Controllable bipolar random signal generation method and device, medium and equipment

Abstract

The invention discloses an adjustable bipolar random signal generation method and device, a medium and equipment. The method comprises the following steps: sending a periodic pulse excitation signal; exciting novel memories connected in parallel based on the periodic pulse excitation signal to generate a bipolar random electric signal; and performing digital processing on the bipolar random electric signal to obtain a bipolar random bit stream. According to the invention, the proportion of different states of the bipolar random signal can be adjusted, so that different application requirements are met.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of integrated circuits, and in particular relates to a method, device, medium and equipment for generating an adjustable bipolar random signal. Background Art

[0002] Existing computing paradigms such as probabilistic computing and neuromorphic computing have broken through the limitations of traditional computing models and can play an important role in various fields such as artificial intelligence, big data analysis, pattern recognition and optimization problems. Among these new computing paradigms, bipolar coding has attracted much attention due to its high information capacity. Compared with unipolar coding, the representation range of bipolar coding extends from [0, 1] to [-1, 1], and can be applied to more computing scenarios, such as implementing scaled subtraction and random tanh functions. In spiking neural networks, bipolar coding can transmit excitation and inhibition signals at the same time, significantly improving the efficiency of neural networks.

[0003] However, the current generation of bipolar random electrical signals mainly relies on special devices, such as four-terminal spin-orbit torque magnetic tunnel junctions with heavy metal layers and carbon nanotube field-effect transistors, which are incompatible with industrial standard CMOS or magnetic tunnel junction manufacturing processes. Although the existing technology uses two standard MTJs and combines magnetic field and electrical bias drive, this method still faces challenges at the integrated circuit level. Therefore, it is difficult for the existing technology to generate stable and controllable bipolar random signals efficiently and with low power consumption, which limits its application in new computing paradigms and pulse neural networks. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present disclosure is to provide a method, device, medium and equipment for generating an adjustable bipolar random signal. The present disclosure can adjust the proportion of different states of the bipolar random signal to meet different application requirements.

[0005] To achieve the above objectives, the present disclosure provides the following technical solutions: A method for generating an adjustable bipolar random signal comprises: sending a periodic pulse excitation signal; stimulating a novel memory connected in parallel based on the periodic pulse excitation signal to generate a bipolar random electrical signal; and digitally processing the bipolar random electrical signal to obtain a bipolar random bit stream.

[0006] Optionally, the periodic pulse excitation signal is a three-stage pulse, including a probability setting pulse, a probability reset pulse and a read pulse in one cycle; wherein the probability setting pulse is used to make the new memory flip with a certain probability; the probability reset pulse is used to make the new memory return to an initial state with a certain probability; and the read pulse is used to read the current state of the new memory.

[0007] Optionally, the periodic pulse excitation signal includes a periodic voltage pulse signal or a periodic current pulse signal.

[0008] Optionally, the new type of memory includes but is not limited to any of the following: magnetic random access memory, phase change memory and ferroelectric memory.

[0009] In addition, the present disclosure also provides an adjustable bipolar random signal generating device, which includes: a pulse excitation unit for sending a periodic pulse excitation signal; a random signal generating unit for stimulating a new type of memory connected in parallel based on the periodic pulse excitation signal to generate a bipolar random electrical signal; and a random bit stream generating unit for digitally processing the bipolar random electrical signal to obtain a bipolar random bit stream.

[0010] In addition, the present disclosure also provides a controllable bipolar random signal generation system, which includes the device as described above.

[0011] In addition, the present disclosure also provides an adjustable bipolar random signal generating device, and the device applies the method described in any of the preceding items.

[0012] In addition, the present disclosure also provides a computer storage medium, wherein the storage medium includes computer instructions, and when the computer is executed on the computer, the computer executes the method as described in any of the above items.

[0013] In addition, the present disclosure also provides an electronic device, the electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method described in any of the preceding items is implemented.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. Use industrial-grade new memory technology: By adopting advanced memory technology, a parallel circuit is designed using pure electrical operations to generate physical bipolar random signals. Compared with the prior art, the technology provided by the present disclosure is compatible with standard memory manufacturing processes and memory write / read circuits, does not require additional process steps or circuits, and saves costs.

[0015] 2. Excellent randomness and security: The generated bipolar signal has excellent randomness and security, which helps to achieve bipolar encoding. The signal has the unique ability to transmit excitatory and inhibitory random signals simultaneously.

[0016] 3. Controllable average residence time: By applying pulses, the average residence time of the two states can be controlled. The ratio of the three states when the bipolar random signal generated by the prior art is fixed and cannot be adjusted according to specific needs. The method disclosed in the present invention provides flexibility and adjustability to meet different application requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of a flow chart of a method for generating an adjustable bipolar random signal provided by an embodiment of the present disclosure; Figure 2 A schematic diagram of a generated bipolar random signal, a bipolar random bit stream and a binary random bit stream provided by another embodiment of the present disclosure; Figure 3 is a schematic diagram of randomness detection results of a bipolar random signal provided by another embodiment of the present disclosure; Figure 4 is a schematic diagram of the dependence of the time parameter and pulse amplitude of a bipolar random signal provided by another embodiment of the present disclosure; Figure 5 A schematic diagram of the structure of an adjustable bipolar random signal generating device provided in another embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] The following will refer to the attached Figures 1 to 5 Specific embodiments of the present disclosure are described in detail. Although specific embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0019] It should be noted that certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that technicians may use different nouns to refer to the same component. This specification and claims do not use the difference in nouns as a way to distinguish components, but use the difference in the functions of the components as the criterion for distinction. As mentioned throughout the specification and claims, "including" or "comprising" is an open term, so it should be interpreted as "including but not limited to". The subsequent description of the specification is a preferred embodiment of the present disclosure, but the description is based on the general principles of the specification and is not used to limit the scope of the present disclosure. The scope of protection of the present disclosure shall be determined by the attached claims.

[0020] To facilitate understanding of the embodiments of the present disclosure, further explanation will be given below using specific embodiments as examples in conjunction with the accompanying drawings, and each of the accompanying drawings does not constitute a limitation on the embodiments of the present disclosure.

[0021] Figure 1 An exemplary embodiment of the present disclosure provides a method for generating a controllable bipolar random signal. Figure 1 As shown, the method includes: S100: sending a periodic pulse excitation signal; S200: Stimulating the new memory connected in parallel based on the periodic pulse stimulation signal to generate Figure 2 The bipolar random electrical signal shown; S300: Digitally process the bipolar random electrical signal to obtain Figure 2 The bipolar random bit stream is shown.

[0022] In this embodiment, the novel memory includes but is not limited to: magnetic random access memory, phase change memory and ferroelectric memory. The novel memory includes a high resistance state (AP) and a low resistance state (P), and the novel memory can change its internal state under the action of an external pulse excitation signal (such as voltage or current pulse excitation), and the high resistance state (AP) and the low resistance state (P) are reversible, that is, the novel memory can be converted between these two different states. When subjected to a periodic pulse excitation signal, the states of the two parallel novel memories will change randomly, and these state changes can form different electrical signal levels, thereby generating a bipolar random electrical signal. For example, if both of the two parallel-connected novel memories are in the AP state, that is, when the combination is expressed as [AP, AP], a low-level signal can be generated; if both of the two parallel-connected novel memories are out of the P state, that is, when the combination is expressed as [P, P], a high-level signal can be generated; if one of the two parallel-connected novel memories is in the AP state and the other is out of the P state, that is, when the state combination is expressed as [AP, P] or [P, AP], a medium-level signal can be generated. Therefore, in this embodiment, based on the physical characteristics of the novel memory and the state change under specific pulse stimulation, a bipolar random electrical signal with three distinguishable states can be generated.

[0023] It should be noted that the reason why the novel memory is connected in parallel in this embodiment, rather than in series or in other ways, is that: 1. By connecting multiple new memories in parallel, multiple electrical signal levels (such as high, medium, and low levels) can be generated according to the state combination of different memories. This combination makes the output of the signal more flexible and the proportion of different states can be adjusted as needed.

[0024] 2. Parallel connection is usually more efficient than series connection in signal transmission and processing, because each component in the parallel circuit can respond to the stimulus signal at the same time without waiting for the previous component to respond, which helps to speed up the signal generation.

[0025] 3. Parallel connection is relatively simple in design and does not require additional logic elements or complex control circuits to achieve signal generation and processing, which helps reduce circuit complexity and cost.

[0026] In another exemplary embodiment, in step S100, the periodic pulse excitation signal is a three-stage pulse, including a probability setting pulse, a probability reset pulse and a read pulse in one cycle; wherein the probability setting pulse is used to make the new memory flip with a certain probability; the probability reset pulse is used to make the new memory return to an initial state with a certain probability; and the read pulse is used to read the current state of the new memory.

[0027] In this embodiment, the cycle refers to the entire process from the start of the probability setting pulse to the completion of the read pulse. During this cycle, the state of the novel memory may or may not be flipped, depending on the parameters of the probability setting pulse and the probability reset pulse, such as the amplitude and width of the pulse.

[0028] Exemplarily, the pulse parameters within one cycle are as follows: Probability setting pulse: 2V, 10ns Probability reset pulse: 1V, 5ns Read pulse: 0.5V, 1ns Example cycle: At time t=0ns, the probability setting pulse starts and lasts for 10ns.

[0029] At time t=10ns, the probability reset pulse starts and lasts for 5ns.

[0030] At time t=15ns, the read pulse starts and lasts for 1ns.

[0031] Therefore, a complete cycle starts at t = 0ns and ends at t = 16ns, a total of 16ns. In each such cycle, the state of the new memory changes and its state is read by a read pulse.

[0032] In the prior art, since the generation of bipolar random signals mainly depends on the physical properties of the four-terminal spin-orbit torque magnetic tunnel junction (SOT-MTJ) with a heavy metal layer and the carbon nanotube field effect transistor (CNTFET), the ratio of the three states (high level, medium level, low level) of the generated bipolar random signal is fixed. For example, the SOT-MTJ uses the spin-orbit torque (SOT) generated when the spin-polarized current passes through the heavy metal layer to switch the magnetization direction of the magnetic material. However, the switching probability and stability of the magnetization state are affected by many factors, including current size, pulse width, temperature, etc. Since these factors are difficult to control accurately and there are physical limitations, the ratio of the magnetization state is relatively fixed. CNTFET uses carbon nanotubes as a conductive channel, and its conductive performance is regulated by the gate voltage. However, when used to generate random signals, the switching state of CNTFET (i.e., whether it is conductive or not) is affected by factors such as manufacturing process, material defects, temperature fluctuations, etc., resulting in a fixed ratio of output states and difficult to regulate.

[0033] In this embodiment, the state change of the new memory can be controlled by adjusting the parameters of the periodic pulse excitation signal (such as pulse width, pulse amplitude, etc.), thereby realizing flexible regulation of the proportion of the three states of the bipolar random electrical signal. Specifically, by changing the amplitude and width of the probability setting pulse and the probability reset pulse, the probability of the new memory flipping and returning to the initial state can be affected, thereby adjusting the proportion of each state in the output signal. Among them, the amplitude and width of the probability setting pulse determine the probability of the new memory flipping (i.e., changing from one state to another). For example, in a magnetic random access memory, a larger amplitude or a longer pulse width may increase the probability of magnetic domain flipping, making it easier for the state of the memory to change from the P state to the AP state, or from the AP state to the P state. In addition, the probability reset pulse is used to reset the new memory back to its initial state with a certain probability. By adjusting the amplitude and width of the reset pulse, the speed and probability of the memory returning to the initial state can be controlled, wherein a larger amplitude or a longer pulse width can increase the probability of the memory resetting back to the initial state. For example, assuming that it is desired to increase the proportion of the high-level signal generated by the memory, it can be achieved by reducing the amplitude or width of the probability reset pulse. Doing so will reduce the probability of the memory being reset back to the initial state (which may correspond to a low state), thereby relatively increasing the proportion of the high state.

[0034] Furthermore, this embodiment uses specific data to illustrate the above control process in a visual way. Assume that there is a bipolar random signal generation system, which includes two new memories connected in parallel. The two memories can be magnetic random access memory (MRAM), phase change memory (PCM) or ferroelectric memory (FRAM), etc. Each memory has two states: high resistance state (AP) and low resistance state (P). In the initial state, it is assumed that the two new memories connected in parallel are both in the AP state. If you want to increase the probability of flipping from the AP state to the P state, you can do so by increasing the amplitude or width of the probability setting pulse. For example, if the original pulse amplitude is 2V and the duration is 10ns, its amplitude can be increased to 3V or the duration can be increased to 15ns. The increased pulse is more likely to trigger the flip of the memory state, for example, increasing the probability of going from the AP state to the P state from the original 50% to 70%. In addition, if the goal is to increase the proportion of high-level signals generated by the memory, the amplitude or width of the probability reset pulse can be appropriately reduced. For example, it is assumed that the amplitude of the original probability reset pulse is 1V and the duration is 5ns. In order to reduce the probability of the memory being reset from the P state to the AP state, the pulse amplitude can be reduced to 0.5V or the duration can be reduced to 3ns. The reduced probability reset pulse reduces the probability of the memory state being reset, for example, from the original 60% to 30%, which can increase the probability of the memory generating a high-level signal.

[0035] It should be noted that the present disclosure has practical application value by adjusting the ratio of the three states (high level, medium level and low level) of the random signal. For example, in the application scenario of the pulse neural network, by adjusting the state ratio of the bipolar random signal, more complex neural behaviors can be simulated, and the expression ability and learning ability of the network can be enhanced, thereby helping to solve complex pattern recognition, image processing and natural language processing problems. In addition, random numbers are also widely used in encryption, simulation and emulation. The bipolar random signal generated by this method can optimize the statistical characteristics of random numbers by adjusting the state ratio, such as distribution uniformity, independence and unpredictability, thereby improving the security and efficiency of the random number generator. Further, when simulating physical systems (such as quantum systems, biological systems, etc.), bipolar random signals can represent the fluctuations and changes in the internal state of the system. By adjusting the state ratio, the behavioral characteristics of the system under different conditions can be simulated, thereby providing strong support for physical science research and engineering design.

[0036] In another exemplary embodiment, in step S100, the periodic pulse excitation signal includes a periodic voltage pulse signal or a periodic current pulse signal.

[0037] In this embodiment, the periodic voltage pulse signal is used to stimulate the new memory by periodically applying voltage. The amplitude and width of the voltage pulse can be precisely controlled to affect the state change probability of the new memory. For example, a larger voltage pulse amplitude or a longer pulse width may cause the new memory to flip from one state to another more easily.

[0038] The periodic current pulse signal stimulates the novel memory by periodically applying current. Similar to the voltage pulse, the amplitude and width of the current pulse can also be adjusted to control the state change of the novel memory.

[0039] In another exemplary embodiment, in step S300, the bipolar random electrical signal is digitally processed, including: The bipolar random electrical signal is compared with a preset threshold value, and a bipolar random bit stream is generated according to the comparison result.

[0040] Threshold A and threshold B may be preset (for example, threshold A may be set to +1V, and threshold B may be set to -1V), wherein threshold A is greater than threshold B. If the signal level of the bipolar random electrical signal is greater than threshold A, it indicates that the signal is at a high level, and 1 is output; if the signal level of the bipolar random electrical signal is less than threshold B, it indicates that the signal is at a low level, and -1 is output; if the signal level of the bipolar random electrical signal is greater than threshold B and less than threshold A, it indicates that the signal is at a medium level, and 0 is output.

[0041] For example, suppose there is a set of values ​​of a bipolar random electrical signal, representing the level state of the signal at different times. The following is a list of example signal level values: 1.5V -0.5V 0.5V -1.5V 0.9V -0.9V 1.1V -1.1V 0.0V Based on the threshold settings above, these signal levels can be converted into a bipolar random bit stream: 1.5V > +1V → High → 1 -0.5V > -1V and -0.5V < +1V → mid-level → 0 0.5V > -1V and 0.5V < +1V → mid-level → 0 -1.5V < -1V → Low level → -1 0.9V > +1V → High level → 1 -0.9V < -1V → Low level → -1 1.1V > +1V → High level → 1 -1.1V < -1V → low level → -1 0.0V > -1V and 0.0V < +1V → mid level → 0 Therefore, based on these signal levels, the bipolar random bit stream is: 1, 0, 0, -1, 1, -1,1, -1, 0.

[0042] This embodiment sets a threshold value so that the bipolar random electrical signal can be effectively converted into a bipolar random bit stream, so that the bipolar random signal can not only achieve random changes in high and low levels at the physical level, but also be processed and transmitted in a clear coding form at the digital level, which not only improves the versatility and flexibility of the signal, but also enables the signal to better adapt to the needs of various computing and application scenarios.

[0043] It should be noted that the above is only exemplary, and the specific values ​​of the thresholds A and B can be adjusted according to actual needs to achieve fine control of signal characteristics.

[0044] In another exemplary embodiment, the bipolar random electrical signal may be selected to generate a binary random bit stream through a Schmitt trigger for randomness verification.

[0045] In this embodiment, the Schmitt trigger has two threshold voltages, including a positive threshold voltage and a negative threshold voltage. When the input signal rises from a low level to above the positive threshold voltage, the trigger outputs a high level; when the input signal drops from a high level to below the negative threshold voltage, the trigger outputs a low level.

[0046] In the application of generating bipolar random signals disclosed in the present disclosure, since the bipolar random electrical signal has three distinguishable states (high level, medium level and low level), direct conversion into a binary random bit stream is not direct enough or there is ambiguity. Therefore, through a Schmitt trigger, the high level and the medium level in the bipolar random electrical signal can be uniformly converted into a high level (or 1), and the low level can be converted into a low level (or 0), thereby generating a clear binary random bit stream.

[0047] Furthermore, after obtaining a binary random bit stream, a variety of statistical tests can be used to verify its randomness. These tests are designed to detect whether there are predictable patterns, deviations or correlations in the bit stream, so as to determine whether it is random enough for applications that require high randomness, such as encryption and simulation.

[0048] Figure 3 is a randomness detection result of a bipolar random signal provided by another embodiment of the present disclosure. Figure 3 As shown, in and A bipolar random signal containing 20,000 bits is generated. In this embodiment, 12 sub-tests in the standardized statistical test suite NIST SP 800-22 of the National Institute of Standards and Technology of the United States are applied to the random signal to detect whether 0 and 1 in the signal are evenly distributed, whether the oscillation speed is appropriate, whether it has periodicity, etc. A P-value greater than 0.001 indicates that the sub-test has been passed. From the results in the table, it can be seen that the generated random signal has passed the 12 NIST sub-tests, which confirms that the bipolar random signal has the characteristics of unpredictability, non-repeating patterns, and no correlation between consecutive bits, and meets the encryption quality requirements.

[0049] In practical applications, after passing the randomness test, the generated binary random bit stream can be safely used in various application scenarios that require randomness, such as key generation for encryption algorithms, random number generators, random event triggering in simulation experiments, etc.

[0050] In another exemplary embodiment, the bipolar random signal is regulated by adjusting the pulse width or pulse amplitude of the periodic voltage pulse signal or the periodic current pulse signal.

[0051] In this embodiment, the average dwell time is a time parameter of the bipolar random signal, specifically the average time that each state stays when fluctuating between the three states. It can be defined by the following formula:

[0052] in, Indicates the dwell time, i.e. the bipolar random signal remains in one state i Until the next transfer time, Indicates status i The number of.

[0053] For example, the amplitude of the probability setting pulse can be increased to make the new memory flip, thereby increasing the frequency of the signal in the high level or medium level state and reducing the residence time in the low level state. Increasing the amplitude of the probability reset pulse will make it easier for the new memory to return to the initial state, increase the residence time in the low level state, and reduce the residence time in the high level or medium level state.

[0054] Similarly, adjusting the pulse width can also affect the state change probability of the new memory, thereby affecting the distribution and average dwell time of the bipolar random signal.

[0055] Figure 4 is the dependency relationship between the time parameter and pulse amplitude of a bipolar random signal provided by another embodiment of the present disclosure. Figure 4 As shown, Affected by positive and negative pulses, it decreases with the increase of probability setting pulse amplitude and increases with the increase of probability reset pulse amplitude; It is independent of the probability setting pulse amplitude and only depends on the probability reset pulse, and decreases with the increase of the probability reset pulse amplitude. Therefore, it can be achieved in two steps and Target combination: First, reset the pulse amplitude setting by adjusting the probability , and then adjust the pulse amplitude by adjusting the probability .

[0056] In another exemplary embodiment, Figure 5 As shown, the present disclosure also provides an adjustable bipolar random signal generating device, wherein the device includes: a pulse excitation unit, used to send a periodic pulse excitation signal; a bipolar signal generating unit, used to stimulate a new type of memory connected in parallel based on the periodic pulse excitation signal to generate a bipolar random electrical signal; a random bit stream generating unit, used to digitally process the bipolar random electrical signal to obtain a bipolar random bit stream.

[0057] Optionally, the adjustable bipolar random signal generating device further comprises a binary random bit stream generating unit, which is used to convert the bipolar random electrical signal into a binary random bit stream. Specifically, the binary random bit stream generating unit adopts a Schmitt trigger.

[0058] In another exemplary embodiment, the present disclosure further provides a controllable bipolar random signal generation system, wherein the system comprises the device as described above.

[0059] In another exemplary embodiment, the present disclosure further provides an adjustable bipolar random signal generating device, wherein the device applies the method as described in any of the preceding items.

[0060] In another exemplary embodiment, the present disclosure further provides a computer storage medium, wherein the storage medium includes computer instructions, and when the computer is executed on the computer, the computer executes the method as described in any of the preceding items.

[0061] In another exemplary embodiment, the present disclosure further provides an electronic device, wherein the electronic device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any of the preceding items when executing the program.

[0062] The above embodiments are only for illustrating the technical concept and features of the present disclosure, and their purpose is to enable people familiar with the technology to understand the content of the present disclosure and implement it accordingly, and they cannot be used to limit the protection scope of the present disclosure. Any equivalent changes or modifications made according to the spirit of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A method for generating a controllable bipolar random signal, comprising: Sending a periodic pulse excitation signal; Based on the periodic pulse excitation signal, the novel memories connected in parallel are stimulated to generate a bipolar random electrical signal; The bipolar random electrical signal is digitally processed to obtain a bipolar random bit stream.

2. The method according to claim 1, wherein: Preferably, the periodic pulse excitation signal includes a periodic voltage pulse signal or a periodic current pulse signal.

3. The method according to claim 1, wherein: The novel memory includes but is not limited to any of the following: magnetic random access memory, phase change memory and ferroelectric memory.

4. A controllable bipolar random signal generating device, wherein: The device comprises: A pulse excitation unit, used for sending a periodic pulse excitation signal; A random signal generating unit, used for stimulating the novel memories connected in parallel based on the periodic pulse excitation signal to generate a bipolar random electrical signal; The random bit stream generating unit is used for digitally processing the bipolar random electrical signal to obtain a bipolar random bit stream.

5. A controllable bipolar random signal generation system, wherein: The system comprises the apparatus of claim 4.

6. A controllable bipolar random signal generating device, wherein: The device applies the method according to any one of claims 1 to 3.

7. A computer storage medium, wherein: The storage medium includes computer instructions, which, when executed on a computer, enable the computer to execute the method according to any one of claims 1 to 3.

8. An electronic device, wherein: The electronic device comprises: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 3 is implemented.

Images