JP7358420B2 - Signal processing device, magnetic tape cartridge, magnetic tape reader, processing method of signal processing device, operating method of magnetic tape reader, and program - Google Patents

Description

The technology of the present disclosure relates to a signal processing device, a magnetic tape cartridge, a magnetic tape reader, a processing method for a signal processing device, an operating method for a magnetic tape reader, and a program.

Patent Document 1 describes a write head that writes data on a recording medium, a read head that reads data written on the recording medium, and a device that digitizes an analog reproduction signal read by the read head and outputs a digital reproduction signal. A digital signal recording and reproducing apparatus is disclosed, which includes an A/D converter that inputs a digital reproduction signal and a signal equalization circuit that inputs a digital reproduction signal. In the digital signal recording and reproducing device described in Patent Document 1, a signal equalization circuit performs nonlinear equalization to reduce the influence of crosstalk that occurs when a write operation of a write head occurs simultaneously with a read operation of a read head. .

Furthermore, in the signal recording/reproducing device described in Patent Document 1, the signal equalization circuit includes a first input layer, a second input layer, an intermediate layer, and an output layer. The first input layer is a plurality of units that input a digital playback signal, delay the playback signal by a predetermined time by a plurality of continuously connected first delay elements, and output the respective delayed digital playback signals. has. The second input layer inputs a recording signal used when the write head writes data, delays the recording signal by a predetermined time by a plurality of continuously connected second delay elements, and records each delayed recording signal. It has multiple units that each output a signal. The intermediate layer receives the outputs of each unit of the first input layer and the second input layer as input, and calculates the sum of products of the nth (n is a natural number) equalization coefficient determined by learning and the output of each unit of the previous layer. It consists of n layers that convert and output using a nonlinear function. The output layer outputs the sum of products of the output of each unit of the intermediate layer and the (n+1)th equalization coefficient determined by learning. The first to n+1 equalization coefficients are determined by learning to minimize the equalization error between the output of the output layer and a predetermined equalization target value.

Patent Document 2 discloses a magnetic disk device having a head for recording and reproducing data on a disk, which is a recording medium, and a data recording and reproducing processing means for processing a recording and reproducing signal to the head. There is.

In the magnetic disk device described in Patent Document 2, the data recording and reproducing processing means includes a storage means for storing recording guarantee information for each predetermined data recording unit corresponding to a physical position on the disk, and a storage means for storing recording guarantee information for each predetermined data recording unit corresponding to a physical position on the disk, Recording compensation processing means for reading recording compensation information corresponding to the data recording area of the data recording unit to be accessed from the storage means and executing recording compensation processing for each data recording unit on the data recording area. do.

Patent Document 3 discloses a multi-track reading circuit equipped with a plurality of adaptive equalizers corresponding to a plurality of tracks, which equalize signals read from a storage medium via a head in accordance with the characteristics of a recording/reproducing system. input signal switching means, which is provided on the input side of the adaptive equalizer, and has a function of connecting the signal read out via the head and each adaptive equalizer in correspondence, and changing this correspondence relationship; A multi-track reading circuit is disclosed.

The multi-track reading circuit described in Patent Document 3 determines whether or not there is periodicity in a signal input to an adaptive equalizer that is performing adaptive learning, and stops adaptive learning if there is periodicity. periodicity determining means for outputting a signal that causes

Patent Document 4 discloses a data reproducing device that reads recorded information recorded on a magnetic recording medium and reproduces the input signal information while adjusting system characteristics through an adaptive equalization operation according to the characteristics of the read input signal information. Extracts the characteristics of a playback input signal of a specific area or a specific pattern area on a recording medium, and generates an equalized output signal of the playback input signal of the specific area or specific pattern area and an expectation of the specific area or specific pattern area. A method for controlling an adaptive equalizer is disclosed in which adaptive learning is performed on an adaptive waveform equalizer using a difference signal from a value signal, and adaptive equalization is performed in accordance with the result of the adaptive learning.

Patent Document 5 includes a magnetic tape, a reading element unit, and an extraction section, the magnetic tape has a magnetic layer containing ferromagnetic powder and a binder on a non-magnetic support, and the magnetic layer includes: The ferromagnetic powder has a servo pattern, and the ferromagnetic powder is a hexagonal ferrite powder, and the peak of the diffraction peak of the (114) plane of the hexagonal ferrite crystal structure determined by X-ray diffraction analysis of the magnetic layer using the in-plane method. The intensity ratio of the peak intensity Int(110) of the diffraction peak of the (110) plane to the intensity Int(114), Int(110)/Int(114), is 0.5 or more and 4.0 or less, and The directional squareness ratio is 0.65 or more and 1.00 or less, and the reading element units each read data from a specific track area including the track to be read out of the track areas included in the magnetic tape using a linear scan method. The extraction unit performs waveform equalization processing on each of the reading results for each reading element according to the amount of positional deviation between the magnetic tape and the reading element unit. , a magnetic tape device is disclosed in which data originating from a track to be read is extracted, and the amount of deviation is determined according to the result obtained by a servo element reading a servo pattern included in a magnetic layer of the magnetic tape. .

Patent Document 6 describes a data reproducing apparatus that reads Nw recording tracks on a recording medium using Nr reproducing heads (Nw≦Nr), and separates and retrieves data from the Nw recording tracks. a signal amplification section that amplifies the amplitude of each of the Nr reproduction signals read from the reproduction head; a quantization section that quantizes each of the amplified Nr reproduction signals; and a quantization section that quantizes each of the Nr reproduction signals that have been amplified; An equalization section that equalizes the frequency response of the signal using a filter with a fixed state and outputs Nw output signals, a synchronization section that synchronizes the bits of each of the Nw equalized output signals, and a synchronization section that synchronizes the bits of the Nw equalized output signals. a detection section that detects each of the Nw output signals, a gain control section that controls the amplitude of each of the reproduction signals read from the Nr reproduction heads based on the detection results in the detection section, and bit synchronization. A data reproducing apparatus is disclosed that includes an adaptive equalization control section that calculates filter coefficients of a filter based on the output signal obtained by the above-described output signal and the detection result of the detection section.

Japanese Patent Application Publication No. 2004-095096 Japanese Patent Application Publication No. 10-091908 Japanese Patent Application Publication No. 04-121803 Japanese Patent Application Publication No. 05-067374 JP2020-009517A JP2008-282477A

However, the technology applied to the digital signal recording and reproducing device described in Patent Document 1 is designed to reduce the influence of crosstalk that occurs when the write operation of the write head occurs simultaneously with the read operation of the read head. Since this is a targeted technology, it is difficult to reduce distortion that nonlinearly occurs in reproduced signals obtained by reading magnetic tapes with various magnetic tape reading devices under various environments.

In addition, since the technology applied to the magnetic disk device described in Patent Document 2 is a technology aimed at suppressing the bias of nonlinear recording distortion, it is possible to use various magnetic tapes under various environments. It is difficult to reduce distortion that occurs nonlinearly in a reproduced signal obtained by reading a magnetic tape with a reading device.

One embodiment of the technology of the present disclosure can reduce distortion that nonlinearly occurs in multiple reproduced signal sequences, compared to the case where waveform equalization of multiple reproduced signal sequences is performed using a linear filter. A signal processing device, a magnetic tape cartridge, a magnetic tape reader, a processing method for a signal processing device, a method for operating a magnetic tape reader, and a program are provided.

A first aspect of the technology of the present disclosure is that a plurality of reading results obtained by reading data from a magnetic tape on which data is recorded by a plurality of reading elements mounted on a reading head are read from a plurality of A/ A receiver that receives a plurality of reproduced signal sequences obtained by being digitized by a D converter, and a plurality of equalizers that perform waveform equalization of the plurality of reproduced signal sequences received by the receiver. The multiple equalizers use multiple nonlinear filters that have been trained to reduce distortion that occurs nonlinearly in multiple playback signal sequences depending on the environmental conditions in which data is read from the magnetic tape. This is a signal processing device in which a plurality of nonlinear filters are optimized to have characteristics suitable for a plurality of reading elements based on a plurality of reading results.

A second aspect of the technology of the present disclosure is related to the first aspect, in which a plurality of reading results are obtained by reading a specific pattern recorded as data in a specific area of a magnetic tape by a plurality of reading elements. It is a signal processing device.

A third aspect of the technology of the present disclosure is that, in parallel with the operation of recording a specific pattern in a specific area by the plurality of recording elements arranged upstream of the plurality of reading elements in the running direction of the magnetic tape, This is a signal processing device according to a second aspect, in which a specific pattern is read by a plurality of reading elements.

A fourth aspect of the technology of the present disclosure is the signal processing device according to any one of the first to third aspects, wherein the conditions include conditions due to individual differences in the reading heads. It is.

A fifth aspect of the technology of the present disclosure is the signal processing device according to any one of the first to fourth aspects, wherein the conditions include conditions due to individual differences in magnetic tapes. It is.

A sixth aspect of the technology of the present disclosure is a signal according to any one of the first to fifth aspects, wherein the conditions include a speed condition regarding the speed at which the magnetic tape is running. It is a processing device.

A seventh aspect according to the technology of the present disclosure is the signal processing device according to the sixth aspect, wherein the speed condition includes a condition regarding the running speed of the magnetic tape when recording is performed on the magnetic tape. .

An eighth aspect according to the technology of the present disclosure is any one of the first to seventh aspects, wherein the conditions include conditions caused by individual differences in processing circuits that affect waveform equalization. 1 is a signal processing device according to one aspect.

A ninth aspect according to the technology of the present disclosure is the signal processing device according to any one of the first to eighth aspects, wherein the nonlinear filter is a filter having a neural network in which learning has been performed. .

A tenth aspect according to the technology of the present disclosure is provided for each reading element, and includes a plurality of storage elements in which the reproduced signal series is stored in time series, and a neural network is configured to store a plurality of storage elements corresponding to the plurality of storage elements. and a subsequent layer, each of the plurality of storage elements outputs the input reproduction signal sequence to a corresponding one of the plurality of previous layer nodes, Each of the plurality of front layer nodes outputs the reproduced signal sequence inputted from the corresponding one of the plurality of storage elements to the subsequent layer, and the latter layer outputs the reproduced signal sequence inputted from the plurality of front layer nodes. The composite value obtained based on the sum of products of The signal processing device according to the ninth aspect is a signal processing device in which the load is determined by performing learning on the neural network to minimize the amount of deviation between the subsequent layer value and the predetermined target value.

In an eleventh aspect of the technology of the present disclosure, the neural network has an input layer as a previous layer, and an intermediate layer and an output layer as subsequent layers, and the plurality of previous layer nodes are connected to the plurality of input layers. node, the intermediate layer has a plurality of intermediate layer nodes, and each of the plurality of input layer nodes outputs a reproduced signal sequence input from a corresponding one of the plurality of storage elements to the intermediate layer. , multiple hidden layer nodes convert the hidden layer value obtained as a composite value based on the product sum of the reproduced signal sequence input from multiple input layer nodes and the hidden layer connection weight using an activation function. The output layer generates a value and outputs it to the output layer, and the output layer outputs the output layer value obtained as the subsequent layer value based on the conversion value input from the intermediate layer, the output layer connection weight, and the product sum, and In the signal processing device according to the tenth aspect, the load and the output layer connection load are determined by performing learning on the neural network to minimize the amount of deviation between the output layer value and a predetermined target value. be.

A twelfth aspect of the technology of the present disclosure is that the intermediate layer value is a value based on the sum of products of the reproduced signal sequence and the intermediate layer connection weight and the first variable, and the first variable is This is a signal processing device according to an eleventh aspect, which is determined by being performed on a signal processing device.

A thirteenth aspect according to the technology of the present disclosure is the signal processing device according to the tenth aspect, in which the neural network includes two layers, a former layer and a later layer.

A fourteenth aspect of the technology of the present disclosure is that the subsequent layer value is a value based on the sum of products of the converted value and the subsequent layer connection weight and the second variable, and the second variable is The signal processing device according to any one of the tenth to thirteenth aspects determined by the following.

A fifteenth aspect according to the technology of the present disclosure is that the plurality of storage elements are a plurality of delay elements into which the reproduced signal sequence is delayed by a predetermined time and input, and the subsequent layer value is stored in the plurality of delay elements. The signal processing device according to any one of the tenth to fourteenth aspects, in which the value is related to the first input reproduction signal sequence among the plurality of reproduction signal sequences input.

In a sixteenth aspect of the technology of the present disclosure, the target value is an ideal reproduction signal regarding known data recorded on the learning magnetic tape in a predetermined recording pattern along the longitudinal direction of the learning magnetic tape. The signal according to any one of the tenth to fifteenth aspects, which is teacher data predetermined based on at least one of the sequence and the ideal reproduction signal sequence derived by computer simulation. It is a processing device.

A seventeenth aspect according to the technology of the present disclosure is a magnetic tape cartridge including a magnetic tape, the magnetic tape being used by the signal processing device according to any one of the first to sixteenth aspects. The magnetic tape cartridge is a magnetic tape cartridge in which parameters related to the plurality of nonlinear filters are recorded.

An eighteenth aspect according to the technology of the present disclosure is a magnetic tape cartridge including a non-contact storage medium, wherein the non-contact storage medium includes any one of the first to sixteenth aspects. The present invention is a magnetic tape cartridge in which parameters related to the plurality of nonlinear filters used by such a signal processing device are recorded.

A nineteenth aspect of the technology of the present disclosure is a read head equipped with a plurality of reading elements that read data from a magnetic tape on which data is recorded, and a read head that is obtained by reading data by the plurality of reading elements. A receiver that receives a plurality of reproduction signal sequences obtained by digitizing a plurality of reading results by a plurality of A/D converters, and a waveform equalization of the plurality of reproduction signal sequences received by the receiver. and a plurality of equalizers that perform learning to reduce distortion that nonlinearly occurs in the plurality of reproduced signal sequences depending on the environmental conditions in which data is read from the magnetic tape. A magnetic tape reader in which waveform equalization is performed using multiple nonlinear filters, and the multiple nonlinear filters are optimized to have characteristics suitable for multiple reading elements based on multiple reading results. It is a device.

A 20th aspect according to the technology of the present disclosure is according to the 19th aspect, wherein a plurality of reading results are obtained by reading a specific pattern recorded as data in a specific area of a magnetic tape by a plurality of reading elements. A magnetic tape reader.

A twenty-first aspect of the technology of the present disclosure is that, in parallel with the operation of recording a specific pattern in a specific area by a plurality of recording elements arranged upstream of a plurality of reading elements in the running direction of the magnetic tape, This is a magnetic tape reading device according to a twentieth aspect, in which a specific pattern is read by a plurality of reading elements.

A twenty-second aspect of the technology of the present disclosure is the magnetic tape reading according to any one of the nineteenth to twenty-first aspects, wherein the conditions include conditions due to individual differences in the reading heads. It is a device.

A twenty-third aspect of the technology of the present disclosure is the magnetic tape reading according to any one of the nineteenth to twenty-second aspects, wherein the conditions include conditions due to individual differences in the magnetic tape. It is a device.

A twenty-fourth aspect according to the technology of the present disclosure is a magnetic tape according to any one of the nineteenth to twenty-third aspects, wherein the conditions include a speed condition regarding the speed at which the magnetic tape is running. It is a tape reader.

A twenty-fifth aspect according to the technology of the present disclosure is the magnetic tape reading device according to the twenty-fourth aspect, wherein the speed condition includes a condition regarding the traveling speed of the magnetic tape when recording is performed on the magnetic tape. be.

A twenty-sixth aspect according to the technology of the present disclosure is any one of the nineteenth to twenty-fifth aspects, wherein the conditions include conditions caused by individual differences in processing circuits that affect waveform equalization. 1 is a magnetic tape reader according to two aspects.

A twenty-seventh aspect of the technology of the present disclosure is the magnetic tape reading device according to any one of the nineteenth to twenty-sixth aspects, wherein the nonlinear filter is a filter having a trained neural network. be.

A twenty-eighth aspect according to the technology of the present disclosure is provided for each reading element, and includes a plurality of storage elements in which the reproduced signal series is stored in time series, and a neural network is configured to store a plurality of storage elements corresponding to the plurality of storage elements. and a subsequent layer, each of the plurality of storage elements outputs the input reproduction signal sequence to a corresponding one of the plurality of previous layer nodes, Each of the plurality of front layer nodes outputs the reproduced signal sequence inputted from the corresponding one of the plurality of storage elements to the subsequent layer, and the latter layer outputs the reproduced signal sequence inputted from the plurality of front layer nodes. The composite value obtained based on the sum of products of A magnetic tape reading device according to a twenty-seventh aspect, in which the load is determined by performing learning on a neural network to minimize the amount of deviation between a subsequent layer value and a predetermined target value.

A twenty-ninth aspect of the technology of the present disclosure is that the neural network has an input layer as a previous layer, and an intermediate layer and an output layer as subsequent layers, and the plurality of previous layer nodes are connected to the plurality of input layers. node, the intermediate layer has a plurality of intermediate layer nodes, and each of the plurality of input layer nodes outputs a reproduced signal sequence input from a corresponding one of the plurality of storage elements to the intermediate layer. , multiple hidden layer nodes convert the hidden layer value obtained as a composite value based on the product sum of the reproduced signal sequence input from multiple input layer nodes and the hidden layer connection weight using an activation function. The output layer generates a value and outputs it to the output layer, and the output layer outputs the output layer value obtained as the subsequent layer value based on the conversion value input from the intermediate layer, the output layer connection weight, and the product sum, and The magnetic tape reading device according to the twenty-eighth aspect, wherein the load and the output layer connection load are determined by performing learning on a neural network to minimize the amount of deviation between the output layer value and a predetermined target value. It is.

In a 30th aspect of the technology of the present disclosure, the intermediate layer value is a value based on the sum of products of the reproduced signal sequence and the intermediate layer connection weight and the first variable, and the first variable is This is a magnetic tape reading device according to a twenty-ninth aspect defined by the method.

A 31st aspect according to the technology of the present disclosure is the magnetic tape reading device according to the 28th aspect, in which the neural network includes two layers, a former layer and a later layer.

A thirty-second aspect of the technology of the present disclosure is that the subsequent layer value is a value based on the sum of products of the converted value and the subsequent layer connection weight and a second variable, and the second variable is a value that is A magnetic tape reading device according to any one of the twenty-eighth to thirty-first aspects determined by the method.

A thirty-third aspect of the technology of the present disclosure is that the plurality of storage elements are a plurality of delay elements into which the reproduced signal sequence is input after being delayed by a predetermined time, and the subsequent layer value is stored in the plurality of delay elements. In the magnetic tape reading device according to any one of the twenty-eighth to thirty-second aspects, the value is the value related to the first input reproduction signal sequence among the plurality of reproduction signal sequences input.

A thirty-fourth aspect of the technology of the present disclosure is that the target value is an ideal reproduction signal regarding known data recorded on the learning magnetic tape in a predetermined recording pattern along the longitudinal direction of the learning magnetic tape. magnetism according to any one of the 28th to 33rd aspects, which is teacher data predetermined based on at least one of the sequence and the ideal reproduction signal sequence derived by computer simulation. It is a tape reader.

A thirty-fifth aspect of the technology of the present disclosure is that a plurality of reading results obtained by reading data from a magnetic tape on which data is recorded by a plurality of reading elements mounted on a reading head are read from a plurality of A/ A receiver that receives a plurality of reproduced signal sequences obtained by being digitized by a D converter, and a plurality of equalizers that perform waveform equalization of the plurality of reproduced signal sequences received by the receiver. A processing method for a signal processing device equipped with a learning method that uses a plurality of equalizers to reduce distortion that nonlinearly occurs in a plurality of reproduced signal sequences depending on the environmental conditions in which data is read from a magnetic tape. Waveform equalization is performed using multiple nonlinear filters, and the multiple nonlinear filters are optimized to have characteristics suitable for multiple reading elements based on multiple reading results. , a processing method for a signal processing device.

A thirty-sixth aspect of the technology of the present disclosure is a read head equipped with a plurality of reading elements that read data from a magnetic tape on which data is recorded, and a read head that is obtained by reading data by the plurality of reading elements. A receiver that receives a plurality of reproduction signal sequences obtained by digitizing a plurality of reading results by a plurality of A/D converters, and a waveform equalization of the plurality of reproduction signal sequences received by the receiver. 1. A method of operating a magnetic tape reading device comprising: a plurality of equalizers that read data from a magnetic tape; This includes performing waveform equalization using multiple nonlinear filters that have been trained to reduce distortion that occurs nonlinearly. This is a method of operating a magnetic tape reader that is optimized to suit the characteristics of the magnetic tape reader.

A thirty-seventh aspect of the technology of the present disclosure is a program for causing a computer applied to a signal processing device to execute processing, wherein the signal processing device retrieves data from a magnetic tape on which the data is recorded. A receiver that receives a plurality of reproduction signal sequences obtained by digitizing a plurality of reading results obtained by reading by a plurality of reading elements mounted on a reading head by a plurality of A/D converters. and a plurality of equalizers that perform waveform equalization of the plurality of reproduction signal sequences received by the receiver, and the processing is performed using a plurality of equalizers according to the environmental conditions in which data is read from the magnetic tape. This includes performing waveform equalization using multiple nonlinear filters that have been trained to reduce distortion that occurs nonlinearly in the reproduced signal sequence. This is a program that is optimized to have characteristics suitable for the device.

A thirty-eighth aspect of the technology of the present disclosure is a program for causing a computer applied to a magnetic tape reader to execute processing, wherein the magnetic tape reader reads data from a magnetic tape on which data is recorded. The data is obtained by using a reading head equipped with multiple reading elements that read data, and by digitizing the multiple reading results obtained by reading the data with the multiple reading elements using multiple A/D converters. a receiver for receiving a plurality of reproduction signal sequences received by the receiver; and a plurality of equalizers for equalizing the waveforms of the plurality of reproduction signal sequences received by the receiver. This includes performing waveform equalization using a nonlinear filter that has been trained to reduce distortion that occurs nonlinearly in the reproduced signal sequence depending on the environmental conditions in which the process is performed. This is a program that is optimized to have characteristics suitable for multiple reading elements based on the following.

According to the embodiment of the technology of the present disclosure, it is possible to reduce distortion that occurs nonlinearly in a plurality of reproduced signal sequences, compared to a case where waveform equalization of a plurality of reproduced signal sequences is performed using a linear filter. This effect can be obtained.

FIG. 1 is a schematic configuration diagram showing an example of the overall configuration of a magnetic tape drive according to a first embodiment. FIG. 2 is a schematic plan view showing an example of a planar configuration of a reading head and a magnetic tape included in the magnetic tape drive according to the first embodiment. FIG. 2 is a block diagram illustrating an example of the mutual relationship between a reading head, a moving mechanism, a motor, and a control device included in the magnetic tape drive according to the first embodiment. FIG. 2 is a block diagram illustrating an example of the mutual relationship between a servo element pair, a control device, a movement mechanism, and a motor included in the magnetic tape drive according to the first embodiment. FIG. 2 is a block diagram showing an example of the mutual relationship among a reading element, an amplifier, an A/D converter, an LPF, a phase synchronization circuit, an equalizer, a decoder, and a computer included in the magnetic tape drive according to the first embodiment. An example of the correlation between the waveform of the current supplied to the recording head when recording 1-bit recording pattern data on the magnetic tape, the waveform of the magnetic field generated by the recording head, and the recording pattern of data recorded on the magnetic tape. This is a graph showing. FIG. 2 is a block diagram showing an example of the electrical hardware configuration of the signal processing device according to the first embodiment. FIG. 2 is a block diagram illustrating an example of the configuration of a computer and its surroundings used for learning the neural network according to the first embodiment. FIG. 2 is a block diagram illustrating an example of main functions of a CPU in a learning stage of the neural network according to the first embodiment. 1 is a conceptual diagram showing an example of the configuration of a test playback signal supply device according to a first embodiment; FIG. FIG. 2 is a conceptual diagram showing an example of a hierarchical structure of a neural network according to the first embodiment. FIG. 2 is a block diagram illustrating an example of main functions of a CPU in a learning stage of the neural network according to the first embodiment. FIG. 2 is a conceptual diagram showing an example of a mode in which a trained model obtained by training the neural network according to the first embodiment is stored in a storage. FIG. 2 is a block diagram illustrating an example of main functions of a CPU in the operation stage of a trained model according to the first embodiment. FIG. 2 is a conceptual diagram showing an example of a hierarchical structure of trained models according to the first embodiment. 7 is a flowchart illustrating an example of the flow of learning execution processing according to the first embodiment. 7 is a flowchart illustrating an example of the flow of waveform equalization execution processing according to the first embodiment. 12 is a graph showing an example of the results of comparing the SNR of the FIR equalization method and the SNR of the neural network equalization method. 2A and 2B are graphs illustrating an example of noise generated in waveform equalization processing using an FIR filter, and graphs illustrating an example of noise generated in waveform equalization processing according to the technology of the present disclosure. It is a conceptual diagram which shows the 1st modification of the hierarchical structure of the neural network based on 1st Embodiment. 21 is a conceptual diagram showing an example of a hierarchical structure of a trained model obtained by training the neural network shown in FIG. 20. FIG. It is a conceptual diagram which shows the 2nd modification of the hierarchical structure of the neural network based on 1st Embodiment. 23 is a conceptual diagram showing an example of a hierarchical structure of a trained model obtained by training the neural network shown in FIG. 22. FIG. FIG. 2 is a block diagram showing a modified example of the configuration of the signal processing device according to the first embodiment. FIG. 7 is a conceptual diagram showing an example of the configuration of a test reproduction signal supply device according to a second embodiment. FIG. 2 is a conceptual diagram showing an example of the configuration of a BOT area of a magnetic tape. FIG. 2 is a conceptual diagram showing an example of the configuration of a magnetic head and the configuration of a specific area of a magnetic tape. FIG. 2 is a conceptual diagram showing an example of a magnetic element unit and a configuration around the magnetic element unit. FIG. 3 is a conceptual diagram illustrating an example of a mode in which parameters related to a trained model are recorded in a BOT area. FIG. 3 is a schematic configuration diagram showing a modified example of the configuration of a magnetic tape drive. FIG. 2 is a block diagram showing an example of a mode in which a waveform equalization execution program is installed in an equalizer in a magnetic tape drive from a storage medium in which the waveform equalization execution program is stored.

Hereinafter, an example of an embodiment of a signal processing device, a magnetic tape reading device, a processing method of the signal processing device, an operating method of the magnetic tape reading device, and a program according to the technology of the present disclosure will be described with reference to the accompanying drawings.

First, the words used in the following explanation will be explained.

CPU is an abbreviation for "Central Processing Unit." RAM is an abbreviation for "Random Access Memory." HDD is an abbreviation for "Hard Disk Drive." EEPROM is an abbreviation for "Electrically Erasable and Programmable Read Only Memory." SSD is an abbreviation for "Solid State Drive." USB is an abbreviation for "Universal Serial Bus." ASIC is an abbreviation for "Application Specific Integrated Circuit." FPGA is an abbreviation for "Field-Programmable Gate Array." PLD is an abbreviation for "Programmable Logic Device." SoC is an abbreviation for "System-on-a-chip." UI is an abbreviation for "User Interface". I/F is an abbreviation for "Interface". A/D is an abbreviation for "Analog/Digital." FIR is an abbreviation for "Finite Impulse Response". IIR is an abbreviation for "Infinite Impulse Response." LPF is an abbreviation for "Low Pass Filter." FIFO is an abbreviation for "First In First Out." SNR is an abbreviation for "signal-to-noise ratio." BOT is an abbreviation for "Beginning Of Tape." EOT is an abbreviation for "End Of Tape". MSE is an abbreviation for "Mean Square Error". Furthermore, in the following description, a range expressed using "~" means a range that includes the elements described before and after "~" as lower and upper limits.

[First embodiment]
As an example, as shown in FIG. 1, a magnetic tape drive 10, which is an example of a "magnetic tape reading device" according to the technology of the present disclosure, includes a magnetic tape cartridge 12, a transport device 14, a reading head 16, a control device 18, and a delivery motor. 20, a take-up reel 22, a take-up motor 24, a UI system device 26, and an external I/F 28. The magnetic tape cartridge 12 accommodates a magnetic tape MT. Data is recorded on the magnetic tape MT. The magnetic tape drive 10 is a device that extracts a magnetic tape MT from a magnetic tape cartridge 12 and reads data from the extracted magnetic tape MT using a read head 16 using a linear scan method.

Note that in the first embodiment, reading data means, in other words, reproducing data. In the following description, the data read by the read head 16 will also be referred to as a "reproduction signal." In addition, if there is no need to distinguish between a test reproduction signal, which will be described later, a real-time reproduction signal, which will be described later, a neural network signal, which will be described later, and a waveform-equalized reproduction signal, which will be described later, these will also be simply referred to as "reproduction signals."

The magnetic tape MT is generally produced by forming a magnetic layer containing ferromagnetic powder and optionally one or more additives on a non-magnetic support. The magnetic layer can be non-oriented, longitudinally oriented, or perpendicularly oriented. The magnetic layer and the like will be explained in detail.

<Magnetic layer>
(Ferromagnetic powder)
The magnetic layer includes ferromagnetic powder. As the ferromagnetic powder contained in the magnetic layer, one type or a combination of two or more types of ferromagnetic powders known as ferromagnetic powders used in the magnetic layers of various magnetic tapes MT can be used. It is preferable to use ferromagnetic powder having a small average particle size from the viewpoint of improving recording density. From this point of view, the average particle size of the ferromagnetic powder is preferably 50 nm or less, more preferably 45 nm or less, even more preferably 40 nm or less, even more preferably 35 nm or less, and 30 nm or less. It is even more preferable that it is, even more preferably that it is 25 nm or less, and even more preferably that it is 20 nm or less. On the other hand, from the viewpoint of magnetization stability, the average particle size of the ferromagnetic powder is preferably 5 nm or more, more preferably 8 nm or more, even more preferably 10 nm or more, and still more preferably 15 nm or more. is more preferable, and even more preferably 20 nm or more.

(Hexagonal ferrite powder)
A preferred example of the ferromagnetic powder is hexagonal ferrite powder. For details on hexagonal ferrite powder, see, for example, paragraphs 0012 to 0030 of JP2011-225417A, paragraphs 0134 to 0136 of JP2011-216149A, paragraphs 0013 to 0030 of JP2012-204726A, and Paragraphs 0029 to 0084 of Japanese Patent Application Publication No. 2015-127985 can be referred to.

In the technology of the present disclosure and this specification, "hexagonal ferrite powder" refers to a ferromagnetic powder in which a hexagonal ferrite type crystal structure is detected as the main phase by X-ray diffraction analysis. The main phase refers to a structure to which the most intense diffraction peak belongs in an X-ray diffraction spectrum obtained by X-ray diffraction analysis. For example, if the most intense diffraction peak in an X-ray diffraction spectrum obtained by X-ray diffraction analysis is attributed to a hexagonal ferrite crystal structure, it is determined that a hexagonal ferrite crystal structure has been detected as the main phase. shall be taken as a thing. When only a single structure is detected by X-ray diffraction analysis, this detected structure is taken as the main phase. The hexagonal ferrite type crystal structure includes at least iron atoms, divalent metal atoms, and oxygen atoms as constituent atoms. The divalent metal atom is a metal atom that can become a divalent cation as an ion, and includes alkaline earth metal atoms such as strontium atom, barium atom, and calcium atom, lead atom, and the like. In the technology of the present disclosure and this specification, hexagonal strontium ferrite powder refers to powder in which the main divalent metal atoms contained in this powder are strontium atoms, and hexagonal barium ferrite powder refers to powder in which main divalent metal atoms contained in this powder are strontium atoms. The main divalent metal atom contained in the metal is a barium atom. The main divalent metal atoms are the divalent metal atoms that account for the largest percentage on an atomic percent basis among the divalent metal atoms contained in this powder. However, the divalent metal atoms mentioned above do not include rare earth atoms. A "rare earth atom" in the present invention and herein is selected from the group consisting of scandium atom (Sc), yttrium atom (Y), and lanthanide atom. Lanthanoid atoms include lanthanum atom (La), cerium atom (Ce), praseodymium atom (Pr), neodymium atom (Nd), promethium atom (Pm), samarium atom (Sm), europium atom (Eu), and gadolinium atom (Gd). ), terbium atom (Tb), dysprosium atom (Dy), holmium atom (Ho), erbium atom (Er), thulium atom (Tm), ytterbium atom (Yb), and lutetium atom (Lu). Ru.

Below, hexagonal strontium ferrite powder, which is one embodiment of hexagonal ferrite powder, will be explained in more detail.

The activation volume of the hexagonal strontium ferrite powder is preferably in the range of 800 to 1500 nm ³ . Finely divided hexagonal strontium ferrite powder exhibiting an activation volume in the above range is suitable for producing a magnetic tape MT that exhibits excellent electromagnetic conversion characteristics. The activation volume of the hexagonal strontium ferrite powder is preferably 800 nm ³ or more, and can also be, for example, 850 nm ³ or more. In addition, from the viewpoint of further improving electromagnetic conversion characteristics, the activated volume of the hexagonal strontium ferrite powder is more preferably 1400 nm ³ or less, even more preferably 1300 nm ³ or less, and even more preferably 1200 nm ³ or less. is more preferable, and even more preferably 1100 nm ³ or less.

"Activation volume" is a unit of magnetization reversal and is an index indicating the magnetic size of a particle. The technology of the present disclosure and the activation volume described in this specification and the anisotropy constant Ku described below were measured using a vibrating sample magnetometer at magnetic field sweep speeds of 3 minutes and 30 minutes in the coercive force Hc measuring section. (Measurement temperature: 23° C.±1° C.), which is a value determined from the following relational expression between Hc and activation volume V. Note that the unit of the anisotropy constant Ku is 1erg/cc=1.0×10 ⁻¹ J/m ³ .

Hc=2Ku/Ms {1-[(kT/KuV)ln(At/0.693)] ^1/2 }

[In the above formula, Ku: anisotropy constant (unit: J/m ³ ), Ms: saturation magnetization (unit: kA/m), k: Boltzmann constant, T: absolute temperature (unit: K), V: activity volume (unit: cm ³ ), A: spin precession frequency (unit: s ⁻¹ ), t: magnetic field reversal time (unit: s)]

The anisotropy constant Ku can be cited as an index of reduction in thermal fluctuation, in other words, improvement in thermal stability. The hexagonal strontium ferrite powder can preferably have a Ku of 1.8×10 ⁵ J/m ³ or more, more preferably 2.0×10 ⁵ J/m ³ or more. Moreover, Ku of the hexagonal strontium ferrite powder can be, for example, 2.5×10 ⁵ J/m ³ or less. However, since a higher Ku means higher thermal stability and is preferable, it is not limited to the values exemplified above.

The hexagonal strontium ferrite powder may or may not contain rare earth atoms. When the hexagonal strontium ferrite powder contains rare earth atoms, it is preferable that the rare earth atoms are contained at a content (bulk content) of 0.5 to 5.0 at % based on 100 at % iron atoms. In one embodiment, the hexagonal strontium ferrite powder containing rare earth atoms can have rare earth atoms unevenly distributed in the surface layer. In the technology of the present disclosure and in this specification, "rare earth atom uneven distribution in the surface layer" refers to the rare earth atom content (%) relative to 100 atom % of iron atoms in a solution obtained by partially dissolving hexagonal strontium ferrite powder with acid. (hereinafter referred to as "rare earth atom surface content" or simply "surface layer content" with respect to rare earth atoms) is 100 iron atoms in the solution obtained by completely dissolving hexagonal strontium ferrite powder with acid. Rare earth atom content relative to atomic % (hereinafter referred to as "rare earth atom bulk content" or simply "bulk content" with respect to rare earth atoms) and earth atom surface layer content/rare earth atom bulk content>1. This means that the ratio of 0 is satisfied. The rare earth atom content of the hexagonal strontium ferrite powder described below is synonymous with the rare earth atom bulk content. On the other hand, partial dissolution using an acid dissolves the surface layer of the particles constituting the hexagonal strontium ferrite powder, so the rare earth atom content in the solution obtained by partial dissolution is This is the rare earth atom content in the surface layer of the particles. The fact that the rare earth atom content in the surface layer satisfies the ratio of "rare earth atom surface layer content/rare earth atom bulk content > 1.0" means that rare earth atoms are present in the surface layer in the particles constituting the hexagonal strontium ferrite powder. It means that it is omnipresent (that is, it is present more than inside). In the technology of the present disclosure and in this specification, the surface layer portion refers to a partial region from the surface of the particles constituting the hexagonal strontium ferrite powder toward the inside.

When the hexagonal strontium ferrite powder contains rare earth atoms, the rare earth atom content (bulk content) is preferably in the range of 0.5 to 5.0 atom % based on 100 atom % of iron atoms. It is believed that the uneven distribution of rare earth atoms in the surface layer of the particles constituting the hexagonal strontium ferrite powder, which contains rare earth atoms at a bulk content in the above range, contributes to suppressing the reduction in reproduction output during repeated reproduction. Conceivable. This is due to the fact that the hexagonal strontium ferrite powder contains rare earth atoms in the bulk content within the above range, and that the rare earth atoms are unevenly distributed in the surface layer of the particles constituting the hexagonal strontium ferrite powder. It is presumed that this is because it can increase the The higher the value of the anisotropy constant Ku, the more the phenomenon called thermal fluctuation can be suppressed (in other words, the thermal stability can be improved). By suppressing the occurrence of thermal fluctuations, it is possible to suppress a decrease in reproduction output during repeated reproduction. The uneven distribution of rare earth atoms in the particle surface of hexagonal strontium ferrite powder contributes to stabilizing the spin of iron (Fe) sites in the crystal lattice in the surface layer, which increases the anisotropy constant Ku. It is speculated that this will increase.

In addition, it is assumed that using hexagonal strontium ferrite powder, which has rare earth atoms unevenly distributed in the surface layer, as the ferromagnetic powder for the magnetic layer also contributes to suppressing the surface of the magnetic layer from being scraped by sliding with the magnetic head. Ru. That is, it is presumed that the hexagonal strontium ferrite powder having rare earth atoms unevenly distributed in the surface layer may also contribute to improving the running durability of the magnetic tape MT. This is due to the uneven distribution of rare earth atoms on the surface of the particles constituting the hexagonal strontium ferrite powder, which improves the interaction between the particle surface and the organic substances (e.g. binders and/or additives) contained in the magnetic layer. It is presumed that this is because the magnetic layer contributes to the strength of the magnetic layer, and as a result, the strength of the magnetic layer is improved.

From the viewpoint of further suppressing the decline in reproduction output during repeated reproduction and/or further improving running durability, the rare earth atom content (bulk content) is in the range of 0.5 to 4.5 at%. It is more preferably in the range of 1.0 to 4.5 atom %, even more preferably in the range of 1.5 to 4.5 atom %.

The above bulk content is the content determined by completely melting the hexagonal strontium ferrite powder. In the technology of the present disclosure and this specification, unless otherwise specified, the atomic content refers to the bulk content determined by completely dissolving the hexagonal strontium ferrite powder. The hexagonal strontium ferrite powder containing rare earth atoms may contain only one kind of rare earth atoms, or may contain two or more kinds of rare earth atoms. The above-mentioned bulk content when two or more types of rare earth atoms are included is calculated for the total of two or more types of rare earth atoms. This point also applies to the technology of the present disclosure and other components in this specification. That is, unless otherwise specified, one kind of a certain component may be used, or two or more kinds thereof may be used. When two or more types are used, the content or content rate refers to the total of the two or more types.

When the hexagonal strontium ferrite powder contains rare earth atoms, the rare earth atoms contained may be at least one kind of rare earth atoms. Preferred rare earth atoms from the viewpoint of further suppressing a decrease in reproduction output during repeated reproduction include neodymium atoms, samarium atoms, yttrium atoms, and dysprosium atoms, with neodymium atoms, samarium atoms, and yttrium atoms being more preferred; Atom is more preferred.

In the hexagonal strontium ferrite powder having rare earth atoms unevenly distributed in the surface layer, it is sufficient that the rare earth atoms are unevenly distributed in the surface layer of the particles constituting the hexagonal strontium ferrite powder, and the degree of uneven distribution is not limited. For example, for a hexagonal strontium ferrite powder that has rare earth atoms unevenly distributed in the surface layer, the surface layer content of rare earth atoms determined by partial melting under the melting conditions described below and the rare earth content determined by completely melting under the melting conditions described below. The ratio of atoms to the bulk content, "surface layer content/bulk content" is more than 1.0, and can be 1.5 or more. When the "surface layer content ratio/bulk content ratio" is larger than 1.0, it means that rare earth atoms are unevenly distributed in the surface layer (that is, more present than in the interior) in the particles that make up the hexagonal strontium ferrite powder. do. In addition, the ratio of the surface layer content of rare earth atoms determined by partial dissolution under the dissolution conditions described below and the bulk content of rare earth atoms determined by total dissolution under the dissolution conditions described below, ``surface layer content/ Bulk content" can be, for example, 10.0 or less, 9.0 or less, 8.0 or less, 7.0 or less, 6.0 or less, 5.0 or less, or 4.0 or less. However, in a hexagonal strontium ferrite powder having rare earth atoms unevenly distributed in the surface layer, it is sufficient that the rare earth atoms are unevenly distributed in the surface layer of the particles constituting the hexagonal strontium ferrite powder, and the above "surface layer content/bulk content" is sufficient. The "rate" is not limited to the illustrated upper or lower limits.

Partial and total dissolution of hexagonal strontium ferrite powder will be explained below. For hexagonal strontium ferrite powders that are present as powders, the partially dissolved and fully dissolved sample powders are taken from the same lot of powder. On the other hand, regarding the hexagonal strontium ferrite powder contained in the magnetic layer of the magnetic tape MT, a part of the hexagonal strontium ferrite powder taken out from the magnetic layer is subjected to partial melting, and the other part is subjected to total melting. . The hexagonal strontium ferrite powder can be taken out from the magnetic layer by, for example, the method described in paragraph 0032 of JP-A-2015-91747.

The above-mentioned partial dissolution means that the hexagonal strontium ferrite powder is dissolved to such an extent that residual hexagonal strontium ferrite powder can be visually confirmed in the liquid at the end of dissolution. For example, by partial dissolution, it is possible to dissolve a region of 10 to 20 mass % of the particles constituting the hexagonal strontium ferrite powder, assuming that the entire particle is 100 mass %. On the other hand, the above-mentioned complete dissolution means that the hexagonal strontium ferrite powder is dissolved to a state where no residual hexagonal strontium ferrite powder is visually confirmed in the liquid upon completion of dissolution.

The above-mentioned partial dissolution and measurement of the surface layer content are performed, for example, by the following method. However, the dissolution conditions such as the amount of sample powder described below are merely examples, and any dissolution conditions that allow partial dissolution and complete dissolution can be adopted as desired.

A container (for example, a beaker) containing 12 mg of sample powder and 10 mL of 1 mol/L hydrochloric acid is held on a hot plate with a set temperature of 70° C. for 1 hour. The obtained solution is filtered through a 0.1 μm membrane filter. Elemental analysis of the filtrate thus obtained is performed using an inductively coupled plasma (ICP) analyzer. In this way, the content of rare earth atoms in the surface layer relative to 100 atom % of iron atoms can be determined. When multiple types of rare earth atoms are detected by elemental analysis, the total content of all rare earth atoms is taken as the surface layer content. This point also applies to the measurement of bulk content.

On the other hand, the measurement of the total dissolution and bulk content is performed, for example, by the following method.

A container (for example, a beaker) containing 12 mg of sample powder and 10 mL of 4 mol/L hydrochloric acid is held on a hot plate with a set temperature of 80° C. for 3 hours. Thereafter, it is possible to determine the bulk content relative to 100 atom % of iron atoms by performing the same steps as the above-described partial dissolution and measurement of the surface layer content.

From the viewpoint of increasing the reproduction output when reproducing data recorded on the magnetic tape MT, it is desirable that the mass magnetization σs of the ferromagnetic powder contained in the magnetic tape MT is high. In this regard, a hexagonal strontium ferrite powder that contains rare earth atoms but does not have rare earth atoms unevenly distributed in the surface layer tends to have a significantly lower σs than a hexagonal strontium ferrite powder that does not contain rare earth atoms. On the other hand, a hexagonal strontium ferrite powder having rare earth atoms unevenly distributed in the surface layer is considered to be preferable in order to suppress such a large decrease in σs. In one aspect, σs of the hexagonal strontium ferrite powder can be 45 A·m ² /kg or more, and can also be 47 A·m ² /kg or more. On the other hand, from the viewpoint of noise reduction, σs is preferably 80 A·m ² /kg or less, more preferably 60 A·m ² /kg or less. σs can be measured using a known measuring device capable of measuring magnetic properties, such as a vibrating sample magnetometer. In the technology of the present disclosure and this specification, unless otherwise specified, mass magnetization σs is a value measured at a magnetic field strength of 1194 kA/m (15 kOe).

Regarding the content of constituent atoms (bulk content) of the hexagonal strontium ferrite powder, the strontium atomic content can be in the range of, for example, 2.0 to 15.0 atomic% with respect to 100 atomic% of iron atoms. . In one embodiment, the hexagonal strontium ferrite powder may contain only strontium atoms as divalent metal atoms. In another aspect, the hexagonal strontium ferrite powder can also contain one or more other divalent metal atoms in addition to strontium atoms. For example, it can contain barium atoms and/or calcium atoms. When divalent metal atoms other than strontium atoms are included, the barium atom content and calcium atom content in the hexagonal strontium ferrite powder are, for example, 0.05 to 5, respectively, relative to 100 at% of iron atoms. It can be in the range of .0 at.%.

As the crystal structure of hexagonal ferrite, magnetoplumbite type (also called "M type"), W type, Y type, and Z type are known. The hexagonal strontium ferrite powder may have any crystal structure. Crystal structure can be confirmed by X-ray diffraction analysis. The hexagonal strontium ferrite powder may have a single crystal structure or two or more types of crystal structures detected by X-ray diffraction analysis. For example, in one embodiment, the hexagonal strontium ferrite powder can have only an M-type crystal structure detected by X-ray diffraction analysis. For example, M-type hexagonal ferrite is represented by the composition formula AFe ₁₂ O ₁₉ . Here, A represents a divalent metal atom, and if the hexagonal strontium ferrite powder is M type, A is only a strontium atom (Sr), or if A contains multiple divalent metal atoms, As mentioned above, strontium atoms (Sr) account for the largest amount on an atomic percent basis. The divalent metal atom content of the hexagonal strontium ferrite powder is usually determined by the type of crystal structure of the hexagonal ferrite, and is not particularly limited. The same applies to the iron atom content and the oxygen atom content. The hexagonal strontium ferrite powder contains at least iron atoms, strontium atoms, and oxygen atoms, and may also contain rare earth atoms. Furthermore, the hexagonal strontium ferrite powder may or may not contain atoms other than these atoms. As an example, the hexagonal strontium ferrite powder may contain aluminum atoms (Al). The content of aluminum atoms can be, for example, 0.5 to 10.0 atomic % with respect to 100 atomic % of iron atoms. From the viewpoint of further suppressing the reduction in playback output during repeated playback, the hexagonal strontium ferrite powder contains iron atoms, strontium atoms, oxygen atoms, and rare earth atoms, and the content of atoms other than these atoms is 100 iron atoms. %, it is preferably 10.0 atom % or less, more preferably in the range of 0 to 5.0 atom %, and may be 0 atom %. That is, in one embodiment, the hexagonal strontium ferrite powder may not contain atoms other than iron atoms, strontium atoms, oxygen atoms, and rare earth atoms. The content expressed in atomic % above is the content of each atom (unit: mass %) obtained by completely melting the hexagonal strontium ferrite powder, and is converted to the value expressed in atomic % using the atomic weight of each atom. It can be calculated by converting. In addition, in the technology of the present disclosure and this specification, "not containing" a certain atom means that the content thereof as measured by an ICP analyzer after being completely dissolved is 0% by mass. The detection limit of an ICP analyzer is usually 0.01 ppm (parts per million) or less on a mass basis. The above-mentioned "does not contain" is used to mean that it is contained in an amount below the detection limit of the ICP analyzer. In one embodiment, the hexagonal strontium ferrite powder can be free of bismuth atoms (Bi).

(metal powder)
A preferable specific example of the ferromagnetic powder is a ferromagnetic metal powder. For details of the ferromagnetic metal powder, reference can be made to, for example, paragraphs 0137 to 0141 of JP-A No. 2011-216149 and paragraphs 0009 to 0023 of JP-A No. 2005-251351.

(ε-iron oxide powder)
A preferred example of the ferromagnetic powder is ε-iron oxide powder. In the technology of the present disclosure and this specification, "ε-iron oxide powder" refers to a ferromagnetic powder in which an ε-iron oxide type crystal structure is detected as the main phase by X-ray diffraction analysis. For example, if the most intense diffraction peak in an X-ray diffraction spectrum obtained by X-ray diffraction analysis is attributed to the ε-iron oxide type crystal structure, it is assumed that the ε-iron oxide type crystal structure has been detected as the main phase. shall judge. Known methods for producing ε-iron oxide powder include a method for producing it from goethite, a reverse micelle method, and the like. All of the above manufacturing methods are publicly known. Further, regarding a method for producing ε-iron oxide powder in which a portion of Fe is substituted with substitution atoms such as Ga, Co, Ti, Al, Rh, etc., see, for example, J. Jpn. Soc. Powder Metallurgy Vol. 61 Supplement, No. S1, pp. S280-S284, J. Mater. Chem. C, 2013, 1, pp. 5200-5206 etc. can be referred to. However, the method for producing the ε-iron oxide powder that can be used as the ferromagnetic powder in the magnetic layer of the magnetic tape MT is not limited to the method mentioned above.

The activated volume of the ε-iron oxide powder is preferably in the range of 300 to 1500 nm ³ . Finely divided ε-iron oxide powder exhibiting an activation volume in the above range is suitable for producing a magnetic tape MT that exhibits excellent electromagnetic conversion characteristics. The activated volume of the ε-iron oxide powder is preferably 300 nm ³ or more, and can also be, for example, 500 nm ³ or more. Further, from the viewpoint of further improving electromagnetic conversion characteristics, the activated volume of the ε-iron oxide powder is more preferably 1400 nm ³ or less, even more preferably 1300 nm ³ or less, and 1200 nm ³ or less. is more preferable, and even more preferably 1100 nm ³ or less.

The anisotropy constant Ku can be cited as an index of reduction in thermal fluctuation, in other words, improvement in thermal stability. The ε-iron oxide powder can preferably have a Ku of 3.0×10 ⁴ J/m ³ or more, more preferably 8.0×10 ⁴ J/m ³ or more. Further, Ku of the ε-iron oxide powder can be, for example, 3.0×10 ⁵ J/m ³ or less. However, since a higher Ku means higher thermal stability, which is preferable, it is not limited to the values exemplified above.

From the viewpoint of increasing the reproduction output when reproducing data recorded on the magnetic tape MT, it is desirable that the mass magnetization σs of the ferromagnetic powder contained in the magnetic tape MT is high. In this regard, in one aspect, the σs of the ε-iron oxide powder can be 8 A·m ² /kg or more, and can also be 12 A·m ² /kg or more. On the other hand, from the viewpoint of noise reduction, σs of the ε-iron oxide powder is preferably 40 A·m ² /kg or less, more preferably 35 A·m ² /kg or less.

In the technology of the present disclosure and this specification, unless otherwise specified, the average particle size of various powders such as ferromagnetic powder is a value measured by the following method using a transmission electron microscope.

Photograph the powder using a transmission electron microscope at a magnification of 100,000 times, and print it on photographic paper at a total magnification of 500,000 times or display it on a display to obtain a photograph of the particles that make up the powder. . Select the target particle from the obtained particle photo, trace the outline of the particle with a digitizer, and measure the size of the particle (primary particle). Primary particles refer to independent particles without agglomeration.

The above measurements are performed on 500 randomly extracted particles. The arithmetic mean of the particle sizes of the 500 particles thus obtained is defined as the average particle size of the powder. As the transmission electron microscope, for example, a transmission electron microscope model H-9000 manufactured by Hitachi may be used. Furthermore, the particle size can be measured using known image analysis software, such as Carl Zeiss image analysis software KS-400. Unless otherwise specified, the average particle size shown in the examples below was measured using a transmission electron microscope H-9000 model manufactured by Hitachi as a transmission electron microscope and image analysis software KS-400 manufactured by Carl Zeiss as image analysis software. It is a value. In the technology of this disclosure and this specification, powder means a collection of multiple particles. For example, ferromagnetic powder means a collection of multiple ferromagnetic particles. Furthermore, an aggregate of multiple particles is not limited to an embodiment in which the particles constituting the aggregate are in direct contact with each other, but also includes an embodiment in which a binder, an additive, etc., which will be described later, are interposed between the particles. Ru. The term particle is sometimes used to refer to powder.

As a method for collecting sample powder from the magnetic tape MT for particle size measurement, for example, the method described in paragraph 0015 of JP-A No. 2011-048878 can be adopted.

In the technology of the present disclosure and this specification, unless otherwise specified, the size of the particles constituting the powder (particle size) is as follows:
(1) If the particle is needle-shaped, spindle-shaped, columnar (however, the height is larger than the maximum major axis of the bottom surface), etc., it is expressed by the length of the major axis that makes up the particle, that is, the major axis length,
(2) If it is plate-shaped or columnar (however, the thickness or height is smaller than the maximum major axis of the plate surface or bottom surface), it is expressed by the maximum major axis of the plate surface or bottom surface,
(3) If the particle is spherical, polyhedral, unspecified, etc., and the long axis constituting the particle cannot be determined from the shape, it is expressed by an equivalent circle diameter. The equivalent circle diameter refers to what is determined by the circular projection method.

In addition, the average acicular ratio of the powder can be determined by measuring the short axis length of the particles in the above measurement, determining the value of (major axis length / short axis length) of each particle, and calculating the average acicular ratio of the 500 particles. Refers to the arithmetic mean of the values obtained for the particles. Here, unless otherwise specified, the short axis length refers to the length of the short axis constituting the particle in the case of (1) in the above particle size definition, and the thickness or height in the case of (2). Similarly, in the case of (3), there is no distinction between the major axis and the minor axis, so (major axis length/minor axis length) is regarded as 1 for convenience.

Unless otherwise specified, when the particle shape is specific, for example, in the case of the above definition (1) of particle size, the average particle size is the average major axis length, and in the case of the same definition (2), the average particle size is This is the average plate diameter. In the case of the same definition (3), the average particle size is the average diameter (also referred to as average particle diameter or average particle diameter).

The content (filling rate) of the ferromagnetic powder in the magnetic layer is preferably in the range of 50 to 90% by mass, more preferably in the range of 60 to 90% by mass. The magnetic layer comprises ferromagnetic powder, may contain a binder, and optionally may also contain one or more further additives. It is preferable that the filling rate of ferromagnetic powder in the magnetic layer is high from the viewpoint of improving recording density.

(Binder, hardening agent)
The magnetic tape MT may be a coated magnetic tape, and may contain a binder in the magnetic layer. The binder is one or more resins. As the binder, various resins commonly used as binders for coated magnetic tapes can be used.

For example, binders include polyurethane resins, polyester resins, polyamide resins, vinyl chloride resins, acrylic resins copolymerized with styrene, acrylonitrile, methyl methacrylate, etc., cellulose resins such as nitrocellulose, epoxy resins, phenoxy resins, polyvinyl acetals, A resin selected from polyvinyl alkyral resins such as polyvinyl butyral can be used alone, or a plurality of resins can be used in combination. Among these, preferred are polyurethane resins, acrylic resins, cellulose resins, and vinyl chloride resins. These resins may be homopolymers or copolymers. These resins can also be used as a binder in the nonmagnetic layer and/or back coat layer described below. Regarding the above binders, reference may also be made to paragraphs 0028 to 0031 of JP-A-2010-24113. The content of the binder in the magnetic layer can be, for example, 1.0 to 30.0 parts by mass based on 100.0 parts by mass of the ferromagnetic powder. The average molecular weight of the resin used as the binder can be, for example, 10,000 or more and 200,000 or less as a weight average molecular weight.

A curing agent can also be used in conjunction with a resin that can be used as a binder. In one embodiment, the curing agent can be a thermosetting compound, which is a compound in which a curing reaction (crosslinking reaction) proceeds by heating, and in another embodiment, a photocuring compound, in which a curing reaction (crosslinking reaction) proceeds by light irradiation. It can be a chemical compound. As a curing reaction progresses during the magnetic layer forming process, at least a portion of the curing agent may be contained in the magnetic layer in a reacted (crosslinked) state with other components such as a binder. This point also applies to layers formed using the composition when the composition used to form the other layer contains a curing agent. Preferred curing agents are thermosetting compounds, with polyisocyanates being preferred. For details of the polyisocyanate, paragraphs 0124 to 0125 of JP-A No. 2011-216149 can be referred to. The content of the curing agent in the composition for forming a magnetic layer can be, for example, 0 to 80.0 parts by mass based on 100.0 parts by mass of the binder, and from the viewpoint of improving the strength of the magnetic layer, it is 50.0 parts by mass. ~80.0 parts by mass.

(Additive)
The magnetic layer may contain one or more additives as necessary. Examples of the additives include the above-mentioned curing agents. Further, examples of additives contained in the magnetic layer include nonmagnetic powder, lubricant, dispersant, dispersion aid, fungicide, antistatic agent, antioxidant, and the like. As lubricants it is possible to use, for example, fatty acid amides which can function as boundary lubricants. Boundary lubricants are considered to be lubricants that can reduce contact friction by adsorbing to the surface of powder (for example, ferromagnetic powder) and forming a strong lubricant film. Examples of fatty acid amides include amides of various fatty acids such as lauric acid, myristic acid, palmitic acid, stearic acid, oleic acid, linoleic acid, linolenic acid, behenic acid, erucic acid, and elaidic acid, specifically lauric acid amide. , myristic acid amide, palmitic acid amide, stearic acid amide, and the like. The fatty acid amide content of the magnetic layer is, for example, 0 to 3.0 parts by mass, preferably 0 to 2.0 parts by mass, and more preferably 0 to 1.0 parts by mass, per 100.0 parts by mass of ferromagnetic powder. Part by mass. Furthermore, the nonmagnetic layer may also contain fatty acid amide. The fatty acid amide content of the nonmagnetic layer is, for example, 0 to 3.0 parts by mass, and preferably 0 to 1.0 parts by mass, per 100.0 parts by mass of the nonmagnetic powder. Regarding the dispersant, paragraphs 0061 and 0071 of JP 2012-133837A can be referred to. A dispersant may be added to the composition for forming a nonmagnetic layer. Regarding the dispersant that can be added to the composition for forming a non-magnetic layer, paragraph 0061 of JP-A-2012-133837 can be referred to. In addition, examples of the non-magnetic powder that can be included in the magnetic layer include non-magnetic powder that can function as an abrasive, and non-magnetic powder that can function as a protrusion-forming agent that forms protrusions that appropriately protrude on the surface of the magnetic layer. etc. Examples of polishing agents include alumina (Al ₂ O ₃ ), silicon carbide, boron carbide (B ₄ C), TiC, chromium oxide (Cr ₂ O ₃ ), and cerium oxide, which are substances commonly used as polishing agents for magnetic layers. , zirconium oxide (ZrO ₂ ), iron oxide, diamond, etc., among which powders of alumina such as α-alumina, silicon carbide, and diamond are preferred. The content of the abrasive in the magnetic layer is preferably 1.0 to 20.0 parts by mass, more preferably 3.0 to 15.0 parts by mass, based on 100.0 parts by mass of the ferromagnetic powder. , more preferably 4.0 to 10.0 parts by mass. The average particle size of the abrasive is, for example, in the range of 30 to 300 nm, preferably in the range of 50 to 200 nm. Examples of the protrusion-forming agent include carbon black and colloid particles. The content of the protrusion forming agent in the magnetic layer is preferably 0.1 to 10.0 parts by mass, more preferably 0.1 to 5.0 parts by mass, based on 100.0 parts by mass of the ferromagnetic powder. The amount is preferably 0.5 to 5.0 parts by mass, and more preferably 0.5 to 5.0 parts by mass. The average particle size of the colloidal particles is, for example, preferably in the range of 90 to 200 nm, more preferably in the range of 100 to 150 nm. The average particle size of carbon black is preferably in the range of 5 to 200 nm, more preferably in the range of 10 to 150 nm.

The magnetic layer described above can be provided directly on the surface of the nonmagnetic support or indirectly via the nonmagnetic layer.

<Nonmagnetic layer>
Next, the nonmagnetic layer will be explained. The magnetic tape MT may have a magnetic layer directly on the surface of a non-magnetic support, or may have a magnetic layer on the surface of a non-magnetic support via a non-magnetic layer containing non-magnetic powder. good. The nonmagnetic powder used in the nonmagnetic layer may be an inorganic powder or an organic powder. Further, carbon black or the like can also be used. Examples of the inorganic powder include powders of metals, metal oxides, metal carbonates, metal sulfates, metal nitrides, metal carbides, metal sulfides, and the like. These nonmagnetic powders are available as commercial products, and can also be produced by known methods. For details thereof, paragraphs 0146 to 0150 of JP-A No. 2011-216149 can be referred to. Regarding carbon black that can be used in the nonmagnetic layer, paragraphs 0040 to 0041 of JP-A No. 2010-24113 can also be referred to. The content (filling rate) of the nonmagnetic powder in the nonmagnetic layer is preferably in the range of 50 to 90% by mass, more preferably in the range of 60 to 90% by mass.

The non-magnetic layer can be a layer containing non-magnetic powder and a binder, and can further include one or more additives. For other details such as the binder and additives of the nonmagnetic layer, known techniques regarding nonmagnetic layers can be applied. Further, for example, regarding the type and content of the binder, the type and content of the additive, and the like, known techniques regarding the magnetic layer can also be applied.

In the disclosed technology and herein, a non-magnetic layer is intended to include a substantially non-magnetic layer containing a small amount of ferromagnetic powder, for example as an impurity or intentionally, along with non-magnetic powder. do. Here, a substantially non-magnetic layer means that this layer has a residual magnetic flux density of 10 mT or less, a coercive force of 100 Oe or less, or a residual magnetic flux density of 10 mT or less and a coercive force of 100 Oe. It refers to the following layers. 1 [kOe]=10 ⁶ /4π [A/m]. Preferably, the nonmagnetic layer has no residual magnetic flux density and no coercive force.

In one embodiment, a compound having an ammonium salt structure of an alkyl ester anion represented by Formula 1 may be included in the nonmagnetic layer. The compound having an ammonium salt structure of an alkyl ester anion represented by Formula 1 is preferably contained in the nonmagnetic layer in an amount of 0.01 part by mass or more based on 100.0 parts by mass of the nonmagnetic powder, and 0.1 part by mass. It is more preferable that the content is more than 0.5 parts by mass, and even more preferably 0.5 parts by mass or more. Further, the content of the above compound in the nonmagnetic layer is preferably 15.0 parts by mass or less, more preferably 10.0 parts by mass or less, and 8 parts by mass or less based on 100.0 parts by mass of the nonmagnetic powder. More preferably, it is .0 part by mass or less. Further, the preferred range of the content of the above compound in the nonmagnetic layer forming composition used to form the nonmagnetic layer is also the same. The above compound contained in the nonmagnetic layer can migrate to the magnetic layer, and can also migrate to the surface of the magnetic layer to form a liquid film. Details of the above-mentioned compound that can be included in the nonmagnetic layer or the composition for forming a nonmagnetic layer are as described above.

<Nonmagnetic support>
Next, the non-magnetic support (hereinafter also simply referred to as "support") will be explained. Examples of the nonmagnetic support include known ones such as biaxially stretched polyethylene terephthalate, polyethylene naphthalate, polyamide, polyamideimide, and aromatic polyamide. Among these, polyethylene terephthalate, polyethylene naphthalate, and polyamide are preferred. These supports may be subjected to corona discharge, plasma treatment, adhesive treatment, heat treatment, etc. in advance.

<Back coat layer>
The above magnetic tape MT may also have a back coat layer containing nonmagnetic powder on the surface side of the nonmagnetic support opposite to the surface side having the magnetic layer. The back coat layer preferably contains one or both of carbon black and inorganic powder. The backcoat layer can be a layer containing non-magnetic powder and a binder, and can further include one or more additives. Regarding the binder and various additives that may be optionally included in the backcoat layer, known techniques related to the backcoat layer can be applied, and known techniques related to the formulation of the magnetic layer and/or nonmagnetic layer can also be applied. . For example, the descriptions in paragraphs 0018 to 0020 of JP-A No. 2006-331625 and in column 4, line 65 to column 5, line 38 of U.S. Patent No. 7,029,774 can be referred to regarding the back coat layer. .

<Various thicknesses>
The thickness of the non-magnetic support is, for example, 3.0 to 80.0 μm, preferably 3.0 to 20.0 μm, more preferably 3.0 to 10.0 μm, even more preferably 3.0 to 6.0 μm. It is.

The thickness of the magnetic layer can be optimized depending on the saturation magnetization amount, head gap length, recording signal band, etc. of the magnetic head used. From the viewpoint of high-density recording, the thickness of the magnetic layer is preferably 10 nm to 150 nm, more preferably 20 nm to 120 nm, and even more preferably 30 nm to 100 nm. It is sufficient that there is at least one magnetic layer, and the magnetic layer may be separated into two or more layers having different magnetic properties, and a structure related to a known multilayer magnetic layer can be applied. The thickness of the magnetic layer when it is separated into two or more layers is the total thickness of these layers.

The thickness of the nonmagnetic layer is, for example, 0.1 to 3.0 μm, preferably 0.1 to 2.0 μm, and more preferably 0.1 to 1.5 μm.

The thickness of the back coat layer is preferably 0.9 μm or less, more preferably in the range of 0.1 to 0.7 μm.

The thickness of each layer of the magnetic tape MT and the nonmagnetic support can be determined by a known film thickness measurement method. As an example, after a cross section of the magnetic tape in the thickness direction is exposed using a known method such as an ion beam or a microtome, the exposed cross section is observed using a scanning electron microscope. Various thicknesses can be determined as the arithmetic mean of the thickness determined at any one location during cross-sectional observation, or the thickness determined at two or more randomly extracted locations, for example, two locations. Alternatively, the thickness of each layer may be determined as a design thickness calculated from manufacturing conditions.

The control device 18 controls the entire magnetic tape drive 10 . The control device 18 is realized by a plurality of hardware resources including a computer including a CPU, memory, and storage, an ASIC, and an FPGA. In the first embodiment, the memory temporarily stores various information and is used as a work memory. An example of the memory is a RAM, but the present invention is not limited to this, and other types of storage devices may be used. The storage stores various parameters and various programs. Storage is a non-volatile storage device. Here, EEPROM is employed as an example of storage. EEPROM is just one example, and instead of or together with EEPROM, an HDD and/or SSD may be used as storage.

Here, a plurality of hardware resources including a computer, an ASIC, and an FPGA are illustrated as the control device 18, but the technology of the present disclosure is not limited thereto. For example, the control device 18 may be realized by hardware resources including a computer, an ASIC, an FPGA, or a PLD. Further, the control device 18 may be realized by a hardware resource that is a combination of one or more of ASIC, FPGA, and PLD and a computer. In this way, the control device 18 may be any device as long as it is realized by hardware resources that have the function of an electronic computer.

The conveyance device 14 is a device that selectively conveys the magnetic tape MT in the forward direction and the reverse direction, and includes a delivery motor 20, a take-up reel 22, a take-up motor 24, a plurality of guide rollers GR, and a control device 18. ing.

A cartridge reel CR is provided inside the magnetic tape cartridge 12. A magnetic tape MT is wound around the cartridge reel CR. The feed motor 20 rotates the cartridge reel CR within the magnetic tape cartridge 12 under the control of the control device 18 . The control device 18 controls the rotation direction, rotation speed, rotation torque, etc. of the cartridge reel CR by controlling the delivery motor 20.

When the magnetic tape MT is taken up by the take-up reel 22, the control device 18 rotates the sending motor 20 so as to run the magnetic tape MT in the forward direction. The rotational speed, rotational torque, etc. of the delivery motor 20 are adjusted according to the speed of the magnetic tape MT being wound by the take-up reel 22.

The take-up motor 24 rotates the take-up reel 22 under the control of the control device 18 . The control device 18 controls the rotation direction, rotation speed, rotation torque, etc. of the take-up reel 22 by controlling the take-up motor 24 .

When the magnetic tape MT is taken up by the take-up reel 22, the control device 18 rotates the take-up motor 24 so as to run the magnetic tape MT in the forward direction. The rotational speed, rotational torque, etc. of the take-up motor 24 are adjusted according to the speed of the magnetic tape MT being wound up by the take-up reel 22.

By adjusting the rotational speed, rotational torque, etc. of each of the delivery motor 20 and the take-up motor 24 in this manner, a tension within a predetermined tension range is applied to the magnetic tape MT. Here, "within the predetermined tension range" means, for example, the range of tension at which data can be read from the magnetic tape MT by the read head 16, from the lower limit to the upper limit of the tension obtained by computer simulation and/or tests using actual equipment. Refers to the range up to the value.

When rewinding the magnetic tape MT onto the cartridge reel CR, the control device 18 rotates the feed motor 20 and the take-up motor 24 so as to run the magnetic tape MT in the opposite direction.

In the first embodiment, the tension of the magnetic tape MT is controlled by controlling the rotational speed, rotational torque, etc. of the delivery motor 20 and the take-up motor 24, but the technology of the present disclosure is not limited to this. For example, the tension of the magnetic tape MT may be controlled using a dancer roller or by drawing the magnetic tape MT into a vacuum chamber.

Each of the plurality of guide rollers GR is a roller that guides the magnetic tape MT. The travel path of the magnetic tape MT is determined by the magnetic tape MT being stretched at several locations (two locations in the example shown in FIG. 1) between the magnetic tape cartridge 12 and the take-up reel 22.

The reading head 16 is arranged in the running direction of the magnetic tape MT (hereinafter also simply referred to as the "running direction"). The running direction is a direction corresponding to the forward direction of the magnetic tape MT. The reading head 16 includes a reading element 16A and a holder 16B. The reading element 16A is, for example, an element having a magnetoresistive element. The reading element 16A is held by a holder 16B in a position where it can read data from the magnetic tape MT. The read head 16 uses the read element 16A to read data from the magnetic tape MT while the magnetic tape MT is running.

A UI system device 26 and an external I/F 28 are connected to the control device 18 . The UI device 26 includes a display and a reception device. The display displays various information such as images under the control of the control device 18. The receiving device has hard keys, a touch panel, etc., and receives instructions from the user of the magnetic tape drive 10 and the like. The control device 18 operates according to instructions received by the receiving device.

The external I/F 28 is in charge of exchanging various information between the control device 18 and a device that exists outside the magnetic tape drive 10 (hereinafter also referred to as an “external device”). An example of the external I/F 28 is a USB interface. External devices (not shown) such as smart devices, personal computers, servers, USB memories, memory cards, and/or printers are directly or indirectly connected to the USB interface.

As an example, as shown in FIG. 2, the magnetic tape MT includes a track area 30 and a servo pattern 32. The servo pattern 32 is a pattern used to detect the position of the read head 16 with respect to the magnetic tape MT. The servo pattern 32 includes a plurality of first diagonal lines 32A, each having a first predetermined angle (for example, 6 degrees), and a plurality of first diagonal lines 32A, each having a first predetermined angle (for example, 6 degrees), at both ends of the magnetic tape MT in the width direction (hereinafter also simply referred to as "tape width direction"). This is a pattern in which a plurality of second diagonal lines 32B having two predetermined angles (for example, 174 degrees) are alternately arranged along the running direction of the magnetic tape MT.

The servo pattern 32 in the magnetic tape MT shown in FIG. 2 is shown in a simplified manner for convenience of explanation. The first diagonal line 32A illustrated in the magnetic tape MT shown in FIG. 2 is the first diagonal line 32A on the most downstream side in the running direction among the plurality of first diagonal lines 32A in one servo pattern 32. The second diagonal line 32B illustrated in the magnetic tape MT shown in FIG. 2 is the second diagonal line 32B on the most downstream side in the running direction among the plurality of second diagonal lines 32B in one servo pattern 32.

Specifically, for example, as shown in the enlarged view of FIG. 2, there are five first diagonal lines 32A and four first diagonal lines 32A as the first diagonal lines 32A, and there are four first diagonal lines 32A as the second diagonal lines 32B. , there are five second diagonal lines 32B and four second diagonal lines 32B. That is, five first diagonal lines 32A, five second diagonal lines 32B, four first diagonal lines 32A, and four second diagonal lines 32B are arranged in this order along the running direction of the magnetic tape MT.

The track area 30 is an area in which data to be read is written, and is formed at the center of the magnetic tape MT in the tape width direction. The "central portion in the tape width direction" herein refers to, for example, a region between the servo pattern 32 at one end of the magnetic tape MT in the tape width direction and the servo pattern 32 at the other end.

The read head 16 includes a pair of servo elements 36 . Servo element pair 36 includes servo elements 36A and 36B. Each of the servo elements 36A and 36B is, for example, an element having a magnetoresistive element. The servo element 36A is arranged at a position facing the servo pattern 32 at one end of the magnetic tape MT in the tape width direction, and the servo element 36B is arranged at a position facing the servo pattern 32 at the other end of the magnetic tape MT in the tape width direction. are placed in opposing positions. Note that although the servo elements 36A and 36B are illustrated here, the technique of the present disclosure can be applied even if only one of the servo elements 36A and 36B is used. That is, it is sufficient that the number of servo elements required for the reading head 16 to read data using the linear scan method is used for the reading head 16.

The reading head 16 includes a plurality of reading elements 16A. The plurality of reading elements 16A are arranged at positions facing the track area 30 when the magnetic tape drive 10 is in a default state.

Here, the default state of the magnetic tape drive 10 refers to a state where the magnetic tape MT is not deformed and the positional relationship between the magnetic tape MT and the reading head 16 is correct. Here, the correct positional relationship refers to, for example, a positional relationship in which the center of the track area 30 in the tape width direction and the center of the reading head 16 in the longitudinal direction match. Note that the meaning of "match" in the first embodiment includes not only a complete match but also a match that includes errors that are generally allowed in the technical field to which the technology of the present disclosure belongs. Point.

The track area 30 includes a plurality of tracks, and the plurality of tracks are arranged at equal intervals in the tape width direction. The reading elements 16A are arranged at equal intervals in the tape width direction, for example, every several or tens of tracks along the tape width direction. In the first embodiment, 32 reading elements 16A are employed.

That is, the reading elements 16A are arranged at positions corresponding to each of the 32 tracks included in the magnetic tape MT. In other words, the reading element 16A is arranged at a position corresponding to each single track of the 32 tracks included in the magnetic tape MT. Here, the number of tracks and reading elements 16A is exemplified as 32, but this is just an example, and the number may be greater or less than 32. In addition, the meaning of "equal spacing" in the first embodiment includes not only completely equal spacing but also errors that are generally allowed in the technical field to which the technology of the present disclosure belongs. Refers to equal intervals.

Note that, hereinafter, for convenience of explanation, one track to which the reading element 16A is assigned among the 32 tracks included in the track area 30 is also referred to as a "single track."

A moving mechanism 40 is provided at the end of the reading head 16 . The moving mechanism 40 moves the reading head 16 in the tape width direction in response to externally applied power. Specifically, the moving mechanism 40 selectively moves the reading head 16 to one side and the other side in the tape width direction in response to externally applied power. In the example shown in FIG. 2, one side and the other side in the tape width direction refer to the direction of arrow A.

In the first embodiment, while the magnetic tape MT is running under the control of the control device 18 (see FIG. 1), the read head 16 reads data from a single track using a linear scan method. be exposed. In the linear scan method, the servo pattern 32 is read by the servo element pair 36 in synchronization with the reading operation of the reading element 16A. That is, in the linear scan method according to the first embodiment, reading on the magnetic tape MT is performed in parallel by the reading element 16A and the servo element pair 36.

As an example, as shown in FIG. 3, the read head 16 is connected to a control device 18. The reproduced signal obtained from a single track by the reading element 16A is output to the control device 18 as a reproduced signal series that is a time-series signal. Since the reading head 16 is provided with a plurality of reading elements 16A, each of the plurality of reading elements 16A obtains a reproduced signal from the track area 30 (for example, each single track corresponding to each reading element 16A). The obtained reproduced signals are outputted to the control device 18 as a plurality of reproduced signal sequences. Further, an analog servo signal (hereinafter referred to as an “analog servo signal”) obtained by reading the servo pattern 32 by the servo element pair 36 is output to the control device 18.

A motor 42 is connected to the control device 18 . An example of the motor 42 is a voice coil motor. A voice coil motor generates power by converting electrical energy based on a current flowing through a coil into kinetic energy using the energy of a magnet as a medium. Motor 42 is connected to moving mechanism 40 . The moving mechanism 40 moves the reading head 16 in the tape width direction by being powered by a motor 42 under the control of the control device 18 .

Note that although a voice coil motor is cited as an example of the motor 42 here, the technology of the present disclosure is not limited to this, and for example, a different type of motor than the voice coil motor may be used. Moreover, a piezoelectric element and/or a solenoid may be used instead of a motor. Further, the power applied to the reading head 16 may be generated by a device that combines a plurality of motors, piezoelectric elements, solenoids, and the like.

As shown in FIG. 4 as an example, the control device 18 includes a controller 44, an amplifier 46, and an A/D converter 48. Servo element pair 36 is connected to controller 44 via amplifier 46 and A/D converter 48 . Controller 44 is connected to motor 42 .

The amplifier 46 receives an analog servo signal from the servo element pair 36, amplifies the input analog servo signal, and outputs the amplified analog servo signal to the A/D converter 48. The A/D converter 48 converts the analog servo signal input from the amplifier 46 into a digital signal. The digital signal obtained by the A/D converter 48 is outputted by the A/D converter 48 to the controller 44 as a digital servo signal (hereinafter simply referred to as a "servo signal").

The amount of positional deviation between the single track and the reading element 16A (hereinafter simply referred to as "the amount of deviation") is determined according to a servo signal obtained by reading the servo pattern 32 by the servo element pair 36.

Note that the misalignment between the single track and the reading element 16A refers to, for example, the misalignment between the center of the single track in the tape width direction and the center of the reading element 16A in the tape width direction.

The controller 44 is a device including a computer, and the computer has a CPU, memory, and storage, as described above. The controller 44 controls the entire magnetic tape drive 10 .

The controller 44 controls the motor 42 to apply power to the moving mechanism 40 according to the amount of deviation. The moving mechanism 40 moves the reading head 16 in the tape width direction according to the power applied from the motor 42, and adjusts the position of the reading head 16 to a normal position. Here, the "normal position" of the reading head 16 refers to, for example, a position where the deviation between the center of a single track in the tape width direction and the center of the reading element 16A in the tape width direction is "0".

The amount of deviation is calculated, for example, based on the ratio of the second distance to the first distance. The second distance means, for example, that the first diagonal line 32A (see FIG. 2) on the most downstream side and the second diagonal line 32B (see FIG. 2) on the most downstream side within one servo pattern 32 are read by the servo element 36A. This refers to the distance calculated from the results obtained. The first distance means, for example, that the second diagonal line 32B on the most downstream side in one of the adjacent servo patterns 32 and the second diagonal line 32B on the most downstream side in the other servo pattern 32 are It refers to the distance calculated from the result obtained by reading by the element 36A.

Specifically, for example, the amount of deviation is calculated using the following equation (1). The above-mentioned first predetermined angle and second predetermined angle are applied as the "oblique line angle α" in Equation (1). The first predetermined angle is an angle between the first diagonal line 32A and a straight line along the tape width direction, and the second predetermined angle is an angle between the second diagonal line 32B and a straight line along the tape width direction. In other words, the first predetermined angle is the angle that the first diagonal line 32A makes with respect to the straight line along the tape width direction in the clockwise direction when viewed from the front, and the second predetermined angle is "180 degrees - 1 predetermined angle”.

As shown in FIG. 5 as an example, the control device 18 includes a signal processing device 50. The signal processing device 50 performs signal processing on an analog reproduction signal (hereinafter also referred to as "analog reproduction signal") which is data read by the reading element 16A from a single track.

The signal processing device 50 includes a plurality of element-specific signal processing devices 50A. One element-specific signal processing device 50A is provided for each of the plurality of reading elements 16A. The element-specific signal processing device 50A includes an amplifier 52, an A/D converter 54, an LPF 56, a phase synchronization circuit 58, an equalizer 60, and a decoder 62.

The reading element 16A is connected to an LPF 56 via an amplifier 52 and an A/D converter 54. The LPF 56 is connected to an equalizer 60 via a phase synchronization circuit 58. Equalizer 60 is connected to decoder 62. Decoder 62 is connected to phase synchronization circuit 58 . A computer 64 is provided outside the control device 18, and the decoder 62 is connected to the computer 64.

The reading element 16A outputs an analog reproduction signal to the amplifier 52. That is, the analog playback signal obtained by reading from a single track by the reading element 16A is input to the signal processing device 50 in real time. Note that the analog playback signal is an example of a "reading result" according to the technology of the present disclosure.

The amplifier 52 amplifies the input analog reproduction signal and outputs the amplified analog reproduction signal to the A/D converter 54. The A/D converter 54 is an example of a "processing circuit" according to the technology of the present disclosure. The A/D converter 54 digitizes the input analog playback signal to convert it into a digital signal. The digital signal obtained by the A/D converter 54 is input to the LPF 56. The LPF 56 generates a real-time reproduction signal and outputs it to the phase synchronization circuit 58. The real-time reproduction signal refers to a signal obtained by removing high frequency components from the digital signal input to the LPF 56 by the LPF 56.

Incidentally, due to deformation of the magnetic tape MT, steep vibrations applied to the magnetic tape MT and/or the reading head 16, etc., jitter during the running of the magnetic tape MT, etc., a phase shift in the running direction may occur in the real-time reproduced signal. may occur.

Therefore, the phase synchronization circuit 58 performs phase synchronization processing on the real-time reproduction signal input from the LPF 56. The phase synchronization process refers to a process of keeping the phase shift of the real-time reproduced signal in the traveling direction within a certain allowable error range based on the decoding result of the decoder 62.

A decoding result (for example, a decoded signal described later) of a past real-time reproduction signal (for example, a past real-time reproduction signal by several bits) at the decoder 62 is fed back to the phase synchronization circuit 58 . Then, the phase synchronization circuit 58 identifies a phase shift that occurred in the past from the fed-back decoding result, and corrects the identified phase shift after a delay of several bits up to the present. In this way, in the phase synchronization circuit 58, the phase shift is maintained within a certain allowable error range by repeating feedback and correction with a delay of several bits.

Note that although an example is given here in which the phase synchronization process using the decoding result of the decoder 62 is executed by the phase synchronization circuit 58, the technology of the present disclosure is not limited to this. For example, a phase shift caused by a minute shift in the running direction due to steep vibrations and/or jitter applied to the magnetic tape MT and/or the reading element 16A, etc., is determined in advance for the control device 18. Processing to synchronize with the phase of a reference clock (hereinafter simply referred to as "reference clock") may be performed.

The equalizer 60 performs waveform equalization of the real-time reproduced signal. That is, the equalizer 60 performs waveform equalization processing on the real-time reproduction signal that has been subjected to phase synchronization processing by the phase synchronization circuit 58. A waveform equalized reproduced signal obtained by performing waveform equalization processing on the real-time reproduced signal by the equalizer 60 is output to a decoder 62 .

The decoder 62 decodes the waveform equalized reproduced signal input from the equalizer 60, and converts the resulting decoded signal (for example, a signal indicating either "0" or "1") into a phase It is output to the synchronization circuit 58 and computer 64. The computer 64 performs various processing on the decoded signal input from the decoder 62.

Incidentally, data is recorded in bit units on a single track in a predetermined recording pattern along the running direction. In recent years, with the increase in the density of data recorded on magnetic tape MT, the bit interval (for example, the interval between positions where data is recorded in units of bits along the running direction of magnetic tape MT) has increased. Due to the reduction in size and the increase in the data transfer rate, distortion occurring in the reproduced signal after waveform equalization is increasing.

As the bit interval becomes narrower, the magnetic field leaking from adjacent bit recording positions during data recording and the magnetic field generated by the recording head interfere with each other, causing the magnetic tape MT to move away from the bit recording position where it should originally be recorded. Data may be recorded at a shifted position.

Further, as shown in FIG. 6 as an example, the current supplied to the recording head (in the example shown in FIG. 6, the "recording current") when recording data of a 1-bit recording pattern on the magnetic tape MT rises. It takes a certain amount of time for the magnetic field to fall , and the fall time of the magnetic field generated by the recording current is longer than the rise time. As the fall time of the magnetic field becomes longer than the rise time of the magnetic field, asymmetry occurs in the magnetization reversal position on the magnetic tape MT, and as a result, bits that should originally be recorded on the magnetic tape MT Data is recorded at a position that is shifted from the recording position.

On the other hand, the distortion of the reproduced signal after waveform equalization is caused not only by the recording of data on the magnetic tape MT but also by the characteristics of the magnetic tape drive 10. The characteristics of the magnetic tape drive 10 include the characteristics of the read head 16 (for example, nonlinearity of the magnetoresistive element), the characteristics of the magnetic tape MT, the characteristics of the running speed of the magnetic tape MT, the characteristics of the A/D converter 54, etc. can be mentioned.

If the distortion of the reproduced signal after waveform equalization occurs only when data is recorded on the magnetic tape MT, it is possible to deal with it by adjusting the recording conditions; However, it is difficult to reduce distortion caused by the characteristics of the magnetic tape drive 10. Further, when adjusting the recording conditions, when data is read from the magnetic tape MT by a good magnetic tape drive 10, there is a risk that the adjustment will result in over-adjustment, which may actually increase distortion. In particular, this can become a big problem when recording and reading data on a replaceable medium such as magnetic tape MT using separate heads. This is because the causes of distortion in the reproduced signal after waveform equalization are different between when data is recorded on the magnetic tape MT and when data is read. Note that the FIR filter works effectively against distortion that occurs linearly, that is, stationary distortion, but depending on the environmental conditions in which data is read from the magnetic tape MT, the playback signal after waveform equalization may be affected. The FIR filter does not work as effectively against distortion that occurs nonlinearly, that is, unsteady distortion (hereinafter also referred to as "nonlinear distortion" or "nonlinear noise"), compared to distortion that occurs linearly.

Therefore, in the magnetic tape drive 10, the equalizer 60 performs waveform equalization of the real-time reproduction signal using a nonlinear filter trained to reduce nonlinear distortion. Examples of the nonlinear filter include a filter having a trained neural network. An example of a neural network that has been trained is a trained model 82 shown in FIG. 15.

As shown in FIG. 7 as an example, the equalizer 60 is an example of a "computer" according to the technology of the present disclosure, and includes a CPU 70, a memory 72, and a storage 74. Note that the CPU 70, memory 72, and storage 74 are the same hardware resources as the CPU, memory, and storage included in the computer included in the controller 44 described above.

The CPU 70 has an internal memory 71. CPU 70, memory 72, and storage 74 are connected to bus 76. Note that in the example shown in FIG. 7, one bus is shown as the bus 76 for convenience of illustration, but a plurality of buses may be used. The bus 76 may be a serial bus or a parallel bus including a data bus, an address bus, a control bus, and the like.

The signal processing device 50 includes an input/output I/F 78, and a phase synchronization circuit 58 and a decoder 62 are connected to the input/output I/F 78. The input/output I/F 78 is connected to the bus 76, and the input/output I/F 78 exchanges various information between the CPU 70 and the phase synchronization circuit 58, and exchanges various information between the CPU 70 and the decoder 62. in charge of

The input/output I/F 78 is an example of a “receiver” according to the technology of the present disclosure, and receives a real-time reproduction signal from the phase synchronization circuit 58. The equalizer 60 performs waveform equalization of the real-time reproduction signal received by the input/output I/F 78. A reproduced signal after waveform equalization obtained by performing waveform equalization by the equalizer 60 is output to the decoder 62 via the input/output I/F 78 .

The storage 74 stores a waveform equalization execution program 80 and a learned model 82. The waveform equalization execution program 80 is an example of a "program" according to the technology of the present disclosure. The CPU 70 reads the waveform equalization execution program 80 from the storage 74 and executes the read waveform equalization execution program 80 on the memory 72 . By executing the waveform equalization execution program 80 by the CPU 70, waveform equalization execution processing (see FIG. 17), which will be described later, is realized.

Here, how to create the trained model 82 and how to use the trained model will be explained with reference to FIGS. 8 to 15. Note that FIGS. 8 to 13 show an example of the configuration of a learning stage (hereinafter also simply referred to as "learning stage") in which the neural network 108 (see FIG. 11) is trained, and FIGS. 1 shows an example of the configuration of the operation stage (hereinafter also simply referred to as "operation stage") in which the learned model 82 is operated.

First, the operation stage will be explained with reference to FIGS. 8 to 13. The trained model 82 is generated by training the neural network 108 (see FIG. 11). As an example, as shown in FIG. 8, learning for the neural network 108 is realized by a computer 90. The computer 90 includes a CPU 92, a memory 94, a storage 96, an input/output I/F 98, and a bus 100, and the CPU 92, the memory 94, the storage 96, and the input/output I/F 98 are connected to the bus 100. The CPU 92 has an internal memory 93. Note that the CPU 92, memory 94, and storage 96 are hardware resources similar to the CPU 70, memory 72, and storage 74 described above.

A test playback signal supply device 102 and a UI device 104 are connected to the input/output I/F 98 . The input/output I/F 98 is in charge of exchanging various information between the CPU 92 and the test playback signal supply device 102 and between the CPU 92 and the UI device 104.

The test playback signal supply device 102 is, for example, a computer having hardware resources similar to the above-mentioned CPU, memory, and storage, and supplies a test playback signal to the computer 90. The test reproduction signal is acquired by the CPU 92 via the input/output I/F 98. As will be described in detail later, the test reproduction signal is a reproduction signal obtained by reading data from the magnetic tape MT under various conditions in various environments in which data is read from the magnetic tape MT.

The UI device 104 includes a display and a reception device. The display displays various information such as images under the control of the CPU 92. The reception device has a keyboard, a mouse, a touch panel, and the like, and receives instructions from the user of the computer 90 and the like. The CPU 92 operates according to instructions received by the receiving device.

The storage 96 stores a learning execution program 106, a neural network 108, and teacher data 110.

As an example, as shown in FIG. 9, the CPU 92 reads the learning execution program 106 from the storage 96 and executes the read learning execution program 106, thereby controlling the learning stage delay storage section 92A, the learning stage calculation section 92B, and the error calculation section. 92C and a variable adjustment section 92D.

The learning stage calculation unit 92B acquires the test reproduction signal bit by bit from the test reproduction signal supply device 102, and stores the test reproduction signal bit by bit in the acquired order in the first storage element group (for example, a plurality of bits in the internal memory 93 shown in FIG. 8). storage element). That is, each time the learning stage calculation unit 92B acquires the test reproduction signal bit by bit from the test reproduction signal supply device 102, it stores it in the first storage element group in chronological order in the acquisition order. Note that the test reproduction signal is stored in the first storage element group in a FIFO format.

The learning stage calculation unit 92B performs calculations using the test reproduction signal stored in time series by the learning stage delay storage unit 92A and the neural network 108 in the storage 96.

The error calculation unit 92C calculates the error between the calculation result by the learning stage calculation unit 92B and the teacher data 110 in the storage 96. Here, the teacher data 110 refers to an ideal reproduction signal regarding known data recorded on the learning magnetic tape in a predetermined recording pattern along the longitudinal direction (for example, running direction) of the learning magnetic tape. . Note that the teacher data 110 is an example of a "predetermined target value" according to the technology of the present disclosure. Further, the error is an example of a "deviation amount" according to the technology of the present disclosure.

The neural network 108 has a plurality of optimization variables (hereinafter also simply referred to as "optimization variables") such as connection weights and offset values (hereinafter also referred to as "thresholds"), and is calculated by the error calculation unit 92C. The neural network 108 is trained by adjusting the optimization variables so that the calculated error is minimized. Therefore, the variable adjustment section 92D adjusts the optimization variables so that the error calculated by the error calculation section 92C is minimized.

As an example, as shown in FIG. 10, the test reproduction signal supply device 102 has a test reproduction signal for each environmental condition (hereinafter also referred to as "reading environment condition") under which data is read from the magnetic tape MT. In the example shown in FIG. 10, the first reading environmental condition to the Nth reading environmental condition are shown, and a test reproduction signal is associated with each of the first reading environmental condition to the Nth reading environmental condition. In the following, for convenience of explanation, the first to Nth reading environmental conditions will be simply referred to as "reading environmental conditions" unless it is necessary to explain them separately.

The reading environment conditions include reading head conditions, magnetic tape conditions, running speed conditions, and A/D converter conditions.

The reading head conditions refer to conditions caused by individual differences in the reading heads 16. Individual differences in the read heads 16 refer to, for example, differences in the characteristics of the read elements 16A (for example, nonlinearity of magnetoresistive elements) for each read head 16. Differences in characteristics between the reading elements 16A are mainly caused by manufacturing errors of the reading elements 16A and/or deterioration of the reading elements 16A over time. Examples of indicators that quantitatively indicate the degree of deterioration over time of the reading element 16A include the number of times the reading head 16 has been used, the average time that the reading head 16 has been continuously used, and the number of times since the reading head 16 was manufactured. Examples include the time it takes to reach a specific point in time (for example, the current moment).

The magnetic tape conditions refer to conditions caused by individual differences in the magnetic tape MT. Individual differences among magnetic tapes MT refer to, for example, differences in characteristics between magnetic tapes MT. Differences in characteristics between magnetic tapes MT are mainly caused by differences in bit spacing, differences in the material of the magnetic tape MT, manufacturing errors in the magnetic tape MT, and/or deterioration of the magnetic tape MT over time. Indices that quantitatively indicate the degree of deterioration over time of the magnetic tape MT include the number of times the magnetic tape MT has been used, the average time that the magnetic tape MT has been continuously used, and the number of times the magnetic tape MT has been continuously used Examples include the time it takes to reach a point in time (for example, the present moment).

The running speed condition refers to a condition regarding the speed at which the magnetic tape MT is running. Examples of the speed at which the magnetic tape MT is running are 3 m/s (meters/second), 4 m/s, and 5 m/s. Note that the speed at which the magnetic tape MT is running may be the average speed when data is read over the entire length of the magnetic tape MT, the median speed, or the most frequent speed. There may be. Further, it may be an average speed, a median speed, or a most frequently used speed when reading data from a partial range of the magnetic tape MT.

The A/D converter condition refers to a condition caused by individual differences in the A/D converter 54. Individual differences in the A/D converters 54 refer to, for example, differences in characteristics of each A/D converter 54. Differences in characteristics among the A/D converters 54 are mainly caused by manufacturing errors of the A/D converters 54 and/or deterioration of the A/D converters 54 over time. Examples of indicators that quantitatively indicate the degree of deterioration over time of the A/D converter 54 include the number of times the A/D converter 54 is used, the average time that the A/D converter 54 is continuously used, and , the time from when the A/D converter 54 is manufactured to a specific point in time (for example, the current time).

The contents of the first to Nth reading environmental conditions are different from each other. That is, the combinations of read head conditions, magnetic tape conditions, running speed conditions, and A/D converter conditions are different between the first reading environment condition and the Nth reading environment condition, and a unique test playback is performed for each reading environment condition. Signals are associated.

The test reproduction signal supplying device 102 supplies a test reproduction signal associated with a reading environmental condition selected from the first to Nth reading environmental conditions to the learning stage delay storage section 92A. Note that the reading environment conditions are selected according to instructions given to the UI device 104, for example.

As shown in FIG. 11 as an example, the learning stage delay storage section 92A includes a plurality of storage elements 92A1 as the above-mentioned first storage element group. The plurality of storage elements 92A1 are realized by, for example, the internal memory 93. Each of the plurality of storage elements 92A1 is a delay element, and a test reproduction signal is input to the plurality of storage elements 92A1 after being delayed by a predetermined time (hereinafter also referred to as "delay time") to be described later. The plurality of storage elements 92A1 store test reproduction signals in time series. That is, each storage element 92A1 delays and stores the test reproduction signal by one bit each time one bit of the test reproduction signal is input. In the example shown in FIG. 11, a plurality of storage elements 92A1 are connected in series, and the test reproduction signal is delayed by one bit and stored in each storage element 92A1. Test reproduction signals are stored in the plurality of storage elements 92A1 in a FIFO format, and the test reproduction signals stored in the terminal storage element 92A1 (hereinafter also referred to as "termination storage element E1") are stored in the learning stage delay storage section. When a new 1-bit test reproduction signal is input to 92A, it is erased from storage element 92A1.

Neural network 108 includes a first layer 107 having a plurality of first layer nodes 107A corresponding to a plurality of storage elements 92A1, and a second layer 109 having a plurality of second layer nodes 109A. Each of the plurality of storage elements 92A1 outputs the input test reproduction signal to a corresponding one of the plurality of previous layer nodes 107A. Each of the plurality of front layer nodes 107A outputs the test reproduction signal inputted from the corresponding storage element 92A1 among the plurality of storage elements 92A1 to the rear layer 109. The subsequent layer 109 outputs the product of the test reproduction signal (“x” in the example shown in FIG. 11) inputted from the plurality of previous layer nodes 107A and the subsequent layer coupling load (“w” in the example shown in FIG. 11). The composite value obtained based on the sum is transformed using an activation function. Then, the subsequent layer 109 outputs a subsequent layer value (in the example shown in FIG. 11, a "neural network signal") based on the converted value obtained by converting the composite value using the activation function. The subsequent layer value output from the subsequent layer 109 is a signal related to the test reproduction signal inputted first among the plurality of test reproduction signals stored in the plurality of storage elements 92A1, that is, stored in the terminal storage element E1. This is a value related to the test playback signal being used. Here, the subsequent layer connection weight is a type of optimization variable that is adjusted by the variable adjustment unit 92D (see FIG. 9), and the neural network 108 performs learning to minimize the error between the subsequent layer value and the teacher data 110. It is determined by what is done to

The subsequent layer value is a value based on the sum of products of the converted value and the subsequent layer connection load, and a threshold value (see Equation (4) below). The threshold value used for the sum of products of the converted value and the subsequent layer coupling load is a value that reduces the sum of products of the converted value and the subsequent layer coupling load. The threshold value used for the product sum of the conversion value and the subsequent layer connection weight is determined by performing learning on the neural network 108 to reduce nonlinear distortion. Note that the threshold value is an example of a "first variable" and a "second variable" according to the technology of the present disclosure.

In the example shown in FIG. 11, an input layer 108A is shown as an example of the former layer 107, and an intermediate layer 108B and an output layer 108C are shown as examples of the latter layer 109. The input layer 108A has a plurality of input layer nodes 108A1. The middle layer 108B has a plurality of middle layer nodes 108B1. The output layer 108C has an output layer node 108C1. In the example shown in FIG _. 11, N1 input layer nodes 108A1 are shown as an example of the plurality of previous layer nodes 107A, and _N2 intermediate layer nodes 108B1, as an example of the plurality of subsequent layer nodes 109A. and N ₃ output layer nodes 108C1 are shown.

Each of the plurality of input layer nodes 108A1 outputs the test reproduction signal input from the corresponding one of the plurality of storage elements 92A1 to the intermediate layer 108B. The plurality of intermediate layer nodes 108B1 convert the intermediate layer value obtained as the above-mentioned composite value into an activation function (e.g. , a sigmoid function shown in Equation (3) below), the above-described converted value is generated and output to the output layer 108C. The output layer 108C outputs the output layer value obtained as the above-mentioned subsequent layer value as a neural network signal based on the product sum of the conversion value input from the intermediate layer 108B and the output layer connection weight. The neural network signal output from the output layer 108C is a signal related to the test reproduction signal inputted first among the plurality of test reproduction signals stored in the plurality of storage elements 92A1, that is, the signal stored in the terminal storage element E1. This is a signal related to the test playback signal being used. The intermediate layer connection weight and the output layer connection weight are determined by performing learning on the neural network 108 to minimize the error between the neural network signal and the teacher data 110.

Further, the intermediate layer value is a value based on the sum of products of the test reproduction signal and the intermediate layer coupling load and a threshold value (see Equation (4) described below). The threshold value used for the sum of products of the test reproduction signal and the intermediate layer coupling load is a value that reduces the sum of products of the test reproduction signal and the intermediate layer coupling load. The threshold value used for the product sum of the test reproduction signal and the intermediate layer connection weight is determined by performing learning on the neural network 108 to reduce nonlinear distortion. Further, the neural network signal is a value based on the sum of products of the conversion value input from the intermediate layer 108B and the output layer connection weight, and a threshold value (see Equation (4) described below). The threshold value used for the sum of products of the conversion value and the output layer connection load is a value that subtracts the sum of products of the conversion value and the output layer connection load. The threshold value used for the product sum of the conversion value and the output layer connection weight is determined by performing learning on the neural network 108 to reduce nonlinear distortion.

As described above, the neural network signal is a signal related to the test reproduction signal inputted first among the plurality of test reproduction signals stored in the plurality of storage elements 92A1. That is, the neural network signal is a signal related to the test reproduction signal stored in the terminal storage element E1. Note that among the plurality of test reproduction signals stored in the plurality of storage elements 92A1, the signal related to the test reproduction signal inputted first is a signal related to the test reproduction signal inputted first, which is stored in the plurality of delay elements stored in the plurality of delay elements. This is an example of "a value related to the first input reproduction signal among the reproduction signals of". Further, the neural network signal is an example of a "later layer value" according to the technology of the present disclosure.

Here, in the following formulas (2) to (8), "m" is a natural number from 1 to 3, "k" is a natural number from 1 to _Nm , and "j" is a natural number from 1 to Nm _-1 . When it is a natural number, the above-mentioned delay time is calculated by the following formula (2). In addition, test playback signals, real-time playback signals, subsequent layer connection weights, composite values, activation functions, conversion values, subsequent layer values, intermediate layer connection loads, intermediate layer values, output layer connection loads, neural network signals, waveforms, etc. The reproduced signal after conversion is expressed by variables shown in the following equations (3) and (4). In addition, below, unless it is necessary to explain separately the latter layer joint load, the intermediate layer joint load, and the output layer joint load, they will simply be referred to as "joint load." Further, the connection weight and the threshold value are examples of the optimization variables described above, and are used to optimize the neural network 108.

As an example, as shown in FIG. 12, the learning stage calculation unit 92B outputs a neural network signal from the output layer 108C of the neural network 108 to the error calculation unit 92C as a calculation result. The error calculation unit 92C calculates the calculation result by the learning stage calculation unit 92B, that is, the error between the neural network signal input from the output layer 108C of the neural network 108 and the teacher data 110, and applies the calculated error to the variable adjustment unit 92D. Output to. Here, the error is calculated by the following formula (5).

The variable adjustment unit 92D adjusts the optimization variables using the error backpropagation method so as to minimize the error calculated by the error calculation unit 92C. Therefore, the variable adjustment unit 92D calculates the adjustment value used for adjusting the optimization variable, that is, the amount of change in the optimization variable according to the following equations (6) and (7). Specifically, the adjustment value of the connection load is calculated using equation (6), and the adjustment value of the threshold value is calculated using equation (7).

The variable adjustment unit 92D adjusts the connection weight of the neural network 108 using the adjustment value calculated from equation (6), and adjusts the threshold value of the neural network 108 using the adjustment value calculated from equation (7). , optimize the neural network 108. The neural network 108 performs learning by being optimized by the variable adjustment unit 92D for each of the above-mentioned reading environment conditions. As an example, as shown in FIG. 13, the neural network 108 that has been optimized by the variable adjustment unit 92D for each reading environment condition is stored as a trained model 82 by the learning stage calculation unit 92B in the storage 96 for each reading environment condition. is memorized.

Next, the operation stage will be explained with reference to FIGS. 14 and 15.

As an example, as shown in FIG. 14, the trained model 82 is transferred from the storage 96 of the computer 90 and stored in the storage 74. The trained model 82 is transferred from the computer 90 to the equalizer 60, for example, while the computer 90 is connected to the external I/F 28 (see FIG. 1), according to instructions received by the UI device 26. Ru. Specifically, the trained model 82 is selected from the storage 96 of the computer 90 in accordance with an instruction received by the UI device 26, and the selected trained model 82 is stored in the storage 74 by the CPU 70 of the equalizer 60. Ru.

Note that the storage 74 may store a trained model that is a combination of a plurality of trained models 82, and may be used by the CPU 70. Here, integrating the plurality of trained models 82 refers to, for example, composing the plurality of trained models 82 by averaging mutually corresponding optimization variables among the plurality of trained models 82. In this case, for example, the CPU 70 stores a learned model 82 of reading environmental conditions including a traveling speed condition of 3 m/s, a learned model 82 of reading environmental conditions including a traveling speed condition of 4 m/s, and a learned model 82 of reading environmental conditions including a traveling speed condition of 5 m/s. By integrating the learned model 82 of the reading environment conditions including the running speed condition, a trained model corresponding to the running speed of 3 m/s to 5 m/s is generated and stored in the storage 74. Good too.

The CPU 70 reads the waveform equalization execution program 80 from the storage 74 and executes the read waveform equalization execution program 80, thereby operating as the operation stage delay storage unit 70A and the operation stage calculation unit 70B.

The operation stage delay storage unit 70A acquires the real-time reproduction signal bit by bit from the phase synchronization circuit 58, and stores the real-time reproduction signal bit by bit in the order of acquisition in a second storage element group (for example, a plurality of storage elements in the internal memory 71 shown in FIG. 7). element). That is, every time the operational stage delay storage unit 70A acquires the real-time reproduction signal bit by bit from the phase synchronization circuit 58, it stores it in the second storage element group in chronological order in the acquisition order. Furthermore, the real-time reproduction signal is stored in the second storage element group in a FIFO format. The operation stage calculation unit 70B performs calculations using the real-time reproduction signal stored in time series by the operation stage delay storage unit 70A and the learned model 82 in the storage 74.

As an example, as shown in FIG. 15, the operation stage delay storage section 70A includes a plurality of storage elements 70A1 as the above-mentioned second storage element group. The plurality of storage elements 70A1 are realized by, for example, the internal memory 71. Each of the plurality of storage elements 70A1 is a delay element, and the real-time reproduction signal is input to the plurality of storage elements 70A1 after being delayed by the above-mentioned delay time. The plurality of storage elements 70A1 store real-time reproduction signals in time series. That is, each storage element 70A1 delays the real-time playback signal by 1 bit and stores the real-time playback signal every time one bit of the real-time playback signal is input. In the example shown in FIG. 15, a plurality of storage elements 70A1 are connected in series, and the real-time reproduction signal is delayed by one bit and stored in each storage element 70A1. Real-time reproduction signals are stored in the plurality of storage elements 70A1 in a FIFO format, and the real-time reproduction signals stored in the terminal storage element 70A1 (hereinafter also referred to as "termination storage element E2") are stored in the operation stage delay storage section. When a new 1-bit real-time playback signal is input to 70A, it is erased from storage element 70A1.

The trained model 82 includes a former layer 107 having a plurality of former layer nodes 107A corresponding to a plurality of storage elements 70A1, and a latter layer 109 having a plurality of later layer nodes 109A. Each of the plurality of storage elements 70A1 outputs the input real-time reproduction signal to a corresponding one of the plurality of previous-layer nodes 107A. Each of the plurality of front layer nodes 107A outputs the real-time reproduction signal input from the corresponding storage element 70A1 among the plurality of storage elements 70A1 to the rear layer 109. The subsequent layer 109 reproduces the product of the real-time reproduction signal (“x” in the example shown in FIG. 15) input from the plurality of previous layer nodes 107A and the subsequent layer coupling load (“w” in the example shown in FIG. 15). The composite value obtained based on the sum is transformed using an activation function. Then, the subsequent layer 109 outputs a subsequent layer value (in the example shown in FIG. 15, a "waveform equalized reproduced signal") based on a converted value obtained by converting the composite value using an activation function. The subsequent layer value output from the subsequent layer 109 is a value related to the real-time reproduction signal stored in the termination storage element E2.

In the example shown in FIG. 15, an input layer 108A is shown as an example of the former layer 107, and an intermediate layer 108B and an output layer 108C are shown as examples of the latter layer 109. The input layer 108A has a plurality of input layer nodes 108A1. The middle layer 108B has a plurality of middle layer nodes 108B1. The output layer 108C has an output layer node 108C1. In the example shown in FIG. 15, _N1 input layer nodes 108A1 are shown as an example of the plurality of previous layer nodes 107A, and _N2 intermediate layer nodes 108B1, as an example of the plurality of subsequent layer nodes 109A. and N ₃ output layer nodes 108C1 are shown.

Each of the plurality of input layer nodes 108A1 outputs the real-time reproduction signal input from the corresponding one of the plurality of storage elements 70A1 to the intermediate layer 108B. The plurality of intermediate layer nodes 108B1 convert the intermediate layer value obtained as the above-mentioned composite value into an activation function (e.g. , a sigmoid function shown in equation (3)), the above-mentioned converted value is generated and output to the output layer 108C. Here, the intermediate layer value is a value based on the sum of products of the real-time reproduction signal and the intermediate layer coupling load and a threshold value (see formula (4)). Specifically, the intermediate layer value is the value obtained by subtracting the threshold value from the sum of products of the real-time reproduction signal and the intermediate layer coupling load, as shown in Equation (4). The output layer 108C outputs the output layer value obtained as the above-described subsequent layer value as a waveform-equalized reproduced signal based on the product sum of the conversion value input from the intermediate layer 108B and the output layer coupling weight.

The waveform-equalized reproduced signal outputted from the output layer 108C is the real-time reproduced signal inputted first among the plurality of real-time reproduced signals stored in the plurality of storage elements 70A1, that is, the one inputted to the terminal storage element E2. This is a signal related to a stored real-time playback signal. Note that among the plurality of real-time reproduction signals stored in the plurality of storage elements 70A1, the signal related to the real-time reproduction signal inputted first is the "multiple delay elements stored in the plurality of delay elements" according to the technology of the present disclosure. This is an example of "a value related to the first input reproduction signal among the reproduction signals of". Further, the reproduced signal after waveform equalization is an example of a "later layer value" according to the technology of the present disclosure.

Further, the reproduced signal after waveform equalization is a value based on the sum of products of the conversion value input from the intermediate layer 108B and the output layer coupling weight, and a threshold value (see formula (4)). The threshold value used for the sum of products of the conversion value and the output layer connection load is a value that subtracts the sum of products of the conversion value and the output layer connection load (see formula (4)).

Next, the operation of the magnetic tape drive 10 will be explained.

First, the learning execution process executed by the CPU 92 of the computer 90 will be described with reference to FIG. 16. The learning execution process is executed by the CPU 92 of the computer 90 according to the learning execution program 106 when the UI device 104 (see FIG. 8) receives an instruction to start execution of the learning execution process. For convenience of explanation, the following description will be made on the assumption that the test reproduction signal is supplied bit by bit from the test reproduction signal supply device 102 to the learning stage delay storage section 92A (see FIGS. 9 to 12).

In the learning execution process shown in FIG. 16, first, in step ST10, the learning stage delay storage section 92A obtains a 1-bit test reproduction signal from the test reproduction signal supply device 102, and then the learning execution process proceeds to step ST12. Transition.

In step ST12, the learning stage delay storage section 92A stores the test reproduction signal acquired in step ST10 in a plurality of storage elements 92A1 in a FIFO format and in time series, and then the learning execution process moves to step ST14.

In step ST14, the learning stage delay storage section 92A determines whether the test reproduction signal is stored in all the storage elements 92A1. In step ST14, if the test reproduction signal is not stored in all the storage elements 92A1, the determination is negative and the learning execution process moves to step ST10. In step ST14, if the test reproduction signal is stored in all the storage elements 92A1, the determination is affirmative and the learning execution process moves to step ST16.

In step ST16, the learning stage calculation unit 92B performs calculation using the test reproduction signal stored in the plurality of storage elements 92A1 in time series by the learning stage delay storage unit 92A and the neural network 108 in the storage 96, and then , the learning execution process moves to step ST18.

In step ST18, the error calculation section 92C determines whether a neural network signal has been output from the learning stage calculation section 92B. In step ST18, if the neural network signal is not output from the learning stage calculation section 92B, the determination is negative and the determination in step ST18 is performed again. In step ST18, if the neural network signal is output from the learning stage calculation unit 92B, the determination is affirmative and the learning execution process moves to step ST20.

In step ST20, the error calculation unit 92C calculates the error between the neural network signal input from the learning stage calculation unit 92B and the teacher data 110 in the storage 96, and then the learning execution process moves to step ST22.

In step ST22, the variable adjustment unit 92D determines whether the error calculated in step ST20 is within a predetermined range. Here, the predetermined range refers to a generally accepted range in the technical field to which the technology of the present disclosure belongs. The predetermined range may be a fixed value or a variable value that is changed depending on a given condition (for example, the content of an instruction received by the UI device 104). In step ST22, if the error calculated in step ST20 is within the predetermined range, the determination is affirmative and the learning execution process moves to step ST28. In step ST22, if the error calculated in step ST20 is outside the predetermined range, the determination is negative and the learning execution process moves to step ST24.

In step ST24, the variable adjustment unit 92D calculates an adjustment value for adjusting the optimization variable so as to minimize the error calculated in step ST20, and then the learning execution process moves to step ST26.

In step ST26, the variable adjustment unit 92D adjusts optimization variables such as connection weights using the adjustment values calculated in step ST28, and then the learning execution process moves to step ST10.

In step ST28, the learning stage calculation unit 92B stores the latest neural network 108 used in the calculation in step ST16 as the trained model 82 in the storage 96, and then the learning execution process ends.

Next, the waveform equalization execution process executed by the CPU 70 of the equalizer 60 will be described with reference to FIG. 17. The waveform equalization execution process shown in FIG. 17 is executed by the CPU 70 of the equalizer 60 when an instruction to start the waveform equalization execution process is accepted by the UI device 26 (see FIG. 1). The program 80 is executed.

Note that the flow of the waveform equalization execution process shown in FIG. 17 is an example of the "processing method of a signal processing device" and the "operating method of a magnetic tape reading device" according to the technology of the present disclosure. Furthermore, for convenience of explanation, the following description will be made on the assumption that the real-time reproduction signal is supplied bit by bit from the phase synchronization circuit 58 to the operational stage delay storage section 70A (see FIGS. 14 and 15). Furthermore, for convenience of explanation, the following description will be made on the premise that the learned model 82 obtained by executing the learning execution process is already stored in the storage 74 of the equalizer 60.

In the waveform equalization execution process shown in FIG. 17, first, in step ST50, the operation stage delay storage unit 70A obtains a 1-bit real-time reproduction signal from the phase synchronization circuit 58, and then the waveform equalization execution process is performed in step ST50. Move to ST52.

In step ST52, the operation stage delay storage unit 70A stores the real-time reproduction signal acquired in step ST50 in the plurality of storage elements 70A1 in FIFO format and in chronological order, and then the waveform equalization execution process moves to step ST54. .

In step ST54, the operation stage delay storage section 70A determines whether the real-time reproduction signal is stored in all the storage elements 70A1. In step ST54, if the real-time reproduction signal is not stored in all of the storage elements 70A1, the determination is negative and the waveform equalization execution process moves to step ST50. In step ST54, if the real-time reproduction signal is stored in all the storage elements 70A1, the determination is affirmative, and the waveform equalization execution processing moves to step ST56.

In step ST56, the operation stage calculation unit 70B performs calculation using the real-time reproduction signal stored in the plurality of storage elements 70A1 in time series by the operation stage delay storage unit 70A and the trained model 82 in the storage 74. Then, a waveform-equalized reproduction signal regarding the real-time reproduction signal stored in the termination storage element E2 is generated, and then the waveform equalization execution processing moves to step ST58.

In step ST58, the operation stage calculation unit 70B outputs the waveform equalized reproduced signal generated in step ST56 to the decoder 62, and then the waveform equalization execution process ends.

Incidentally, a waveform equalization method using a conventionally known linear filter is generally known as a method of waveform equalization for a real-time reproduced signal. An FIR filter is known as a linear filter. Figure 18 shows the SNR of the signal obtained by waveform equalization of the real-time reproduced signal using the FIR equalization method (hereinafter also referred to as "FIR equalization method SNR"), and the real-time An example of the results of comparison with the SNR of a signal obtained by waveform equalization of a reproduced signal (hereinafter also referred to as "neural network equalization SNR") is shown. Here, the FIR equalization method refers to a waveform equalization method using a conventionally known FIR filter. Further, the neural network equalization method refers to a waveform equalization method (for example, , a waveform equalization method using the waveform equalization execution process shown in FIG. 17). In the graph shown in FIG. 18, the horizontal axis is the running speed of the magnetic tape MT, and the vertical axis is the SNR.

As an example, as shown in FIG. 18, both the FIR equalization method SNR and the neural network equalization method SNR decrease as the running speed of the magnetic tape MT increases. However, regardless of the running speed of the magnetic tape MT, the SNR of the neural network equalization method is higher than the SNR of the FIR equalization method. Furthermore, in the example shown in FIG. 18, the higher the running speed of the magnetic tape MT, the greater the ratio of the neural network equalization method SNR to the FIR equalization method SNR. In other words, from the example shown in FIG. 18, it is clear that the SNR is better when the waveform of the real-time reproduced signal is equalized using the neural network equalization method than when the waveform of the real-time reproduced signal is equalized using the FIR equalization method. It can be seen that this contributes to

FIG. 19 also shows that waveform equalization processing using an FIR filter is performed on the real-time playback signal when the running speed of the magnetic tape MT is 6 m/s, 4 m/s, and 2 m/s. Waveform equalization processing (for example, the waveform shown in FIG. An example of the results of comparison with noise generated when equalization execution processing is performed on the real-time reproduced signal is shown. In the graph shown in FIG. 19, the horizontal axis is the position in the longitudinal direction of the magnetic tape MT, and the vertical axis is the signal value of the real-time reproduction signal output from the phase synchronization circuit 58.

As an example, as shown in FIG. 19, in waveform equalization processing using an FIR filter, regardless of the running speed of the magnetic tape MT, a specific arrangement of data recorded on the magnetic tape MT, that is, a specific recording pattern Nonlinear noise that has a dependence on On the other hand, in the waveform equalization process according to the technique of the present disclosure, nonlinear noise that is dependent on a specific recording pattern is reduced compared to the waveform equalization process using an FIR filter.

As described above, in the magnetic tape drive 10, the equalizer 60 performs waveform equalization of the real-time reproduction signal. Waveform equalization is performed using a trained model 82 obtained by performing training on the neural network 108 to reduce nonlinear distortion that occurs depending on the reading environment conditions (see FIG. 10). Therefore, according to this configuration, it is possible to reduce nonlinear distortion occurring in the real-time reproduced signal, compared to the case where the waveform equalization of the real-time reproduced signal is performed using a linear filter. Furthermore, in the first embodiment, as shown in FIG. 5, an element-specific signal processing device 50A is provided for each of the plurality of reading elements 16A, so an equalizer 60 is provided for each reading element 16A. The waveform equalization of the reproduced signal sequence is performed by. Therefore, according to this configuration, it is possible to reduce distortion that nonlinearly occurs in a plurality of reproduced signal sequences, compared to a case where waveform equalization of a plurality of reproduced signal sequences is performed using a linear filter.

Further, in the magnetic tape drive 10, waveform equalization is performed using a trained model 82 obtained by performing training on the neural network 108 to reduce nonlinear distortion that occurs depending on the read head conditions. Therefore, according to this configuration, nonlinear distortion occurring in the real-time reproduced signal due to individual differences in the reading heads 16 can be reduced, compared to the case where the waveform equalization of the real-time reproduced signal is performed using a linear filter.

Further, in the magnetic tape drive 10, waveform equalization is performed using a trained model 82 obtained by performing training on the neural network 108 to reduce nonlinear distortion that occurs depending on magnetic tape conditions. Therefore, with this configuration, compared to the case where waveform equalization of the real-time playback signal is performed using a linear filter, it is possible to reduce nonlinear distortion that occurs in the real-time playback signal due to individual differences in the magnetic tape MT.

Further, in the magnetic tape drive 10, waveform equalization is performed using a trained model 82 obtained by performing learning on the neural network 108 to reduce nonlinear distortion that occurs depending on the running speed condition. Therefore, according to this configuration, compared to the case where the waveform equalization of the real-time playback signal is performed using a linear filter, it is possible to reduce the nonlinear distortion that occurs in the real-time playback signal due to the running speed of the magnetic tape MT.

In addition, in the magnetic tape drive 10, waveform equalization is performed using a trained model 82 obtained by performing training on the neural network 108 to reduce nonlinear distortion that occurs depending on A/D converter conditions. be exposed. Therefore, according to this configuration, nonlinear distortion occurring in the real-time reproduced signal due to individual differences in the A/D converter 54 can be reduced, compared to the case where waveform equalization of the real-time reproduced signal is performed using a linear filter. I can do it.

Further, in the magnetic tape drive 10, the waveform equalization of the real-time reproduction signal is performed by the equalizer 60 using the trained model 82 as a nonlinear filter trained to reduce nonlinear distortion. Therefore, according to this configuration, it is possible to reduce nonlinear distortion occurring in the real-time reproduced signal, compared to the case where the waveform equalization of the real-time reproduced signal is performed using a linear filter.

Further, in the magnetic tape drive 10, the learned model 82 has a former layer 107 and a latter layer 109. The pre-layer 107 has a plurality of pre-layer nodes corresponding to the plurality of storage elements 70A1. Each of the plurality of storage elements 70A1 outputs the input real-time reproduction signal to a corresponding one of the plurality of previous-layer nodes 107A. Each of the plurality of front layer nodes 107A outputs the real-time reproduction signal input from the corresponding storage element 70A1 among the plurality of storage elements 70A1 to the rear layer 109. The subsequent layer 109 uses an activation function to convert a composite value obtained based on the sum of products of the real-time playback signals input from the plurality of previous layer nodes 107A and the subsequent layer coupling loads. The subsequent layer connection weight is determined by performing learning on the neural network 108 to reduce nonlinear distortion. Further, the subsequent layer 109 outputs a waveform-equalized reproduced signal as a subsequent layer value based on a converted value obtained by converting the composite value using an activation function. Therefore, according to this configuration, it is possible to reduce nonlinear distortion occurring in the real-time reproduced signal, compared to the case where the waveform equalization of the real-time reproduced signal is performed using a linear filter.

Further, in the magnetic tape drive 10, a value obtained by subtracting a threshold value from the sum of products of the converted value and the subsequent layer coupling load is used as the subsequent layer value. The threshold value used for the product sum of the conversion value and the subsequent layer connection weight is determined by performing learning on the neural network 108 to reduce nonlinear distortion. Therefore, according to this configuration, the nonlinear distortion occurring in the real-time reproduced signal is greater than the case where the subsequent layer value is determined without using the threshold value determined by performing learning to reduce nonlinear distortion on the neural network 108. can be reduced with high precision.

Further, in the magnetic tape drive 10, real-time playback signals are input to the plurality of storage elements 70A1 after being delayed by a delay time. The plurality of storage elements 70A1 are a plurality of delay elements into which the real-time reproduction signal is input after being delayed by the delay time, and the waveform-equalized reproduction signal outputted from the latter layer 109 as a later layer value is inputted to the plurality of storage elements 70A1. This value is related to the real-time playback signal inputted first among the plurality of real-time playback signals stored in 70A1, that is, the real-time playback signal stored in the termination storage element E2. Therefore, according to this configuration, waveform-equalized playback signals corresponding to real-time playback signals can be obtained from the subsequent layer 109 in the order of the real-time playback signals input to the plurality of storage elements 70A1.

Further, in the magnetic tape drive 10, the learned model 82 has an input layer 108A as a former layer 107, and an intermediate layer 108B and an output layer 108C as a latter layer 109. Further, the input layer 108A includes a plurality of input layer nodes 108A1 as a plurality of previous layer nodes 107A. The intermediate layer 108B has a plurality of intermediate layer nodes 108B1 as a plurality of subsequent layer nodes 109A. The subsequent layer 109 has an output layer node 108C1 as a subsequent layer node 109A. Each of the plurality of input layer nodes 108A1 outputs the real-time reproduction signal input from the corresponding storage element 92A1 among the plurality of storage elements 92A1 to the intermediate layer 108B. The plurality of intermediate layer nodes 108B1 converts the intermediate layer value obtained as a composite value based on the product sum of the real-time reproduction signal inputted from the plurality of input layer nodes 108A1 and the intermediate layer coupling weight using an activation function. A converted value is generated and output to the output layer 108C. The intermediate layer connection weight is determined by performing learning on the neural network 108 to minimize the error between the neural network signal and the teacher data 110. The output layer 108C outputs the waveform-equalized reproduced signal as an output layer value based on the sum of products of the conversion value input from the intermediate layer 108B and the output layer coupling weight. The output layer connection weight is determined by performing learning on the neural network 108 to minimize the error between the neural network signal and the teacher data 110. Therefore, according to this configuration, it is possible to reduce nonlinear distortion occurring in the real-time reproduced signal, compared to the case where the waveform equalization of the real-time reproduced signal is performed using a linear filter.

Further, in the magnetic tape drive 10, a value obtained by subtracting a threshold value from the sum of products of the real-time reproduction signal and the interlayer coupling load is adopted as the intermediate layer value. The threshold value used for the product sum of the real-time reproduction signal and the intermediate layer connection weight is determined by performing learning on the neural network 108 to reduce nonlinear distortion. Therefore, according to this configuration, the nonlinear distortion that occurs in the real-time reproduced signal is greater than when the intermediate layer value is determined without using the threshold value determined by performing learning to reduce nonlinear distortion on the neural network 108. can be reduced with high precision.

Furthermore, in the magnetic tape drive 10, the teacher data 110 is an ideal reproduction signal regarding known data recorded on the learning magnetic tape in a predetermined recording pattern along the longitudinal direction of the learning magnetic tape. Therefore, according to this configuration, nonlinear distortion occurring in the real-time playback signal can be reduced compared to the case where teaching data unrelated to known data recorded in a predetermined recording pattern on the learning magnetic tape is used. I can do it.

In the first embodiment, the teacher data 110 is an ideal reproduction signal (hereinafter referred to as an ideal reproduction signal) regarding known data recorded on the learning magnetic tape in a predetermined recording pattern along the longitudinal direction of the learning magnetic tape. , also referred to as "first ideal reproduction signal"), the technology of the present disclosure is not limited thereto. For example, it may be an ideal reproduction signal (hereinafter also referred to as "second ideal reproduction signal") derived by computer simulation. Further, a signal obtained by combining the first ideal reproduction signal and the second ideal reproduction signal may be used as the teacher data 110. Here, the combination of the first ideal reproduction signal and the second ideal reproduction signal refers to, for example, the average of the first ideal reproduction signal and the second ideal reproduction signal. In this way, the second ideal reproduction signal may be used as the teacher data 110, or the first ideal reproduction signal and the second ideal reproduction signal may be combined and used. As a result, nonlinear distortion occurring in the real-time reproduction signal can be reduced with high precision compared to the case where teacher data unrelated to either the first ideal reproduction signal or the second ideal reproduction signal is used.

Further, in the first embodiment described above, each of the neural network 108 and the trained model 82 is explained by giving an example of a form consisting of three layers: the input layer 108A, the intermediate layer 108B, and the output layer 108C. The technology is not limited to this. For example, in the learning stage, as shown in FIG. 20, a neural network 208 is applied instead of the neural network 108, and in the operation stage, as shown in FIG. apply. Each of the neural network 208 and the trained model 182 consists of two layers: an input layer 108A and an output layer 108C. Even when waveform equalization of a real-time reproduced signal is performed using the neural network 208 and learned model 182 configured in this way, the real-time reproduced signal is equalized using a linear filter as in the first embodiment. Compared to the case where waveform equalization is performed, nonlinear distortion occurring in the real-time reproduced signal can be reduced.

Further, in the first embodiment, the case where the intermediate layer 108B is a single layer has been described, but the technology of the present disclosure is not limited to this. For example, in the learning stage, as shown in FIG. 22, a neural network 308 having a plurality of intermediate layers 108B is applied instead of the neural network 108, and in the operational stage, as shown in FIG. Instead, a trained model 282 having multiple hidden layers 108B is applied. The number of layers in the intermediate layer 108B may be determined according to the number of nodes included in the input layer 108A and the number of nodes included in the output layer 108C.

Furthermore, in the first embodiment, the output layer 108C has been described using an example in which the output layer 108C outputs a reproduced signal after waveform equalization regarding only the real-time reproduced signal stored in the termination storage element E2 (see FIG. 15). , the technology of the present disclosure is not limited thereto. For example, the output layer 108C may have a plurality of output layer nodes 108C1 (eg, the same number of output layer nodes 108C1 as input layer nodes 108A1). In this case, for example, a two-dimensional real-time playback signal that is likened to a two-dimensional image is input to a plurality of input layer nodes 108A1, and a convolutional neural network having a convolution layer and a pooling layer is used to create a plurality of output layers. The node 108C1 may output a waveform-equalized reproduced signal that looks like a two-dimensional image.

Further, in the first embodiment, a combination of the read head condition, magnetic tape condition, running speed condition, and A/D converter condition is exemplified as the read environment condition, but the technology of the present disclosure is not limited to this. , the read environment condition may be at least one of read head conditions, magnetic tape conditions, running speed conditions, and A/D converter conditions.

Further, the reading environment conditions are not limited to at least one of the reading head conditions, magnetic tape conditions, running speed conditions, and A/D converter conditions, and may be used in place of or in place of these conditions. Other conditions may be applied together with at least one of the conditions. Other conditions include, for example, conditions resulting from individual differences (types) of recording patterns determined according to running speed and bit interval. In addition, other conditions include, for example, a plurality of processing circuits (hereinafter simply referred to as "processing Also included are conditions (hereinafter referred to as "processing circuit conditions") caused by individual differences in at least one of the circuits (hereinafter referred to as "processing circuit conditions"). Processing circuit conditions refer to conditions caused by individual differences in processing circuits. Individual differences in processing circuits refer to, for example, differences in characteristics between processing circuits. Differences in characteristics between processing circuits are mainly caused by manufacturing errors in the processing circuits and/or deterioration over time of the processing circuits. Indices that quantitatively indicate the degree of deterioration of a processing circuit over time include, for example, the number of times the processing circuit is used, the average time that the processing circuit is continuously used, and the number of points at a specific point in time after the processing circuit was manufactured. For example, the time taken to reach the current point in time can be cited.

Further, in the first embodiment, the equalizer 60 uses the learned model 82, but the technology of the present disclosure is not limited to this, and the learned model 82 of the equalizer 60 may be finely adjusted by the CPU 70 of the equalizer 60. In this case, for example, the CPU 70 of the equalizer 60 may execute the learning execution process according to the learning execution program 106 using the real-time reproduction signal as the above-mentioned test reproduction signal.

Further, in the first embodiment, the sigmoid function was illustrated as an example of the "activation function" according to the technology of the present disclosure, but the technology of the present disclosure is not limited to this, and instead of the sigmoid function, or Along with the function, other activation functions may be applied, such as a hyperbolic tangent function, a ramp function, and/or a softmax function.

Further, in the first embodiment, the trained model 82 is illustrated as a nonlinear filter, but the technology of the present disclosure is not limited to this. For example, the learning to reduce the nonlinear distortion described in the first embodiment can be performed. A modified IIR filter may also be used.

Further, in the first embodiment, an example is given in which a real-time reproduced signal subjected to phase synchronization processing is inputted from the phase synchronization circuit 58 to the equalizer 60, but the technology of the present disclosure is limited to this. Not done. For example, as shown in FIG. 24, the position of the phase synchronization circuit 58 and the position of the equalizer 60 may be exchanged. That is, the real-time reproduced signal is input from the LPF 56 to the equalizer 60 without going through the phase synchronization circuit 58, and the equalizer 60 performs waveform equalization of the real-time reproduced signal to obtain a waveform-equalized reproduced signal. is input to the phase synchronization circuit 58, and the phase synchronization circuit 58 performs phase synchronization processing on the reproduced signal after waveform equalization based on the decoding result of the decoder 62.

Further, in the first embodiment, the plurality of storage elements 70A1 (see FIG. 15) are realized by the internal memory 71 (see FIG. 7), and the plurality of storage elements 92A1 (see FIG. 11) are realized by the internal memory 93 (see FIG. 8). ), the technology of the present disclosure is not limited thereto. For example, instead of the operation stage delay storage unit 70A, a delay circuit realized by connecting a plurality of storage elements 70A1 in series may be used as a circuit separate from the CPU 70. Further, for example, instead of the learning stage delay storage section 92A, a delay circuit realized by connecting a plurality of storage elements 92A1 in series may be used as a circuit separate from the CPU 92.

[Second embodiment]
In the first embodiment described above, the read head 16 was illustrated, but in the second embodiment, multi-channel recording and reading (i.e., reproduction) are performed using a tape head (for example, the magnetic head 112 shown in FIG. 27). An example of the form in which this is performed will be explained. In the second embodiment, the same components as those described in the first embodiment are given the same reference numerals, explanations thereof are omitted, and only parts different from the first embodiment will be explained.

For example, the tape head is equipped with 32 magnetic elements for recording and for reading (for example, the data magnetic element DRW shown in FIGS. 27 and 28). This makes it possible to record and read data at a higher data transfer rate than a tape head equipped with only a few magnetic elements. However, there is a problem in that there are large variations in performance between magnetic elements. Possible causes of large variations in performance between magnetic elements include, for example, manufacturing tolerances during tape head manufacture and variations in the degree of deterioration over time resulting from repeated use of tape heads.

Deterioration over time caused by repeated use of a tape head includes, for example, wear and scratches on the magnetic element caused by abrasive particles contained in the magnetic tape MT, and physical damage caused by contact with protruding particles on the magnetic tape MT. Examples include deterioration of magnetic performance due to repeated exposure to collisions.

These deteriorations are influenced by the accumulation of accidental events (for example, collisions with foreign objects and/or protrusions that are irregularly present in the magnetic tape MT) during use of the magnetic tape MT. Therefore, particularly in tape heads that have deteriorated over time due to repeated use, large variations in performance occur among the plurality of magnetic elements. In a deteriorated magnetic element, the signal-to-noise ratio decreases due to increased nonlinearity of the reproduced signal. Therefore, especially in tape heads that have deteriorated over time, the difference in signal-to-noise ratio between magnetic elements becomes large.

In order for the magnetic tape drive 10 to function properly as a storage system, it is necessary to ensure the minimum signal-to-noise ratio allowable in the storage system for all recording magnetic elements and all reading magnetic elements. becomes. Therefore, in order to improve the reliability of large-capacity tape systems, it is especially important to use playback signals from magnetic elements that have a large degree of deterioration and increased nonlinearity (for example, FIR filters, which are effective for linear waveform equalization, It is important to improve the quality of nonlinear reproduced signals that cannot be handled.

As described above, since the deterioration of the magnetic element is largely due to the increase in nonlinearity of the reproduced signal, even if recording and reading are performed in multiple channels, for example, as shown in FIG. Similarly to the first embodiment, the neural network 108 (see FIG. 9), which is an example of a nonlinear filter, detects the magnetic element based on the test reproduction signal generated according to each of the first to Nth reading environmental conditions. It is effective to optimize the characteristics for each case. This makes it possible to reduce variations in signal quality between magnetic elements.

Furthermore, the degree of deterioration in the quality of the reproduced signal varies depending on the combination of the magnetic element and the media. For example, if a tape head is used with a magnetic element whose recording performance has deteriorated due to repeated use, the quality of the playback signal will not be as high with media (e.g. magnetic tape) that uses magnetic particles with a small coercive force. does not change, but in media made of magnetic particles with high coercive force, recording defects cause noticeable quality deterioration of the reproduced signal.

Therefore, when media data is reproduced using a different combination of tape head and media, the above-mentioned nonlinear filter (eg, neural network 108) needs to be optimized each time. To do this, an area for optimizing the nonlinear filter (for example, the specific area 116 shown in FIG. 26) is secured on the magnetic tape MT separately from the area in which data used by the user is recorded, and then the nonlinear filter is A known pattern (for example, the specific pattern 116A shown in FIGS. 26 and 27) that is the basis of the teacher data 110 (see FIG. 9) is recorded in the area for optimization, and a reproduced signal (for example, the specific pattern 116A shown in FIG. 28) is recorded. An effective method is to optimize the nonlinear filter each time based on the characteristics of the test reproduction signal sequence.

As a result, for example, the quality of the playback signal can be improved compared to the case where a nonlinear filter optimized for a specific tape head and media combination is also used for other tape head and media combinations. becomes possible.

Additionally, as mentioned above, a tape head is equipped with multiple magnetic elements, which allows it to record and read data at a higher data transfer rate than a tape head that is equipped with only a few magnetic elements. is possible. In order to realize this, a magnetic tape drive 10 that transports the magnetic tape MT at high speed is required.

On the other hand, as shown in FIG. 6, when the magnetic tape MT is conveyed at high speed and data is recorded on the magnetic tape MT, the rise time of the recording magnetic field generated by the reading element 16A is delayed, so that the bit recording position may shift. arise. The greater the amount of deviation, the greater the nonlinearity of the reproduced signal. Therefore, in the second embodiment, the neural network 108 (see FIG. 9), which is the above-mentioned nonlinear filter, is optimized according to the running speed of the magnetic tape MT. An example of this case will be specifically described below.

In the second embodiment, as shown in FIG. 25 as an example, as an example of the running speed condition, the magnetic tape MT is Conditions regarding the running speed (that is, the speed at which the magnetic tape MT is running) are applied. The conditions related to the running speed may be, for example, the running speed itself, or the signals that control the rotational drive of the sending motor 20 and the take-up motor 24 (i.e., the conditions of the sending motor 20 and It may also be a signal for controlling the take-up motor 24). An example of the running speed of the magnetic tape MT is, for example, a speed of 2 m/s to 7 m/s. For example, each traveling speed condition included in the first reading environmental condition to the Nth reading environmental condition may be a condition related to a mutually different traveling speed.

Note that the first reading environmental condition to the Nth reading environmental condition may be changed depending on the usage status and/or usage environment of the plurality of magnetic heads 112 and/or the plurality of magnetic tape drives 10, or may be changed over time. It may be changed or may be changed in accordance with an instruction given by a user or the like. Further, the first to Nth reading environmental conditions may be associated with each magnetic head 112, each magnetic element unit 120 (see FIG. 27), or each magnetic tape drive 10, for example. The first to Nth reading environmental conditions are stored, for example, in a storage device of a host computer (not shown) such as a mainframe or a cloud server that can communicate with at least one magnetic tape drive 10. Good too. In this case, an instruction received by the UI device 26 (see FIG. 1) of the magnetic tape drive 10, an instruction given directly or indirectly to the test playback signal supply device 102, or an instruction given to the host computer Test playback is performed based on the read environment conditions corresponding to the magnetic head 112 or the magnetic tape drive 10 from among a plurality of read environment conditions stored in the storage device of the host computer, according to instructions given directly or indirectly. The test reproduction signal may be generated by the signal supply device 102. Also, among a plurality of reading environment conditions stored in the storage device of the host computer, instructions received by the UI system device 26 (see FIG. 1) of the magnetic tape drive 10, and instructions directly transmitted to the test reproduction signal supply device 102. A test playback signal is generated by the test playback signal supply device 102 based on an instruction given directly or indirectly to the host computer, or based on reading environment conditions selected according to an instruction given directly or indirectly to the host computer. It is also possible to do so. Further, the reading environment conditions obtained from the host computer may be written to a storage medium (for example, the cartridge memory 122 (see FIG. 30) and/or the BOT area 114 of the magnetic tape MT (see FIG. 26), etc.). The test reproduction signal supply device 102 may generate a test reproduction signal according to the reading environment conditions stored on the medium.

Further, the plurality of running speed conditions that are different from each other (for example, the plurality of running speed conditions included in the first to Nth reading environment conditions) are the conditions for using the plurality of magnetic heads 112 and/or the plurality of magnetic tape drives 10. It may be changed according to the situation, it may be changed over time, or it may be changed according to an instruction given by a user or the like. Further, the plurality of running speed conditions may be associated with each magnetic head 112, each magnetic element unit 120 (see FIG. 27), or each magnetic tape drive 10. The plurality of traveling speed conditions may be stored, for example, in a storage device of a host computer (not shown) such as a mainframe or a cloud server that can communicate with at least one magnetic tape drive 10. In this case, an instruction received by the UI device 26 (see FIG. 1) of the magnetic tape drive 10, an instruction given directly or indirectly to the test playback signal supply device 102, or an instruction given to the host computer Test playback is performed based on the running speed condition corresponding to the magnetic head 112 or the magnetic tape drive 10 from among the plurality of running speed conditions stored in the storage device of the host computer, according to instructions given directly or indirectly. The test reproduction signal may be generated by the signal supply device 102. Also, among the plurality of running speed conditions stored in the storage device of the host computer, instructions received by the UI system device 26 (see FIG. 1) of the magnetic tape drive 10, and instructions directly transmitted to the test reproduction signal supply device 102 A test reproduction signal is generated by the test reproduction signal supply device 102 based on an instruction given directly or indirectly to the host computer, or a traveling speed condition selected according to an instruction given directly or indirectly to the host computer. It is also possible to do so. Further, the running speed condition acquired from the host computer may be written to a storage medium (for example, the cartridge memory 122 (see FIG. 30) and/or the BOT area 114 of the magnetic tape MT (see FIG. 26), etc.). The test reproduction signal supply device 102 may generate a test reproduction signal according to the reading environment conditions stored on the medium.

Furthermore, in the example shown in FIG. 25, read head conditions, magnetic tape conditions, and A/D converter conditions are also shown, but in the second embodiment, these conditions may or may not exist. . In the following, to avoid complications, a case will be described in which a test reproduction signal is generated only according to a running speed condition without read head conditions, magnetic tape conditions, and A/D converter conditions.

As an example, as shown in FIG. 26, in the second embodiment, a magnetic head 112 is used in place of the reading head 16 described in the first embodiment. The magnetic head 112 reads data from the magnetic tape MT. The magnetic tape MT has a BOT area 114. The BOT area 114 is a data band provided at the beginning of the magnetic tape MT. The BOT area 114 has a specific area 116 and a magnetic tape cartridge related information area 118. The specific area 116 and the magnetic tape cartridge related information area 118 are adjacent to each other, and are arranged in this order from the head of the magnetic tape MT.

A specific pattern 116A is recorded in the specific area 116 as data read by the magnetic head 112. The data recorded in the specific pattern 116A is an ideal reproduction signal sequence. The ideal reproduced signal sequence is also used, for example, as teacher data 110 (see FIG. 9).

The magnetic tape cartridge related information area 118 contains information regarding the magnetic tape cartridge 12 (for example, information indicating the type of magnetic tape MT, a summary of information stored in the main body of the magnetic tape MT, and/or information about the magnetic tape cartridge 12). 12) are recorded.

As an example, as shown in FIG. 27, the magnetic head 36 includes a holder 118 and a magnetic element unit 120. The magnetic element unit 120 is held by the holder 118 so as to be in contact with the running magnetic tape MT. The magnetic element unit 120 includes a servo reading element SR and a plurality of data magnetic elements DRW. In the example shown in FIG. 27, servo reading elements SR1 and SR2 are illustrated as the servo reading elements SR. In the following, for convenience of explanation, the servo reading elements SR1 and SR2 will be referred to as servo reading elements SR unless it is necessary to specifically explain them separately.

Servo reading element SR is provided at a position corresponding to servo band SB. In the example shown in FIG. 27, the servo reading element SR1 is arranged at a position facing the servo pattern 32 at one end of the magnetic tape MT in the tape width direction. The servo reading element SR2 is arranged at a position facing the servo pattern 32 at the other end of the magnetic tape MT in the tape width direction. The servo reading element SR1 reads the servo pattern 32 at one end in the tape width direction of the magnetic tape MT running in the forward or reverse direction, and the servo reading element SR2 reads the servo pattern 32 at one end in the tape width direction of the magnetic tape MT running in the forward or reverse direction. The servo pattern 32 at the other end of the tape MT in the tape width direction is read. Note that the magnetic head 112 moves in the tape width direction according to the servo pattern 32 read by the servo reading element SR. As described in the first embodiment, the movement of the magnetic head 112 in the tape width direction is realized by the movement mechanism 40 (see FIG. 3).

The plurality of data magnetic elements DRW are arranged at positions facing the track area 30 when the magnetic tape drive 10 is in a default state. Further, as described above, by moving the magnetic head 112 in the tape width direction according to the servo pattern 32, the plurality of data magnetic elements DRW are arranged at designated positions on the track area 30.

The plurality of data magnetic elements DRW record data on the magnetic tape MT running in the forward or reverse direction, or read data from the magnetic tape MT running in the forward or reverse direction.

As an example, as shown in FIG. 28, the magnetic element unit 120 includes a first data recording element group DWG1, a second data recording element group DWG2, and a data reading element group DRG. A servo reading element SR1 is located at one end of the magnetic element unit 120, and a servo reading element SR2 is located at the other end of the magnetic element unit 120.

In the example shown in FIG. 28, a servo reading element SR and a plurality of data magnetic elements DRW are shown as the plurality of magnetic elements included in the magnetic element unit 120. The data magnetic element DRW includes a first data recording element DW1, a second data recording element DW2, and a data reading element DR.

The first data recording element group DWG1 includes a plurality of first data recording elements DW1. The second data recording element group DWG2 includes a plurality of second data recording elements DW2. The data reading element group DRG includes a plurality of data reading elements DR.

Each of the first data recording element DW1 and the second data recording element DW2 records data in the track area 30 (see FIG. 27). Data reading element DR reads data from track area 30 (see FIG. 27). Note that, hereinafter, the first data recording element DW1 and the second data recording element DW2 will be referred to as a data recording element DW unless there is a need to specifically explain them separately.

The first data recording element group DWG1, the second data recording element group DWG2, and the data reading element group DRG are arranged from the take-up reel 22 (see FIG. 1) side to the cartridge reel CR (see FIG. 1) along the entire length direction of the magnetic tape MT. (see) side, a first data recording element group DWG1, a data reading element group DRG, and a second data recording element group DWG2 are arranged at regular intervals in this order. Here, the fixed interval refers to, for example, an interval predetermined by tests using actual equipment and/or computer simulation as an interval at which no crosstalk occurs between the data reading element DR and the data reading element DW. In addition, "constant" here means not only a completely constant error but also an error within a range that is allowable in the technical field to which the technology of the present disclosure belongs and does not deviate from the spirit of the technology of the present disclosure. It also means approximately constant, including.

The servo reading element SR includes a first servo reading element SRa, a second servo reading element SRb, and a third servo reading element SRc. The first servo reading element SRa, the second servo reading element SRb, and the third servo reading element SRc extend from the take-up reel 22 side to the cartridge reel CR side in the overall length direction of the magnetic tape MT. A servo reading element SRb and a third servo reading element SRc are provided in this order.

Note that although the first servo reading element SRa, the second servo reading element SRb, and the third servo reading element SRc are illustrated here, the technology of the present disclosure is not limited to this, and the first servo reading element SRa , the second servo reading element SRb, and the third servo reading element SRc.

The first data recording element group DWG1 includes a first servo reading element SRa of the servo reading element SR1, a first servo reading element SRa of the servo reading element SR2, and a plurality of first data recording elements DW1 (for example, 32 first It has a data recording element DW1). The plurality of first data recording elements DW1 are arranged in a straight line from the first servo reading element SRa side of the servo reading element SR1 to the first servo reading element SRa side of the servo reading element SR2.

The second data recording element group DWG2 includes a third servo reading element SRc of the servo reading element SR1, a third servo reading element SRc of the servo reading element SR2, and a plurality of second data recording elements DW2 (for example, 32 second It has a data recording element DW2). The plurality of second data recording elements DW2 are arranged in a straight line from the third servo reading element SRc side of the servo reading element SR1 to the third servo reading element SRc side of the servo reading element SR2.

The data reading element group DRG includes a second servo reading element SRb of the servo reading element SR1, a second servo reading element SRb of the servo reading element SR2, and a plurality of data reading elements DR (for example, 32 data reading elements DR). have The plurality of data reading elements DR are arranged in a straight line from the second servo reading element SRb side of the servo reading element SR1 to the second servo reading element SRb side of the servo reading element SR2.

In the magnetic element unit 46, the data reading element DR is sandwiched between the first data recording element DW1 and the second data recording element DW2 along the entire length direction of the magnetic tape MT. On the other hand, this is not only to read data from the track area 30 (see FIG. 27), but also to perform verification. For example, when pulling out the magnetic tape MT from the magnetic tape cartridge 10 (when the running direction of the magnetic tape MT is the forward direction), after the second data recording element DW2 records data on the data track DT, the data reading element DR Then, the data recorded on the data track DT by the second data recording element DW2 is read for error checking. Further, when returning the magnetic tape MT to the tape cartridge 10 (when the running direction of the magnetic tape MT is in the opposite direction), after the first data recording element DW1 records data on the data track DT, the data reading element DR , the data recorded in the track area 30 (see FIG. 27) is read by the first data reading element DW1 for error checking.

In the second embodiment, the running speed of the magnetic tape MT is adjusted by controlling the rotational drive of the feed motor 20 and the take-up motor 24 by the control device 18. In the example shown in FIG. 28, the control device 18 is connected to the test reproduction signal supply device 102, and controls the running speed of the magnetic tape MT according to instructions given from the test reproduction signal supply device 102.
For example, when the specific pattern 116A is recorded in the specific area 116, the test reproduction signal supply device 102 can provide the control device 18 with an instruction according to the reading environment conditions, that is, an instruction according to the traveling speed condition, etc. , causes the control device 18 to control the running speed of the magnetic tape MT. The traveling speed conditions used here relate to a predetermined traveling speed of the magnetic tape MT (for example, a specified speed from 2 m/s to 7 m/s) when recording is performed on the magnetic tape MT. It is a condition. The control device 18 controls the magnetic tape MT so that the running speed of the magnetic tape MT reaches a predetermined speed when recording is performed on the magnetic tape MT, according to instructions given from the test playback signal supply device 102. Control running speed.

The control device 18 controls a plurality of first data recording elements DW1 included in the first data recording element group DWG1 (for example, all the first data By operating the recording element DW1), the specific pattern 116A is recorded in the specific area 116 on the plurality of first data recording elements DW1. The specific pattern 116A is recorded in the specific area 116 at each position corresponding to each of the plurality of data magnetic elements DRW. Then, the specific pattern 116A is recorded in the specific area 116 by the plurality of first data recording elements DW1 included in the first data recording element group DWG1 arranged upstream of the data reading element group DRG in the forward direction. In parallel, the control device 18 causes the plurality of data reading elements DR to read the specific pattern 116A from the specific region 116 by operating the plurality of data reading elements DR included in the data reading element group DRG.

The magnetic element unit 120 is connected to the test reproduction signal supply device 102. The test reproduction signal supply device 102 obtains a plurality of reading results in which the specific pattern 116A is read by the plurality of data reading elements DR (that is, reading results read by each of the plurality of data reading elements DR). The reading result of the specific pattern 116A by the data reading element DR is acquired by the test reproduction signal supply device 102 as a test reproduction signal series which is time-series data (that is, a time-series reproduction signal).

The test reproduction signal supply device 102 obtains a test reproduction signal sequence for each traveling speed condition and supplies the acquired test reproduction signal sequence to the computer 90, thereby providing a test reproduction signal sequence for each traveling speed condition, as in the first embodiment. The neural network 108 is trained to generate a trained model 82 for each traveling speed condition. That is, the neural network 108 is optimized to have characteristics suitable for the plurality of data reading elements DR for the traveling speed condition based on the plurality of reading results.

Here, the characteristic suitable for the traveling speed condition is, for example, a characteristic that maximizes the SNR in a waveform-equalized reproduction signal sequence, or a characteristic that maximizes the SNR in a waveform-equalized reproduction signal sequence and an ideal test reproduction signal sequence. refers to the characteristic that minimizes the root mean square error (i.e., MSE) corresponding to the difference between .

In the above example, the specific pattern 116A is recorded in the specific area 116 on the plurality of first data recording elements DW1 while the magnetic tape MT is running in the forward direction according to the running speed conditions, etc. The technology of the present disclosure is not limited to this. For example, the specific pattern 116A may be recorded in the specific area 116 by the plurality of second data recording elements DW2 while the magnetic tape MT is running in the opposite direction according to the running speed conditions.

In this case, the specific pattern 116A is recorded in the specific area 116 by the plurality of second data recording elements DW2 included in the second data recording element group DWG2 arranged upstream of the data reading element group DRG in the opposite direction. In parallel, the control device 18 causes the plurality of data reading elements DR to read the specific pattern 116A from the specific area 116 by activating the plurality of data reading elements DR included in the data reading element group DRG. Just do it.

Parameters related to the trained model 82 obtained by optimizing the neural network 108 to characteristics suitable for the plurality of data reading elements DR based on the plurality of reading results (for example, the combinations included in the trained model 82 (load and/or threshold value, etc.) may be recorded on the magnetic tape MT by the magnetic head 112. In this case, for example, as shown in FIG. 29, the magnetic head 112 records parameters regarding the learned model 82 in the magnetic tape cartridge related information area 118.

Note that in the example shown in FIG. 29, parameters related to the learned model 82 are recorded in the magnetic tape cartridge related information area 118 of the BOT area 114, but this is just an example; In addition to or in place of the BOT area 114, parameters regarding the learned model 82 may be recorded in an EOT area (not shown) provided at the rear of the magnetic tape MT.

Further, for example, as shown in FIG. 30, when the magnetic tape cartridge 12 is equipped with a cartridge memory 122 as a non-contact storage medium, parameters regarding the learned model 82 may be stored in the cartridge memory 122. Generally, the cartridge memory 122 stores information regarding the magnetic tape MT. The information regarding the magnetic tape MT refers to, for example, management information for managing the magnetic tape cartridge 12. The management information includes, for example, information regarding the cartridge memory 122, information that allows identification of the magnetic tape cartridge 12, recording capacity of the magnetic tape MT, summary of data recorded on the magnetic tape MT, data items, and data recording. Contains information indicating the format, etc.

Cartridge memory 122 performs contactless communication with contactless reading/writing device 124 . Examples of the non-contact reading/writing device 124 include a non-contact reading/writing device used in the manufacturing process of the magnetic tape cartridge 12 and a non-contact reading/writing device used within the magnetic tape drive 10.

As explained above, according to the second embodiment, optimization is performed under conditions of different running speeds of the magnetic tape MT (here, as an example, running speeds from 2 m/s to 7 m/s). Since a trained model 82 (see FIG. 13) is generated in advance, by using the trained model 82 that corresponds to the running speed of the magnetic tape MT, the entire speed range (for example, from 2 m/s to 7 m/s) can be used. speed range), the nonlinearity of the reproduced signal can be reduced. As a result, it is possible to reduce variations in the quality of reproduced signals between the data magnetic elements DRW (that is, between the data reading elements DR).

Note that in the second embodiment, the first data recording element group DWG1, the second data recording element group DWG2, and the data reading element group DRG are illustrated, but the technology of the present disclosure is not limited thereto. For example, if the specific pattern 116A is recorded and read only along the forward direction, the second data recording element group DWG1, the second data recording element group DWG2, and the data reading element group DRG, Data recording element group DWG2 is unnecessary. For example, if the specific pattern 116A is recorded and read only along the reverse direction, among the first data recording element group DWG1, the second data recording element group DWG2, and the data reading element group DRG, The first data recording element group DWG1 is unnecessary.

Further, in the second embodiment, the first data recording element group DWG1, the second data recording element group DWG2, and the data reading element group DRG are collectively mounted on the magnetic head 112. , the technology of the present disclosure is not limited thereto. The first data recording element group DWG1, the second data recording element group DWG2, and the data reading element group DRG may be mounted on separate heads. Also in this case, the order in which the first data recording element group DWG1, the second data recording element group DWG2, and the data reading element group DRG are arranged along the running direction of the magnetic tape MT is the same as in the second embodiment. .

Further, in each of the above embodiments, an example in which the waveform equalization execution program 80 is stored in the storage 74 has been described, but the technology of the present disclosure is not limited to this. For example, as shown in FIG. 25, a waveform equalization execution program 80 may be stored in the storage medium 200. Storage medium 200 is a non-transitory storage medium. An example of the storage medium 200 is any portable storage medium such as an SSD or a USB memory.

The waveform equalization execution program 80 stored in the storage medium 200 is installed in the equalizer 60. The CPU 70 executes waveform equalization execution processing according to the waveform equalization execution program 80. Note that in the example shown in FIG. 25, the CPU 70 is a single CPU, but it may be a plurality of CPUs.

In addition, the waveform equalization execution program 80 is stored in a storage unit of another computer or server device connected to the magnetic tape drive 10 via a communication network (not shown), and the waveform equalization execution program 80 is stored in response to a request from the magnetic tape drive 10. The waveform equalization execution program 80 may be downloaded and installed in the equalizer 60.

Note that it is not necessary to store the entire waveform equalization execution program 80 in the storage unit of another computer or server device connected to the equalizer 60, or in the storage 74; may be stored in memory.

As hardware resources for executing the waveform equalization execution process described in the first embodiment, the following various processors can be used. Examples of the processor include a CPU, which is a general-purpose processor that functions as a hardware resource to execute waveform equalization execution processing by executing software, that is, a program. Examples of the processor include a dedicated electric circuit such as an FPGA, PLD, or ASIC, which is a processor having a circuit configuration specifically designed to execute a specific process. Each processor has a built-in or connected memory, and each processor uses the memory to execute waveform equalization execution processing.

The hardware resources that execute waveform equalization execution processing may be configured with one of these various types of processors, or may be configured with a combination of two or more processors of the same type or different types (for example, a combination of multiple FPGAs). , or a combination of a CPU and an FPGA). Furthermore, the hardware resource that executes the waveform equalization execution process may be one processor.

As an example of configuration using one processor, first, one processor is configured by a combination of one or more CPUs and software, and this processor functions as a hardware resource for executing waveform equalization execution processing. There is. Second, as typified by SoC, there is a form of using a processor that implements the functions of the entire system including a plurality of hardware resources that execute waveform equalization execution processing with one IC chip. In this way, the waveform equalization execution process is realized using one or more of the various processors described above as hardware resources.

Furthermore, as the hardware structure of these various processors, more specifically, an electric circuit that is a combination of circuit elements such as semiconductor elements can be used. Furthermore, the above-described waveform equalization execution process is just an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be changed within the scope of the main idea.

The descriptions and illustrations described above are detailed explanations of portions related to the technology of the present disclosure, and are merely examples of the technology of the present disclosure. For example, the above description regarding the configuration, function, operation, and effect is an example of the configuration, function, operation, and effect of the part related to the technology of the present disclosure. Therefore, unnecessary parts may be deleted, new elements may be added, or replacements may be made to the written and illustrated contents described above without departing from the gist of the technology of the present disclosure. Needless to say. In addition, in order to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the descriptions and illustrations shown above do not include parts that require particular explanation in order to enable implementation of the technology of the present disclosure. Explanations regarding common technical knowledge, etc. that do not apply are omitted.

In this specification, "A and/or B" is synonymous with "at least one of A and B." That is, "A and/or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, even when three or more items are expressed by connecting them with "and/or", the same concept as "A and/or B" is applied.

All documents, patent applications, and technical standards mentioned herein are incorporated herein by reference to the same extent as if each individual document, patent application, and technical standard was specifically and individually indicated to be incorporated by reference. Incorporated by reference into this book.

10 Magnetic tape drive 12 Magnetic tape cartridge 14 Conveying device 16 Reading head 16A Reading element 16B Holder 18 Control device 20 Sending motor 22 Take-up reel 24 Take-up motor 26 UI device 28 External I/F
30 Track area 32 Servo pattern 32A First diagonal line 32B Second diagonal line 36 Servo element pair 36A Servo element 36B Servo element 40 Movement mechanism 42 Motor 44 Controller 46 Amplifier 48 A/D converter 50 Signal processing device 50A Element-specific signal processing device 52 Amplifier 54 A/D converter 56 LPF
58 Phase synchronized circuit 60 Equalizer 62 Decoder 64 Computer 70 CPU
70A Operation stage delay storage unit 70A1 Storage element 70B Operation stage calculation unit 71 Internal memory 72 Memory 74 Storage 76 Bus 78 Input/output I/F
80 Waveform equalization execution program 82 Learned model 90 Computer 92 CPU
92A Learning stage delay storage unit 92A1 Storage element 92B Learning stage calculation unit 92C Error calculation unit 92D Variable adjustment unit 93 Internal memory 94 Memory 96 Storage 98 Input/output I/F
100 Bus 102 Test reproduction signal supply device 104 UI device 106 Learning execution program 107 Previous layer 107A Previous layer node 108 Neural network 108A Input layer 108A1 Input layer node 108B Middle layer 108B1 Middle layer node 108C Output layer 108C1 Output layer node 109 Later layer 109A Later layer node 110 Teacher data 112 Magnetic head 114 BOT area 116 Specific area 116A Specific pattern 118 Magnetic tape cartridge related information area 122 Cartridge memory 124 Non-contact reading/writing device 182 Learned model 200 Storage medium 282 Learned model 208 Neural network 308 Neural network A Arrow E1 End storage element E2 End storage element CR Cartridge reel GR Guide roller MT Magnetic tape SR, SR1, SR2 Servo reading element SRa First servo reading element SRb Second servo reading element SRc Third servo reading element

Claims (30)

A plurality of reading results obtained by reading the data from a magnetic tape on which data is recorded by a plurality of reading elements mounted on a reading head are digitized by a plurality of A/D converters. a receiver for receiving a plurality of reproduced signal sequences;
a plurality of equalizers that perform waveform equalization of the plurality of reproduced signal sequences received by the receiver,
The plurality of equalizers include a plurality of nonlinear filters that are trained to reduce distortion that nonlinearly occurs in the plurality of reproduction signal sequences depending on the environmental conditions in which the data is read from the magnetic tape. Perform the waveform equalization using
The plurality of nonlinear filters are optimized to have characteristics suitable for the plurality of reading elements based on the plurality of reading results,
The conditions include a read head individual difference condition resulting from individual differences in the read head, a magnetic tape individual difference condition resulting from individual differences in the magnetic tape, a speed condition regarding the speed at which the magnetic tape is running, and the Contains at least one of processing circuit individual difference conditions resulting from individual differences in processing circuits that affect waveform equalization
Signal processing device. The signal processing device according to claim 1, wherein the plurality of reading results are obtained by reading a specific pattern recorded as the data in a specific area of the magnetic tape by the plurality of reading elements. In parallel with the operation in which the specific pattern is recorded in the specific area by a plurality of recording elements arranged upstream of the plurality of reading elements in the running direction of the magnetic tape, the specific pattern is read by the plurality of reading elements. The signal processing device according to claim 2, wherein the signal processing device is read by an element. The speed conditions include conditions regarding the running speed of the magnetic tape when recording is performed on the magnetic tape.
The signal processing device according to any one of claims 1 to 3 . The signal processing device according to any one of claims 1 to 4, wherein the nonlinear filter is a filter having a neural network in which the learning has been performed. comprising a plurality of storage elements provided for each of the reading elements and storing the reproduced signal series in time series;
The neural network includes a front layer having a plurality of front layer nodes corresponding to the plurality of storage elements, and a rear layer,
Each of the plurality of storage elements outputs the input reproduction signal sequence to a corresponding one of the plurality of previous layer nodes,
Each of the plurality of front layer nodes outputs the reproduction signal sequence inputted from a corresponding one of the plurality of storage elements to the rear layer;
The latter layer is
converting a composite value obtained based on the sum of products of the reproduction signal sequence input from the plurality of front layer nodes and the rear layer coupling weight using an activation function;
outputting a subsequent layer value based on a converted value obtained by converting the composite value with the activation function;
The subsequent layer connection weight is determined by performing learning on the neural network to minimize the amount of deviation between the subsequent layer value and a predetermined target value as the learning.
The signal processing device according to claim 5 . The neural network has an input layer as the preceding layer, and an intermediate layer and an output layer as the subsequent layer,
The plurality of previous layer nodes are a plurality of input layer nodes,
The middle layer has a plurality of middle layer nodes,
Each of the plurality of input layer nodes outputs the reproduction signal sequence input from a corresponding one of the plurality of storage elements to the intermediate layer,
The plurality of intermediate layer nodes converts the intermediate layer value obtained as the composite value based on the product sum of the reproduced signal sequence inputted from the plurality of input layer nodes and the intermediate layer coupling weight using the activation function. to generate the converted value and output it to the output layer,
The output layer outputs an output layer value obtained as the subsequent layer value based on the conversion value input from the intermediate layer, the output layer connection weight, and the sum of products,
The intermediate layer connection weight and the output layer connection weight are determined by performing learning on the neural network to minimize the amount of deviation between the output layer value and a predetermined target value.
The signal processing device according to claim 6 . The intermediate layer value is a value based on the sum of products of the reproduced signal sequence and the intermediate layer coupling load and a first variable,
The first variable is determined by the learning being performed on the neural network.
The signal processing device according to claim 7 . The neural network consists of two layers: the preceding layer and the subsequent layer.
The signal processing device according to claim 6 . The latter layer value is a value based on the sum of products of the converted value and the latter layer connection load and a second variable,
The second variable is determined by the learning being performed on the neural network.
The signal processing device according to any one of claims 6 to 9 . The plurality of storage elements are a plurality of delay elements into which the reproduction signal sequence is input after being delayed by a predetermined time,
The latter layer value is a value related to the reproduction signal sequence inputted first among the plurality of reproduction signal sequences stored in the plurality of delay elements.
The signal processing device according to any one of claims 6 to 10 . The target value is an ideal reproduction signal sequence related to known data recorded on the learning magnetic tape in a predetermined recording pattern along the longitudinal direction of the learning magnetic tape, and an ideal reproduction signal sequence derived by computer simulation. The training data is predetermined based on at least one of the reproduced signal sequences.
The signal processing device according to any one of claims 6 to 11 . A magnetic tape cartridge comprising a magnetic tape,
A magnetic tape cartridge. Parameters related to the plurality of nonlinear filters used by the signal processing device according to any one of claims 1 to 12 are recorded on the magnetic tape. A magnetic tape cartridge comprising a non-contact storage medium, the cartridge comprising:
A magnetic tape cartridge, wherein the non-contact storage medium stores parameters regarding the plurality of nonlinear filters used by the signal processing device according to any one of claims 1 to 12 . a read head equipped with a plurality of read elements for reading data from a magnetic tape on which the data is recorded;
a receiver that receives a plurality of reproduction signal sequences obtained by digitizing a plurality of reading results obtained by reading the data by the plurality of reading elements by a plurality of A/D converters;
a plurality of equalizers that perform waveform equalization of the plurality of reproduced signal sequences received by the receiver,
The plurality of equalizers include a plurality of nonlinear filters that are trained to reduce distortion that nonlinearly occurs in the plurality of reproduction signal sequences depending on the environmental conditions in which the data is read from the magnetic tape. Perform the waveform equalization using
The plurality of nonlinear filters are optimized to have characteristics suitable for the plurality of reading elements based on the plurality of reading results,
The conditions include a read head individual difference condition resulting from individual differences in the read head, a magnetic tape individual difference condition resulting from individual differences in the magnetic tape, a speed condition regarding the speed at which the magnetic tape is running, and the Contains at least one of processing circuit individual difference conditions resulting from individual differences in processing circuits that affect waveform equalization
Magnetic tape reader. The plurality of reading results are obtained by reading a specific pattern recorded as the data in a specific area of the magnetic tape by the plurality of reading elements.
The magnetic tape reader according to claim 15 . In parallel with the operation in which the specific pattern is recorded in the specific area by a plurality of recording elements arranged upstream of the plurality of reading elements in the running direction of the magnetic tape, the specific pattern is read by the plurality of reading elements. read by element
A magnetic tape reader according to claim 16 . The speed conditions include conditions regarding the running speed of the magnetic tape when recording is performed on the magnetic tape.
A magnetic tape reader according to any one of claims 15 to 17 . The nonlinear filter is a filter having a neural network in which the learning has been performed.
A magnetic tape reader according to any one of claims 15 to 18 . comprising a plurality of storage elements provided for each of the reading elements and storing the reproduced signal series in time series;
The neural network includes a front layer having a plurality of front layer nodes corresponding to the plurality of storage elements, and a rear layer,
Each of the plurality of storage elements outputs the input reproduction signal sequence to a corresponding one of the plurality of previous layer nodes,
Each of the plurality of front layer nodes outputs the reproduction signal sequence inputted from a corresponding one of the plurality of storage elements to the rear layer;
The latter layer is
converting a composite value obtained based on the sum of products of the reproduction signal sequence input from the plurality of front layer nodes and the rear layer coupling weight using an activation function;
outputting a subsequent layer value based on a converted value obtained by converting the composite value with the activation function;
The subsequent layer connection weight is determined by performing learning on the neural network to minimize the amount of deviation between the subsequent layer value and a predetermined target value as the learning.
A magnetic tape reader according to claim 19 . The neural network has an input layer as the preceding layer, and an intermediate layer and an output layer as the subsequent layer,
The plurality of previous layer nodes are a plurality of input layer nodes,
The middle layer has a plurality of middle layer nodes,
Each of the plurality of input layer nodes outputs the reproduction signal sequence input from a corresponding one of the plurality of storage elements to the intermediate layer,
The plurality of intermediate layer nodes converts the intermediate layer value obtained as the composite value based on the product sum of the reproduced signal sequence inputted from the plurality of input layer nodes and the intermediate layer coupling weight using the activation function. to generate the converted value and output it to the output layer,
The output layer outputs an output layer value obtained as the subsequent layer value based on the conversion value input from the intermediate layer, the output layer connection weight, and the sum of products,
The intermediate layer connection weight and the output layer connection weight are determined by performing learning on the neural network to minimize the amount of deviation between the output layer value and a predetermined target value.
A magnetic tape reader according to claim 20 . The intermediate layer value is a value based on the sum of products of the reproduced signal sequence and the intermediate layer coupling load and a first variable,
The first variable is determined by the learning being performed on the neural network.
A magnetic tape reader according to claim 21 . The neural network consists of two layers: the preceding layer and the subsequent layer.
A magnetic tape reader according to claim 20 . The latter layer value is a value based on the sum of products of the converted value and the latter layer connection load and a second variable,
The second variable is determined by the learning being performed on the neural network.
A magnetic tape reader according to any one of claims 20 to 23 . The plurality of storage elements are a plurality of delay elements into which the reproduction signal sequence is input after being delayed by a predetermined time,
The latter layer value is a value related to the reproduction signal sequence inputted first among the plurality of reproduction signal sequences stored in the plurality of delay elements.
A magnetic tape reader according to any one of claims 20 to 24 . The target value is an ideal reproduction signal sequence related to known data recorded on the learning magnetic tape in a predetermined recording pattern along the longitudinal direction of the learning magnetic tape, and an ideal reproduction signal sequence derived by computer simulation. The training data is predetermined based on at least one of the reproduced signal sequences.
A magnetic tape reader according to any one of claims 20 to 25 . A plurality of reading results obtained by reading the data from a magnetic tape on which data is recorded by a plurality of reading elements mounted on a reading head are digitized by a plurality of A/D converters. A processing method for a signal processing device, comprising: a receiver that receives a plurality of reproduced signal sequences received by the receiver; and a plurality of equalizers that perform waveform equalization of the plurality of reproduced signal sequences received by the receiver. There it is,
a plurality of nonlinear filters that are trained by the plurality of equalizers to reduce distortion nonlinearly generated in the plurality of reproduction signal sequences according to environmental conditions in which the data is read from the magnetic tape; performing the waveform equalization using
The plurality of nonlinear filters are optimized to have characteristics suitable for the plurality of reading elements based on the plurality of reading results,
The conditions include a read head individual difference condition resulting from individual differences in the read head, a magnetic tape individual difference condition resulting from individual differences in the magnetic tape, a speed condition regarding the speed at which the magnetic tape is running, and the Contains at least one of processing circuit individual difference conditions resulting from individual differences in processing circuits that affect waveform equalization
Processing method of signal processing device. a read head equipped with a plurality of read elements for reading data from a magnetic tape on which the data is recorded;
a receiver that receives a plurality of reproduction signal sequences obtained by digitizing a plurality of reading results obtained by reading the data by the plurality of reading elements by a plurality of A/D converters;
A method of operating a magnetic tape reading device, comprising: a plurality of equalizers that equalize the waveforms of the plurality of reproduction signal sequences received by the receiver, the method comprising:
a plurality of nonlinear filters that are trained by the plurality of equalizers to reduce distortion nonlinearly generated in the plurality of reproduction signal sequences according to environmental conditions in which the data is read from the magnetic tape; performing the waveform equalization using
The plurality of nonlinear filters are optimized to have characteristics suitable for the plurality of reading elements based on the plurality of reading results,
The conditions include a read head individual difference condition resulting from individual differences in the read head, a magnetic tape individual difference condition resulting from individual differences in the magnetic tape, a speed condition regarding the speed at which the magnetic tape is running, and the Contains at least one of processing circuit individual difference conditions resulting from individual differences in processing circuits that affect waveform equalization
How magnetic tape readers work. A program for causing a computer to execute processing applied to a signal processing device,
The signal processing device includes:
A plurality of reading results obtained by reading the data from a magnetic tape on which data is recorded by a plurality of reading elements mounted on a reading head are digitized by a plurality of A/D converters. a receiver for receiving a plurality of reproduced signal sequences;
a plurality of equalizers that perform waveform equalization of the plurality of reproduced signal sequences received by the receiver,
The processing is
The waveform equalization is performed using a plurality of nonlinear filters that are trained to reduce distortion nonlinearly generated in the plurality of reproduction signal sequences according to environmental conditions in which the data is read from the magnetic tape. including doing;
The plurality of nonlinear filters are optimized to have characteristics suitable for the plurality of reading elements based on the plurality of reading results,
The conditions include a read head individual difference condition resulting from individual differences in the read head, a magnetic tape individual difference condition resulting from individual differences in the magnetic tape, a speed condition regarding the speed at which the magnetic tape is running, and the Contains at least one of processing circuit individual difference conditions resulting from individual differences in processing circuits that affect waveform equalization
program. A program for causing a computer to perform processing applied to a magnetic tape reader,
The magnetic tape reader includes:
a read head equipped with a plurality of read elements for reading data from a magnetic tape on which the data is recorded;
a receiver that receives a plurality of reproduction signal sequences obtained by digitizing a plurality of reading results obtained by reading the data by the plurality of reading elements by a plurality of A/D converters; a plurality of equalizers that perform waveform equalization of the plurality of reproduced signal sequences received by the receiver,
The processing is
Performing the waveform equalization using a plurality of nonlinear filters trained to reduce distortion that nonlinearly occurs in the reproduced signal sequence according to environmental conditions in which the data is read from the magnetic tape. including;
The plurality of nonlinear filters are optimized to have characteristics suitable for the plurality of reading elements based on the plurality of reading results,
The conditions include a read head individual difference condition resulting from individual differences in the read head, a magnetic tape individual difference condition resulting from individual differences in the magnetic tape, a speed condition regarding the speed at which the magnetic tape is running, and the Contains at least one of processing circuit individual difference conditions resulting from individual differences in processing circuits that affect waveform equalization
program.