JPS60182056A - System for detecting deterioration of signal - Google Patents

Abstract

PURPOSE:To attain simple automatic equalizing system with high accuracy by detecting the deterioration of an amplitude characteristic of a recording and reproducing system form a level difference between a peak level of an envelope of a frequency in a signal reproduced from a burst signal and a level at a DC period. CONSTITUTION:Circuits comprising an integrator 3 - an AD converter 7 process a signal and display the deterioration of the frequency characteristic of the recording and reproducing system in a binary number. A data processing section 8 generates a control signal required to control a frequency characteristic compensating device 2 based on the binary data. The basic operation of the data processing section 8 is as follows: A false sample value is eliminated and the mean value of samples of frequency corresponding to the same frequency of a multi-burst signal is obtained. Then each frequency ratio of the ideal characteristic satisfying the Nyquist condition is obtained in advance and incorporated in the memory. The ratio of each frequency is obtained by using a data obtained from the recording and reproducing system and subject to the said processing. The error (square mean error) between the ratio and the said ideal value is obtained and the frequency characteristic compensating circuit 2 is set so as to minimize the mean square error E.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a method for automatically compensating for deterioration in the frequency characteristics of a reproduced signal in digital magnetic recording.

[Background of the invention]

When a digital signal, that is, a pulse, is transmitted through a communication line or recorded on a recording medium such as a magnetic tape, deterioration in frequency characteristics usually occurs. Therefore, compensation is required.

This compensation method has traditionally been a major research topic in the field of communications, and several effective methods have been developed. First, the basic frequency characteristics necessary for pulse transmission are given as Nyquist conditions, as is well known. The pulse transmission rate is f
Then, the ideal band is 0 to f as shown in Figure 1 (a).
/2. Further, as shown in FIG. 1(b), any roll-off characteristic (for example, linear or cosine characteristic) can be provided as long as it is oddly symmetrical about f/2.

Satisfying this Nyquist condition is a sufficient condition for pulse transmission using non-symbol interference. Further, an operation that satisfies such conditions is called waveform equalization.

Regarding the so-called automatic equalization method for automatically performing this waveform equalization, a method using a transversal filter has been proposed. (Advances in PCM communication: Miyayo et al. 1:
Product information, please refer to the degree. ) This method has the following characteristics: (1) Highly accurate equalization is possible.

(2) Complex calculations such as square calculations are required.

(3) The circuit scale increases.

etc. In particular, in the case of high-speed data transmission of several M pitches or more, each squaring operation must be performed in an analog manner, making it difficult to ensure calculation accuracy. In addition, the circuit scale becomes large.

In general, in magnetic tape devices, magnetic disk devices, etc., the frequency characteristics during reproduction change due to head wear, changes in the amount of spacing between the tape and head, or differences in the amount of recording current.

In particular, as the recording density increases, the deterioration of frequency characteristics due to the above-mentioned causes becomes more significant. Therefore, processing such as waveform equalization and automatic equalization is required. However, in general, the transmission rate of magnetic recording/reproducing devices is high, such as several megapips to 1/s or more, and they are often multi-channel. Therefore, it is desired that the automatic equalization circuit be as small as possible, inexpensive, and have high performance.

[Purpose of the invention]

An object of the present invention is to provide a simple and high-performance waveform deterioration detection method and automatic equalization method for compensating for waveform deterioration occurring in a magnetic recording/reproducing system.

[Summary of the invention]

First, in order to detect waveform deterioration, a multi-burst waveform is used in which burst waves of the Nyquist frequency and several types of frequencies before and after the Nyquist frequency are sequentially arranged over a predetermined time period. Furthermore, a DC period is provided for these multi-burst signals.

This waveform is shown in FIG. 2 (a). In the same figure, f1~
f4 are the repetition frequencies f, , f, , f, , f, respectively.
is a burst pulse train (a sine wave burst may be used), and 1 is a DC period. Next, this waveform is recorded and then played back. The waveform after reproduction is shown in FIG. 2 (b). Generally, in a wound head, the output has a differential waveform as shown in FIG. 2 (b). Therefore, this waveform integrator performs integration as shown in FIG. 2 (c). Next, this waveform is rectified in both waves, and its peak value is sampled and held. This waveform is shown in FIG. 2(d). This waveform corresponds to the deterioration of the frequency characteristics of the recording/reproducing system. Next, values corresponding to each frequency level are successively sampled using the level of DC period 1 of waveform (d) in FIG. 2 as a reference. The waveform obtained as a result is shown in FIG. 2 (E). Now, the more types of frequencies that make up a multiburst signal, the more accurate the detection of frequency deterioration will be. However, as the types of frequencies increase, the circuit scale or the time required for frequency characteristic correction described below increases. For this reason, it is necessary to know the minimum number required.

To construct a Nyquist characteristic with linear roll-off (shoulder reduction), first define a frequency f below f2 where the roll-off characteristic does not apply, a roll-off starting point frequency f2, a Nyquist frequency f3, and a high-frequency characteristic above Nyquist. If you select the 4 points of f4 required for this, fortune telling.

fl may be set to as low a frequency as possible. Further, f4 may be selected between the Nyquist frequency and the frequency at which the amplitude is 0. Furthermore, in order to construct a Nyquist characteristic with a roll-off due to a curve such as a cosine wave, it is necessary to newly select a frequency between the roll-off point and the Nyquist frequency. As an example, assume that a straight line rolloff is to be performed. As shown in Figure 3, the values of each frequency in an ideal state where the frequency characteristics satisfy the Nyquist condition are expressed as
xfx.

IJ'atIf4. On the other hand, the sample value of the frequency characteristic obtained in the recording/reproducing system is Gf, , G: F2°af, , a
Let it be f4. Generally, when the frequency characteristics deteriorate, I f2/
I f, > Gf, /Gf,, I f3/I f, > Gf
, /Gf, , I f4/I f,>Gf4/Gf, holds true. Therefore, making the corresponding Rakuhoku values the same compensates for the frequency characteristics. That is, If
, /If,=Gf2/Gf, etc. It is usually difficult to perform such calculations using unlog signals as they are from the viewpoint of calculation accuracy. For this reason, it is desirable to quantize the above-mentioned sampling values using an AD converter and convert them into binary numbers. FIG. 4 shows a block diagram of the automatic waveform equalization method of the present invention. The integrator 3, double-wave rectifier 4, peak detector 5, sample hold circuit 6, and AD converter 7 are used to perform the signal processing described above and to display the deterioration of the frequency characteristics of the recording/reproducing system in binary numbers. be. The data processing section 8 generates a control signal necessary to control the frequency characteristic compensator 2 based on this binary data. The basic operation of the data processing section 8 is as follows.

(1) Data Determination In general, in tape head systems, signal loss, so-called dropout, occurs due to dust or scratches on the tape. For this reason, the signal level of the sample belt shown in FIG. 2 (e) also drops extremely when dropout occurs. It is necessary to remove such false sample values.

(2) The averaged sample value of data is not necessarily constant due to changes in the contact state of the tape head. For this reason, it is necessary to calculate the average value of sample values of portions corresponding to the same frequency of the multiburst signal.

(3) Comparison of data Frequency ratios of ideal characteristics that satisfy the Nyquist condition are determined in advance and stored in a memory (for example, ROM).

Furthermore, the ratio of each frequency is calculated using the above-mentioned processed data obtained from the recording/reproducing system. Calculate the error (mean square error or absolute value) between this value and the above ideal value. Note that it is necessary to weight the error at each frequency point in consideration of the convergence conditions. Letting the total mean squared error be E, E=α1If2 /INF, -Gf,, /Gf,
+2+βI■:F3/If, -Gf3/Gf, +2+
γIIf4 /I:F, -Gf4/Gf, +2α〉β〉
γ α, β, and γ are weights.

The frequency characteristic compensation circuit 2 is set so that the mean square error E is minimized.

Now, there are several types of frequency characteristic compensation circuits, including transversal filters and LCR resonant circuits, but here we will use a simple compensation circuit.
The invention will be described in detail regarding the case where a resonator made of CR is used.

Let us assume a case where the recording current changes in an all-magnetic recording/reproducing device. Reproduction 411 recorded with a certain appropriate recording current
Assume that the frequency characteristics of the signal are corrected so that it satisfies the Nyquist condition. This state is shown in FIG. 5(A). When the recording current increases from this state, the signal output decreases in the high frequency region, and the frequency characteristics of the reproduced signal deteriorate as shown in FIG. 5(C). Conversely, when the current decreases, the signal becomes emphasized in the high range, as shown in FIG. 5(B). In order to compensate for such changes in characteristics, a series resonant circuit as shown in FIG. 6 is used. In other words, by changing the capacitance C, it is possible to control the resonant frequency, and by changing the resistance R, it is possible to control the effective Q value.Using this change in transfer characteristics, it is possible to control the resonant frequency as shown in Figure 5. Changes in the frequency characteristics of the signal can be compensated for.

Now, the optimal value of this compensation amount can be determined by sequentially changing the resonance frequency and the Q value at that frequency, and determining the value that minimizes the value of the squared error E.

[Embodiments of the invention]

The present invention will be explained in detail below using examples. In Figure 7,
A specific example of the configuration of the main part of the present invention will be shown.

The capacity I of the resonant circuit in the frequency characteristic compensation circuit 2 is controlled by varying the voltage applied to the varactor diode 9. This control voltage is obtained from the output of the I/A converter 10, and required binary data is given to the D/A converter 10 from the microcomputer 18 via the interface 15. Similarly, the resistance value within the resonant circuit can be varied by turning on and off the analog switches 11, 12, and 13. Further, a signal for controlling the on/off of the analog switch is given from the microcomputer 18 via the interface 16.

Next, the signal after frequency characteristic compensation is applied to the input terminal of the integrator 3. The integrator 3 is composed of an amplifier 14, a capacitor, and a resistor. The integration time constant must be selected so that sag due to resistive cutoff is not a practical problem.

This integrator output is then connected by switch 19 to a series of signal processing circuits, such as a double-wave rectifier, even during multi-burst signal periods. Further, after setting the constants of the frequency characteristic correction circuit, by switching the switch 19, a waveform-equalized pulse is output. The signal processing circuits after the integrating circuit are commonly used, and their explanation will be omitted here.

Now, the output signal of the sample hold circuit 6 is transferred to the AD converter 7.
The selection of the number of bits is important. In other words, the amount of code interference remaining after waveform equalization is determined by the number of quantization bits. However, as the number of bits increases, the calculation time and circuit scale increase. Therefore, it is necessary to select a pitch number suitable for the magnetic recording/reproducing system. Magnetic recording/reproducing systems inherently have nonlinear characteristics. In other words, waveform distortion caused by the fact that the superimposition side does not hold remains about 1/10 to 1/20 of the signal level. This amount of nonlinear distortion serves as a guideline for determining the number of quantization bits.

That is, it is sufficient to ensure accuracy of approximately 4 bits. However, if rounding errors during calculation are taken into consideration, about 5 to 6 bits are required. FIG. 8 shows flowchart 1 of the calculation process. The previous sample value and the squared error are successively compared and the minimum value is found. Initially, the resonance frequency is set as high as possible and the Q value is set as low as possible. This is a state in which the amount of correction is at its lowest. After this state is revised, the resonant frequency is successively shifted to a lower state, and the Q value is changed each time to obtain a square error. If this operation is continued, the error will decrease and eventually reach the minimum value, and if you go one step too far, the error will increase. This allows the minimum value to be identified. Furthermore, in order to bring the resonant frequency to a state where -b is low, the reverse bias voltage of the varactor diode 9 should be made the lowest. For this purpose, for example, an all II OII signal is applied to the D/A converter 10 from the microcomputer, and the DC voltage is adjusted so that the output voltage of the D/A converter 10 in this state satisfies the above-mentioned conditions. Regarding the Q value, the value can be controlled from a low state to a high state by sequentially turning on similar analog switches from an off state.

Further, the multi-burst signal is repeatedly recorded and reproduced for a sufficient period of time to allow the above calculation to converge. Therefore, a training period in which multi-burst signals are recorded and reproduced is required before the actual data begins. Furthermore, by intermittently recording and reproducing the multiverse +-4g signal during the data period, it is possible to always set the constant of the frequency compensation circuit to the optimum value.

〔Effect of the invention〕

As described above, the present waveform detection method and the automatic equalization method using it have a simple and highly accurate equalization ability, and are therefore particularly suitable for compensating the frequency characteristics of a magnetic recording/reproducing system.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing Nyquist conditions, Fig. 2 is a waveform diagram showing the multiburst waveform used in the present invention and its signal processing, Fig. 3 is a diagram showing selection points of frequency characteristics, and Fig. 4 is a diagram showing the present invention. FIG. 2 is a block diagram of an automatic equalization method according to FIG. FIG. 5 is a schematic diagram of the frequency characteristics of the magnetic recording/reproducing system. FIG. 6 is a circuit diagram showing a configuration example of a frequency characteristic compensation circuit used in the present invention, FIG. 7 is a circuit diagram showing a specific example of an automatic equalization method according to the present invention, and FIG. 8 is a flowchart of calculation processing 1- shows. f, ~f4... Frequency of multiburst signal, 1...
・DC period of multi-burst signal, 2... Frequency characteristic compensator, 3... Integrator, 4... Double wave rectification? 1), 5・
...Peak detector, 6...Sample hold circuit, 7
・・A/Ura 1 Figure 11 Tears @ (Map/, l Hill No. 3 Illustrated stream winding 4 Figure 5 (2) No. Otsu Figure 3 Gigyo)

Claims (1)

[Claims] 1. A burst-like signal consisting of a burst of two or more frequencies including the Nyquist frequency and a DC period is recorded in a recording/reproducing device, and the envelope of the frequency in the signal reproduced is obtained. 1. A signal deterioration detection method comprising: detecting a peak level of a signal, and detecting deterioration of an amplitude characteristic of a recording/reproducing system at the frequency based on a level difference between this peak level and a DC period. 2. In the detection method described in item 1, the burst frequency constituting the burst signal may be a frequency at or near the Nyquist frequency, a frequency at which roll-off begins or any frequency below this frequency, and a Nyquist frequency and an amplitude level. A signal deterioration detection method characterized in that a minimum number of frequencies located between a frequency of 0 and a frequency of 0 are included.