Variable bit number pipeline analog-to-digital converter bit number reconfiguration method and implementation circuit
By using a reconfigurable pipelined ADC architecture and digital registers to control the multi-input selector and the shutdown of the sub-ADC stage, the bit depth of the pipelined analog-to-digital converter can be reconfigured. This solves the problem of high design costs for different bit depths, reduces R&D and management costs, and improves flexibility and power efficiency.
Patent Information
- Application Number
- CN202611586279.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-14
- Publication Date
- 2026-11-10
AI Technical Summary
Existing pipelined analog-to-digital converters (ADCs) require independent design and tape-out for different quantization bit depths, resulting in high R&D costs, high management costs, and slow response speeds.
A reconfigurable pipelined ADC architecture is adopted, which controls the shutdown of the multi-input selector and the sub-ADC stage through digital registers to achieve the conversion of different quantization bit widths. By utilizing the multi-bit output of the high-weight sub-ADC and the single output of the low-weight sub-ADC, combined with differential clock and alternating clock design, bit width reconstruction is achieved.
By enabling the conversion of different quantization bits on the same chip design, we can reduce R&D and tape-out costs, reduce power consumption, simplify control logic, maintain signal amplitude consistency, and avoid modifications to complex analog circuits.
Smart Images

Figure CN123074504A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of integrated circuit design technology, specifically a method and implementation circuit for reconfiguring the bit depth of a pipelined analog-to-digital converter with variable bit depth. Background Technology
[0002] Pipeline analog-to-digital converters (ADCs) are the most common architecture for achieving high-speed, high-precision analog-to-digital conversion. Their two most critical metrics are conversion speed and quantization bit depth. Existing commercial product catalogs reveal that pipelined ADCs, for a given conversion speed, typically offer a range of models covering different bit depths. For instance, when the conversion speed is consistently 100 MSps, the quantization precision usually exhibits an even-numbered distribution, such as 12-bit, 14-bit, and 16-bit. Due to the inherent characteristics of pipelined architectures, mature products generally have a quantization bit depth ranging from 8 to 16 bits.
[0003] In the traditional integrated circuit design model, products with different quantization bits at the same conversion speed usually need to be designed and fabricated independently. This approach has the following drawbacks: (1) Each bit specification requires independent design, verification and fabrication, resulting in a long design cycle and high R&D costs; (2) Products with different bit specifications need to maintain separate product lines and inventory, resulting in high management costs; (3) The response speed is slow when facing customers' flexible needs for different bit specifications. Summary of the Invention
[0004] On the one hand, the present invention provides a variable bit-number pipelined analog-to-digital converter bit-number reconstruction method to solve the problem that existing pipelined ADCs with different quantization bit numbers need to be designed independently and have high R&D costs.
[0005] The technical solution to achieve the objective of this invention is: a variable-bit pipelined analog-to-digital converter bit reconfiguration method, comprising: A basic architecture for a reconfigurable pipelined ADC is constructed, comprising M stages of sub-ADCs with different weights, a final stage of L-bit sub-ADCs, and a clock distribution module. Each sub-ADC stage with high weights uses several bits of quantization output, while each sub-ADC stage with low weights uses 1 bit of quantization output. The clock distribution module outputs differential clocks to each sub-ADC, with adjacent sub-ADC stages using opposite clocks. A multi-input gate is set at the input terminal of the last L-position sub-ADC. Each input of the multi-input gate is connected to the residual output terminal of the sub-ADC of different levels in the low-weight bit, and a switch is set on each path. When it is necessary to reduce the overall quantization bit depth of the ADC, the low-weight bits are used in reverse order to turn off the adjacent H-level low-weight sub-ADCs (where H is an even number). At the same time, the switches on each path of the multi-input selector are controlled so that only the residual output of the previous sub-ADC of the turned-off stage is directly selected to the input of the last sub-ADC.
[0006] On the other hand, the present invention provides an implementation circuit for a variable-bit pipelined analog-to-digital converter bit reconfiguration method, comprising: M is a sub-ADC with different weights, where the i-th stage is a high-weight sub-ADC, each stage uses multi-bit quantization output, and the Mi-th stage is a low-weight sub-ADC, each stage uses 1-bit quantization output. The final L-position ADC; The clock distribution module is used to output differential clocks to each sub-ADC, with adjacent sub-ADCs using opposite clocks; A multi-input gate is set at the input of the last stage sub-ADC, and each input is connected to the residual output of a different stage of the sub-ADC in the low-weight bit. The digital register stores configuration words used to control the on / off state of the multi-input selector and the off state of the sub-ADC stage. When it is necessary to reduce the overall quantization bit depth, the adjacent H-stage low-weight sub-ADC is turned off by the digital register and the corresponding multi-input selector input is turned on. H is an even number, and the residual output of the stage before the stage to be turned off is directly connected to the input of the last stage sub-ADC.
[0007] Compared with the prior art, the significant advantages of this invention are: In this invention, the high-weight sub-ADC adopts a single-stage multi-bit output to reduce the overall number of stages and ensure low power consumption; the low-weight sub-ADC adopts a single-stage 1-bit output to facilitate bit reconstruction.
[0008] When reconstructing the number of bits, this invention uses 2 bits as the actual reconstruction step, and does not require changing the clock phase of each stage during bit reconstruction.
[0009] Based on the same chip design, different quantization bit widths can be achieved by configuring digital registers, avoiding the need to design separate chips for each bit width specification and significantly reducing R&D and tape-out costs; In the lower bit-level operating mode, the disabled sub-ADC stages do not consume power, and the overall power consumption is reduced accordingly.
[0010] Bit reconstruction is achieved solely through configuring the MUX switch and turn-off control via digital registers, requiring no complex modifications to analog circuits; the digital control logic is simple and reliable. The amplitude range of the residual output signals at each stage remains consistent, allowing the multi-input selector to select outputs at different stages arbitrarily without additional signal conditioning circuitry.
[0011] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description
[0012] Figure 1 This is a schematic diagram of a pipelined ADC architecture.
[0013] Figure 2 This is a schematic diagram of the bit reconstruction architecture.
[0014] Figure 3 This is a schematic diagram of the MUX architecture.
[0015] Figure 4 It is a 14-bit ADC architecture with reconfigurable bit depth.
[0016] Figure 5 To reconstruct a 12-bit ADC architecture.
[0017] Figure 6 To reconstruct a 10-bit ADC architecture.
[0018] Figure 7 A schematic diagram illustrating the clock phase problem when reconstructing 2 bits per stage. Detailed Implementation
[0019] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.
[0020] The specific concept of a variable-bit-width pipelined analog-to-digital converter (ADC) bit-width reconfiguration method is as follows: Based on a pipelined ADC architecture, and using the same design, circuit conversion between different bit widths is achieved by configuring digital registers, without requiring a completely new chip design for different bit widths. Based on this invention, it is possible to achieve the goal of one design with multiple outputs within a certain bit width range.
[0021] A method for reconfiguring the bit depth of a pipelined analog-to-digital converter with variable bit depth includes: S1: Construct the basic architecture of a reconfigurable pipelined ADC. For example... Figure 1 As shown, the basic architecture of the reconfigurable pipelined ADC includes M sub-ADCs with different weights, a final L-bit sub-ADC, and a clock distribution module. After the input analog signal passes through the sample-and-hold circuit (S / H) of the first-stage sub-ADC, it is coarsely quantized by the sub-ADC to obtain a number of digital codes. At the same time, the sub-DAC reconstructs the quantization result into an analog signal, which is subtracted from the original input signal to obtain the residual signal. The residual signal is amplified by the residual amplifier and then passed to the next stage sub-ADC for further processing.
[0022] In the M-level sub-ADC with different weights, level i is the high-weighted bit (corresponding to the higher bits in the digital output, closer to the most significant bit (MSB), and level Mi is the low-weighted bit (corresponding to the lower bits in the digital output, closer to the least significant bit (LSB)). Each high-weighted sub-ADC level uses multi-bit quantization output, such as 2 or 3 bits per level, while each low-weighted sub-ADC level uses 1-bit quantization output. The high-weighted sub-ADC level determines the high-bit precision of the conversion result, and the low-weighted sub-ADC level determines the low-bit precision. It should be noted that "high weight" and "low weight" are relative to the weighted bits of the output result of each sub-ADC level in the digital output.
[0023] The clock distribution module outputs differential clocks clkp and clkn to each sub-ADC. Adjacent sub-ADC stages in the pipeline use opposite clocks: odd-numbered stages use clkp, and even-numbered stages use clkn. This alternating clock design is a fundamental timing requirement for pipelined ADCs—adjacent sub-ADC stages operate in the sampling and hold phases respectively. By alternating the use of clocks with opposite phases, the signal is ensured to be pipelined sequentially. When a sub-ADC stage is in the sampling phase, the next stage is in the hold phase, thus achieving continuous pipelined conversion.
[0024] S2: Set a multi-input gate (MUX) at the input of the last stage L-position sub-ADC, such as... Figure 3 As shown, each input of the multi-input selector is connected to the residual output of a different stage of the sub-ADC in the low-weighted bit. A digital register controls the switches corresponding to the multiple inputs, ensuring that only one switch is active at any given time. This selects the corresponding input to a single output and connects it to the input of the last stage sub-ADC. Because the amplitude range of the residual output signals at each stage is consistent, and the input signals of the multi-input selector have the same amplitude range, the last stage sub-ADC will function normally when any input is selected to the last stage.
[0025] Specifically, the digital register is configured to the default value of the highest bit operating mode when powered on and reset, so that all sub-ADC stages are in working state.
[0026] In a further embodiment, the last-stage sub-ADC has the lowest accuracy requirement among all pipeline sub-ADCs; therefore, the multi-input selector is implemented using a CMOS transmission gate switch consisting of a PMOS and an NMOS transistor connected in parallel. If higher accuracy is required, a gate-voltage bootstrap switch can be used instead.
[0027] S3: As Figure 2As shown, when it's necessary to reduce the overall quantization bit depth of the ADC, the low-weight bits are turned off from back to front using corresponding control bits, with H being an even number. This is because the later the sub-ADC stage in a pipelined ADC, the lower its impact on accuracy. Therefore, when it's necessary to reconstruct the overall bit depth, the low-weight bits are turned off from back to front. Figures 4-6 In this case, shutting down the ADC in the order of levels 8, 7, 6, and 5 ensures that the overall performance of the ADC is not affected.
[0028] The shutdown operation includes: cutting off the clock input of the shutdown sub-ADC by clock gating to stop its operation; shutting off the bias current source of the shutdown sub-ADC by power gating to prevent it from consuming static power; and disconnecting the analog signal path between the shutdown sub-ADC and other sub-ADC stages. Simultaneously, a digital register controls a multi-input selector to directly select the residual output of the preceding sub-ADC stage to the input of the final sub-ADC stage. Since shutting down two stages maintains the clock cycle between the remaining operating stages, the pipeline timing remains normal. The shutdown sub-ADC stage no longer consumes any power, and the overall quantization bit depth is reduced by 2 bits. Simultaneously, the digital correction logic automatically adjusts the bit alignment of the output data according to the configuration word in the digital register, and the digital output bits corresponding to the shutdown stage are masked or set to zero. For example, stages M and M-1, while directly connecting the output of stage M-2 to the input of the final stage, reduce the overall quantization bit depth of the ADC by 2 bits. At this time, the M and M-1 sub-ADC stages are in a shutdown state and no longer consume any power.
[0029] Pipeline ADCs require that the clocks of adjacent working stages be out of phase. For example... Figure 7 As shown, if only one sub-ADC stage is turned off, the adjacent sub-ADC stages before and after it will become directly adjacent, and their clock phases will be the same (both clkp or both clkn), causing the sample / hold operations of the two stages to no longer alternate, thus disrupting the pipeline timing. Therefore, in this invention, the low-weight bits are designed to be 1 bit per stage, and 2 stages are turned off each time (reducing the number of bits by 2), which precisely ensures that the clock alternation relationship between the remaining working stages is not disrupted.
[0030] This invention also proposes an implementation circuit for a variable-bit pipelined analog-to-digital converter bit reconfiguration method, comprising: M is a sub-ADC with different weights, where the i-th stage is a high-weight sub-ADC, each stage uses multi-bit quantization output, and the Mi-th stage is a low-weight sub-ADC, each stage uses 1-bit quantization output. The final L-position ADC; The clock distribution module is used to output differential clocks to each sub-ADC, with adjacent sub-ADCs using opposite clocks; A multi-input gate is set at the input of the last stage sub-ADC, and each input is connected to the residual output of a different stage of the sub-ADC in the low-weight bit. The digital register stores configuration words used to control the on / off state of the multi-input selector and the off state of the sub-ADC stage. When it is necessary to reduce the overall quantization bit depth, the adjacent H-stage low-weight sub-ADC is turned off by the digital register and the corresponding multi-input selector input is turned on. H is an even number, and the residual output of the stage before the stage to be turned off is directly connected to the input of the last stage sub-ADC. Example
[0031] Taking a pipelined ADC with a maximum bit depth of 14 bits, which can be reconfigured to 12 bits or 10 bits as an example, the implementation method of the present invention will be described in detail.
[0032] like Figure 4 As shown, the top two stages are high-weighted bits, with 3 bits per stage; stages 3-8 each have 1 bit; and the last stage has 2 bits, for a total of 14 bits. The last stage has a 3-input MUX, connected to the outputs of stages 4, 6, and 8 respectively. In the default 14-bit configuration, sw3 is on, while sw1 and sw2 are off. Regarding the clock, starting from the first stage, adjacent stages alternate between clkp and clkn to ensure that each sub-ADC in the pipeline operates in the prescribed sequence.
[0033] When reconfiguring into a 12-bit ADC, turn off the 7th and 8th stage sub-ADCs and only turn on SW2, as shown below. Figure 5 At this point, the overall quantization bit depth is reduced by 2 bits compared to the initial value, resulting in a 12-bit ADC.
[0034] When reconfiguring into a 10-bit ADC, turn off the 5th, 6th, 7th, and 8th sub-ADCs, and only turn on SW1, as shown below. Figure 6 At this point, the whole system forms a 10-digit ADC.
Claims
1. A method for reconstructing the bit depth of a pipelined analog-to-digital converter with variable bit depth, characterized in that, include: A basic architecture for a reconfigurable pipelined ADC is constructed, comprising M stages of sub-ADCs with different weights, a final stage of L-bit sub-ADCs, and a clock distribution module. Each sub-ADC stage with high weights uses several bits of quantization output, while each sub-ADC stage with low weights uses 1 bit of quantization output. The clock distribution module outputs differential clocks to each sub-ADC, with adjacent sub-ADC stages using opposite clocks. A multi-input gate is set at the input terminal of the last L-position sub-ADC. Each input of the multi-input gate is connected to the residual output terminal of the sub-ADC of different levels in the low-weight bit, and a switch is set on each path. When it is necessary to reduce the overall quantization bit depth of the ADC, the low-weight bits are used in reverse order to turn off the adjacent H-level low-weight sub-ADCs (where H is an even number). At the same time, the switches on each path of the multi-input selector are controlled so that only the residual output of the previous sub-ADC of the turned-off stage is directly selected to the input of the last sub-ADC.
2. The variable-bit-width pipelined analog-to-digital converter bit-width reconstruction method according to claim 1, characterized in that, The interstage gain of each sub-ADC is 2, ensuring that the amplitude range of the residual output signal remains consistent across all stages.
3. The variable-bit pipelined analog-to-digital converter bit reconstruction method according to claim 1, characterized in that, The specific method for shutting down the adjacent H-level low-weight sub-ADC is as follows: cut off the clock input of the sub-ADC to be shut down by clock gating, so that the sub-ADC to be shut down stops working; cut off the bias current source of the sub-ADC to be shut down by power gating, so that the sub-ADC to be shut down does not consume static power. And disconnect the analog signal path between the shut-down sub-ADC and other sub-ADCs.
4. The variable-bit-count pipelined analog-to-digital converter bit reconstruction method according to claim 1, characterized in that, The multi-input gate is implemented using a CMOS transmission gate switch consisting of a PMOS and an NMOS connected in parallel.
5. The variable-bit pipelined analog-to-digital converter bit reconfiguration method according to claim 1, characterized in that, The digital register controls the switches corresponding to the multiple inputs, enabling the corresponding inputs to be selected to a single output terminal and connected to the input terminal of the last sub-ADC stage. The digital register stores a configuration word, which simultaneously controls the switching state of the multi-input selector and the shutdown control bit of the disabled sub-ADC stage, and only one switch is in the on state at any given time.
6. The variable-bit pipelined analog-to-digital converter bit reconfiguration method according to claim 1, characterized in that, When the sub-ADC is turned off, the digital output bits corresponding to the turn-off stage are masked or set to zero. The digital correction logic automatically adjusts the bit alignment of the output data according to the configuration word in the digital register to match the current effective stage and bit allocation.
7. The variable-bit-width pipelined analog-to-digital converter bit-width reconstruction method according to claim 1, characterized in that, The number of input channels of the multi-input selector determines the number of reconfigurable bit depths. The minimum granularity of bit depth reconstruction is 2 bits, meaning that each time the 2-stage 1-bit sub-ADC is turned off, the quantization accuracy is reduced by 2 bits.
8. The variable-bit pipelined analog-to-digital converter bit reconstruction method according to claim 1, characterized in that, In the M-level sub-ADC with different weights, the high-weight bits are at level i, and the low-weight bits are at level Mi.
9. The implementation circuit of the variable-bit-count pipelined analog-to-digital converter bit-reconfiguration method according to any one of claims 1 to 8, characterized in that, include: M is a sub-ADC with different weights, where the i-th stage is a high-weight sub-ADC, each stage uses multi-bit quantization output, and the Mi-th stage is a low-weight sub-ADC, each stage uses 1-bit quantization output. The final L-position ADC; The clock distribution module is used to output differential clocks to each sub-ADC, with adjacent sub-ADCs using opposite clocks; A multi-input gate is set at the input of the last stage sub-ADC, and each input is connected to the residual output of a different stage of the sub-ADC in the low-weight bit. The digital register stores configuration words used to control the on / off state of the multi-input selector and the off state of the sub-ADC stage. When it is necessary to reduce the overall quantization bit depth, the adjacent H-stage low-weight sub-ADC is turned off by the digital register and the corresponding multi-input selector input is turned on. H is an even number, and the residual output of the stage before the stage to be turned off is directly connected to the input of the last stage sub-ADC.
10. The implementation circuit of the variable-bit-count pipelined analog-to-digital converter bit-count reconstruction method according to claim 9, characterized in that, The high-weighted bits have 2 levels, with 3-bit quantization output per level; the low-weighted bits have 6 levels, with 1-bit quantization output per level; the last level is a 2-bit flash ADC; the multi-input selector is a 3-input selector, which is connected to the residual output of the low-weighted bit sub-ADCs of the 4th, 6th and 8th levels respectively; the total quantization precision is 14 bits, which can be reconstructed into 12 bits or 10 bits.
