DE3943795C2 - interpolation - Google Patents

Abstract

The invention relates to an interpolation filter for processing input data which have been sampled at a sampling frequency fs, the sampling frequency being assigned a corresponding sampling period. A filter arrangement consisting of m sets of nth order filters is used to find digital data for m interpolation points that lie within the sampling period, thereby generating interpolation data with a resolution of m fs, the m sets of nth filters Order each have the same amplitude characteristic and an integer is not less than two.

Description

The invention relates to an interpolation filter according to the Preamble of claim 1.

Sound sources such as those found in electronic musical instruments are used, for. B. analog sound sources z. B. from VCO, VCA and VCF exist, and digital sound sources, one have programmable tone generator or a ROM  which sound data are read out. As a digital sound source also used one made from the tones of Musikin instruments source sound data determined by sampling and this Stores data in a memory.

Because a large storage capacity is generally required to sample and store source audio data Various techniques have been proposed to increase storage capacity save. This includes the formation of loops, the Periodicity of signal patterns in the musical tones exploited , and this includes bit compression by non-li linear quantization.

The aforementioned loop formation is also used to generate tö longer than the musical sound originally sampled take. A musical tone has e.g. B. a formant area right after the start of sound generation, in which none there are clearly repeating signal patterns. it is followed by an area in which signal patterns with a Basic interval can be repeated. By arranging n perio the repeating signal pattern, where n is an integer is, as a loop domain, and by repeatedly reading the Loop domain can keep tones sustained for longer Utilization of a memory of relatively small capacity generated become.

For bit compression of common audio PCM signals, a pre Hardness type filter generally used on the coding side. This System transmits sub-data or sub-data, i. H. Data related to compression, along with the com primed data, being for the filter in decoding direction an IIR (Infinite Impulse Response) filter or recursive digital filter is used. Such one System is z. B. also with standards for digital optical Plates used.  

The scanning of musical tones and the formation of loops from clay components corresponds to the repeated joining of Grinding of connecting the end and start points of the loop fendomänen. So that not due to discontinuities sets loop noise at the starting points, the The start and end points each have approximately the same level sen. However, it is difficult to start and loop To select the loop end point so that it is essentially the same Chen level, which is related to the sampling period that usually doesn't exactly match the loop period matches. Therefore, it has not been satisfactory so far Procedure for loop formation with low loop noise discovered.

Different tones, especially drum tones, have none Loop domain.

Another problem with processing digital data with Help from loops is that when swelling tone data are read from a memory, data start addresses and loop start addresses from a table of contents can be read out, which is in the memory with the source sound data is located. The two address data usually have un different content. If the address data is constantly in an address register of a device for processing Audio signals are stored is an increased number of Memory access is required, which is especially important to him increased load on the arrangement, if in the case of time multiplex processing for generating multiple tones on the Memory needs to be accessed.

As already stated, digital data exist without loops fendomäne. Therefore, source sound data exist with as well without loop domains. Stop playing Source sound data depends on whether loop domains in the  Source sound data is present or not. When playing source sound data To end with a loop domain is usually a flag used, which is contained in the source sound data for the loop domain and that Indicates loop end. When you stop playing a sound that is not Has loop domain, this must be done in a different way. usual a separate address for the playback termination signal is used for this purpose used.

So, when the source sound data has to be stopped, it depends whether or not there is a loop domain. The data processing must depending on the case at hand with different addresses, to stop playback. This leads to increased required food capacity and a complicated structure of the device.

Source sound data that may have repeating patterns are said to be the smallest possible memory requirement can be processed.

For this purpose, data is interpolated in the end areas of a loop domain and from the original samples and the interpolated values used for the start and end points that are closest in level different. This ensures that at every transition from a domain ne the next as small a level jump as possible and thus as small as possible Level noise occurs. Because of the interpolation, it is possible despite the ratio moderately less stored sample data, satisfactory transitions into one another to achieve subsequent loop areas.

A tone signal generator is known from EP 0 273 447 A2, in which tone signals are included tels filter coefficients processed and interpolated by a filter arrangement the. Different filters deliver different filter coefficients, which too have different characteristics.

It is an object of the invention to create a simple interpolation filter fen, which leads to particularly low-noise signal processing and for example probably for the interpolation mentioned above, as well as a filter in a bit compression coding device can be used.  

To achieve this object, the invention provides an interpolation filter with the Features of claim 1 before. An advantageous further training of the Erfin dung is specified in claim 2.

Data can contain immediate PCM words for an initial compression block be prefixed. This measure eliminates the need for special initial values for the Save signal processing.

With as few memory accesses as possible, source sound data can be read out, by only starting addresses for source sound data and starting addresses for loops data are stored. To read the data, only the start addresses are from ei read out memory, which are then changed by a counter. It must So not every address can be read out separately.

A flag can indicate whether source sound data contains a loop domain or Not. Indicates the flag that a loop domain is missing and the end of source sound data is reached, a damping device is activated with the help of the flag fourth.

A memory can have two memory areas. Whenever out of one Storage area source sound data are read out in the other area Source sound data imported. Through this alternate writing and reading it is possible to extract longer-lasting signals with a small storage capacity ben.

The invention is illustrated below with the aid of figures illustrated embodiment described in more detail. It demonstrate:

Fig. 1 is a block diagram for explaining an apparatus for forming, recording and reproducing source sound data;

Fig. 2 is a musical tone signal;

Fig. 3 is a block diagram for explaining an execution form for processing Quelltondaten;

Fig. 4 is a block diagram for explaining how an interval is determined;

Fig. 5 is another block diagram for explaining the interval determination;

Fig. 6 is a musical tone signal with envelope;

Fig. 7 is a signal illustrating the decay of a musical tone signal;

Fig. 8 is a block diagram for explaining how an envelope is detected;

Fig. 9 is a characteristic of an FIR filter;

FIG. 10 is a signal having a constant signal level to Hüllkur venkorrektur;

Fig. 11 shows a comb filter characteristic;

Figure 12 is a signal illustrating how the location of a loop area is determined;

Fig. 13 signals representing a musical tone signal before (A) and after (B) time-base corrector;

FIG. 14A and B are diagrams for explaining how a Bitkompres sion block for amplitude data by Zeitbasiskor adapted rection;

Fig. 15 shows a signal consisting of lined up domain domains;

FIG. 16 is a signal for a formant region by Hüllkur venkorrektur based on decay data;

Fig. 17 is a flowchart for explaining steps that take place before and after looping;

FIG. 18 is a block diagram for explaining the flow of a bitkodierenden bit compression;

Fig. 19 is a schematic diagram illustrating a data block as obtained with quasi-instantaneous bit compression;

FIG. 20 is a schematic diagram anschaulicht the contents of a block in the initial region of a musical tone signal ver;

Fig. 21 shows a signal illustrating samples at loop points;

Figure 22 is a signal illustrating how loops are appended to each other;

FIG. 23 is a signal for explaining conversion interval;

Figure 24 is a block diagram illustrating how an interpolation process operates;

FIG. 25 is a schematic diagram illustrating the location of loop-start and Schleifenendadressen;

Fig. 26 is a block diagram showing the basic structure of an interpolation filter;

Figure 27 is a block diagram for a low pass filter used to find the coefficients for the interpolation filter shown in Figure 26;

FIG. 28 is a diagram for illustrating how straight PCM data are arranged at the starting area of a loop domain;

Fig. 29 is a block diagram illustrating a source sound reproduction page;

Fig. 30 is a diagram illustrating the content of a memory;

Fig. 31 is a timing chart according to illustrate the main function of the circuit. Fig. 29;

Fig. 32 is a block diagram illustrating the construction of an audio processing unit and peripheral devices thereof;

Fig. 33 is a functional block diagram for explaining a source sound processing device which requires relatively little memory access;

Fig. 34 is a functional block diagram for a sound data processing device which distinguishes between sound data with loops and without loops;

FIG. 35 is a functional block diagram for a swelling clay processing means for generating kontinuier Licher Quelltondaten; and

Fig. 36 is a flow chart for a device with other structure of the Quelltondatenwiedergabeseite.

It should be noted that the following description exercise on a device for generating, recording and Playing source sound data refers to that described Device, however, generally for processing digital sig nale can be used.

Fig. 1 shows in block form a device for generating, recording and reproducing source sound data. The device is e.g. B. used as an adapter for a PC in a scanning source sound unit, or it forms the Quelltonbe rich for an electronic musical instrument, as it is sometimes referred to as a "sampler".

When generating and storing source sound data, an analog audio signal of a source sound is fed via an input connection 121 to a preamplifier 122 and subsequent low-pass filter 123 . From a subsequent A / D converter 124 , it is converted into a 16-bit sampling signal Sd with a sampling frequency of 31.5 kHz. This signal Sd is processed by a digital signal processor (DSP) 125 , which outputs source sound data to a memory 126 .

In the exemplary embodiment, memory 126 has an address range of 2M words, 512K words being a 16-bit / address buffer area, while the remaining area of 1.5M words represents a 12-bit / address memory area for storing the signals Sd.

The DSP 125 performs arithmetic operations using the buffer area of the memory 126 . It is e.g. B. corrected or normalized the plitude of the signal Sd so that it becomes constant and fills the full range of the 12 bits mentioned. The 12-bit signal Sd, which is thus obtained after correction, is stored in a part of the memory area of the memory 126 .

Parameters such as B. constants for correcting the signal Sd or how start or end addresses, which are used in storing the signal Sd in the memory 126 , are supplied by the DSP 125 to an 8-bit CPU 111 of the aforementioned Tonquellenein unit. The values are stored in a RAM 113 for work area values and parameters. The Quelltonein unit also has a ROM 112 in which a system control program is stored. The RAM 113 and the ROM 112 are connected to the CPU 111 via a bus line 119 .

In this way, e.g. B. 32 types of musical tones, normalized to a predetermined amplitude, are stored in the memory 126 , while parameters relating to these signals are stored in the RAM 113 .

When a keyboard 114 for MIDI standard is operated, for playing back source sound data for editing or musical evaluation, the associated parameters are read out by the CPU 111 from the RAM 113 , depending on settings on an operating panel 115 . They are then transmitted to the DSP 125 and the digital signal Sd for the associated sound is read out from the memory 126 . However, the read-out signal is a 12-bit scanning signal with an amplitude that is normalized to a constant value. The sampling frequency for the signal Sd is still 31.5 kHz, as when storing. The keyboard 114 and the control panel 115 are connected via an interface 116 to the bus line 119 , to which a display device 118 is also connected via a driver 117 .

If the original sound is longer than the storage period corresponds, the signal Sd only shows the duration of the memory period on. It then becomes a predetermined part of the period, which lies towards the end of the signal Sd is repeated. There corrected the amplitude of the signal Sd to a constant value  there is no pitch at connection points between Signals SD on.

The signal Sd is supplied to an interval converter 131 , while sound data are taken out by pressing keys on the keyboard 114 and transmitted to the converter 131 via the CPU 117 and the DSP 125 .

The interval converter 131 has an interpolating FIR digital filter, by means of which the signal Sd is subjected to a 256-fold superordinate sampling, which is followed by a post-sampling. In this way, the interval of the analog signal that it had when converting from the signal Sd is converted back into the interval that is assigned to the actuated key. The sampling frequency of the signal Sd remains unchanged.

The signal Sd from the converter 131 is transmitted to a DSP 132 , while associated parameters are read out from the RAM 113 and supplied to the DSP 132 to convert the signal Sd back to its original bit length and to correct it back so that the digital signal Sd for he will keep the original sound. ADSR parameters, as will be explained and illustrated below with reference to FIG. 6, are selected depending on the settings on the control panel 115 , so that the signal Sd is processed by the DSP 132 in accordance with these parameters.

In this way, the DSP 132 outputs a digital signal Sd which has a constant frequency and in which the interval, the volume and the timbre are set in accordance with the settings on the control panel 115 . This signal Sd is given via a D / A converter 133 , a low-pass filter 134 and an output amplifier 135 to an output terminal 136 .

The signal Sd is read out from the memory 126 in time division multiplexing and then processed, in up to 16 channels simultaneously, so that up to 16 voices or tones are output simultaneously at the output terminal 136 .

In the above, the basic function for generating, saving Saving and playing source sound data in a sampling Sound source unit described. When the input audio signal corresponds to the sound of a musical instrument, the Sig nal normally vibrations with a fundamental frequency. In this case the signal contains repetitive signal template. At least one period of the repeating sig nalpattern is stored in memory and during playback repeated several times in order to prolonged output of the musical tone to achieve. This repetition of the signal pattern is called Denotes loop formation; it serves storage capacity to save. Another technique to save storage space action is the compression of data when it is recorded ben. In the exemplary embodiment, the compression takes place with With the help of a filter selection technique, in which several Ab Sample values are grouped in a block and for each block that filter is output from several specified filters that leads to optimal data compression.

The above-mentioned loop formation is briefly explained below with reference to FIG. 2. Immediately after the start of a musical tone, there is usually a non-tone component, such as e.g. B. a noise that comes from a keystroke in a piano or a blowing sound from a wind instrument. This part of the signal forms a formant area without clearly recognizable signal pattern periodicity. The formant area is followed by an area in which a signal pattern with a fundamental frequency, ie in a certain interval, is repeated. n Repetitions of the signal pattern are treated as a loop domain LP, which extends between a loop start point LP _S and a loop end point LP _E. A musical tone can be generated for any desired length by first storing data for the formant area FR and the loop domain LP in a memory and when playing back the data for the formant area FR and then repeating the data for the loop domain LP.

With reference to FIGS. 3-20 will now be explained in more detail, are generated as source audio data and as a recording-side system is constructed and operates.

Figure 3 is a block diagram illustrating what functions are performed between sampling data and storing it.

An input music signal supplied to an input terminal 10 may either be a signal that is picked up directly by a microphone, or it may be one that is input in the form of digital or analog audio signals from a recording medium.

The input musical tone signal is sampled in a sampling block 11 , e.g. B. at 38 kHz. A 16-bit signal is obtained before sampling. The sampling corresponds to the A / D conversion of analog input signals.

In a subsequent interval detector 12 , the fundamental frequency, ie the frequency f _{0 of} the fundamental, is determined.

The principle of the interval detector 12 will now be explained. The musical tone signal sometimes has a fundamental frequency that is significantly lower than the sampling frequency f _s , so that it is difficult to determine the interval by only sampling the peaks in the musical tone along the frequency axis. It is therefore necessary to use the spectrum of harmonic overtones.

The musical tone signal f (t), whose interval is to be determined, can be expressed by Fourier expansion as follows:

f (t) = Σ | ωa (ω) cos [ωt + ϕ (ω)] (1)

where a (ω) and ϕ (ω) mean the amplitude or phase of the overtones. If the phase shift ϕ (ω) is set to zero for each overtone, the above formula can be rewritten as follows:

(t) = Σ | ωa (ω) cosωt (2)

The peaks of the signal f (t) thus adjusted correspond entirely number multiples of the periods of all overtones of the signal at t = 0. The peaks only show the period of the fundamental.

Based on this principle, the sequence of steps for detecting the interval will now be described with reference to FIG. 3.

Musical sound data and a value "0" are fed to a real part input 31 and an imaginary part 33 of a transformation block 33 for fast Fourier transformation.

For the fast Fourier transform, if the interval is denoted by x (t):

a _n cos (2πf _n t + θ) (3),

x (t) can be given by

This can be written in complex notation as

where the following equation

cosθ = (exp (jθ) + exp (-jθ)) / 2 (6)

is used. The following equation is obtained by Fourier transformation:

in which δ (ω - ω _n ) describes a delta function.

In a subsequent block 34 , the absolute value, ie the root of the sum of the squares of the real and the imaginary part of the data after the fast Fourier transformation, is calculated.

By using the absolute value Y (ω) of X (ω), the phase component stands out, so that the following applies:

This becomes the phase adjustment for all high frequency compos elements in the musical sound data. The phase components can be adjusted to 0 by setting the imaginary parts.

The normalization value calculated in this way is supplied as a real data part to a further block 36 for fast Fourier transformation, which in this case is designed as a block for inverse fast Fourier transformation. While this data reaches an input for real data, an input for imaginary data is supplied with the value "0". The inverse fast Fourier transform can be expressed by

The ge after the inverse fast Fourier transformation Won music data are output as a signal that the Synthesis of the cosine waves corresponds to the proportions higher frequency are adjusted in phase.

The peak values of the source sound data thus produced are determined in a peak detector 37 . The peak points are the points at which the peaks coincide for all frequency components. In a subsequent block 38, the peak values are sorted according to descending values. The interval follows by measuring the periods of the peaks determined.

FIG. 5 illustrates an arrangement of the peak detector block 37 from FIG. 4 for determining the peaks, ie the maximum values of the musical tone data.

It is noted that within musical sound data a large number of peaks with different values  is available. An interval can be found that the distance between peaks is determined.

FIG. 5 is a sequence of musical tone data that result from the inverse Fourier transform, connecting through an input 41 a (N + 1) -stage shift register 42 is supplied, and it will, via register a _{-N / 2.} , ., a _0,. , ., a _{N / 2} supplied to an output terminal 43 . The shift register 42 acts as a window of wide (N + 1) samples with respect to the sequence of musical sound data. The sampled values are fed to a maximum value detector 44 via this window. Since the musical sound data first reaches the register a _{-N / 2} and is then transported to the register a _{N / 2} , (N + 1) scanning musical data is transferred from the registers to the maximum value detector 44 .

The maximum value detector 44 is designed so that when z. B. the value from the middle register a _{0 became} maximum of all values, gives this value as maximum value to the output terminal 45 . The width (N + 1) of the window can be set as desired.

The envelope of the sampled digital musical tone signal is obtained in an envelope detector 13 using the interval data. The envelope signal, designated B in Fig. 6, is obtained by connecting the peaks in the musical tone signal A in sequence. It indicates a change in volume since the time the sound was generated. The envelope is described by parameter ADSR (Attack Time / Decay Time / Sustain Level / Release Time). In the case of a piano tone, part A (Attack Time T _A ) describes the time that elapses between pressing a key and the time until the desired volume is reached. The second part D (Decay Time T _D ) is the time that elapses between the time the desired volume is reached and the time a volume is reached that is maintained. The third portion S (Sustain Level L _S ) indicates the volume held, while the last portion R (Release Time T _R ) indicates the time between the release of a key and the decay of a tone. The times T _A , T _D and T _R can also be understood as changes in the volume. Shell parameters other than the above four parameters can also be used.

It is noted that the overall decay rate of the signal is obtained in the envelope detector 13 , at the same time as envelope data as it is e.g. B. are given by the above parameters ADSR. This fall time data takes a reference value "1" from the time of sound generation during the attack time T _A and then falls smoothly as illustrated by FIG. 7.

The function of the envelope detector 13 according to FIG. 1 will now be illustrated with reference to the flow diagram of FIG. 8.

The principle used is similar to that used for the Er average of the envelope of an amplitude modulated (AM) sig nals is used. The envelope is determined by it is assumed that the determined interval of the carrier fre corresponds to an AM signal. The envelope data who used when playing musical tones, the reason position of the envelope data and the interval data the.

The musical tone data is fed via an input connection 51 to an absolute value output block 52 , which determines the absolute value of the signal height data of the musical tones. This absolute value data is fed to a digital FIR (Finite Impulse Response) filter block 55 . This FIR block 55 serves as a low-pass filter, the clipping characteristic of which is determined by the fact that filter coefficients are supplied to it, which were first generated in an LPF coefficient generator 54 on the basis of interval data which are supplied via an input terminal 53 .

The filter characteristic according to the example of FIG. 9 has zero points at multiples of the fundamental frequency (f ₀ ). The envelope data illustrated with B in FIG. 6 can be obtained from the musical tone data signal A in FIG. 6 by weakening the frequencies of the fundamental and the overtones by the FIR filter. The formula applies to the filter characteristics

H (f) = k. (Sin (πf / f ₀ )) / f (11)

in which f _{0 represents} the fundamental frequency, ie the interval of the musical tone signal.

Referring to Fig. 1 will now be explained how the pitch data for the formant portion FR and the loop LP domain from the tone height data of the sampled musical tone signal to be generated.

In a first block 14 for generating the loop data, the sampled musical sound data is divided, which was obtained with the aid of the previously determined envelope curve 8 according to FIG. 6 (or it is multiplied by the reciprocal value of the data). This serves to generate an envelope correction which ensures a constant amplitude, as is shown in FIG. 10. This corrected signal, more precisely the corresponding signal height data, is filtered in order to also generate pitch data in which other components besides the sound components are weakened, or in other words in which the sound components are emphasized. Sound components are frequency components with integer multiples of the basic frequency f ₀ . The data is passed through a high pass filter (HPF) to low-frequency components such. B. remove vibrato frequencies that are present in the signal corrected with the envelope. This is followed by a comb filter with a frequency characteristic as indicated by dash-dotted lines in FIG. 11, ie with a frequency characteristic with frequency bands that are integer multiples of the fundamental frequency f ₀ . As a result, only the signals present in the HPF signal are passed through where they are weakened by non-tone components or noise components. If necessary, the data is also sent through a low pass filter (LPF) to remove noise components that are superimposed on the output signal from the comb filter.

Since musical sound data generally have a constant interval, ie a constant fundamental frequency, energy concentration occurs in the vicinity of the fundamental frequency f ₀ and integral multiples thereof. In contrast, noise components have an even frequency distribution. When the inputted musical sound data is processed with the aid of a comb filter with the frequency characteristic as shown in broken lines in FIG. 11, only frequency components of the fundamental frequency or integer multiples thereof pass through the filter essentially unattenuated, while other components are weakened, which improves the signal-to-noise ratio. The frequency characteristic of a comb filter shown in FIG. 10 can be represented by the formula

H (f) = [(cos (2πf / f ₀ ) + 1) / 2] ^N (12)

are illustrated, in which f _{0 is} the fundamental frequency of the input signal, ie the frequency of the fundamental tone with a certain interval. N is the number of stages of the comb filter.

The musical tone signal, the noise components of which have been reduced in the manner mentioned, is fed to a circuit which detects repeated waveforms. Such waveforms, such as. B. the loop domain LP of FIG. 2 is separated abge and the remaining part is recorded, for. B. in a semiconductor memory. The stored musical sound data signal contains the non-sound component and part of the noise. This ensures that when the repeated signal part is played back, the loop noise is reduced.

The frequency characteristics of the HPF, the comb filter and the LPF are determined on the basis of the fundamental frequency f ₀ , ie on the basis of the interval data as obtained from block 12 .

In a block 16 for recognizing the loop domain, the musical tone signal is examined for repeated waveforms in order to determine from tone components that differ from the aforementioned tone components that were emphasized by the filter method described. With these, the loop start point LP _S and the loop end point LP _{E are} formed.

In the determination block 16 , loop points are selected that are spaced apart from one another by an integer multiple of the interval of the musical tone signal. The selection principle is explained below. As just mentioned, the loop point spacing must be an integer multiple of the interval as given by the fundamental frequency of the musical tone signal. By precisely determining the interval, the loop point distance can be determined.

Since the loop spacing is predetermined after the interval has been determined, two points are selected which are separated from one another by the named spacing. The signals in the vicinity of these points are then examined and a correlation is formed. An evaluation function will now be explained which uses the convolution or summation of products with respect to sampling points in the area of the two points mentioned. The convolution is carried out several times for sets of all points in order to determine the correlation of the signals. The musical sound data are successively fed to a unit for adding up products, e.g. B. a digital signal processing device, which is described below. The convolution is calculated and issued there. The set of two points for which the convolution becomes maximum is regarded as the set of the correct loop start point LP _S and the correct loop end point LP _E.

In Fig. 12, a _{0 is} a candidate point for the loop starting point LP _S. b ₀ is accordingly a candidate for the loop end point LP _E. Further amplitude data a _-N,. , ., a _-2 , a _-1 , a ₀ , a ₁ , a _2,. , ., a _N in several points, here (2N + 1) points before and after the candidate point a ₀ are also shown. Correspondingly, amplitude points b _-N,. , ., b _-2 , b _-1 , b ₀ , b ₁ , b ₂ ,. , ., b _N exist before and after the candidate point b ₀ . The estimation function E (a ₀ , b ₀ ) then becomes:

The convolution around points a ₀ and b ₀ follows from equation (13). The sets of candidates a ₀ and b ₀ are changed step by step until all loop point candidates are examined. Those for which the evaluation function E becomes maximum are regarded as loop points.

In addition to the convolution method, e.g. B. the method of mean squares can be used to find loop points. The following equation then applies:

In this case, the correct points are available if the evaluation function becomes minimal.  

In the loop point determination block 16 according to FIG. 3, the loop start point LP _S and the loop end point LP _{E are determined} on the basis of the interval. From the spacing of the points, which can comprise several intervals of the fundamental tone, there follows an interval conversion ratio, which is used in a time base correction block 17 .

The time base correction is carried out in order to adapt the intervals of different source sound data to one another when these data are stored in a memory. Instead of the interval conversion ratio, the interval data can be used for these purposes, as it provides the interval detector 12th

The interval normalization process carried out in the time base correction block will now be explained with reference to FIG. 13.

Fig. 13A shows the musical tone signal before companding the time base, while Fig. 13B shows the corresponding signal afterwards. The timelines are divided into blocks for quasi-direct bit compression and coding, as further explained below.

With signal A before the time base correction, the loop domain LP is generally not in direct relation to a block. In the signal of FIG. 13B, the loops LP domain is companded in the time base so as to (m-fold) is an integer multiple of a block length or block period. The loop domain is also moved along the time axis, so that the loop start point LP _S and the loop end point LP _E match the boundaries of blocks. The time base correction, that is to say companding the time base and shifting in the manner mentioned, ensures that loops can be formed for an integer number (m) of blocks, as a result of which the interval of the source sound data is normalized during recording.

With an offset ΔT from the foremost block boundary of the musical signal, a pitch signal value "0" can be inserted. This data value "0" serves as a pseudo data value, so that filters which require an initial value for the data compression can be selected. This is explained below using a blockwise compression illustrated by FIG. 20.

Fig. 14 shows the structure for a block for signal amplitude values after the time base correction, which data is then subjected to bit compression and coding, as explained later. There is h amplitude data in one block (number of samples or words). In the present case, the interval normalization consists of basic companding in a time in which the number of words within n periods (the constant time period T _W according to FIG. 2), ie within the loop period LP, is an integer multiple (m times) of the number of the words h are in the block. The interval normalization is preferably that the loop start point LP _S and the loop end point LP _{E are} each shifted to block boundaries. If this is the case, it becomes possible to avoid errors when switching blocks, which is the case when decoding with the aid of a bit compression and coding system.

In FIG. 15A, words WLP _S and WLP _E in each block are the samples at the loop start and end points LP _S and LP _E , more precisely the samples in each case directly before a loop end point LP _E in the corrected signal. If not shifted, the start and end points mentioned do not necessarily coincide with the block boundaries, so that, as shown in Fig. 15B, the words WLP _S and WLP _{E are} at any position within the blocks. In any case, however, the number of words between the words WLP _S and WLP _{E is} m times the number of words h in a block, where m is an integer, which means that the interval is normalized.

Companding the time base of a music signal with the The normalization just mentioned can be done in different ways consequences. For example, it can be done that the amplitudes values of the sampled signals are interpolated, with a Parent scan filter is used.

If the loop period for a music signal is not an integer multiple of the sampling period, so that there is an offset in the amplitude values in the loop start point LP _S and in the loop end point LP _E , the amplitude corresponding to the amplitude value in the start point can be found adjacent to the loop end point, and in that z. B. parent scanning is used to implement the loops period.

Such a loop period, which is not an integer multiple of the sampling period, can be set to be an integer multiple of the block period, which is done by the time base correction method described above. If time base companding by e.g. B. 256-fold higher-level sampling, the amplitude error between the values for the start point LP _S and the end point LP _E can be reduced to 1/256.

After the loop domain LP is determined and corrected in time basis, loop domains LP are appended to one another, as illustrated by FIG. 15, to generate loop data. The signal according to FIG. 15 is obtained by a single loop domain LP according to FIG. 13B being picked out and this one domain being linked several times. The concatenation of loop domains takes place in a loop signal generation block 21 ( FIG. 1).

Since the loop data is formed by repeating a loop domain LP several times, a start block with the word WLP _{S is} immediately followed by an end block with the word WLP _E , which corresponds to the loop end point WLP _E , more precisely the point immediately before this point. So that coding for bit compression can take place, at least the end block must be stored directly before the start block of the loop domain LP. More generally speaking, at the time of bit compression and block-based coding, the parameters for the start block, ie the data for bit compression and coding for each block, e.g. B. range or filter selection data, which will be described below, can only be formed on the basis of data of the start and end blocks. This technique can also be used in the case where the music data consists only of loop data without formants.

With this procedure, the same data are available for several sampling points before and after the loop start point LP _S and the loop end point LP _E. Therefore, the parameters for bit compression and coding in the blocks immediately before these points are the same, which can reduce errors and noise when reproducing the loops after decoding. The music data during playback are then stable and free of transition noise. In the embodiment, there are about 500 sample data in a loop area LP before the start block.

For the signal generation for the formant region FR, an envelope correction is carried out in a block 18 ( FIG. 3), corresponding to the correction for loop data in block 14 . The envelope correction is carried out by dividing the sampled musical sound data by envelope data ( Fig. 7) consisting only of the data for falling signals, thereby obtaining amplitude values for a signal as shown in Fig. 16. When Ausgangssig nal according to Fig. 16 makes accordingly just the envelope of currency end of the time period T _A noticeable, while in the other rich Be constant amplitude is present.

If necessary, the envelope-corrected signal is filtered in a block 19 . This is done with a Kammfil ter with a frequency characteristic z. B. corresponds to that as shown by the dash-dotted line in Fig. 11. The comb filter data are formed on the basis of the interval data for the fundamental frequency, which data are supplied by the interval detector 12 . The data are used to generate signal data from the formant area in the source sound data, which are finally stored in a memory.

In a subsequent time base correction block 20 , a time base correction is carried out, as described above with reference to block 17 . The time base correction is used to normalize the intervals for the sound data by companding the time base based on the interval conversion ratio found in block 16 or on the interval data as provided by block 12 .

In a mixer 22 , the formant data and the loop data, which have been corrected with the same interval data or interval change ratios, are mixed. For this purpose, a Hamming window is applied to the formant area formation signal from block 20 . A fade-out type signal to be mixed with the loop data is generated which decays. A similar Hamming window is applied to the loop data from block 21 . A fade-out type signal to be mixed with the formant data is generated which grows. The two signals are mixed (or blended) to produce a musical tone signal that represents the source sound data. As loop data to be stored, data can be picked out, which originate from a loop domain which is somewhat away from the cross-faded region, in order to thereby reduce the loop noise when reproducing loop signals. In this way, pitch data of a source tone signal is generated which has a formant area (non-tone components) and loop data (repeated, regular data).

The starting point for the loop data signal can also be with the Loop start point for the formant-forming signal verbun be that.

To determine the loop domain and mix the forman area and the loop data is first ge manually mixed. You then listen to the result and it becomes ver based on the loop start and end points is working.

Before the more precise determination of the loop area takes place in block 16 , this determination and the mixing are carried out by hand, and a trial listening takes place, as explained in the following FIG. 17. Thereafter, the more detailed process described above is carried out from step S26 in FIG. 18.

In the flowchart according to FIG. 17, the loop points are determined with a low accuracy in a step S21 by determining zero crossings or by optically checking the signal. In a step S22, the signal between the loop points is reproduced several times. In a step S23, it is judged by listening whether the loop formation is carried out properly. If this is not the case, step S21 follows in order to determine loop points again. This process continues until a satisfactory result is obtained. If this is the case, a step S24 follows in which the signal z. B. is mixed by fading with the For mantensignal. In a step S25, it is checked whether the auditory impression is good at the transition from the formant to the loops. If this is not the case, step S24 is repeated to mix again. If the hearing pressure is good, a step S26 follows in which the loop points are determined according to block 16 with high accuracy. There is scanning z. B. with a resolution of 1/256 over the sampling period with z. B. 256-fold superordinate scanning. In a subsequent step S27, the interval conversion ratio for the interval normalization is calculated. In a step S28, the time base correction is carried out in accordance with blocks 17 or 20 . In a step S29, the loop formation according to block 21 ( FIG. 3) is carried out. The mixing according to block 22 takes place in a subsequent step S30. The procedure from step S26 is carried out with the aid of the loop points as obtained in steps S21-S25. Steps S21-S25 can be omitted for full automation of looping.

The amplitude values thus obtained, which consist of the formant region FR and the loop domain LP, are bit-compressed and encoded in a block 23 following the mixer 22 .

Various systems can be used for compression and encoding. One is suitable as proposed in JP-62-008629 and JP-62-003516. There, a predetermined number of h sample words with amplitude values are grouped in a block, and bit compression is carried out block by block. This highly effective system is briefly explained below with reference to FIG. 19.

In the block diagram of FIG. 18 is the recording side, an encoder 70, and the reproduction side, a decoder 90 present. The amplitude values x (n) of the source sound data signal are fed to an input terminal 71 of the encoder 70 .

The amplitude values x (n) of the input signal are fed to an FIR digital filter 74 , which is formed by a prediction element 72 and an adder 73 . A respective amplitude value x (n) of the prediction signal from the prediction element 72 is supplied to the adder 73 as a subtraction signal. There it is subtracted from the input signal x (n) in order to generate a prediction error signal or a deviation signal d (n). The prediction element 72 calculates the prediction value x (n) from a primary combination of the last p input values x (n - p), x (n - p + 1),. , ., x (n - 1). The FIR filter 74 is referred to below as the coding filter.

With the highly effective system for bit compression and coding specified above, the source sound data falling within a predetermined period of time, ie a predetermined number h of words, which are grouped into blocks, are selected for each block. The coding filter 74 has optimal characteristics for this. It can be formed that several, in the case of four, different characteristics are defined in advance and then the optimal characteristic is selected, ie the one with which the highest compression ratio can be achieved. For this purpose, a set of coefficients is stored in the predictor 72 in a plurality, in the example four, of sets of coefficient memories. A set of coefficients is selected in time multiplex.

The deviation signal d (n), which represents a predicted error, is fed via a summing point 81 to a bit compressor, which has a level shifter 75 (G) and a quantizer 76 . There the compression takes place in such a way that the index part and the mantissa correspond to the gain G or the output signal from the quantizer 76 in the case of floating-point command. A requantization is thus carried out, in which the input data are shifted by the level shifter 75 by a number of bits, which number corresponds to the gain G. This shifts the range and a predetermined number of bits of the bit-shifted data is taken out by the quantizer 76 . A noise shaping circuit 77 is used to form the quantized error between the output signal and the input signal of the quantizer 76 at a summing point 78 and to give this signal via a G ^-1 level shifter 79 to a predictive element 80 . The predicted signal for the quantization error is given as a subtraction signal to the summing point 81 , in order to obtain a so-called error feedback. After this requantization by the quantizer 76 and the error feedback by the noise shaping circuit 77 , an output signal (s) is obtained at an output terminal 82 .

The output signal d '(n) from the summing point 81 is the difference d (n) from the summing point 73 , minus the prediction signal (n) for the quantization error as formed by the noise shaping circuit 77 . The output signal d "(n) from the G level shifter 75 corresponds to the output signal d '(n) from the summing point 81 multiplied by the gain G. The output signal (n) from the quantizer 76 is the sum of the output signals d" (n) from the G- Level shifter 75 and the quantization error signal e (n), as is given by the summing point 78 . After passing through the G ^-1 level shifter 79 and the predictor 80 using the primary combination of the last r input signals, the quantization error e (n) is converted into the prediction signal (s) for the quantization error.

After encoding as described above, the source sound data is converted into the output signal (s) from the quantizer 76 and is output at an output terminal 82 .

Mode selection data for the optimal filter selection are output by an adaptive range prediction circuit 84 and e.g. B. given to the prediction element 72 of the coding filter 74 and an output terminal 87 . Area data for determining the bit shift number or the gains G and G ^-1 who are also given to the level shifters 75 and 79 and to an output terminal 86 .

The playback-side decoder 90 has an input terminal 91 to which a signal '(n) is supplied, which is formed by transmitting or recording and reproducing the output signal (n) from the output terminal 82 from the encoder 70 ge. This input signal is fed to a summing point 93 via a G ^-1 level shifter 92 . The output signal x '(n) from the summing point 93 is fed to a prediction element 94 and thereby converted into a prediction signal (n), which is then passed to a summing point 93 , to which the output signal "(n) from the slide 92 is added. The sum signal is given as a decoder output signal '(n) to an output terminal 95 .

The range data and the mode selection data output at the output terminals 86 and 87 of the encoder 70 are input to the input terminals 96 and 97 of the decoder 90 , respectively. The range data from the input terminal 96 is transmitted to the slider 92 to set its gain G ^-1 , while the mode selection data is input from the input terminal 97 to the predictor 94 to set the prediction characteristic to be that of the predictor 72 of the encoder 70 matches.

In the decoder 90 thus constructed, the output signal "(n) from the slide 92 is the product of the input signal '(n) and the gain G ^-1 . The output signal' (n) from the summing point 93 is the sum of the said output signal from Slider 92 and from the prediction signal '(n).

Fig. 19 shows a block of output data from the encoder bitkomprimie leaders 70th There are 1-byte header data (parameter data relating to compression or sub-data) RF and 8-byte scan data D _A0 -D _B3 . The header data RF is 4-bit area data, 2-bit mode selection data or filter selection data and two 1-bit flag data, such as. B. a data value LI, which indicates the presence or absence of a loop, and a data value EI, which indicates whether the last block of the signal is negative. Each sample of the amplitude data is reproduced by four bits after bit compression, while 16 samples of 4-bit data D _A0H -D _{B3L are contained} in the data D _A0 -D _B3 .

FIG. 20 shows a block of bit-compressed and encoded amplitude data corresponding to the beginning of the musical tone signal shown in FIG. 2. Only the amplitude data are shown without the head. For the sake of clarity, each block consists of only eight sample data. B. 16 sample data are available. This applies to the case of FIG. 14.

The quasi-instantaneous bit compression described above and encoding selects either immediate PCM mode that directly outputs the entered musical tone signal, or one First order differential filter mode or such second order, each of which filtered the musical tone signal output, resulting in output signals with the highest compression relationship leads.

When music tones are sampled and read from a memory starts the input of the musical tone signal in one tone generation start point KS. If a differential filter mode first or second order that needs an initial value selected in the first block after the tone generation starting point KS it would be necessary to deduce an initialization value to save. However, it is desirable that this not be required is. For this reason, in the initial period the tone generation starting point KS pseudo input signals testifies that lead to the immediate PCM mode ge is chosen. The data processing then takes place in such a way that with these pseudo signals the input signals are processed.

This is done e.g. B. in that, as shown in Fig. 20, a block with loud values "0" is arranged as a pseudo input signal block in front of the tone generation start point KS. These data "0" are treated as amplitude values and bit-compressed. This can be done by storing a block with the values "0" in a memory or by initially placing values "0" before the starting point KS when scanning musical tones, ie a quiet area before the sound is generated. At least one block of pseudo input signals is required.

The musical sound data including the pseudo input signals thus formed are bit-compressed and encoded by the system illustrated in FIG. 18 and then recorded, e.g. B. in a memory. The compressed signal is played back.

When playing music sound data with pseudo input signal the immediate PCM mode is selected first, which makes it is not necessary, initial values for differential filters to determine first or second order in advance.

The delay in the start of the sound signals due to of the pseudo input signal appears to interfere with the Tone signal because there is silence at first. However, this bothers not actually, since the sampling frequency is 32 kHz, with 16 samples each in a block. This makes the ver delay about 0.5 msec, which is not noticeable in the auditory impression.

When building loop data, it can get worse Continuity at transitions between successive ones Loops come, which is due to the fact that the sampling frequency is too high compared to the repetition period of the Signal pattern.

Fig. 21 shows a loop domain LP 'with a repetitive signal pattern (shown with circles) by sampling the signal at a predetermined period is obtained. Scans are carried out at the loop start point LP _S and at the loop end point LP _E. If the sampling period does not exactly match the pattern repetition period, the loop end point LP _E 'found at the sampling does not coincide with the actual loop end point LP _E , which means that the actual loop end point and the sampled loop end point do not generally have the same signal level. This leads to an offset when loops are connected to one another, as is shown in FIG. 22. To help with this, an interpolation of samples is carried out in order to determine the loop start point LP _S and the loop end point LP _E through the interpolation and to form as closely as possible values which are then selected as connection points. This can be done in the time base correction block 17 in order to generate the interval normalization with a time base resolution that is shorter than the sampling period. If no interval normalization is carried out, the above-mentioned actual samples can be stored in a memory, and interpolation is carried out during data readout or reproduction, in order to improve the continuity of signal pattern transitions by this interpolation. If the signal of a musical tone with the interval of an occasionally pressed key on a keyboard is to be reproduced on the basis of source sound data with a predetermined interval as stored in a memory of a so-called sampler, it becomes necessary to carry out interval conversion in order to include notes to generate different intervals. If interpolation with superordinate samples is used for this, the technical concept mentioned above can be carried out without having to increase the hardware in order to obtain interpolated samples. Fig. 23 illustrates an interval conversion system which operates with this interpolation. The pick-out interval for interpolated samples, e.g. B. is obtained by quadruple superordinate scanning is changed to achieve interval conversion. In the example of FIG. 23, signal A represents the original signal in which actual samples are represented by dots, while interpolated samples are identified by x marks. The interpolated values are achieved by four-fold superordinate sampling for each actual sample value. The circle-marked actual samples which follow one another with the original sampling period Ts are then taken out in order to produce a signal B converted in the interval. This changes the frequency to 5/4 of the original frequency. In the embodiment according to FIG. 23, the frequency is increased, but it can also be lowered. The resolution can also be improved by increasing the number of multiple scans in superordinate scanning. For example, 256-fold higher-level scanning can be used for practical applications.

FIG. 24 illustrates, as a block diagram, a circuit for converting intervals in accordance with the method explained with reference to FIG. 23.

A memory 101 , e.g. For example, a ROM in which source sound data is stored outputs this data based on address data as stored in an address generator 102 . This address generator receives interval data, as they are used to determine the interval conversion ratio, from an interval data generator 103 , and it also indirectly receives data from a register 107 a for subordinate data, a loop start address register 107 b and a loop end address register 107 c. The above-mentioned access data for access to the memory 101 are generated on the basis of this data. The register 107 a for subordinate data, the loop start address register 107 b and the loop end address register 107 c are supplied with corresponding data from the source audio data. The register 107 a for subordinate data serves to store in Fig. 19 illustrated header data FR on a block by block basis, while the loop start address register 107 b and the Schleifenendadreßregister 107 c used to point to addresses for the loop start LP _S and the Save loop end point LP _E. The output data from the aforementioned three registers are fed to the address generator 102 and a coefficient address converter 106 . This converter also receives data from the interval data generator 103 . Based on output data from the coefficient address converter 106 , a coefficient ROM 105 supplies coefficient values stored therein to an interpolation filter 104 to determine the filter characteristic of this filter. In the exemplary embodiment, the interpolation filter consists of n delay units DL1-DLn and multiplication units P1-Pn. It receives source sound data from ROM 101 . These are converted by the interpolation filter 104 in the interval and converted by a D / A converter 108 into an analog signal, which is output as a source tone signal at an output terminal 109 .

Fig. 25 illustrates the loop start address and the loop end, as it is read from the registers 107 b and 107 c. It is assumed that a block is formed by several successive samples. If the sample value for the loop start point LP _{S is} stored at the beginning of a block, it is sufficient to specify the block number as the start address. However, if a loop begins at a sample within a block, a sample number must still be specified within a block, which is indicated by a broken line in FIG. 25. The loop end address contains not only the number of a sample within a block, but also data relating to interpolation points between samples. In this way, high resolution looping, including the interpolated samples, can be achieved with a method similar to that of the interval conversion discussed above. At z. B. 256-fold superordinate sampling, the loop domain LP can have a resolution of 1/256 of the sampling period PS, so that the resolution at starting points on the following signal pattern can be improved to about 1/256. In addition, there is the advantage that when using a source sound system that uses an interval conversion with superordinate scanning, the described method can be carried out in essence without additional hardware, but smooth transitions can be achieved with reduced loop noise.

The method described above for generating source sound data with improved properties at transitions from repetitive patterns uses such interpolation values for the transition points in the interpolation according to FIG. 21 that are as close as possible to the value “0”. In Fig. 21, the loop start point LP _{S of} a loop domain LP is a current sample, while the loop end point LP _{E is} the interpolated value that is closest to the actual loop end point. This is because the amplitude of the interpolation value called is closer to the amplitude of the starting point LP _S than the amplitude of the actual loop end point LP _E 'of the loop domain LP', as was recognized in actual scanning.

As already explained, successive signal patterns are connected to one another with the aid of the interpolated values, as a result of which the transition points connect to one another with great accuracy in terms of amplitude, which is illustrated by FIG. 22.

The processing described above such. B. the interval conversion, the sampling rate conversion or the superordinate Ab can feel as linear conversion of first order signals be understood. With such linear conversion there is conventional practice in bridging the gap between discrete data by filling that interpolation filter or FIR filter be used. The number of interpolations required filter corresponds to the number of interpolation points between samples. For 256-fold higher-level scanning 256 filter arrangements are accordingly required. The filter structure with 256 filter sets must in the filter characteristics be well coordinated. Mismatched amplitude characteristic leads between different filter sets of the filter arrangement to conversion noise, which is then from the selected fil replacement is generated. This type of digital noise has special features  Frequency characteristic on that of the usual white Noise is different. This makes this noise fall in the auditory impression even at a very low level. The Sound impression of an overall system therefore depends in many cases len especially from the interval noise. It is therefore special important to build an interval conversion filter so that the change noise becomes minimal.

In the audio area, interval changing is done by superordinate Ab buttons or frequently used due to sample rate conversion. This type of processing belongs to the linear transformation of First order signals and can be used as a first order version be viewed by affine transformation as they are often is used in image processing. In the case of signals There are different degrees of freedom in the first order Computing time. Gaps between individual dates are therefore pretty well filled by arithmetic calculation. It is possible, Interpolation of samples with high accuracy lead, e.g. B. parent scanning at a rate that corresponds to 256 times the sampling rate, e.g. B. the sampling rate of a source sound unit.

However, there are higher demands in terms of sound quality than in terms of image quality. It should also be noted that it is difficult to do because of cost reasons manufacture a device for extensive interpolation, z. B. with so many taps that 256 times parent Scanning can be done.

The interval conversion noise represents a special one Problem. However, it is not yet a particularly useful one Procedures for an interval conversion filter have been specified, this is clear for a given number of branches has better properties than a FIR filter.  

The inventors tried to remove various filters wrap that despite using interpolation filters are free of digital noise for the interval conversion. You have found points of view that are good Properties of interpolation filters for interval conversion are of particular importance.

It was found through trial-and-error procedures that more importantly, the filter characteristics of each fil ter in a filter arrangement as ho value on the characteristics of each individual To put filters. Second, they found that Adjust the filter characteristics to the properties in the Ab cutting area of the FIR low-pass filter is affected, d. H. that lower ripple in the cut-off area is particularly desirable is worth if a filter arrangement with good overall dough should be preserved.

In the following, an interpolation filter is explained which has m filters of the n-th order in order to find digital data in m interpolation points within a sampling period at a sampling period f _s, to obtain interpolation data from a digital input signal of a sampling frequency f _s with m times the resolution to generate the frequency mf _s , the number m of the nth order filters all having similar amplitude characteristics.

While the interpolation filters are all in the amplitudes characteristics are practically the same, they are in phase changeable, which makes it possible to get noise when switching to avoid the filter.

The typical basic structure of an interpolation filter with the properties just mentioned will now be explained in more detail with reference to FIG. 26.

Fig. 26 shows an n-th order filter formed by n coefficient multipliers 151 , (n-1) delay units 152 and a summing unit 153 for summing products. A digital input signal with a sampling frequency f _s is fed to an input terminal 154 and then delayed sample by sample by the (n-1) delay units 152 and multiplied sample by sample by multipliers 151 by coefficients. The output signals are summed by the summing unit 153 and then output to an output terminal 156 .

To generate interpolated data from the aforementioned digital input signal with a resolution of m times the sampling frequency, that is to say with mf _s , digital data of m interpolation samples as they exist within a sampling period of the sampling frequency f _s are found, ie the m In Nth-order terpolation filters are selected so that they have the same amplitude characteristic, as a result of which digital data generated by these filters are interpolation data which are free from digital noise. The m filters include m sets of coefficients, each set of coefficients containing m coefficients.

The principle of building an interpolation filter for In Interval conversion is explained below.

Interval conversion is understood to mean a linear transformation of a signal along the time axis, ie

y (t) = x (at + b) (15)

where x (t) the original signal and y (t) changed in the interval designated signal. Equation (15) can be used as a version first order of an affine transformation the. If x (t) and y (t) are continuous quantities, that is Transform x to y using this formula easily. In the In practice, however, the values of x (t) are discrete sizes, where  these values do not necessarily increase with the sampling points coincide, which are required to calculate y (t). In this case it is necessary to interpolate values from meh to calculate neighboring neighboring sampling points.

When converting levels, sample values are therefore made up of several twos points interpolated. With 8-fold higher Ab buttons are two points with the help of eight in between points interpolated. At the above source sound unit is carried out 256 times interpolation in order to suffice ensure the appropriate interval setting. the number of Interpolations are referred to below as resolution R. In the case just described, the resolution is R = 256. Interpolation data of resolution R are therefore made up of digital ones Generates input data that have the sampling frequency fs. The interpolation data have the sampling frequency m.fs.

The interpolated values can be calculated by providing an FIR filter or an n-th order non-cyclic filter for each of the m interpolation points. So there are m sets of FIR filters of fi (t) n order, with i = 0, 1, 2,. , ., m - 1. t is a variable that can change with a number that is equal to the number of the order. For the kth interpolation point, the convolution with the input source signal x (t) is calculated as follows:

The k-th FIR filter fk (t) from the aforementioned m Sets of FIR filters of the nth order are used to determine the value for the interpolation point x (t + k / m).

A filter arrangement is made up of m sets of FIR filters fi (t) nth order formed. Designing an interpolation filter  for interval conversion is synonymous with that Design a filter arrangement. The condition that required Lich, so a filter arrangement as an interval conversion filter works will now be explained.

To simplify the explanation, assume that the sampling interval is one in length (ts = 1 / fs). In order for the kth FIR filter fk (t) to function as an interpolation filter, the response behavior of this filter must satisfy the following formula:

By Fourier transform for both sides of this formula it follows:

Y (ω) exp (jωk / m) = F _k (ω) .X (ω) (18)

The imaginary number (-1) ^{1/2 is} not denoted by i, but by j in order to distinguish it from the index i. Equation (18) converted:

Y (ω) = F _k (ω) exp (-jωk / m) .X (ω) (19)

For each individual nth-order FIR filter, the following then applies when

F ₀ (ω) = F _k (ω) exp (-jωk / m) (20)

the following equation:

F _k (ω) = F ₀ (ω) exp (jωk / m) (21)

This equation (21) is the basic equation for the arbitrarily selected kth filter of the filter arrangement. As long as the FIR filter of the n-th order of the filter arrangement satisfies the equation (21) in its entirety, no interval conversion noise can be generated. The characteristics of the filter arrangement itself, such as ripple or cut-off properties, depend on the property of f ₀ (t).

Equation (21) above is converted as a time-based formula. It should be noted that the kth FIR filter fk (t) is actually a discrete quantity. If the coefficient of the qth tap of the kth FIR filter fk (+) nth order is expressed as fk (q), the following applies

When developing a Fourier transform you get:

Folding effects are not taken into account. Substituting equation (23) into equation (21) gives:

The tap coefficients of all FIR filters of the filters arrangement can be obtained by solving this equation the.

It is very difficult to obtain the properties of the filter arrangement by solving equation (24). An actual filter arrangement is therefore developed from a single prototype low-pass filter. An FIR low-pass filter g (t) of (n × m) th order is assumed with a cut-off frequency that is not higher than 1 / m of the sampling frequency, ie not greater than fs / m, as by Figure 27 illustrates.

. The filter of Figure 27 is composed of (m × n - 1) delay units 163, (m × n) multipliers 164 and an in power amplifier Summie 165th A digital input signal of the sampling frequency fs is supplied to an input terminal 161 and then delayed by the delay units 163 . It is also multiplied by coefficients in the multipliers 164 . The output signals of multipliers 164 are summed in summing amplifier 165 . Thus, a set of m coefficients for the n-th order filter shown in FIG. 26 can be obtained by the filter shown in FIG. 27.

With the filter according to FIG. 27, it is also possible to generate the number m of sets of filters of the nth order just mentioned by selecting m filters with the same coefficient in each case and the filters of the nth order from these m filters be formed.

With the m sets of filters fh (t) nth order, which are formed by taking out every nth coefficient from the above-mentioned FIR low-pass filter g (t), with h = 0, 1,. , , (n - 1), m sets of filters of the nth order are obtained which satisfy the following equation:

fh (t) = g (mt + h) (25)

The filter arrangement with these m sets of FIR filters of the fh (t) n-order fulfills the above-mentioned conditions for an interval conversion filter. The following applies:

(1 / m) G (ω / m) (26)

An interval conversion filter can be done using equation (26) can be implemented relatively easily. Occasionally you can however, sufficient characteristics cannot be obtained by testing of a practically designed filter. This may be due to the fact that it is difficult to FIR low-pass filter with a fixed number of taps too design. This error gets with increasingly higher resolution R of the FIR filter more critical.

The frequency characteristic of a filter arrangement with FIR low-pass filters g (t) can be expressed by:

f _h (ω) = (1 / m) G (ω / m) exp (jωh / m) (27)

From this it follows that the characteristic of f ₀ (ω) is that of g (ω) by m times along the frequency axis.

It follows that if the cut-off frequency of the FIR Low-pass filter that has undergone Fourier transformation should not correspond to more than half the sampling frequency, d. H. no more than fs / 2, the cut-off frequency of the FIR low pass filter G (ω), which has been subjected to Fourier transformation de, must not correspond to more than 1 / 2m of the sampling frequency, d. H. not more than fs / 2m.

On the other hand, frequency components of the FIR low-pass filter G (ω) with a cut-off frequency that is not less than 1 / 2m times the sampling frequency, ie not less than fs, clearly appear as companions on the signal F ₀ (ω) / 2m. These accompanying components have different phase components from filter to filter. The differences caused in this way can lead to differences in the filter characteristic, which lead to interval conversion noise.

The higher the resolution R of the filter arrangement, the more the cut-off frequency becomes lower for those mentioned FIR low-pass filter is required. With an insufficient rating Number of orders n for the filter can be fully satisfied filter characteristics are not obtained if the filter is not designed very cleverly because it is in is usually difficult to filter with extremely low or develop extremely high cut-off frequency.

Although the effect of the accompanying components in the above filter development was not taken into account, can such components up to the first order according to fol explanatory note. Because the frequency characteristic of the FIR filter fh (t) usually a is an even number function, assuming a real number, the frequency characteristic of the FIR filter F (ω) is also one even function, assuming a real number.  

If only the range 0 <ω is taken into account for the sake of simplicity, the frequency characteristic of the above-mentioned FIR filter fh (t) can be expressed as follows if first-order accompanying components are taken into account:

Substituting equation (27) into equation (28) gives:

where R (ω) is chosen such that:

R (ω) = (1 / m) G ((ω + 2π) / m) (30)

The term R (ω) exp (2πjh / m) in equation (29) represents an error component that is nothing other than the frequency characteristic of the FIR low-pass filter. This term is responsible for noise because it causes the filter characteristic to change from filter to filter. Assuming that the FIR filters F (ω) and R (ω) represent real numbers, the maximum value for the error is:

where A = F ₀ (ω) + R (ω) exp (2πjk / m). This means that the maximum value for the error is twice the normal value for R (ω). If Nb denotes the bit length of the filter coefficient, the following applies:

It follows that the interval conversion noise can be within the range of the quantization noise. Since Nb = 12 in the sound source unit according to the exemplary embodiment, it is necessary that the following applies to the amplification in the cut-off range:

∥R (ω) ∥ <(2 ^-Nb / 2) = -78 [dB] (33)

The maximum shift in filter characteristics is between two sets of filters that are separated by m / 2 because then the phase of R () reverses. At interval conversion can occur when the interval is 5 points is increased or decreased by an octave, so that it is at most likely is that the filters above alternate to be selected.

These conditions for the inhibit range are quite strict compared to what when developing an ordinary Filters must be observed. When developing a habit filter is not attempted, ripples in the inhibit suppress area to this extent, but it will at the same time tries to improve the cut-off characteristic fibers. However, if a signal to noise ratio is created should be that the bit length of the filter coefficients of Interval conversion filter corresponds to the above conditions at the expense of other conditions the.

The following is the development technique described above summarized for an interval conversion filter.

In the case of an interval conversion filter fh (t) with the resolution R and with n orders, each filter of the filter unit must satisfy the following characteristic:

F _h (ω) = F ₀ (ω) exp (jωh / m) (34)

The filter unit must be created by m-th Coefficient from a low-pass filter (n × m) -th order is taken out and the taken out coefficients be arranged in succession.

For the characteristics of the filter sets to coincide and for the suppression of interval conversion noise, it is sufficient to suppress the gain in the inhibit range of a prototype of a low-pass filter, so that:

∥R (ω) ∥ <(2 ^-Nb / 2) (35)

If this principle is applied to the source sound unit according to FIG. 1, an interpolation filter is used for the interval converter 131 , which filter is created by the development method specified above.

The interval converter 131 converts the interval or the pitch of the analog signal when it is converted from the signal Sd so that the interval corresponds to that as specified by the key pressed on a keyboard. This is done by 256-fold superordinate sampling of the signal Sd, followed by a back sampling, and thereby without changing the sampling frequency for the signal Sd.

The interpolation filter described above includes a thinning filter that thins sample data.

With the interpolation filter described above the filter sets of the arrangement have the same amplitude characteristics ten so that noise, which otherwise by switching filters is generated, is excluded. So it will be a scan reproduced with a very high signal / noise ratio.

To improve the transition between signals on one Loop start point and a loop end point are at embodiment described above, the parameters for the loop start point based on the data for the Loop start point and the loop end point are formed. Instead, PCM data for multiple blocks of the Loop start blocks are used.  

Generally speaking, there is a method of compression and encoding sound data into compressed data words and parameters that apply to compression from digital Data are formed using an analog signal pattern for correspond to a predetermined number of periods. Ent each block holds a predetermined number of samples as Unit, and there will be one or more compression code blocks are formed, each of which has a predetermined number of contains compressed data words and the parameters for them. These compression code blocks are stored in a storage medium stored. One improvement is that PCM Words in a predetermined number as starting words at least in first compression code block can be saved.

Referring to Fig. 28 demonstrates that a loop domain-LP corresponds to a predetermined number of periods of the analog signal. The initial words of a predetermined number in the first compression code block BL of the loop domain LP are direct PCM words W _ST . The number of direct PCM words should not be less than the number of orders when compressing the coding.

At least the first block indicates that all blocks in the to be processed above.

As explained, at least the above are initial Words direct PCM words. Then when grinding together can hang these direct PCM words directly as Loop start point data are used so that it is not is required predictions from data in the neighborhood of the loop end point, which causes the effects can be inferred from past data be called.  

A structure and a function have been described above, how they are used to generate and record source sound data is required, d. H. for storing in a memory. The Structure and the function for the playback page is shown in fol explained, d. H. the functions for reading out, decoding ren and z. B. Loops. The playback page can can only be used as source sound arrangement.

Fig. 29 is a block diagram illustrating a display device for reading out source sound data, which were generated with the above-mentioned method and then stored in a source sound data memory 211 . The decoding takes place with the decoder 90 according to FIG. 18, or the above-mentioned loop formation takes place.

Source sound data are supplied from a device 210 for supplying source sound data to a memory 213 which has the source sound data memory 211 and an address data memory 212 . The means 210 for supplying source sound data may be the above-mentioned means for generating and / or recording source sound data as a whole, or it may be a device for reproducing data from a recording medium, e.g. B. from an optical disk, a magnetic disk or a magnetic tape.

The slider 92 , the summing point 93 and the prediction element 94 of the decoder 90 of FIG. 18 correspond to a slider 232 , a summing point 233 and a prediction element 234 of FIG. 29. The circuit of FIG. 29 essentially performs the function of the decoder 90 of FIG Fig. 18. The decoded source sound data are by envelope addition and z. B. Hall effect generation processed and supplied via a damping noise lock 236 to a D / A converter 237 . The analog signal output by this converter is reproduced by a loudspeaker 238 as a musical tone signal.

The circuit of FIG. 29 has an address generator 220 which is responsive to a key-on signal in order to read out source sound data from the memory 213 , in particular for loop formation. The address generator 220 has an address register 221 to hold the data start address SA for the address data memory 212 . There is also an address counter 222 which is read with the address data value and which is supplied with clocks in order to carry out a counting procedure. A multiplexer 223 receives the address data from the address counter 222 .

The charge control input of the address counter 222 is supplied with timing pulses CP _A from a connection 224 via an AND gate 225 , which is controlled by the output signal from an OR gate 226 . A table of contents address generator 228 is also provided in the address generator 220 and supplies output data to the multiplexer 223 . The multiplexer 223 either selects address data from the address counter 222 or directory address data and controls the memory 213 with the respective address value.

A subordinate data register 214 stores the header data, ie the parameter data relating to the compression, or also sub-data. This takes place when a synchronization pulse CP _B described below is fed to a connection 215 . Loop data L1, ie data indicating whether there is loop data in the sub-data which the register 214 accesses, is passed to a flip-flop 218 via an inverter 216 and an AND gate 217 . End data EI, ie data indicating whether a block is an end block for a signal, are supplied to the AND gate 217 and the OR gate 226 . The output signal from the AND gate 217 reaches the set input S of the flip-flop 218 . The reset input R receives the tone generation start signal, ie the signal key-on K0N via a connection 219 . This key-on signal K0N is also applied to the OR gate 226 and the directory address generator 228 . It includes not only key-on data for the electronic musical instrument, but also a sound generation start trigger signal for starting the software for automatic music generation.

Fig. 30 shows an example of the content of the memory 213 , that 29529 00070 552 001000280000000200012000285912941800040 0002003943795 00004 29410 is 64K byte RAM. It is divided into the source sound data memory 211 and the address data memory 212 . The address data memory 212 is part of the so-called directory area in the memory in which the aforementioned start address data SA and the loop start address data LA are stored, which are accessed with the help of the directory addresses from the directory address generator 228 . The leading address area SDF consists of addresses for consecutive samples which correspond to the formant area FR of the audio signal. Leading addresses of a data area SDL are designated by LSA in FIG. 30. This data also corresponds to successive samples which are assigned to a loop area LP. The example of FIG. 30 shows source sound data SDS1, which consist of formant data SDF1 and loop data SDL2, denoted by SA1 and LSA1, and source sound data SD2 by SDF2 and SDL2, denoted by SA2 and LSA2, and data SDF2, denoted by SA3. However, there can only be data with values for loop areas. The address data SA and LSA in the practical embodiment, for example, designate only the addresses of the header data or sub-data RF of Fig. 19, with compression coding blocks as units. More detailed address display, such as B. address indications on a byte basis, is performed by the address counter 222 .

How the arrangement described above works will be explained based on the timing chart of FIG. 21.

A multiplexing process is illustrated in FIG. 31, the sampling period being designated T _S. At a sampling frequency of 32 kHz, the sampling period T _{S is} 1/32 ms. Each sampling period T _S is divided for the reproduction of a predetermined number of voices. In the exemplary embodiment, there are 8 voices (voices 0-7). The time period allocated to each voice depends on the multiplex procedure. The minimum time unit for the multiplexing process is the time unit τ. Time intervals τ ₀ and τ ₁ serve to access the addresses SA or LSA in the directory area. Spans of time τ ₂ -τ ₅ serve to access bit compression coding data. The address counter 222 is updated in a time interval τ ₆ . In the time interval τ ₂ , the sub-data (header data RF according to FIG. 19) are accessed, while in the time intervals τ ₂ -τ ₅ the scanning data (data D _A0 -D _B3 from FIG. 19) are read. The syn chronizing pulse CP _A is read out in the time period τ ₅ , while a synchronizing pulse C _{B is} output in the time period τ ₂ .

The key-on signal K0N for starting sound generation is output during a sampling period T _S , ie it is at a high level in the interval between the times t ₀ and t ₁ . A standby signal STBY drops with the leading edge of the signal K0N and rises to a high level again after a few sampling periods, in the example shown after five sampling periods, i.e. at time t ₅ .

When the key-on signal K0N is applied to terminal 219 , directory address generator 228 generates a directory address based on the offset address for the memory specified by the CPU. In addition, a source number is displayed, which indicates the type of sound source. The directory address is given to the multiplexer 223 . During the time-division multiplexing intervals τ ₀ and τ ₁ , the multiplexer 223 accesses the addresses which are supplied by the directory address generator 228 in order to access the memory 223 in order to read out predetermined address data in the address data memory 212 , ie data SA which contain the source audio data. Show the start address for the given tone number. This makes it possible to read the data SA into the address register 221 via a data bus. The signal K0N is also supplied via the OR gate 226 to the AND gate 225 , whereby this switches through and the pulse CP _{A is} fed to the charge control input of the address counter 222 at the time of the time window. As a result, the start address data SA stored in the address register 221 is loaded into the address counter 222 in order to set it. The address counter 222 counts from this data value SA, so that those source sound data SDF for which the data value SA represents the start address are read out in the specified address sequence. If loop data SDL follows the data SDF, this loop data SDL is automatically accessed after the data SDF.

On the sampling period that follows that in which the signal K0N is output, ie after the sampling period t ₁ , the signal K0N returns to its initial state of low level. The directory address generator 228 then outputs the start address LSA for the loop data SDL. The address generator 228 provides the loop start address LSA. However, as long as the input counter 228 is not supplied with an input signal at its charge control connection, it does not receive the new address but continues to count. This does not apply if the input data is formed only by data SDF of the formant area.

When the flag for end data EI is set by the subordinate data register 214 , that is, when the end block for loop data SDL or the end block for data SDF of the shape area is reached, the AND gate 225 switches through the OR gate 226 , whereby the loop start address LSA in address register 221 is supplied to address counter 222 when the synchronization pulse CP _{A occurs} . As mentioned above, the addresses are SA or LSA addresses for a bit compression block as a unit, whereby the start block for source sound data is first accessed when the signals for the next block are processed.

The above-mentioned end data value EI is also supplied to the AND gate 217 . The negated output signal of the data value LI, which indicates whether looping should take place or not, is also fed to the AND gate 217 , so that the output signal from the inverter 216 goes high when the source sound data only the first type of data SDF3 ( Data corresponding to the formant area), but loop domain data SDL are missing. When the end block of the source sound data SD3 is reached, the output signal from the AND gate 217 goes high and sets the flip-flop 218 , whereby this drives the damping squelch so that this cuts audio signals. This is the squelch function when there is no looping. If this happens, the loop is played until the next key-on signal occurs. The sound attenuation then takes place with the aid of an envelope characteristic. If the key-on signal K0N is input, it arrives at the reset input R of the flip-flop 218 , which is thereby reset regardless of the previous state, as a result of which the blocking effect of the attenuator 236 is canceled.

When using the two Quelltondaten SD1 and SD2 of Fig. 30, particularly when using the loop data areas SDL1, SDL2 thereof Quelltondaten may for supplying Quelltondaten alternately areas in the memory for SDL1 and are read SDL2 and read out from the external device 210, so that they be decoded in a decoder 230 to perform the decoding of the source sound data for an extended period of time. This means that when reading and decoding source sound data from a storage area for SDL1, SDL2, source sound data from the outer source sound data supply device 210 are written into other storage areas, so that the writing and reading in relation to these storage areas takes place alternately.

This can be achieved very simply by alternating the start addresses LSA1 and LSA2 with one another in order to carry out a loop function. In the memory 213, as is illustrated by Fig. 30, the address data of a registered the Schleifenstartadreßdaten LSA1 in the Schleifenstartadreßdaten LSA2 rewritten in the memory area 212. As a result, the loop start address LSA2 is written into the memory area 212 while the source sound data SDL1 is read out and decoded. The written address LSA2 is transferred to the address register 221 . At the end of the data SDL1, ie when the loop end point is reached, source sound data SDL2 is accessed from the start address LSA2 in that the last-mentioned start address is loaded into the address counter 222 . In the period in which the source sound data SDL2 is then read out and decoded, the loop start address LSA1 is written into the memory area 212 . When the end of the data SDL2 occurs, ie when a loop end point is reached again, the said address LSA1 is loaded into the address counter 222 , so that source sound data SDL1 is now accessed from this address. In this way, source sound data can be decoded for extended periods of time without increasing the hardware expenditure.

It should be noted that digital signal processing with the above-mentioned bit compression and coding or other digital signal processing for generating source sound data is often performed by software that uses digital signal processing. The playback of recorded source sound data is also often accomplished with digital signal processing. Fig. 23 shows an example of an overall system that consists of an audio signal processing device 307 and peripheral devices.

A host computer 304 , e.g. B. a conventional PC, a digital electronic musical instrument or a TV game device is connected to an audio signal processing device 307 , which operates as a sound source unit. Source sound data is loaded into the unit 307 from the host computer 304 . The device 210 for supplying source sound data, which is mentioned in more detail above, is thus provided by the host computer 304 .

The audio signal processing unit 307 consists essentially of a CPU 303 such as a microprocessor, a DSP (digital signal processor) 301 and a memory 302 for storing source sound data. A variety of processing operations, including readout control of the source audio data, looping, bit expansion or restoration, interval conversion, envelope addition, or Hall effect generation, are performed by the DSP 301 . The memory 302 also serves as a buffer for data generated in these processes. The CPU 303 controls the manner in which the operations performed by the DSP 301 are carried out. In addition, the CPU 303 provides for rewriting of address data LSA in the above-mentioned memory 213 (here memory 302 ) or for writing source sound data from the device 210 for supplying source sound data (within the host computer 304 ) into the memory 213 (here 302 ).

The digital musical sound data generated by the DSP 301 after these various processes, resulting from the source sound data from the memory 302 , are converted by a D / A converter 305 (corresponding to the D / A converter 237 ) and output via a loudspeaker 306 .

The following is a generalized explanation of various characteristics of the described device and the described method. One can think of a generalized source sound device as shown schematically in FIG. 33. With her, the number of accesses to a memory is reduced if addresses for various start data are loaded into an address register of an audio signal processing device.

The source sound processing device according to FIG. 33 has a source sound data memory 241 (corresponding to 221 in FIG. 29) for storing source sound data which have a data area SDF with a first sequence of successive sample data and a data area SDL which contains a second number of successive sample values. There is also a start address data memory 242 (corresponding to 212 in FIG. 29) for storing the data start address SA which belongs to the source sound data and for storing a loop start address LSA. An address generator 243 (corresponding to 220 in FIG. 29) serves to generate the read-out addresses for the source sound data memory 241 on the basis of the data start address SA and the loop start address LSA. The above-mentioned data start address SA from the start address data memory 242 is e.g. B. loaded into an address register 244 (corresponding to 221 in FIG. 29) within the address generator 243 . The SDF data from the first number of consecutive samples is read from the memory area, starting with the start address for the source sound data memory 241 . The loop start address LSA from the start address data memory 242 is loaded into the address register 244 , and the data SDL of the second number of consecutive samples are repeatedly read out to reproduce analog or digital audio signals.

In the Quelltonverarbeitungseinrichtung shown in FIG. 33, the data start addresses are provided in the address generator to be able to read out the first number of consecutive samples. The loop start address is then loaded into the address generator in order to be able to read out the second number of consecutive samples. The first number of consecutive samples is then no longer read out until the next source tone generation. In other words, when the key-on signal is generated, either the loop start address or the data start address is loaded. In this way, the data start address is loaded with the key-on signal while the loop start address is loaded differently, whereby the number of accesses to the memory can be reduced, which enables simplified multiplex processing.

Fig. 34 schematically shows a source sound processing device with which it is simplified to judge whether a loop domain is present or whether to stop playing source sound data which is missing a loop domain.

The Quelltonverarbeitungseinrichtung shown in FIG. 34 has a Quelltondatenspeicher 251 (corresponding to 211 in Fig. 29), for selectively storing Quelltondaten that a first type of successive samples Sa including have to be repeated loop domain, and a second type multiple successive samples Sb, where a loop domain is missing. A subordinate data register 252 (corresponding to 214 in Fig. 29) serves to separate out data that is coupled to the source sound data. A Flaggenprüfschaltung 253 (corresponding to 216-218 in Fig. 29) is used to determine of flags, which may be the presence or absence indicating a loop domain in the Quelltondaten, and to determine the end of Quelltondaten. An address generator 254 (corresponding to 220 in FIG. 34) is used to generate addresses for reading out the first type of successive samples Sa and the second type of successive samples Sb from the source sound data memory 251 , with the aid of the source sound data and those subordinate with the register 252 Data. An audio signal processing device 255 (corresponding to 230 in Fig. 29) is used for processing, e.g. B. decoding, the source sound data based on the discarded data, to generate reproducible data. An attenuator 256 (corresponding to 236 in FIG. 29) serves to attenuate the source sound data obtained by processing with the aid of the flag mentioned above within the flag checking circuit 253 . The first or the second type of successive samples Sa or S is repeatedly read from the source audio data memory 251 to produce analog or digital audio signals, while attenuating whenever the associated flag indicates that the source audio data lacks a loop domain , or that the reading of the source sound data has ended.

The Quelltondatenverarbeitungseinrichtung shown in FIG. 34 bil det a flag indicating the end of data or the lack of a loop domain. The output signal is then attenuated. This can simplify loop control without the need for additional data for loop formation. Similar effects can also be achieved by using, in addition to the flag indicating the presence or absence of a loop domain, a flag indicating an end with a loop or an end without a loop.

Fig. 35 is a block diagram showing the manner in which source sound data is continuously read out and reproduced.

The arrangement of FIG. 35 has a Quelltonspei cher 261 (corresponding to the memory 212 of Fig. 29) having a first Quelltonspeicherbereich 261 a and a second Quelltonspeicherbereich 261 b. An address forming circuit 236 forms readout addresses based on the start address as supplied by an address register 262 . A control device 264 ensures that data are read out alternately from the two named memory areas 261 a and 261 b on the basis of the delivered address. A device for supplying source sound data 265 supplies data which are written into one of the two memory areas 261 a, 261 b each time data is being read from the other area. A device 266 for supplying start addresses is provided, which supplies the address register 262 with start addresses for the two memory areas mentioned. A signal processing device 267 serves to process the source sound data read from the two memory areas 261 a and 261 b.

With the arrangement according to FIG. 35, it is possible due to the alternating reading and writing of data into the memory areas or from the memory areas to reproduce source sound data continuously without noise. Loop playback of source sound data that includes unsynchronized signal patterns or repeated signal patterns is enabled.

This data playback is possible by changing the start addresses in the address register without adding hardware he must and without a special timing control is required.

The arrangements according to FIGS . 33-35 served to illustrate three properties of a device for reproducing source sound data.

If the technology explained with reference to FIG. 28 is used on the source sound data generation and storage side, ie if direct PCM data in a predetermined number of leading words are used in a start block for the loop domain, an arrangement as shown in FIG Fig. 36 is shown. A block for the bit compression data is substantially the same as that of Fig. 19. Instead of the flag indicating the presence or absence of a loop and the flag indicating the end of a loop, two 1-bit flags are used, namely Lich data indicating a block with a loop start point, ie a loop start flag LSF, and data indicating a block with a loop end point, ie a loop end point flag LEF.

Compressed and encoded sample data and parameter data (header data RF and subordinate data, as shown in FIG. 19) and sound data are stored in a memory 271 within the block diagram of FIG. 36. The loop start address (address for the loop start block in the memory) is also stored there, specifically as a directory data value. At least the loop start data is transferred to the address register 272 via a data bus from the memory and stored there so that it can be entered into an address counter 273 directly from the end of a loop. Depending on the output signal from the address counter 273 and the loop start flag LSF, a decoder 274 outputs a control signal for direct PCM signals, specifically for the period of time within which a predetermined number of start words lies in the loop start block, ie in the period of time in which the direct PCM data are output. If source sound data for different tones are stored in the memory 271 , the start addresses of the source sound data and the start addresses of the loops are stored there. In response to a signal, e.g. B. the signal key-on, the start addresses for those sound signal data are read out, which are assigned to a preselected sound source. These addresses are read out from the directory address area of the memory 271 and read into the address counter 273 via the address register 272 in order to then read out data which are stored from the first address. If only one type of source sound data is stored, the counter 273 only has to count from a predetermined value (start address of the source sound data) when the key-on signal is present.

From the source sound data, as they are read out from the memory 271 , the aforementioned parameter data, ie the subordinate data and the header data RF according to FIG. 19 are transferred to a sub-data register 275 and temporarily stored there. The scan data D _A0H -D _B3L according to FIG. 19, ie 4-bit compressed data, are transferred to a data register 276 and _buffered there. The direct PCM data likewise stored in the memory 271 for the predetermined number of initial words in the loop start block are transmitted to a multiplexer 278 .

The data temporarily stored in the data register 276 is transferred to a bit shifter 277 , which corresponds to the shifter 92 of the decoder 90 of FIG. 18. The output signal from the bit shifter 277 is supplied via the multiplexer 278 to a summing point 280 which corresponds to the summing point 93 of FIG. 18. The output signal from the summing point 280 is supplied to an output register 286 and at the same time to a prediction circuit 291 , which corresponds to the prediction element 94 of FIG. 18. The prediction circuit 291 has two delay registers 284 and 285 , coefficient multipliers 282 and 283 assigned to these registers, and a summing point 281 . At the summing point, the products are formed from the output signals and from the delay registers with coefficients, as are supplied by a coefficient generator 279 . The prediction circuit 291 , the summing point 280 and the bit shifter 277 form a decoder 290 for decoding the bit-compressed, encoded data.

From the header data FR, which are temporarily stored in the sub-data register 275 , the area data and the window selection data are transmitted to the bit shifter 277 and to the coefficient generator 279 , respectively. The loop end flag LEF and the loop start flag LSF are applied to the set control input for the address counter 273 and to the decoder 274, respectively.

The 4-bit compressed data mentioned is decoded in the decoder 290 of FIG. 36, in accordance with what has been explained with reference to the decoder of FIG. 18. This outputs 16-bit amplitude data, which is supplied to a D / A converter 287 via the output register 286 . The converter 287 outputs an analog musical tone signal to an output terminal 288 . The following explains how loop playback returns from a loop end point to a loop start point. When source sound data for the block containing the loop end point is read out, the loop end flag LEF of the subordinate data is set. When the loop end point is reached, the loop start point address, as temporarily stored in the address register 272 , is transferred to the address counter 273 , since the loop end flag LEF gives the instruction for this. Since the address counter 273 ensures that the loop start block in the memory 271 is accessed via an address bus. As a result, the direct PCM data for the specified number of leading words are read out. While this direct PCM data is being read out, the decoder 274 executes a control in which it gives the switching control signal for the direct PCM data to the multiplexer 278 and the coefficient generator 279 to the direct PCM data directly to the output register 286 to deliver. The multiplexer 278 therefore selects the direct PCM data from the memory 271, while the coefficient generator 279 such coefficients to the multipliers 282 and 283 delivery, which form a filter 0-order, so that the straight PCM data as an output signal from the summing junction 280 be preserved.

Since the direct PCM data directly as loop start point data There is no need to use Before say to hit for the loop endpoint data. As a result, len with good effect mistakes, which otherwise through discount tinctures occur at the loop points.

In the case described, only the direct PCM data for the loop start point LP _{S are} added to the source sound data, while the overall compression rate remains unchanged. This means that the storage capacity does not have to be increased.

The fact that immediate PCM data in the initial area of a Loop domain are arranged, which is done in that a predetermined number of initial words in a block, the follows the first block of a loop domain from immediately possible PCM data words, it becomes possible for errors to avoid when reading, if the data from a belie other block can be read out and reproduced. The number of bits of immediate PCM data can be compared to the number of bits for the compression data be the same, for example the Number also with the number of bits of the original sampled Amplitude data is the same. Immediate PCM data included can be compressed with the area data of a block to be used.

Claims (2)

1. An interpolation filter for processing input data supplied by digital sound sources, which are sampled at a sampling frequency fs, the sampling frequency being assigned a corresponding sampling period, comprising:
a filter arrangement consisting of m sets of filters of the n-th order for finding digital data for m interpolation points which lie within the sampling period, as a result of which interpolation data are generated with a resolution of mf _s , where m ≧ 2 applies,
characterized in that
the m sets of nth order filters each have the same amplitude characteristic. 2. Interpolation filter according to claim 1, characterized by n coefficient multipliers ( 151 ), (n-1) delay units ( 152 ) and a summing unit ( 153 ).