Successive approximation flash memory analog-to-digital converter and operation method thereof

Through the coarse quantization and fine quantization methods of successive approximation flash analog-to-digital converter, the problem of area and power consumption growth in traditional ADC in vector matrix multiplication calculation is solved, and more efficient analog-to-digital conversion is achieved, reducing the number and area of ​​sensitive amplifiers , improves the energy efficiency of the system.

CN119093937AActive Publication Date: 2024-12-06PEKING UNIV

Patent Information

Application Number
CN202411239092.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-05
Publication Date
2024-12-06
Estimated Expiration
2044-09-05

AI Technical Summary

Technical Problem

Traditional SAR-ADC and Flash-ADC have an exponential increase in area and power consumption due to the increase in readout accuracy in vector matrix multiplication calculations, which has become a bottleneck in digital-analog mixed memory computing, limiting the improvement of the system's area and energy efficiency.

Method used

Using successive approximation flash analog-to-digital converter, through coarse quantization and multiple fine quantization, using sensitive amplifiers, Boost capacitors and 2N1-1 select 1 multiplexer, gradually shorten the quantization step size, reduce the number of sensitive amplifiers, and simplify digital control logic. , avoid binary capacitor arrays.

Benefits of technology

The quantization of N(N=N1×N2) bits is realized, which reduces the number of sensitive amplifiers, reduces area and power consumption. It is more efficient than Flash ADCs and SAR ADCs, avoiding the use of binary capacitor arrays.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119093937A_ABST
    Figure CN119093937A_ABST
Patent Text Reader

Abstract

The invention discloses a successive approximation flash memory analog-to-digital converter and an operation method thereof, and belongs to the technical field of in-memory calculation in a semiconductor CMOS (Complementary Metal Oxide Semiconductor) super-large-scale integrated circuit. According to the successive approximation flash memory analog-to-digital converter, rough quantization of N1 bits is completed through one-time operation, rough quantization and then fine quantization are carried out on an input signal VIN through N2-step quantization, the quantization step length is gradually shortened, the output precision of each time of quantization is N1 bits until the offset voltage limit of a sensitive amplifier is reached, and quantization of N (N = N1 * N2) bits is completed. The digital control logic is simple, the use of a binary capacitor array is avoided, and the area and the power consumption are smaller than those of a Flash ADC (Analog to Digital Converter) and an SAR (Synthetic Aperture Radar) ADC with the same precision.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of Compute-In-Memory technology in semiconductors and CMOS ultra-large-scale integrated circuits (ULSI), and in particular relates to an analog-to-digital converter (ADC) used in vector matrix multiplication (VMM) calculations based on a memory array. Background Art

[0002] With the continuous advancement of artificial intelligence and deep learning technologies, artificial neural networks have been widely used in natural language processing, image recognition, autonomous driving, graph neural networks and other fields. However, the expansion of the network scale has led to a large amount of energy consumption in the transmission of data between memory and traditional computing devices (such as CPUs and GPUs). This phenomenon is called the von Neumann bottleneck. In artificial neural network algorithms, the most important computing task is vector matrix multiplication. Compute-In-Memory avoids the frequent transmission of data between memory and computing units by storing weights in memory units and performing vector matrix multiplication calculations in arrays. It is considered to be an effective method to solve the von Neumann bottleneck.

[0003] like Figure 1 As shown in the figure, in-memory computing usually adopts a mixed digital-analog computing method. The memory unit can be a volatile memory such as SRAM, DRAM, or a non-volatile memory such as Flash, RRAM, PCRAM, MRAM, etc. The weights of the vector-matrix multiplication calculation are stored in the memory, and the input is implemented through a digital-to-analog converter (DAC) or a buffer (Buffer). The calculation result is represented by the current or voltage on the bit line (BL). The calculation result needs to be read through an analog-to-digital converter (ADC). The commonly used ADC structures in in-memory computing include successive approximation ADC (SAR-ADC) and flash ADC (Flash-ADC).

[0004] In order to achieve high-precision vector-matrix multiplication calculations, a high-precision ADC is required for analog-to-digital conversion reading. The structure of traditional SAR-ADC and Flash-ADC is as follows: Figure 2As shown in the figure. The area and power consumption of the capacitor array of SAR-ADC will increase exponentially with the increase of readout accuracy. The number of sensitive amplifiers (SA) and total power consumption required by Flash-ADC will also increase exponentially with the increase of readout accuracy. Since the hybrid in-memory computing requires the parallel use of a large number of ADCs for reading, the increase in the area and power consumption of a single ADC greatly increases the total area and power consumption of the system, which becomes the bottleneck of the hybrid in-memory computing in terms of area and energy efficiency improvement. Summary of the invention

[0005] In view of the above problems, the present invention proposes a successive approximation flash analog-to-digital converter, which first performs coarse quantization on the input signal, then performs multiple fine quantizations, and finally realizes the conversion of the signal.

[0006] The technical solution of the present invention is as follows:

[0007] A successive approximation flash analog-to-digital converter, characterized in that it comprises a V REF Producing unit, A successive approximation flash memory module consists of a sense amplifier, a boost capacitor and a 2 N1 -1 to 1 multiplexer, the positive terminal of the sense amplifier is connected to the input voltage, the reference voltage V REF Connect directly to the negative terminal of the sense amplifier through a switch, or connect to the left side of the boost capacitor through a switch. N1 The -1 to 1 multiplexer connects the voltage to be boosted to the right side of the Boost capacitor to charge the Boost capacitor. In the comparison phase, the connection with the Boost capacitor is cut off. The right side of the Boost capacitor is connected to the negative terminal of the sensitive amplifier. The sensitive amplifier is used to compare the two voltages and output the comparison result. If the voltage to be measured is greater than the current V REF [i] is greater, the output is high level "1". If the voltage to be measured is greater than the current V REF [i] is small, then the output is low level "0"; 2 N1 The -1-to-1 multiplexer selects the appropriate next-step quantization basic voltage based on the compared OUT[N1:1], stores this voltage in the Boost capacitor, and raises the reference voltage V of the sense amplifier in the next comparison. REF , A successive approximation flash analog-to-digital module is compared once Thermometer code ( "0" or "1"); the digital control module includes a thermometer decoder and a bit splicing module. The thermometer decoder converts the single operation output The thermometer codes are converted into binary codes OUT[N1:1], and the bit concatenation module integrates the binary codes obtained by the N2 step operation to obtain the final N (N=N1×N2) bit output.

[0008] Furthermore, a method for operating a successive approximation flash analog-to-digital converter is provided, which completes N1-bit coarse quantization in one operation and completes N-bit fine quantization through N2 steps of operation, wherein N=N1×N2, and the steps include:

[0009] 1) Cumulative signal V of CIM array IN Direct input connection to A successive approximation flash memory module, the input signal is connected to the positive terminal of the sense amplifier, The negative terminals of the sense amplifiers are connected in sequence To V REF _1 st [1] Over the entire reference voltage V REF The output accuracy of a single comparison of the flash memory module is N1, and the thermometer decoder obtains the output accuracy of a single comparison. Thermometer code, convert it into binary code OUT_1 st [N1:1];

[0010] 2)2 N1 -1 to 1 multiplexer according to binary encoding result OUT_1 st [N1:1] Determine the quantization reference voltage for the next step The total quantization size is Connect the left end of each Boost capacitor to VSS and the right end to V REF _2 nd , store the quantized basic voltage of this conversion;

[0011] 3) The left end of each Boost capacitor is connected to The level is evenly distributed between 0 and V REF / N1, The right end of the boost capacitor is disconnected from V REF _2 nd The connection is directly connected to the input of the sense amplifier, which in turn raises the reference voltage to V REF _2 nd [i]+V REF _2 nd The thermometer decoder compares the input signal with this reference voltage and obtains Thermometer code and converted into binary code OUT_2 nd [N1:1];

[0012] 4) Repeat steps 2) to 3) to change the voltage value stored in Boost in sequence, so that its voltage value gradually approaches the input voltage V IN , until the last step output OUT_N2 is obtained th [N1:1].

[0013] The beneficial effects of the present invention are as follows:

[0014] The invention provides a successive approximation flash analog-to-digital converter (Multi-Step N bit SAR-Flash-ADC) that quantizes the input signal V by multiple steps (N2). IN The quantization is performed coarsely and then finely, and the quantization step is gradually shortened. The output accuracy of each quantization is N1 bits, until the offset voltage limit of the sense amplifier is reached, that is, the quantization of N (N = N1 × N2) bits is completed. Compared with the N-bit Flash ADC, the number of sense amplifiers in the present invention is reduced from 2 N -1 Reduce to Taking N=16 and N1=4 as an example, the number of sensitive amplifiers is reduced from 65535 to 15; compared with the N-bit SAR ADC, the digital control logic is simpler and the use of binary capacitor arrays is avoided. Its area and power consumption are smaller than those of Flash ADC and SAR ADC with the same accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Schematic diagram of vector-matrix multiplication based on mixed digital-analog in-memory computing;

[0016] Figure 2 It is the structure diagram of SAR-ADC and Flash-ADC;

[0017] Figure 3 This is a structural diagram of the Multi-Step N bit SAR-Flash-ADC of the present invention;

[0018] Figure 4 This is the connection diagram of the first step of rough quantization (step 1) of the present invention;

[0019] Figure 5 Storing quantized basic voltage (step 2) connection diagram for the present invention;

[0020] Figure 6 The connection diagram of the second step refinement (step 3) of the present invention;

[0021] Figure 7 is V of the multi-step quantization process of the present invention REF Relationship diagram; Specific embodiments

[0022] The invention provides a successive approximation flash analog-to-digital converter (Multi-Step N-bit SAR-Flash-ADC), such as Figure 3 As shown, it contains a V REF Producing unit, A successive approximation flash analog-to-digital module, a digital control module consisting of a thermometer decoder and a bit stitching module, V REF The generation unit can be shared in the ADC of CIM. The thermometer decoder converts the single output thermometer code into binary code OUT[N1:1]. The bit splicing module integrates the multi-step (N2) binary code to obtain the final N (N=N1×N2) bit output. The successive approximation flash memory module consists of a sensitive amplifier, a boost capacitor and a 2 N1 -1 to 1 multiplexer. The sensitive amplifier is used to compare the two voltages and output the comparison result. If the voltage to be measured is greater than the current V REF [i] is greater, the output is high level "1". If the voltage to be measured is greater than the current V REF [i] is small, the output is low level "0". N1 -1 to 1 multiplexer based on the OUT[2 N1 -1:1] Select the appropriate quantization basic voltage for the next step, store this voltage in the Boost capacitor, and raise the reference voltage V of the sense amplifier in the next comparison step RFF . V REF The generating unit can be shared among the ADCs of the CIM.

[0023] The operation method of the successive approximation flash memory analog-to-digital converter of the present invention comprises the following steps:

[0024] 1) Cumulative signal V of CIM array IN Direct input connection to 2 N1 -1 successive approximation flash memory module, the input signal is connected to the positive terminal of the sense amplifier, and the negative terminal of the sense amplifier is directly connected to V REF _1 st [2 N1 -1:1],V REF _1 st [2 N1 -1:1] across the reference voltage V REF The output precision of a single comparison of the flash memory module is N1, and the thermometer decoder obtains 2 through a single comparison. N1 -1 thermometer code, convert the thermometer code into binary code OUT_1 st [N1:1], complete the first step of coarse quantization, such as Figure 4 As shown;

[0025] 2) According to the coarse quantization result OUT_1 st [N1:1] Determine the quantization basic voltage V for the second step REF _2 nd , V REF _2 nd Value in In the example, connect the left end of each Boost capacitor to VSS and the right end to V REF _2 nd , storing the quantized basic voltage of this conversion, such as Figure 5 Shown

[0026] 3) 2 N1 -1 Boost capacitor's left end is connected in sequence The level is evenly distributed between 0 and V REF / N1, 2 N1 -1 Boost capacitor right end disconnected from V REF _2 nd The connection is directly connected to the input of the sense amplifier, which in turn raises the reference voltage to V REF _2 nd [i]+V REF _2 nd The thermometer decoder compares the input signal with this reference voltage and obtains 2 N1 -1 thermometer code and converted to binary code OUT_2 nd [N1:1], complete the second step of fine quantization, such as Figure 6 As shown,.

[0027] 4) Repeat steps 2) to 3), that is, according to the refined quantization result OUT_2 nd [N1:1], determine the quantization basic voltage V REF _3 rd , V REF _3 rd The value is V REF _2 nd [i]+V REF _2 nd In the example, connect the left end of each Boost capacitor to VSS and the right end to V REF _3 rd , store the quantized basic voltage of this conversion;

[0028] 5) 2 N1 -1 The left end of the boost capacitor is connected to the reference voltage V REF _3 rd [2 N1 -1:1],V REF _3 rd [2 N1-1:1] is evenly distributed between 0 and V REF / N1 2 Between, 2 N1 -1 Boost capacitor right end disconnected from V REF _3 rd The connection is directly connected to the input of the sense amplifier, which in turn raises the reference voltage to V REF _3 rd [i]+V REF _3 rd The thermometer decoder compares the input signal with this reference voltage and obtains 2 N1 -1 thermometer code and converted to binary code OUT_3 rd [N1:1], complete the third step of refinement;

[0029] 6) Repeat steps 2) to 3), change the voltage value stored in Boost in turn, so that its voltage value gradually approaches the input voltage VIN, and divide the quantization level into finer levels on this basis until the final output OUT_N2 is obtained. th [N1:1]. Ideally, Figure 7 As shown, the input voltage V can be infinitely approached by an infinite number of operation steps. IN , then the actual structure is limited by the offset voltage (Offset) of the sense amplifier, so it is necessary to ensure the minimum quantization step size LSB (V REF / 2 N ) is greater than the offset voltage of the sense amplifier.

[0030] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. A person skilled in the art may modify or make equivalent substitutions for the technical solutions of the present invention without departing from the spirit and scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. A successive approximation flash analog-to-digital converter, characterized in that: Contains a V REF Producing unit, A successive approximation flash memory module consists of a sense amplifier, a boost capacitor and a 2 N1 -1 to 1 multiplexer, the positive terminal of the sense amplifier is connected to the input voltage, the reference voltage V REF Connect directly to the negative terminal of the sense amplifier through a switch, or connect to the left side of the boost capacitor through a switch. N1 The -1 to 1 multiplexer connects the voltage to be boosted to the right side of the Boost capacitor to charge the Boost capacitor, and cuts off the connection with the Boost capacitor in the phase of comparison. The right side of the Boost capacitor is connected to the negative terminal of the ratio-sensitive amplifier. The sensitive amplifier is used to compare the two voltages and output the comparison result. REF [i] is greater, the output is high level "1", if the voltage to be measured is greater than the current V REF [i] is small, then the output is low level "0"; 2 N1 The -1-to-1 multiplexer selects the appropriate next-step quantization basic voltage based on the compared OUT[N1:1], stores this voltage in the Boost capacitor, and raises the reference voltage V of the sense amplifier in the next comparison. REF , A successive approximation flash analog-to-digital module is compared once Thermometer code ( "0" or "1"); the digital control module includes a thermometer decoder and a bit splicing module. The thermometer decoder converts the single operation output The thermometer codes are converted into binary codes OUT[N1:1], and the bit concatenation module integrates the binary codes obtained by the N2 step operation to obtain the final N (N=N1×N2) bit output.

2. The successive approximation register flash analog-to-digital converter according to claim 1, wherein: The V REF The generating unit is arranged in the ADC of the CIM.

3. The successive approximation register flash analog-to-digital converter according to claim 1, wherein: Reference voltage V REF Connect directly to the negative terminal of the sense amplifier through a switch, or connect to the boost capacitor through a switch.

4. The operation method of the successive approximation flash analog-to-digital converter as claimed in claim 1, wherein the coarse quantization of N0 bits is completed in one operation, and the fine quantization of N bits is completed by N2 steps of operation, wherein N=N1×N2, and the steps include: 1) Cumulative signal V of CIM array IN Direct input connection to A successive approximation flash memory module, the input signal is connected to the positive terminal of the sense amplifier, The negative terminals of the sense amplifiers are connected in sequence Over the entire reference voltage V REF The output accuracy of a single comparison of the flash memory module is N1, and the thermometer decoder obtains the output accuracy of a single comparison. Thermometer code, convert it into binary code OUT_1 st [N1:1]; 2)2 N1 -1 to 1 multiplexer according to binary encoding result OUT_1 st [N1:1] Determine the next step of quantization reference voltage The total quantization size is Connect the left end of each Boost capacitor to VSS and the right end to V REF _2 nd , store the quantized basic voltage of this conversion; 3) The left end of each Boost capacitor is connected to The level is evenly distributed between 0 and V REF / N1, The right end of the boost capacitor is disconnected from V REF _2 nd The connection is directly connected to the input of the sense amplifier, which in turn raises the reference voltage to V REF _2 nd [i]+V REF _2 nd The thermometer decoder compares the input signal with this reference voltage and obtains Thermometer code and converted into binary code OUT_2 nd [N1:1]; 4) Repeat steps 2) to 3) to change the voltage value stored in Boost in sequence, so that its voltage value gradually approaches the input voltage V IN , until the last step output OUT_N2 is obtained th [N1:1].

Citation Information

Patent Citations

  • Low-power-consumption successive approximation type analog-to-digital converter and analog-to-digital conversion method

    CN111181563A

  • Distributed successive approximation analog-to-digital converter and operation method thereof

    CN115664422A

  • Analog-to-digital converter and working method

    CN118118018A

  • Self-calibration method and device for pipeline successive approximation type analogue to digital convertor

    WO2015154671A1

Cited By

  • Multi-step analog-to-digital converter sharing reference voltage and operation method of multi-step analog-to-digital converter

    CN119171907A

  • A multi-step analog-to-digital converter sharing a reference voltage and an operation method thereof

    CN119171907B