High-voltage linear LED driving circuit and electronic equipment

Abstract

The invention provides a high-voltage linear LED drive circuit and electronic equipment. The high-voltage linear LED drive circuit is composed of a clamping diode D1, a current reference unit, a voltage reference unit, an operational amplifier AMP, a low-voltage power tube MN1, a high-voltage power tube MN2 and the like. The cathode of the D1 is connected with an external power supply VIN and the anode of the LED, and is connected with the current reference unit, the AMP voltage driving end, the cathode of the D2 and the grid electrode of the MN2; the output end of the AMP controls the grid electrode of the MN1, the drain electrode of the MN1 is connected with the anode of the D2 and the source electrode of the MN2, and the source electrode of the MN1 is grounded through the sampling resistor R1; the drain electrode of the MN2 is connected with the cathode of the LED; the threshold voltage Vth1 of the MN1 is lower than the Vth2 of the MN2, and the gate parasitic capacitance of the MN1 is smaller. A constant current loop is formed based on the AMP, the MN1 and the R1, the conduction state of the MN2 is indirectly controlled by adjusting the MN1, delay caused by direct control of a grid electrode of the MN2 is avoided, overvoltage clamping protection is achieved through cooperation of the D2 and the D1, and meanwhile LED dimming can be intelligently controlled according to audio frequency.

Description

Technical Field

[0001] This application relates to the field of electronic power technology, and in particular, to a high-voltage linear LED driving circuit and an electronic device. Background Art

[0002] Currently, in the lighting system for film and television atmosphere, ordinary linear LED driving chips are difficult to meet the requirements of LED dimming, and there are the following two problems: 1. The dimming linearity is insufficient. When using the same PWM signal to drive multiple groups of LEDs, the response time deviation between chips is relatively large, resulting in a large current deviation between the lamp groups driven by different chips at low brightness, and there will be a large color difference; 2. The response speed corresponding to the film and television atmosphere is relatively slow, and it is difficult to perform real-time dimming on the LEDs according to the film and television atmosphere. Summary of the Invention

[0003] This application provides a high-voltage linear LED driving circuit and an electronic device. When adjusting low-brightness LEDs, by improving the response speed of the LEDs corresponding to the film and television atmosphere and simultaneously improving the dimming linearity, the user experience is improved.

[0004] In a first aspect, an embodiment of this application provides a high-voltage linear LED driving circuit. The high-voltage linear LED driving circuit includes: a clamping diode D1, a current reference unit, a voltage reference unit, a voltage reference driving stage unit, an operational amplifier AMP, a low-voltage power transistor MN1, a high-voltage power transistor MN2, a protection diode D2, and a sampling resistor R1; The cathode of the clamping diode D1 is used to be respectively connected to an external power supply V IN , the anode of an external LED. The cathode of the clamping diode D1 is also respectively connected to the input end of the current reference unit, the voltage driving end of the operational amplifier AMP, the cathode of the protection diode D2, and the Gate end of the high-voltage power transistor MN2; the anode of the clamping diode D1 is grounded; the output end of the current reference unit is connected to the bias current input end of the operational amplifier AMP; the output end of the voltage reference unit is connected to the input end of the voltage reference driving stage unit; the output end of the voltage reference driving stage unit is connected to the positive electrode of the operational amplifier AMP to provide a reference voltage V REF to the operational amplifier AMP; the output end of the operational amplifier AMP is connected to the Gate end of the low-voltage power transistor MN1; the Drain end of the low-voltage power transistor MN1 is respectively connected to the anode of the protection diode D2 and the Source end of the high-voltage power transistor MN2, and the Source end of the low-voltage power transistor MN1 is grounded through the sampling resistor R1; the Drain end of the high-voltage power transistor MN2 is used to be connected to the cathode of the external LED; Among them, the first threshold voltage V of the Gate terminal of the low-voltage power transistor MN1 th1 is less than the second threshold voltage V of the Gate terminal of the high-voltage power transistor MN2 th2 , the parasitic capacitance of the Gate terminal of the low-voltage power transistor MN1 is less than the parasitic capacitance of the Gate terminal of the high-voltage power transistor MN2. The operational amplifier AMP, the low-voltage power transistor MN1, and the sampling resistor R1 form a constant-current loop, and the conduction of the high-voltage power transistor MN2 is indirectly controlled through the constant-current loop. The voltage reference unit is used to determine the level change information of the reference voltage V REF according to a preset target audio file, and generate the reference voltage V REF .

[0005] In a second aspect, an embodiment of the present application provides an electronic device, and the electronic device includes the high-voltage linear LED driving circuit according to any one of the embodiments of the present application.

[0006] An embodiment of the present application provides a high-voltage linear LED driving circuit. The high-voltage linear LED driving circuit includes: a clamping diode D1, a current reference unit, a voltage reference unit, a voltage reference driving stage unit, an operational amplifier AMP, a low-voltage power transistor MN1, a high-voltage power transistor MN2, a protection diode D2, and a sampling resistor R1; the cathode of the clamping diode D1 is used to connect to an external power supply V IN , the anode of an external LED, and the cathode of the clamping diode D1 is also respectively connected to the input terminal of the current reference unit, the voltage driving terminal of the operational amplifier AMP, the cathode of the protection diode D2, and the Gate terminal of the high-voltage power transistor MN2; the anode of the clamping diode D1 is grounded; the output terminal of the current reference unit is connected to the bias current input terminal of the operational amplifier AMP; the output terminal of the voltage reference unit is connected to the input terminal of the voltage reference driving stage unit; the output terminal of the voltage reference driving stage unit is connected to the positive electrode of the operational amplifier AMP to provide the reference voltage V REF to the operational amplifier AMP; the output terminal of the operational amplifier AMP is connected to the Gate terminal of the low-voltage power transistor MN1; the Drain terminal of the low-voltage power transistor MN1 is respectively connected to the anode of the protection diode D2 and the Source terminal of the high-voltage power transistor MN2, and the Source terminal of the low-voltage power transistor MN1 is grounded through the sampling resistor R1; the Drain terminal of the high-voltage power transistor MN2 is used to connect to the cathode of the external LED; among them, the first threshold voltage V of the Gate terminal of the low-voltage power transistor MN1 th1 is less than the second threshold voltage V of the Gate terminal of the high-voltage power transistor MN2 th2, the parasitic capacitance of the Gate terminal of the low-voltage power transistor MN1 is smaller than that of the Gate terminal of the high-voltage power transistor MN2. The operational amplifier AMP, the low-voltage power transistor MN1, and the sampling resistor R1 form a constant-current loop, and the conduction of the high-voltage power transistor MN2 is indirectly controlled through the constant-current loop. The voltage reference unit is used to determine the reference voltage V according to the preset target audio file REF of the level change information, and generate the reference voltage V according to the level change information REF . Through the above method, the following technical effects are achieved: 1. In this application, the audio file analyzing the film and television atmosphere is controlled to determine the current required lighting atmosphere, and it is converted into executable level change information to generate the reference voltage V that can be regulated in the circuit REF , so as to realize the rapid change of the LED lighting with the film and television atmosphere; 2. In this application, the low-voltage power transistor MN1 and the high-voltage power transistor MN2 are connected in series. The high-voltage withstand characteristic of the high-voltage power transistor MN2 is used to enable the entire circuit to operate in the LED-related circuit driven by high voltage. The output current of the operational amplifier AMP is used to drive the low-voltage power transistor MN1 to control the loop current, so as to obtain the characteristics of both high-voltage withstand and rapid start of the loop current, improving the dimming linearity in the case of low brightness; 3. The bias current of the operational amplifier AMP adopts the bias current related to the process corner of the low-voltage power transistor MN1, so as to offset the current response difference between chips brought by different process corners, and improve the consistency in the case of low brightness in the multi-chip system. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0008] Figure 1 FIG. 15 is a first LED current change schematic diagram of a traditional high-voltage linear LED driving circuit; Figure 2 FIG. 18 is a second LED current change schematic diagram of a traditional high-voltage linear LED driving circuit; Figure 3 FIG. 21 is a circuit schematic diagram of a high-voltage linear LED driving circuit provided by an embodiment of the present application; Figure 4 FIG. 24 is a circuit schematic diagram of a traditional high-voltage linear LED driving circuit; Figure 5 FIG. 27 is a circuit schematic diagram of a current reference unit provided by an embodiment of the present application; Figure 6Schematic diagram of current response of a chip with process corner deviation provided by an embodiment of the present application; Figure 7 Schematic diagram of current response of a chip for improving process corner deviation provided by an embodiment of the present application; Figure 8 Schematic flowchart of a method for adjusting level change information provided by an embodiment of the present application. Detailed implementation manners

[0009] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0010] The flowchart shown in the accompanying drawings is only an example illustration, and does not necessarily include all contents and operations / steps, nor does it necessarily execute in the described order. For example, some operations / steps can also be decomposed, combined, or partially merged. Therefore, the actual execution order may be changed according to the actual situation.

[0011] It should also be understood that the terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification of the present application and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0012] It should be further understood that the term " / and" used in the specification of the present application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0013] In the current technology, at high brightness (as shown in Figure 1 ), due to the very small error caused by current startup, which can almost be ignored, ordinary linear LED driver chips can meet the linearity requirements of LED dimming. However, at low brightness (as shown in Figure 2 ), since it takes a relatively long time (about 700 ns) for current startup and the current presents a triangle, when adjusting the same duty cycle, the increment of brightness is inconsistent, resulting in poor dimming linearity.

[0014] As shown in Figure 2As shown, at low brightness (short duty cycle), since it takes a long time for the current to start (about 700ns), the current is triangular. When the same duty cycle is adjusted, the brightness increment is inconsistent, resulting in poor dimming linearity. When the same PWM signal is used to drive multiple groups of LEDs, the response time deviation between chips is large, resulting in large current deviations between light groups driven by different chips at low brightness, and large color differences.

[0015] See also Figure 3 , Figure 3 1 is a circuit diagram of a high voltage linear LED driving circuit provided by an embodiment of the present application. Figure 3 As shown, the high-voltage linear LED driving circuit 100 includes: a clamping diode D1, a current reference unit 11, a voltage reference unit 12, a voltage reference driving stage unit 13, an operational amplifier AMP, a low-voltage power tube MN1, a high-voltage power tube MN2, a protection diode D2 and a sampling resistor R1.

[0016] The cathode of the clamping diode D1 is used to connect the external power supply V IN , the anode of the external LED, and the cathode of the clamping diode D1 are also connected to the input end of the current reference unit 11, the voltage driving end of the operational amplifier AMP, the cathode of the protection diode D2, and the Gate end of the high-voltage power tube MN2. The anode of the clamping diode D1 is grounded. The output end of the current reference unit 11 is connected to the bias current input end of the operational amplifier AMP to output the reference current I to the operational amplifier AMP. Bias The output end of the voltage reference unit 12 is connected to the input end of the voltage reference driving stage unit 13 to provide the initial voltage V0 to the voltage reference driving stage unit 13. The output end of the voltage reference driving stage unit 13 is connected to the positive electrode of the operational amplifier AMP to provide the reference voltage V REF The output end of the operational amplifier AMP is connected to the Gate end of the low-voltage power tube MN1. The Drain end of the low-voltage power tube MN1 is respectively connected to the anode of the protection diode D2 and the Source end of the high-voltage power tube MN2. The Source end of the low-voltage power tube MN1 is grounded through the sampling resistor R1. The Drain end of the high-voltage power tube MN2 is used to connect to the cathode of the external LED. Among them, the first threshold voltage V of the Gate end of the low-voltage power tube MN1 th1 Less than the second threshold voltage V of the Gate terminal of the high-voltage power tube MN2 th2, the parasitic capacitance of the Gate terminal of the low-voltage power transistor MN1 is smaller than that of the Gate terminal of the high-voltage power transistor MN2. The operational amplifier AMP, the low-voltage power transistor MN1, and the sampling resistor R1 form a constant-current loop, and the conduction of the high-voltage power transistor MN2 is indirectly controlled through the constant-current loop to shorten the response time of the high-voltage power transistor MN2. The voltage reference unit is used to determine the reference voltage V according to the preset target audio file REF of the level change information, and generate the reference voltage V according to the level change information REF .

[0017] The voltage reference unit 12 provides the initial voltage V0 for the voltage reference driving stage unit 13, and the voltage reference driving stage unit 13 performs voltage following on the initial voltage V0 to generate the reference voltage V REF , and the voltage reference driving stage unit 13 provides the reference voltage V for the positive electrode of the operational amplifier AMP REF . Since at the moment when the constant-current loop is turned on, a feedback voltage signal that jumps rapidly at the nanosecond level will be generated at the negative electrode of the operational amplifier AMP, the input voltage of the positive electrode of the operational amplifier AMP must have a certain driving ability to maintain a stable reference voltage V for the constant-current loop REF . If there is no voltage reference driving stage unit 13 and the initial voltage V0 is directly input to the positive electrode of the operational amplifier AMP, due to the influence of the parasitic capacitance between the positive electrodes of the operational amplifier AMP, the initial voltage V0 is easily interfered by the constant-current loop, resulting in a longer loop establishment time.

[0018] The operational amplifier AMP provides gain for the chip constant-current loop and provides a transient driving current for the Gate terminal of the low-voltage power transistor MN1. The operational amplifier AMP, the low-voltage power transistor MN1, and the sampling resistor R1 together form a constant-current loop, and the current flowing through the low-voltage power transistor MN1 is determined by the reference voltage V REF and the sampling resistor R1 together.

[0019] The operational amplifier AMP adjusts the Gate terminal voltage of the low-voltage power transistor MN1 according to the magnitude of the current flowing through the low-voltage power transistor MN1, thereby adjusting the current flowing through the low-voltage power transistor MN1 to make the current flowing through the low-voltage power transistor MN1 remain constant. The low-voltage power transistor MN1 is an ordinary 5V NMOS transistor.

[0020] As Figure 4 shown, in the traditional high-voltage linear LED driving circuit, the operational amplifier AMP directly drives the high-voltage NMOS transistor (equivalent to Figure 3 the high-voltage power transistor MN2 in), and the threshold voltage of the 5V NMOS transistor is lower and the Gate terminal parasitic capacitance is smaller, so that the loop response speed is faster. The current rise time of the traditional high-voltage linear LED driving is about 700ns.

[0021] The technical solution of this application can shorten the rise time of the loop current to about 140 ns. The shortening of the rise time makes the LED current closer to an ideal square wave, improving the dimming linearity under low brightness conditions.

[0022] The Gate terminal of the high-voltage power transistor MN2 is directly connected to the bias voltage VDD. The current flowing through the high-voltage power transistor MN2 is equal to the current flowing through the low-voltage power transistor MN1. In this way, the magnitude of the current flowing through the high-voltage power transistor MN2 is controlled by the constant-current loop.

[0023] The positive electrode of the protection diode D2 is connected to the Source terminal of the high-voltage power transistor MN2 and the Drain terminal of the low-voltage power transistor MN1. The negative electrode of the protection diode D2 is connected to the bias voltage VDD. The protection diode D2 and the clamping diode D1 form a series connection to jointly clamp and clamp the voltage at the Drain terminal of the low-voltage power transistor MN1 to prevent a high voltage from being generated at the Drain terminal of the low-voltage power transistor MN1 and damaging the low-voltage power transistor MN1.

[0024] The embodiment of this application provides a high-voltage linear LED driving circuit. The high-voltage linear LED driving circuit includes: a clamping diode D1, a current reference unit, a voltage reference unit, a voltage reference driving stage unit, an operational amplifier AMP, a low-voltage power transistor MN1, a high-voltage power transistor MN2, a protection diode D2, and a sampling resistor R1; the cathode of the clamping diode D1 is used to connect to the external power supply V IN , the anode of the external LED, and the cathode of the clamping diode D1 is also respectively connected to the input terminal of the current reference unit, the voltage driving terminal of the operational amplifier AMP, the cathode of the protection diode D2, and the Gate terminal of the high-voltage power transistor MN2; the anode of the clamping diode D1 is grounded; the output terminal of the current reference unit is connected to the bias current input terminal of the operational amplifier AMP; the output terminal of the voltage reference unit is connected to the input terminal of the voltage reference driving stage unit; the output terminal of the voltage reference driving stage unit is connected to the positive electrode of the operational amplifier AMP to provide a reference voltage V REF to the operational amplifier AMP; the output terminal of the operational amplifier AMP is connected to the Gate terminal of the low-voltage power transistor MN1; the Drain terminal of the low-voltage power transistor MN1 is respectively connected to the anode of the protection diode D2 and the Source terminal of the high-voltage power transistor MN2, and the Source terminal of the low-voltage power transistor MN1 is grounded through the sampling resistor R1; the Drain terminal of the high-voltage power transistor MN2 is used to connect to the cathode of the external LED; wherein, the first threshold voltage V th1 of the Gate terminal of the low-voltage power transistor MN1 is less than the second threshold voltage V th2, the parasitic capacitance of the Gate terminal of the low-voltage power transistor MN1 is less than that of the Gate terminal of the high-voltage power transistor MN2. The operational amplifier AMP, the low-voltage power transistor MN1, and the sampling resistor R1 form a constant-current loop, and the conduction of the high-voltage power transistor MN2 is indirectly controlled through the constant-current loop. The voltage reference unit is used to determine the reference voltage V according to the preset target audio file REF of the level change information, and generate the reference voltage V according to the level change information REF . Through the above method, the following technical effects are achieved: 1. In this application, by controlling and analyzing the audio file of the film and television atmosphere, the current required lighting atmosphere is determined and converted into executable level change information to generate a reference voltage V that can be regulated in the circuit REF , so as to realize the rapid change of the LED light with the film and television atmosphere. 2. This application adopts the series connection method of the low-voltage power transistor MN1 and the high-voltage power transistor MN2. Utilize the high-voltage withstand characteristic of the high-voltage power transistor MN2 to enable the entire circuit to operate in the LED-related circuit driven by high voltage. Use the output current of the operational amplifier AMP to drive the low-voltage power transistor MN1 to control the loop current, so as to obtain the characteristics of both high-voltage withstand and rapid start of the loop current, and improve the dimming linearity under low brightness conditions. 3. The bias current of the operational amplifier AMP adopts the bias current related to the process angle of the low-voltage power transistor MN1, so as to offset the current response difference between chips brought by different process angles, and improve the consistency under low brightness conditions in the multi-chip system.

[0025] To more clearly introduce the technical solution of this application, the technical solution of this application will also be introduced through specific embodiments below. It should be noted that the specific embodiment is used to expand the description of the technical solution of this application, rather than limiting this application.

[0026] In some embodiments, as Figure 5 shown, the current reference unit 11 includes: a current source I1, a first current power transistor MN3, a second current power transistor MN4, and a second sampling resistor R3.

[0027] As Figure 3 and Figure 5 shown, the input terminal of the current source I1 is connected to the cathode of the clamping diode D1, and the output terminal of the current source I1 is respectively connected to the Gate terminal of the first current power transistor MN3 and the Drain terminal of the second current power transistor MN4. The Drain terminal of the first current power transistor MN3 is connected to the positive electrode of the operational amplifier AMP, and the Source terminal of the first current power transistor MN3 is respectively connected to the Gate terminal of the second current power transistor MN4 and the first terminal of the second sampling resistor R3. The second terminal of the second sampling resistor R3 and the Source terminal of the second current power transistor MN4 are both grounded.

[0028] In some embodiments, the process corners of the low-voltage power transistor MN1, the first current power transistor MN3, and the second current power transistor MN4 are the same.

[0029] Exemplarily, the inter-chip differences of the chip are mainly caused by factors such as process corner deviation. Different process corners result in different magnitudes of the first threshold voltage V of the low-voltage power transistor MN1 th1 and different magnitudes of the parasitic capacitance at the Gate terminal, and different loop response speeds, resulting in a time difference of about 40 ns in the current response between chips (as Figure 6 shown), which causes color differences between different LEDs in the low-brightness case.

[0030] The bias current I in this application Bias = V gs4 / R3, where V gs4 is the voltage at the Gate terminal of the second current power transistor MN4. The second current power transistor MN4 and the low-voltage power transistor MN1 are of the same type of device and are placed closely on the layout, so the process corners of the second current power transistor MN4 and the low-voltage power transistor MN1 on each chip are consistent, and the bias current of the operational amplifier AMP changes with the change of the process corners of different chips.

[0031] In different process corners, when the first threshold voltage V of the low-voltage power transistor MN1 th1 is small, the bias current of the operational amplifier AMP and the drive current of the low-voltage power transistor MN1 are also small. When the first threshold voltage V of the low-voltage power transistor MN1 th1 is large, the bias current of the operational amplifier and the drive current of MN1 are also large. Such a design makes the consistency of the response times of different chips more convergent.

[0032] The improved inter-chip current difference response is as Figure 7 shown. For chips with different process corners, the rising slope of the slower current response is larger, and the rising slope of the faster current response is smaller. Under the same duty cycle, the difference in the area covered by the current waveform (corresponding to the brightness of the LED) is smaller.

[0033] In some embodiments, as Figure 3 shown, the high-voltage linear LED driving circuit 100 further includes: a current-limiting resistor R2 and a filtering capacitor C1.

[0034] The first end of the current-limiting resistor R2 is respectively used to connect an external power supply V IN , the anode of the external LED. The second end of the current-limiting resistor R2 is connected to the cathode of the clamping diode D1. The first end of the filtering capacitor C1 is connected to the cathode of the clamping diode D1, and the second end of the filtering capacitor C1 is grounded.

[0035] Exemplarily, the external power supply VIN A lower bias voltage VDD is provided through a current-limiting resistor R2 and a clamping diode D1, and the series-connected current-limiting resistor R2 and filter capacitor C1 provide transient energy for the driving stage of the operational amplifier AMP, so as to achieve the purpose of fast response and stable startup. Without the cooperation of the filter capacitor C1, when the constant-current loop starts up, there will be a large ringing in the bias voltage VDD, which will slow down the response speed of the constant-current loop.

[0036] The high-voltage linear LED driving circuit 100 provided by the embodiment of the present application is also used to adjust the brightness of the usage scenario in cooperation with the video content to complete the atmosphere rendering, and is usually applied to occasions such as movie theaters or music scenes, and is used to improve the user experience by dimming.

[0037] In some embodiments, please refer to Figure 8 , Figure 8 is a schematic flowchart of a method for adjusting level change information provided by an embodiment of the present application. As shown in Figure 8 In the method for adjusting level change information shown, the control unit (not shown in the figure) of the high-voltage linear LED driving circuit 100 is the execution subject, and is specifically used to execute: S101-S106.

[0038] S101. Obtain a target audio file.

[0039] S102. Perform context analysis on the target audio file through a voice analysis model to obtain context description information, including: human voice part-of-speech, human voice energy value, scene part-of-speech, and scene energy value.

[0040] It should be noted that currently, the dimming schemes for the usage scenarios of synchronized video content are basically adjusted by professionals, which have the difficulties of high cost and non-replicability. For prior video content or video content with low popularity, there is usually no corresponding dimming scheme, and synchronized dimming cannot be performed in relevant playback occasions.

[0041] With the rise of artificial intelligence, some intelligent dimming schemes based on human voice content have also emerged. However, the dimming indicators of these schemes are monotonous, it is difficult to achieve precise dimming, and they often bring a worse user experience to users.

[0042] The present application starts from human voice and scene sound, combines their energy values, and performs multi-dimensional scoring to find out the specific factors that determine the current atmosphere. The emotions output by human voice and scene sound may be synchronized (non-different intervals) or fragmented (conflict intervals). There may be intense scene expressions in calm human voice expressions, and there may be happy scene expressions in crying human voice expressions. These conflicting information often leads to misjudgment of emotion classification by the model, and the corresponding evaluation criteria are different in different situations and need to be distinguished.

[0043] S103. If the human voice part-of-speech and the scene part-of-speech are in the conflict range, calculate the first fluctuation amplitude of the human voice energy value and the second fluctuation amplitude of the scene energy value, compare the first fluctuation amplitude and the second fluctuation amplitude, and the larger one is the target fluctuation amplitude. Determine the target part-of-speech from the human voice part-of-speech and the scene part-of-speech according to the target fluctuation amplitude, and determine the target energy value from the human voice energy value and the scene energy value.

[0044] Exemplarily, in the conflict range, following the principle of contrast, an object with a stable fluctuation amplitude is often not suitable as the main object because its relevant emotional expression may be affected by an object with a large fluctuation amplitude. Therefore, in the conflict range, if the human voice part-of-speech has a large fluctuation, use the human voice part-of-speech as the target part-of-speech; if the scene part-of-speech has a large fluctuation, use the scene part-of-speech as the target part-of-speech, and the energy value corresponding to the target part-of-speech is the target energy value.

[0045] S104. If the human voice part-of-speech and the scene part-of-speech are in the non-conflicting range, compare the interval peak density of the human voice energy value and the interval peak density of the scene energy value, and the larger one is the target energy value. Determine the target part-of-speech from the human voice part-of-speech and the scene part-of-speech according to the target energy value.

[0046] Exemplarily, in the non-conflicting range, the larger the interval peak density, the greater the information density in the corresponding interval. In the absence of conflict, the greater the information density, the more suitable it is as the main object, and thus the target part-of-speech and the target energy value are determined.

[0047] S105. Determine the speech context score according to the target part-of-speech and the target energy value.

[0048] Exemplarily, this step needs to call a preset scoring rule, and determine the speech context score according to the preset scoring rule, the target part-of-speech, and the target energy value.

[0049] S106. Match the speech context score with a preset energy duty cycle form to obtain the level change information.

[0050] Exemplarily, the preset energy duty cycle form can calculate the brightness value based on the speech data and the picture brightness of the training data (classical video), and then convert it into the corresponding duty cycle, so as to obtain the LED brightness that should be set for the current atmosphere. After making the above data into an energy duty cycle form and obtaining the speech context score, matching can be performed to obtain the level change information.

[0051] It should be noted that the level change information includes a high-level interval and a low-level interval. The voltage reference unit outputs the reference voltage V in the high-level interval REF , and the voltage reference unit stops outputting the reference voltage V in the low-level interval REF, thereby indirectly controlling the duration of the high output level of the operational amplifier AMP, and further controlling the duty cycle of the low-voltage power transistor MN1.

[0052] In some embodiments, the speech analysis model includes: a speech coding layer, an attention layer, a multi-modal projection layer, and a decoding layer. The target audio file is subjected to context analysis through the speech analysis model to obtain context description information, including: S1021-S1026.

[0053] S1021. Perform time-frequency analysis and text alignment processing on the target audio file to generate a multi-modal input matrix, which includes: a Mel spectrogram, a text part-of-speech tagging sequence, and a time alignment mapping matrix.

[0054] Exemplarily, the target audio file is resampled at a sampling rate of 16 kHz, and framed using a Hamming window (window length 25 ms, step size 10 ms). In frequency-domain processing, a 128-dimensional Mel filter bank is used to extract the spectrogram, and dynamic range compression (log(1 + amplitude)) is used to enhance low-frequency features. At the same time, the audio is transcribed into text through a pre-trained speech recognition model to generate a text sequence aligned with timestamps. The cosine similarity between each speech frame and the morphemes of the corresponding text sequence is calculated. When the similarity is lower than the threshold of 0.75, secondary calibration is triggered for secondary time alignment, and a multi-modal input matrix containing three sub-matrices is output, including: 1. A 64×128-dimensional Mel spectrogram; 2. A text part-of-speech tagging sequence; 3. A time alignment mapping matrix.

[0055] S1022. Based on the speech coding layer, perform hierarchical acoustic feature extraction on the multi-modal input matrix to generate a human voice feature matrix, a scene feature matrix, and an interaction weight matrix.

[0056] The multi-modal input matrix is fed into a speech coding layer containing a dual-path CNN. In the human voice feature path, a 5-layer depthwise separable convolutional network is used, with an SE attention module connected after each layer, focusing on the human voice frequency band of 300-3000 Hz. The scene feature path uses a dilated convolution structure, configured with a spectral masking layer to suppress the energy of the human voice frequency band, and focuses on extracting low-frequency ambient sound features of 20-300 Hz. An innovative cross-path feature gating mechanism is introduced: the feature suppression coefficient of the human voice path for the scene path is generated through the sigmoid function, and a feature cube composed of a 64×32-dimensional human voice feature matrix, a 64×32-dimensional scene feature matrix, and their interaction weight matrix is output. This design realizes the feature decoupling of human voice and ambient sound, and experiments show that the purity of human voice features in a noisy environment is improved by 42%.

[0057] S1023. Based on the attention layer, perform dynamic weighting processing on the human voice feature matrix, the scene feature matrix, and the interaction weight matrix to generate a context compression matrix.

[0058] In some embodiments, based on the attention layer, dynamically weighting the human voice feature matrix, the scene feature matrix, and the interaction weight matrix to generate a context compression matrix, including: S231 - S234.

[0059] S231. Perform multi-head attention processing on the human voice feature matrix to generate a human voice attention feature matrix.

[0060] S232. Perform sparse attention processing on the scene feature matrix to generate a scene attention feature matrix.

[0061] S233. Use the human voice attention feature matrix as a position bias term and inject it into the process of sparse attention processing to generate a cross-modal interaction matrix.

[0062] S234. Perform feature fusion and compression on the human voice attention feature matrix, the scene attention feature matrix, and the cross-modal interaction matrix to generate a context compression matrix.

[0063] Exemplarily, construct parallel dual attention channels in the attention layer. The human voice channel uses multi-head attention (8 heads) with learnable query vectors, where the Q matrix is initialized by trainable parameters, and the K and V matrices are from the human voice feature matrix. The scene channel uses sparse attention based on an energy threshold, defining an energy weight coefficient α, and masking the attention connection of the current frame when α < 0.6. The innovation lies in the cross-feature modulation module: using the output of human voice attention as the position bias term of the scene channel, and performing cross-modal interaction according to the position bias term. After layer normalization of the output, a 32×64-dimensional context compression matrix is obtained, including intermediate representations such as the human voice energy distribution histogram and the scene feature activation heat map.

[0064] S1024. Based on the multi-modal projection layer, perform feature space mapping and optimization on the context compression matrix to generate a human voice energy vector, a scene category vector, and a cross-modal association matrix.

[0065] Exemplarily, the fully connected layer is designed with a two-stream projection structure. In the human voice projection stream, the 32×64-dimensional feature matrix is reduced in dimension to 16×16 through a linear transformation, and an improved NT-Xent contrastive loss is applied, where the positive samples are cross-modal samples from the same speaker, and the negative samples include scene feature interference terms. The scene projection stream introduces an adversarial training mechanism, making the scene features indistinguishable from the random noise vector through a gradient reversal layer to enhance feature robustness. The key innovation lies in the construction of the cross-modal association matrix: establishing an association matrix between the human voice energy value and the scene part-of-speech, and outputting through modeling the conditional probability distribution, including: 1. A 16-dimensional human voice energy vector; 2. A 16-dimensional scene category vector; 3. A cross-modal association matrix.

[0066] S1025. Based on the decoding layer, perform dynamic fusion and decoding processing on the vocal energy vector, scene category vector, and cross-modal correlation matrix to generate a sequence of situational description information.

[0067] In some embodiments, based on the decoding layer, perform dynamic fusion and decoding processing on the vocal energy vector, scene category vector, and cross-modal correlation matrix to generate a sequence of situational description information, including: S251 - S254.

[0068] S251. Initialize the context state of the vocal energy vector and the scene category vector to generate a gating coefficient matrix.

[0069] Exemplarily, based on the conditional gated recurrent unit, at time step zero, align the dimensions of the vocal energy vector and the scene category vector, map them to the hidden space through linear transformation respectively, and calculate the initial hidden state. At time steps greater than or equal to 1, based on the previous hidden state, the current vocal context vector (calculated by the dot product of the cross-modal correlation matrix and the previous hidden state) and the scene context vector (calculated by the attention weight of the scene category vector and the previous hidden state), generate the gating coefficient matrix.

[0070] S252. Perform dual pointer network processing on the gating coefficient matrix and the cross-modal correlation matrix to generate the vocal part-of-speech selection probability and the scene word sampling probability.

[0071] Exemplarily, in the vocal pointer network, concatenate the hidden state with the current vocal context vector, generate the part-of-speech score vector through linear transformation, and perform Softmax normalization to obtain the vocal part-of-speech selection probability. In the scene pointer network, concatenate the hidden state with the scene context vector, generate the scene word score vector through linear transformation, and perform dynamic scaling based on the scene energy value to generate the scene word sampling probability.

[0072] S253. Perform energy-driven mixture distribution calculation on the vocal part-of-speech selection probability and the scene word sampling probability to generate an output word probability vector.

[0073] According to the vocal energy value and the scene energy value at the current time step, calculate the mixture weight. Perform weighted fusion on the vocal part-of-speech selection probability and the scene word sampling probability to generate a mixture probability distribution. At the same time, through the row-column constraint conditions of the cross-modal correlation matrix, sparsify the mixture probability distribution to mask the part-of-speech combination probabilities that do not conform to the non-zero element positions in the cross-modal correlation matrix.

[0074] S254. Perform sequential decoding and candidate pruning on the output word probability vector to generate a sequence of situational description information.

[0075] Exemplarily, a beam search algorithm is used to perform multi-step prediction on the mixed probability distribution, and candidate sequences are retained at each step; for each candidate sequence, its cumulative log-likelihood score is calculated, and combined with the temporal smoothness constraint term of the scene energy value, a situation description score is generated; when the length of the candidate sequence reaches the preset maximum value or the end symbol is generated, the sequence with the highest situation description score is selected as the situation description information sequence, and the merging and deduplication operation is performed on the consecutive repeated part-of-speech tags in the sequence.

[0076] S1026. Based on the decoding layer, perform multi-objective joint optimization and adaptive post-processing on the situation description information sequence to generate situation description data.

[0077] Perform multi-objective joint optimization and adaptive post-processing on the situation description information sequence to generate the final structured situation description data, including: constructing a joint optimization objective including situation classification cross-entropy loss, contrastive learning loss, mutual information minimization loss, and decoding reconstruction loss, and dynamically allocating loss weights through an adaptive weight adjustment strategy; when it is detected that the vocal energy value of consecutive frames exceeds the preset threshold, trigger an energy reallocation algorithm to adjust the energy distribution, and output JSON format data including timestamp-aligned vocal part-of-speech, energy curve, and scene label.

[0078] The embodiment of the present application provides an electronic device, and the electronic device includes the high-voltage linear LED driving circuit as described in any one of the embodiments of the present application.

[0079] As described above, only the specific implementation manners of the present application are provided, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A high-voltage linear LED driving circuit, characterized in that, The high-voltage linear LED driver circuit includes: a clamping diode D1, a current reference unit, a voltage reference unit, a voltage reference driving stage unit, an operational amplifier AMP, a low-voltage power transistor MN1, a high-voltage power transistor MN2, a protection diode D2, and a sampling resistor R1; The cathode of the clamping diode D1 is used to be connected to the external power supply V IN and the anode of the external LED respectively. The cathode of the clamping diode D1 is also connected to the input end of the current reference unit, the voltage driving end of the operational amplifier AMP, the cathode of the protection diode D2, and the Gate end of the high-voltage power transistor MN2 respectively; the anode of the clamping diode D1 is grounded; the output end of the current reference unit is connected to the bias current input end of the operational amplifier AMP; the output end of the voltage reference unit is connected to the input end of the voltage reference driving stage unit; the output end of the voltage reference driving stage unit is connected to the positive electrode of the operational amplifier AMP to provide the reference voltage V REF to the operational amplifier AMP; the output end of the operational amplifier AMP is connected to the Gate end of the low-voltage power transistor MN1; the Drain end of the low-voltage power transistor MN1 is connected to the anode of the protection diode D2 and the Source end of the high-voltage power transistor MN2 respectively, and the Source end of the low-voltage power transistor MN1 is grounded through the sampling resistor R1; the Drain end of the high-voltage power transistor MN2 is used to be connected to the cathode of the external LED; Among them, the first threshold voltage V of the Gate terminal of the low-voltage power transistor MN1 th1 is less than the second threshold voltage V of the Gate terminal of the high-voltage power transistor MN2 th2 . The parasitic capacitance of the Gate terminal of the low-voltage power transistor MN1 is less than that of the Gate terminal of the high-voltage power transistor MN2. The operational amplifier AMP, the low-voltage power transistor MN1, and the sampling resistor R1 form a constant-current loop, and the conduction of the high-voltage power transistor MN2 is indirectly controlled through the constant-current loop. The voltage reference unit is used to determine the level change information of the reference voltage V REF and generate the reference voltage V according to the level change information REF .

2. The high-voltage linear LED driving circuit according to claim 1, wherein The current reference unit includes: a current source I1, a first current power transistor MN3, a second current power transistor MN4, and a second sampling resistor R3; The input end of the current source I1 is connected to the cathode of the clamping diode D1, and the output end of the current source I1 is respectively connected to the Gate end of the first current power transistor MN3 and the Drain end of the second current power transistor MN4; the Drain end of the first current power transistor MN3 is connected to the positive electrode of the operational amplifier AMP, and the Source end of the first current power transistor MN3 is respectively connected to the Gate end of the second current power transistor MN4 and the first end of the second sampling resistor R3; the second end of the second sampling resistor R3 and the Source end of the second current power transistor MN4 are both grounded.

3. The high-voltage linear LED driving circuit according to claim 2, wherein, The process corners of the low-voltage power transistor MN1, the first current power transistor MN3, and the second current power transistor MN4 are the same.

4. The high-voltage linear LED driving circuit according to claim 1, wherein, The high-voltage linear LED driver circuit further includes: a current-limiting resistor R2 and a filter capacitor C1; The first end of the current-limiting resistor R2 is respectively used for connecting to an external power supply V IN , the anode of an external LED. The second end of the current-limiting resistor R2 is connected to the cathode of the clamping diode D1; the first end of the filtering capacitor C1 is connected to the cathode of the clamping diode D1, and the second end of the filtering capacitor C1 is grounded.

5. The high-voltage linear LED driving circuit according to claim 4, wherein, When the voltage reference unit is used to execute the determination of the reference voltage V according to a preset target audio file REF for the level change information, it is specifically used to execute: Obtain a target audio file; Perform a context analysis on the target audio file through a preset speech analysis model to obtain context description information, the context description information including: human voice part-of-speech, human voice energy value, scene part-of-speech, and scene energy value; If the human voice part-of-speech and the scene part-of-speech are in a conflict interval, calculate a first fluctuation amplitude of the human voice energy value and calculate a second fluctuation amplitude of the scene energy value, compare the first fluctuation amplitude and the second fluctuation amplitude, the larger fluctuation amplitude is the target fluctuation amplitude, determine a target part-of-speech from the human voice part-of-speech and the scene part-of-speech according to the target fluctuation amplitude, and determine a target energy value from the human voice energy value and the scene energy value; If the human voice part-of-speech and the scene part-of-speech are in a non-different interval, compare the interval peak density of the human voice energy value and the interval peak density of the scene energy value, the larger interval peak density is the target energy value, and determine a target part-of-speech from the human voice part-of-speech and the scene part-of-speech according to the target energy value; Determine a speech context score according to the target part-of-speech and the target energy value; Match the speech context score with a preset energy duty cycle form to obtain the level change information.

6. The high-voltage linear LED driving circuit according to claim 5, wherein The speech analysis model includes: a speech encoding layer, an attention layer, a multi-modal projection layer, and a decoding layer. Performing a context analysis on the target audio file through the speech analysis model to obtain context description information includes: Perform time-frequency analysis and text alignment processing on the target audio file to generate a multi-modal input matrix, the multi-modal input matrix including: a Mel spectrogram, a text part-of-speech tagging sequence, and a time alignment mapping matrix; Based on the voice coding layer, hierarchical acoustic feature extraction is performed on the multi-modal input matrix to generate a human voice feature matrix, a scene feature matrix, and an interaction weight matrix; Based on the attention layer, dynamic weighting processing is performed on the human voice feature matrix, the scene feature matrix, and the interaction weight matrix to generate a situation compression matrix; Based on the multi-modal projection layer, feature space mapping and optimization are performed on the situation compression matrix to generate a human voice energy vector, a scene category vector, and a cross-modal correlation matrix; Based on the decoding layer, dynamic fusion and decoding processing are performed on the human voice energy vector, the scene category vector, and the cross-modal correlation matrix to generate a situation description information sequence; Based on the decoding layer, multi-objective joint optimization and adaptive post-processing are performed on the situation description information sequence to generate situation description data.

7. The high-voltage linear LED driving circuit according to claim 6, wherein The step of based on the attention layer, performing dynamic weighting processing on the human voice feature matrix, the scene feature matrix, and the interaction weight matrix to generate a situation compression matrix includes: Performing multi-head attention processing on the human voice feature matrix to generate a human voice attention feature matrix; Performing sparse attention processing on the scene feature matrix to generate a scene attention feature matrix; Using the human voice attention feature matrix as a position bias term and injecting it into the process of the sparse attention processing to generate a cross-modal interaction matrix; Performing feature fusion and compression on the human voice attention feature matrix, the scene attention feature matrix, and the cross-modal interaction matrix to generate a situation compression matrix.

8. The high-voltage linear LED driving circuit according to claim 6, wherein The step of based on the decoding layer, performing dynamic fusion and decoding processing on the human voice energy vector, the scene category vector, and the cross-modal correlation matrix to generate a situation description information sequence includes: Initializing the context state of the human voice energy vector and the scene category vector to generate a gating coefficient matrix; Performing a dual-pointer network process on the gating coefficient matrix and the cross-modal correlation matrix to generate a human voice part-of-speech selection probability and a scene word sampling probability; Performing an energy-driven mixture distribution calculation on the human voice part-of-speech selection probability and the scene word sampling probability to generate an output word probability vector; Performing sequential decoding and candidate pruning on the output word probability vector to generate a situation description information sequence.

9. An electronic device, characterized in that, The electronic device includes the high-voltage linear LED driving circuit according to any one of claims 1-8.

Images