Method for the bandsplitting of a signal being sampled with a fundamental sampling frequency in a transmitter and for bringing the bands together in a receiver.

The invention relates to a method for band splitting a signal sampled in a transmitter with a basic sampling frequency and for merging in a receiver with the features given in the preamble of claim 1 a as well as a transmitter and a receiver circuit for performing the method.

A generic method is known from DE 38 34 188 A1. Such methods are used for the symmetrical splitting of signals and have filters which consist of complementary low-pass and high-pass filters for dividing into two frequency bands and at least one further filter arrangement for combining the two frequency bands. The first filter arrangement comprises a coder with a halving of the sampling and the second filter arrangement comprises a decoder with a doubling of the sampling rate. Both symmetrical filter arrangements complement each other in such a way that the output signal corresponds to the input signal. The filter arrangements can be designed in several stages. The filters are constructed in such a way that no errors occur due to the sampling rate reduction if the low-pass and high-pass transmission are error-free. Such filters are e.g. B. Quadrature mirror filter (QMF filter), conjugate quadrature filter, non-recursive wave digital filter or Lattice quadrature mirror filter and bridge wave digital filter.

In the earlier patent application P 39 39 829, a transmitter for generating video signals of different image width to image height ratios and a receiver for the compatible display of the video signals with television receivers with 16: 9 and 4: 3 picture tubes, in which the transmission of the Signals after the so-called Letterbox method is used, in which vertical low-frequency signals are transmitted in the central part of the image and vertical higher-frequency signals are transmitted in lines at the top and bottom of the image. The signals are filtered in a diagonal filter, which acts in the direction of horizontal and vertical frequencies, and subjected to an offset modulator for offset modulation of the output signal and transmitted interleaved with one another via a channel. The received signal is split again and, after filtering and summing to a signal with the original video spectrum after offset demodulation and after passing through a diagonal filter, is displayed on a screen with either an increased vertical number of lines or a lower number of lines.

Furthermore, from EP 0 081 223 A2 a method for transmitting digital color image signals is known, in which the scanning frequency is increased by an integral multiple on the transmitter side, after the A / D conversion a digital filter is provided, which filters out interference in which only the original sampling rate is transmitted and interference on the receiving side is eliminated by increasing the sampling rate and using a further digital filter using a digital D / A converter. This procedure is specified for use in all television standards.

The invention is based on the object of specifying a method for transmitting a signal sampled with a basic sampling frequency in frequency bands obtained on the transmitter side by splitting the input signal spectrum and merged again on the receiver side, with which band splitting is possible in order to transmit signals via different band-limited channels and to merge them again on the receiver side can. Furthermore, suitable transmitters and receivers are to be specified for carrying out the transmitter and receiver-side method steps, in particular for use for the compatible transmission of video signals, e.g. B. those for compatible transmission and display of PAL signals in the aspect ratio 16: 9 by means of a screen with the same aspect ratio or on a screen with an aspect ratio 4: 3 by using the letterbox method.

The object is achieved by the method steps specified in claim 1 and by the suitable training of transmitters and receivers specified in claim 5 ff for carrying out the method.

The method according to the invention specifies pre-and post-filtering for the band splitting and the error-free joining of the bands. Linear-phase, digital filters (FIR filters) are particularly suitable for pre-filtering and post-filtering in splitting branches, in order to enable an alias-free reconstruction of the output spectrum for any band splitting in the specified ratio. The inventive method is for the transmission of a broadband signal via band-limited channels, for. B. for sound transmission, via telephone networks, image transmission over limited data channels, and compared to the prior art even when tapes are not the same size, that is, asymmetrical. The channels can thus have different bandwidths, which means that more transmission capacity is available.

The specified subsampling and oversampling of any signals, that is to say omitting sampled values, is a time-variant operation which can be represented by means of the Fourier transformation as a modulation with suitable cosine functions. An M-fold subsampling with subsequent filling of the sampling sequence with (M-1) zeros can be expressed as modulation according to the following equation known from digital signal processing.<maths id="math0001" num=""><img file="EP0477632A2_D0001.tif" /></maths> mean:<dl id="dl0001"><dt>ξ (iT) =</dt><dd>the scan sequence resulting from the scan</dd><dt>(iT) =</dt><dd>an integer multiple of the sampling time interval</dd><dt>f_s =</dt><dd>the basic sampling frequency of the input signal</dd><dt>s (iT) =</dt><dd>the sampled signal value (amplitude sample)</dd><dt>i =</dt><dd>integer run variable, in principle t</dd><dt>n =</dt><dd>a summation run variable that defines the number of required summation terms with its scope.</dd></dl><maths id="math0002" num=""><img file="EP0477632A2_D0002.tif" /></maths>

Using known undersampling and oversampling conditions, the method according to the invention specifies in general how a sampled signal is to be split up on the transmitter side in order to be available again after the reconstruction on the receiver side as a reconstruction free of aliases and phase errors, the signals being transmissible via band-limited asymmetrical channels are. The FIR filters proposed as particularly suitable for this purpose are known as quadrature mirror filters (QMF). In order to achieve a band splitting in the ratio specified in claim 1, a sampling rate conversion (upsampling and downsampling with interpolation filtering) is carried out to M × N times the original sampling frequency. Due to the sampling rate conversion carried out, instead of a mirror-symmetrical low-pass high-pass pair according to the known methods, a low-pass bandpass pair is used which fulfills the conditions specified in claim 1. The two filters on the receiver side must have the same properties as the filters on the transmitter side. When implementing the filter with an even number of coefficients, however, it is necessary to use bandpasses which differ from one another by a factor of -1 in the bandpass branch. If the signals on the channel are alias-free with respect to the original sampling frequency, then the filters on the receiver side also guarantee an alias-free and phase-error-free reconstruction of the output spectrum. The quality of the reconstruction is determined by the amplitude deviation of the signals in the channels and by the sampling rate conversion (quality of the interpolation filter). With such a pre and post filtering z. B. A compatible transmission of signals for different resolution devices, such as television playback devices, possible. The invention therefore provides the use of the method for the reproduction by means of a video signal generator, e.g. B. a camera, recorded images with an aspect ratio of 16: 9 for compatible playback on a screen with an aspect ratio of 4: 3 explicitly in claim 5. Taking into account the known letterbox method, in which the entire widescreen original is stored in the conventional 4: 3 format in the compatible 16: 9 transmission, filters are used according to the method according to the invention which have an M: N ratio of 3: 1 . The maximum amplitude deviation of the signals in the channels should not be greater than 0.3 dB, whereby the deviation can be improved with the aid of optimization methods. In the letterbox method, the 16: 9 PAL image with 432 active lines is reproduced on the receiver in standard format. This means that the entire height of the screen is not used, but black stripes remain in the top and bottom of the receiver. In order to be able to produce the original vertical resolution of 576 active lines in the PAL process on the receiver side, the letterbox process provides for the transmission of additional information to a receiver with a 16: 9 format. In order to make the additional information invisible to the standard device (4: 3), the signal can be transmitted with a correspondingly reduced amplitude between the black level and the synchronous level or subsequently limited at the receiving end. As is known, the letterbox method has the disadvantage that the vertical resolution is reduced by a factor of 3: 4 for compatible images. In addition, if the brightness level is not set correctly, interference can result from the subordinate additional signal in the image strips. On the other hand, there is an essential advantage that the compatible receiver also reproduces all of the image information, so that the viewer is not lost to any action that is important to the image. The geometry of the receiver with the 16: 9 aspect ratio and the compatible receiver with the 4: 3 aspect ratio is automatically correct, since the entire image information is available in the horizontal direction per line in 52 · 10⁻⁶ seconds. In order to achieve the same horizontal spatial resolution with the widescreen receiver (16: 9) as you are used to with the standard device of 4: 3 images, a transmission bandwidth increased by a factor of 4: 3 compared to the standard image is required, the implementation of which also ensures the specified method.

Further advantageous method steps, in particular with regard to the compatible video signal transmission of signals for displaying different image formats, are specified in subclaims 2 to 4.

Circuit arrangements for carrying out the method are specified in claims 5 ff for transmitters and receivers.

In principle, the signal that is undersampled on the transmitter side via the low-pass branch can be modulated with a carrier frequency and received by a standard receiver 4: 3. Since the signals on the transmitter side are no longer merged and interleaved, this is made possible, so that programs recorded and transmitted in widescreen format can be received and displayed without additional filtering effort. In order to display a 16: 9 picture with a corresponding receiver, on the other hand, it is necessary for this receiver to combine the signals received via the different band-limited channels and to reconstruct the original spectrum using the inventive method and the specified circuit arrangements.

Each of the two branches of a circuit arrangement for carrying out the method can in turn be divided into any number, that is to say the circuit can be cascaded if necessary. A split into several low-pass and band-pass branches is also possible if this is necessary. Since usually only one band splitting is required for transmission in two channels, the advantageous circuit examples relating to the band splitting of the video signals for a compatible transmission of video signals for the representation of aspect ratios of 16: 9 on conventional receivers are explicitly stated.

The method according to the invention has the advantage over the prior art that it can also be used to implement quadrature mirror filters with odd coefficients (Z) if delay lines are used in the two branches in the manner specified. In order to be able to combine an original spectrum on the receiver side again, the useful signal must be added to 1 if possible and the resulting alias must be deleted by non-ideal filters. This is possible in the simplest way by subjecting the samples of one signal branch to a phase shift of 180 ° with respect to the phase position of the samples of the other signal branch, which leads to the cancellation of the alias effects when the samples of both signal branches are added. The properties of a quadrature mirror filter are hereby realized. If a filter with even coefficients is implemented, the phase-free and alias-free reconstruction of the signal is automatically fulfilled because these filters have a transfer function with either a sine or a cosine function, so that the alias is always negative in one branch and always in the other branch is positive. Since this rotation is not given when the filter has an odd number of coefficients, the specified delay lines are necessary for the rotation of the alias in one branch.

The invention is explained in more detail below with reference to the block diagram shown in the drawing using exemplary embodiments.

The block diagram shows: Block 10 is an oversampling circuit in the form of an interpolator, followed by an interpolating low-pass filter 11. After the low pass 11, the band splitting takes place by means of the first filter arrangement comprising a preferably modified QMF low pass filter 20 and a modified QMF band pass filter 21. A delay line 30 is connected downstream of the low pass filter 20. For undersampling, a decimator 31 is provided as undersampling circuit, which is followed by modulation stages, not shown, of a transmitter in order to transmit the spectrum obtained in this way via the band-limited channel.

A delay line 32 is provided in the second transmission channel, which is connected downstream of the bandpass filter 21. The signal is undersampled with a decimator 33 which is also provided. Also not shown in this channel are the necessary modulation devices for modulating the spectrum with a carrier frequency for transmission over the band-limited second channel. The division of the signal sampled at the fundamental frequency and then converted by oversampling is carried out asymmetrically by the two filter branches.

On the receiving side, the received and demodulated signal of the first transmission channel is fed to an interpolator 34, in which the undersampled signal is again oversampled. This interpolator 34 is synchronized with the decimator 31 effecting the undersampling on the transmitter side. A delay line 35 is connected downstream of the interpolator 34, which is followed by a modified QMF low-pass filter 40, which has the same properties as the low-pass filter 20. An interpolator 36, which is followed by a delay line 37, is provided in the receiver circuit for the second transmission channel (bandpass branch). The output of the delay line 37 is connected to the bandpass filter 41, which has the same properties as the bandpass filter 21, but is multiplied by the factor -1 when the filters are implemented with even coefficients (Z). The output signals of the low-pass filter 40 and the band-pass filter 41 are fed to an adder 60, the output of which is connected to a low-pass filter 50. This is followed by an undersampling decimator 51, which is synchronized with the oversampling interpolator 10 on the receiver side. The reconstructed output signal thus corresponds to the original spectrum of the input signal.

The high and bandpass filters 20, 21, 40, 41 used can be implemented with an even number of coefficients as well as with an odd number of coefficients. Linear-phase, digital filters, so-called FIR filters, are preferably used as filters, which enable an alias-free reconstruction of the output spectrum for any band splitting in the ratio specified according to the invention.<ul id="ul0001" list-style="none"><li>1. Embodiment a general band splitting with the following condition:<maths id="math0003" num=""><math display="inline"><mrow><mtext>TP / HP = M / H, M, N ∈ Z, M ≧ N</mtext></mrow></math><img file="EP0477632A2_D0003.tif" /></maths><ul id="ul0002" list-style="none"><li>a) Assuming that the filters are implemented with an even number of coefficients, the delay lines 30, 32, 35 and 37 can be omitted. The low pass 20 should then perform the following transfer functions:<maths id="math0004" num=""><img file="EP0477632A2_D0004.tif" /></maths> The signal should therefore be attenuated according to this transmission function. The bandpass 21 used must consequently meet the following requirement:<maths id="math0005" num=""><img file="EP0477632A2_D0005.tif" /></maths> There is also identity between the low-pass filter 20 and the low-pass filter 40. The bandpass filter 21 differs from the bandpass filter 41 on the receiver side by a factor of -1. The interpolator 34 or the low-pass filter 40 are intended to carry out an amplification of N * (M + N). The interpolator 36 or the bandpass 41, however, a gain M · (M + N). From the conditions mentioned it follows that the interpolator 10 has a new sampling rate<maths id="math0006" num=""><img file="EP0477632A2_D0006.tif" /></maths><ul id="ul0003" list-style="dash"><li>where T_ORI the basic sampling rate is -</li></ul> Fulfills. The cut-off frequency of the interpolating low-pass filter 11 should be:<maths id="math0007" num=""><img file="EP0477632A2_D0007.tif" /></maths>The low pass 50 on the receiver side must have a cutoff frequency of<maths id="math0008" num=""><img file="EP0477632A2_D0008.tif" /></maths> so that the pre- and postfiltering using even and coefficient numbers enables the desired alias-free signal reconstruction. Instead of the specified attenuation (-6dB) for the interpolating low-pass filter, other attenuation values ​​can also be implemented. It has been shown that alias-free merging of the signals is possible even with attenuation values ​​of -3dB.</li><li>b) When implementing filters with odd coefficient numbers, the same conditions as described under a) apply to the low-pass filters 20 and 40. The transfer function of the bandpass as a modulation of the transfer function of the lowpass with the cosine fulfills the following condition according to the example:<maths id="math0009" num=""><img file="EP0477632A2_D0009.tif" /></maths> The interpolator 34 or the low-pass filter 40 should carry out an amplification by N * (M + N). The transfer functions of the low-pass filters used in the two branches are the same in each case. In circuit implementation, it is now necessary to use either delay lines 30 and 37 or 32 and 35 in the two signal branches. When implementing the filters with even coefficients, these have either a sine or cosine in the transfer function, so that the alias is negative in one branch and positive in the other branch. When merging, the alias caused by the nonlinear filter is automatically deleted so that the alias-free additive spectrum is present at the output. This is not the case for filters with an odd number of coefficients. The alias is in phase in both branches. The circuit according to the exemplary embodiment now provides that delay lines are provided in the branches, which rotate the scanning grid with the alias-related interference components contained therein by 180 ° and thus the alias by 180 °. During the addition, the alias that arises in the two branches is thus deleted, so that the principle of the quadrature mirror filter is also possible when filter circuits with odd coefficients are implemented, which was not previously thought possible.</li></ul></li></ul>

Regardless of whether the filters are implemented with even-line or odd-line coefficients, the low-pass component with a reduced bandwidth of M / (M + N) · bandwidth of the original signal and the high-pass component with a corresponding bandwidth of N / ( M + N) · Bandwidth of the original signal transmitted. The transmission bandwidths of both channels add up to give exactly the bandwidth of the original signal. In order to achieve the desired conversion of the sampling rates, the signal is already sampled at a fundamental frequency before it goes into the actual splitting filter path, oversampled on the transmitter side with a common multiple of M and N and filtered by interpolation. After the transmission, the operation for splitting up the signals is reversed in a mirror-inverted manner and, after filtering by means of the low-pass filter 50, is applied alias-free to the decimator 51 for undersampling for the reconstruction.

2nd Embodiment

The following example shows a circuit design for a compatible 16: 9 transmission using vertical signal splitting in a television system according to the PAL system using the letterbox method. In the description, reference is also made to the block diagram.

In order to split the 576 active lines of a 16: 9 format original, which was recorded by means of a camera, into two differently sized parts in the vertical direction in order to achieve an aspect ratio of 16: 9 on a standard screen (4: 3), the screen becomes divided vertically into three areas: an upper stripe of 72 lines, a middle area of ​​the visible image with 432 lines, and a lower stripe of also 72 lines.

This results in a division of 432/144 = 3/1. Signal portions are now accommodated in the 144 lines of the edge strips, which the improved receiver with a picture tube of 16: 9 needs for the reconstruction of the original picture. The resolution in the horizontal direction can be chosen arbitrarily within the framework of the transmission channel, since it has no influence on the vertical splitting. So that the additional information in the stripes on the 4: 3 screen in the upper and lower area remains invisible to the viewer, the signal is suitably limited in amplitude and transmitted with an amplitude below the black level of the BAS signal. This method can be used when it comes to full-screen displays with 576 active lines and z. B. 25 movement phases per second (e.g. conventional film scanner), or to signals that are transmitted using the real interlace method with two fields forming a frame with 625 lines and a field frequency of 50 Hz.

In accordance with the description of the first example, the same basic conditions also apply to this special method. The first transmission channel forms the middle lines of an image:<dl id="dl0002"><dt>Progressive:</dt><dd>432 Lines, corresponding to lines 73 to 504;</dd><dt>Interlace method:</dt><dd>216 Lines, corresponding to lines 37 to 252.</dd></dl>

The additional information required for the letterbox process is transmitted via the second transmission channel (bandpass):<dl id="dl0003"><dt>Progressive:</dt><dd>144 Lines corresponding to lines 1 to 72 and 505 to 576 of the active image;</dd><dt>Interlacing:</dt><dd>72 Lines corresponding to lines 1 to 36 and 253 to 288.</dd></dl>

The use of the method according to the invention can thus be used both in the progressive image scanning with 25 full-frame movement phases and in the real interlacing method with 50 movement phases and filtering in the partial image. The same vertical resolution is given in both cases. With the interlace method, the annoying line flickering is only superimposed. However, if the signal is filtered using the real interlace method, the resolution of a 16: 9 display on a compatible receiver with a picture tube 4: 3 deteriorates by about a factor of 2 compared to the improved receiver, which can be tolerated and no noticeable impairment in one Represents image reproduction. With a 16: 9 receiver, on the other hand, full resolution is achieved through signal reconstruction. The filters and functional circuits used must meet the following conditions: The interpolator 10 performs a triple oversampling, so that the new sampling rate<maths id="math0010" num=""><img file="EP0477632A2_D0010.tif" /></maths> is, which corresponds to a line frequency of 3 · 15 625 Hz based on the PAL system.

The following condition applies to the interpolating low-pass filter 11:<maths id="math0011" num=""><img file="EP0477632A2_D0011.tif" /></maths> which corresponds to the factor M. The specified damping value can be different, e.g. B. -3dB.

The modified QMF low pass 20 in the low pass branch must fulfill the following transfer functions:<maths id="math0012" num=""><img file="EP0477632A2_D0012.tif" /></maths> The modified QMF bandpass filter 21 should have a cutoff frequency of<maths id="math0013" num=""><math display="inline"><mrow><mtext>fg = 432 L / PH (-3dB)</mtext></mrow></math><img file="EP0477632A2_D0013.tif" /></maths> have.

The delay lines 30, 32, 35 and 37 are only necessary on a case-by-case basis if the conditions already described under example 1b) are to be met. In this example, assume that delay line 30 delays the signal by two lines, delay line 32 delays the signal by six lines, delay line 35 delays the signal by six lines, and delay line 37 delays the signal by two lines. The decimator 31 undersampled by a factor of 4, the decimator 33 undersampled by a factor of 12. The sampling was carried out synchronously.

Conversely, oversampling is required on the receiver side. The interpolator 34 therefore oversamples by a factor of 4, the interpolator 36 by a factor of 12, the synchronization being ensured in each case. The low pass 50 has a cutoff frequency<maths id="math0014" num=""><img file="EP0477632A2_D0014.tif" /></maths> on. The undersampling decimator 51 samples the filtered signal by a factor of 3, in synchronism with the oversampling by the interpolator 3 on the transmitter side. Filters 20 and 21 must meet the same requirements as previously described.

The interpolating low-pass filter 11 can of course be integrated in a filter arrangement together with the filters 20 and 21. The same also applies to the low pass 50 on the receiver side, which can be integrated with the filters 40 and 41 in a common filter arrangement.

The circuit works in relation to the compatible 16: 9 transmission in the PAL system as follows:

1. Low pass branch

The low-pass branch 20 limits the signal to 432 L / PH (-3dB). The delay line 30 is only required when using FIR filters with an odd number of coefficients, if these filters are used as low-pass filters 20, 40 and band-pass filters 21 and 41. After subsampling in the decimator 31 by a factor of 4 there are 432 lines, which in the transmission channel 1 correspond to lines 73 to 504 at 625 lines / 1: 1/25 Hz, or 37 ... 225 and 325 ... 540 at 625 lines / 2: 1/50 Hz are transmitted and represent the image signal for the compatible receiver. On the side of the improved 16: 9 receiver, an upward conversion by a factor of 4 takes place in the interpolator 34, which, like the interpolator circuit 10, can be equipped with a line memory in order to be able to carry out the oversampling. The condition described in Example 1b) applies to the delay line 35. The reconstructing low-pass filter 21 is constructed identically to the low-pass filter 20. A signal with a cutoff frequency 432 L / PH (-6 dB) is thus present at the output of the low pass 40. This signal can be used for a receiver with 4: 3 image reproduction for direct display. The transmission and processing of the additional information transmitted via the second transmission channel and their processing is not necessary for this, nor is the reconstruction of the received signal from the TP branch necessary. However, if the method is also to be applied to an improved image receiver, the second channel with the bandpass branch must be provided and the two signals must be combined.

2nd Bandpass branch

The processing in the bandpass branch is carried out analogously to that in the lowpass branch. The bandpass filter 21 has a cutoff frequency 432 L / PH (-3 dB). The agreements described in Example 1b) apply to the delay line, ie they are only necessary in combination if the FIR filters are formed as low and bandpass filters with odd coefficients. The decimator 33 performs a downward conversion by a factor of 12. There are 144 lines at its output. These lines represent the high-pass information and are in the letterbox bars corresponding to lines 1 ... 72 and 505 ... 576 (625 lines / 1: 1/25 Hz) or 1 ... 33, 253 ... 288, 289 to 324 and 541 ... 576 (625 lines / 2: 2/50 Hz) in the sub-black range (compression by approx. A factor of 4). In the improved 16: 9 receiver, an interpolation by a factor of 12 is first carried out with the interpolator 36. The reconstructing bandpass filter 41 is identical to the bandpass filter 21 in the bandpass branch. The band and low-pass signals are then added in circuit 60 and subjected to low-pass post-filtering in low-pass 50, which has a cut-off frequency of 546 L / PH (-6 dB). With this measure, alias interference in the subsequent subsampling by means of the decimator 51 is avoided. At the output of the decimator 57, the signal originally supplied by the camera is available again, so that the image can be displayed with the full resolution in the case of an improved receiver.