier such as the amplifier 820 as shown in Figure 8A. Also, since in the composite signal the conditioned sum signal is used at 50% of the modulation, module 296 only amplifies its input signal by 18 dB so that the output signal generated by module 296 is halfway through. the amplitude of the output signal generated by module 298.

Figure 9 shows a block diagram of an embodiment of the composite modulator 822. The latter receives signals S and D and generates from them a digital version of the composite signal. The modulator 822 includes two interpolators 910, 912, two digital low-pass filters 914, 916, a digital signal multiplier 918, and two digital signal additives 920, 922. The signals S and D are applied to the respective inputs of the ..

<img file="AR024581A2_D0083.tif" />

interpolators 910 and 912. Interpolators 910, 912, which are alternatively referred to as x circuits.

ascending sampling, interpolate a new sample between every two consecutive samples applied to their inputs, thereby generating output signals that have twice the sampling frequency of signals S and D. The output signals generated by interpolators 910 and 912 they are applied to the respective inputs of the low pass filters 914 and 916. The latter eliminate the images introduced in the S and D signals by interpolators 910, 912. The filtered output signal generated by the filter 916 is applied to an input of the signal multiplier 918, and a digital oscillation signal as a function of cos [4% (f_H/F₃) n] is applied to the other input of the multiplier 918. The multiplier 918 thus generates the version of the difference signal (with carrier suppression, double sideband and amplitude modulated) that is used in the composite signal. The output signal generated by the multiplier 918 is applied to an input of the signal aggregator 920, and the filtered output signal generated by the filter 914 is applied to the other input of the signal aggregator 920. The latter generates an output signal by adding the two signals present in its inputs and applies this signal to the signal additive 920. A pilot tone signal that

<img file="AR024581A2_D0084.tif" />

oscillates as a function of Acos [2% (f_H/F₃) n] (where A is a representative constant of 10% of the modulation at * x scale) is applied to the other input of the signal aggregator 922, which generates the composite digital signal by adding the two signals present at its inputs.

Composite modulator 822 includes interpolators 910, 912 because the highest frequency component in the composite signal is slightly less than 3f „(as shown in Figure 2), and therefore the signals applied to the signal multiplier inputs 918 and the signal additive 920 should have sampling frequencies of at least 6f „to meet the Nyquist criteria. Because the sampling frequency at the output of the composite modulator 822 is typically higher than the sampling frequency of the S or D signals, the D / A converter 818 must be able to operate at such high sampling frequencies. If the input signals S and D applied to the composite modulator 822 have sampling frequencies of 3f „, some form of interpolation (as provided by interpolators 910, 912) should be provided to double the sampling frequency. Of course, if sufficiently high sampling rates are used throughout the encoder 200b, then interpolators 910,

912 and the low pass filters 914, 916 can be removed

<img file="AR024581A2_D0085.tif" />

of the 822 modulator.

Figure 8C shows a block diagram of another embodiment of a BTSC 200c encoder constructed in accordance with the invention. The encoder 200c is similar to the encoder 200b (shown in Figure 8B); however, in the encoder 200c the module 298 is eliminated so that the serial generated by the processing section of the difference channel, 230, is the signal D and is applied directly to the composite modulator 822. In addition, in the encoder 200c the module 296 amplifies its input signal by 12 dB (multiplying by 4) instead of 18 dB as it is done in the encoder 200b. Then in the encoder 200c the signals S and D are 6 dB below the levels of those signals in the encoder 200b. The composite modulator 822 therefore generates from these signals a version of the composite signal that is attenuated by 6 dB. This attenuated version of the composite signal is converted into an analog signal by the analog-to-analog converter 818 and is then scaled up by the 6 dB analog amplifier, 820. As with encoder 200b, encoder 200c is preferably implemented using a Only digital signal processing chip.

The differences between encoders 200b and 200c

<img file="AR024581A2_D0086.tif" />

Those skilled in the art will appreciate, when a digital signal is converted into an analog signal with a digital-analog converter, ensuring that the digital signal is at full scale tends to minimize any loss in the signal / noise ratio that could occur as a result of the conversion. The encoder 200b minimizes the loss in the signal / noise ratio produced as a result of the operation of the converter 818 using the modules 296, 298, to ensure that the digital version of the composite signal (generated by the modulator 822) that is applied to the converter 818 is at full scale. However, although converter 200b minimizes any loss of signal / noise ratio that could occur as a result of converter 818, encoder 200b also increases the likelihood that limitation could occur in the composite signal. Since the processing section of the difference channel, 230, uses the relatively large gain provided by the fixed pre-stress filter 232 (shown in Figure 3), it is possible that some clipping may occur in the path of the coded signal difference. . The encoder 200b uses the module 298 to carry the signal D to full scale and this essentially eliminates all free space of the signal path of the signal D and increases from that

<img file="AR024581A2_D0087.tif" />

so the possibility that some cut occurs.

Then the encoder 200b minimizes the loss of any signal / noise ratio that occurs as a result of the converter 818, at the cost of increasing the probability of clipping in the path of the coded difference signal. In contrast, the encoder 200c preserves free space in the path of the coded difference signal and thus reduces the probability of clipping at the cost of increasing the loss of signal / noise ratio that occurs as a result of the operation of the converter 818.

Figure 8D shows a block diagram of another embodiment of a BTSC 200d encoder constructed in accordance with the invention. The encoder 200d is similar to the encoder 200a (shown in Figure 8A); however, encoder 200d additionally includes a portion 822a of a composite modulator. Portion 822a includes two interpolators 910, 912, two low pass filters 914, 916, the digital signal multiplier 918 and a digital signal additive 930. The signal S generated by the module 296 is applied to the interpolator 910, which samples upwardly the signal S and applies the sampled signal ascendingly to the low pass filter 914. The latter filters this signal and applies the filtered signal to an input terminal of the additive 930. A pilot tone

<img file="AR024581A2_D0088.tif" />

digital that has twice the normal amplitude (i.e. 2Acos27r (f „/ f_s) n) is applied to the other terminal of. Adder 930 input, which generates an output signal by adding the two signals present at its input terminals. The signal D generated by the processing section of the difference channel, 230, is applied to the interpolator 912, which generates an ascending sampled signal that is applied to the low pass filter 916. The latter filters this signal and applies the filtered signal to a multiplier terminal 918. A signal that oscillates according to cos47r (f_or/F_s) n is applied to the other terminal of multiplier 918, which generates an output signal by multiplying the two signals present at its input terminals. As with encoders 200a-c, encoder 200d is preferably implemented using a single digital signal processing chip.

The encoder 200d is preferably used together with two digital-analog converters 932, 934, an analogue attenuator of -6 dB, 936, an analogue amplifier of 6 dB, 938, and an analog additive 940. The output signal generated by the additive 930 is applied to converter 932, which generates an analog signal that is applied to attenuator 936. The output signal generated by the multiplier 918 is applied to the converter 934, which generates an analog signal that is applied to the amplifier

<img file="AR024581A2_D0089.tif" />

938 The signals generated by the attenuator 936 and the amplifier 938 are applied to the input terminals of the signal aggregator 940, which adds these signals to generate the composite analog signal. The D / A converters 932 and 934 are designed to be complete converters that include the analogue anti-image filters mentioned above as part of their functionality. It is assumed that converters 932 and 934 are substantially identical to each other, operating at the same sampling frequency and containing substantially the same anti-image filtering. Such converters can be obtained commonly in commercial embodiments such as the Burr Brown PCM1710.

It is also possible to remove the interpolator 910 and the low pass filter 914 from Figure 8D, and operate the D / A converter 932 at a sampling frequency equal to that of the processing section of the sum 220 channel. However, do such This is generally not practical because the cheap D / A converters that are usually available usually come in pairs housed within a single integrated circuit. Such D / A converters in pairs operate naturally at the same sampling frequency. Although it is possible to reduce the complexity of DSP (digital signal processing) by eliminating interpolator 910 and low pass filter 914 of Figure 8D, do such a thing.

<img file="AR024581A2_D0090.tif" />

it probably also increases the cost and complexity çie all the design because a simple stereo D / A converter could no longer be used for the two D / A 932 converters and

934.

The encoder 200d represents a combination of the characteristics of the encoders 200b and 200c. The encoder 200d uses the module 296 to carry the S signal to full scale so as to minimize any loss of signal / noise ratio that could occur as a result of the operation of the converter 932. The encoder 200d also preserves 6 dB of free space in the signal path of signal D and consequently reduces the probability of any loss of accuracy due to clipping. Although the encoder 200d includes more components than one or the other of the encoders 200b and 200c, the encoder 200d minimizes both the loss of signal / noise ratio and the probability of clipping.

Figure 10 shows a block diagram of a preferred embodiment of the processing channel of the sum channel, 220a, and the processing section of the difference channel, 230a, for use in the encoder 200 (and these sections 220a, 230a may be used of course in encoders 200a-d). Processing sections 220a, 230a are similar to sections 220, 230 before

<img file="AR024581A2_D0091.tif" />

rooted; however, section 220a additionally includes the dynamic phase equalization filter 1010, and section 230a additionally includes a dynamic phase equalization filter 1012. In the illustrated embodiment, the output signals generated by the static phase equalization filter 228 and the fixed pre-emphasis filter 232a are applied to the input terminals of the dynamic phase equalization filters 1010 and 1012, respectively, and the signal Output generated by the logarithmic generator 558 on line 558a is applied to the control terminals of the filters 1010, 1012. The output signals generated by filters 1010 and 1012 are applied to the low pass filter 224 and the broadband compression unit 280, respectively.

Dynamic phase matching filters 1010, 1012 are used to compensate for phase errors introduced by variable accent filter 560 which is used in spectral compression unit 290. The phase response of variable accent filter 560 is preferably matched. as closely as possible to the variable accent filter 136 (shown in Figure 1). However, due to the variable and dependent nature of the signal presented by the variable accent filter 136, it is quite difficult to design the filter

<img file="AR024581A2_D0092.tif" />

-78 variable accentuation 136 so that its ΐ-dé'-fàse response is matched to that of variable accentuation filter 136 for all pre-accent / de-accent characteristics, which in turn vary with the signal level. Accordingly, in typical embodiments of the encoder 200, the phase responses of the variable accent filter 560 and the variable accent filter 136 diverge as a function of the signal level. Dynamic phase equalization filters 1010, 1012 preferably introduce compensatory phase errors in the processing sections of the sum and difference channel to compensate for the divergence between the variable accent filter 560 and the variable accent filter 136.

The dynamic phase equalization filters 1010, 1012 therefore perform a function that is similar to that performed by the static phase equalization filter 228. However, while the filter 228 compensates for errors that are independent of the level of the difference signal. coded, filters 1010, 1012 compensate for errors that are dependent on the level of this signal. Filters 1010, 1012 are preferably implemented as all-pass filters that have relatively flat amplitude responses and selected phase responses. Dynamic phase matching filters are included in the sections of

<img file="AR024581A2_D0093.tif" />

processing of both the sum channel and the difference channel because a phase delay may be required in either the * x sum or difference channel to compensate for the phase error introduced by the variable accent filter 560. In preferred embodiments, filters 1010, 1012 are similarly implemented as a variable accentuation unit 536 and include a filter having a variable coefficient transfer function and a LUT to select filter coefficient values ​​during any interval in particular. The signal generated by the logarithmic generator 558 on line 558a is preferably applied to the control terminals of the filters 1010, 1012 and selects the filter coefficients used by those filters.

The digital encoder 200 has been analyzed in relation to certain particular embodiments; however, those skilled in the art will appreciate that variations of these embodiments are also included within the invention. For example, the variable accent unit 536 (shown in Figure 6) has been analyzed in terms of being implemented using a variable accent filter 560 and a LUT 562. However, instead of precalculating all possible coefficients for filter 560 and storing them in LUT 562, it may be preferable to

<img file="AR024581A2_D0094.tif" />

eliminate LUT 562 and include components instead to calculate the filter coefficients in real time. Those skilled in the art will appreciate that such considerations represent an intermediate solution between memory resources (as used by a LUT to store filter coefficients) and calculation resources (as used by components to calculate coefficients in real time filter), and can be solved differently in each particular embodiment of encoder 200. Similar considerations apply to square root extraction devices 456 and 556, reciprocal generator 458 and logarithmic generator 558 (shown in Figures 5 and 6), which may alternatively use memory resources (for example, a LUT to store all values) or processing resources (for example to calculate a polynomial approximation by Taylor series). In other embodiments, any or all of the components in the encoder 200 may be implemented using individual hardware components or alternatively as logical component modules applied in a computer of general or specific use.

Another example of variations of the encoder 200 that are comprised within the invention relates to the

<img file="AR024581A2_D0095.tif" />

demultiplication modules 292, 294 (shown in Figure 8B). These modules are particularly relevant for fixed point embodiments of encoder 200. In fixed point embodiments there is no need to fill each sample with zeros and sign bits to prevent overflow, and these modules can therefore be removed from the embodiments. of fixing point. As another example, the static phase equalization filter 228 (shown in Figure 10) has been analyzed in terms of compensation of phase errors introduced by the filter 232a; however, filter 228 may alternatively be used to compensate for other phase errors introduced by other components in the processing section of the difference channel, 230a. Moreover, filters 228 and 1010 can be implemented as a single filter.

Therefore, since certain changes can be made in the aforementioned area without departing from the scope of the invention discussed herein, it is thought that all material contained in the above description or shown in the attached drawings will be interpreted in an illustrative sense and not limiting.

<img file="AR024581A2_D0096.tif" />

Having thus specially described and determined the nature of the invention and the form as it has to be put into practice, it is claimed to claim as property and exclusive right: