A multi-channel noise shaping data converter based on code division multiplexing

Abstract

This invention discloses a multi-channel noise-shaping data converter based on code division multiplexing, comprising an encoder, a chopper switch array, an adder, a data converter, and a decoder. This invention modulates the multi-channel input signal using the encoder, enabling the analog multi-channel input signal to be converted into data by a single-channel analog-to-digital converter. Through the coherent oversampling strategy proposed in this invention, high-precision analog-to-digital conversion and demodulation of each channel signal can be achieved. The data converter in this invention can achieve noise shaping characteristics, suppressing noise within the bandwidth. To achieve noise shaping, the data converter oversamples the analog signal, suppressing noise within the effective bandwidth and shaping it to other frequency bands. Simultaneously, the oversampling rate of the control circuit is 2. N This configuration can effectively avoid aliasing between channels and interference from out-of-band noise, enabling the system to be applied to multi-channel code division multiplexing noise shaping data converters.

Description

Technical Field

[0001] This invention belongs to the field of CMOS bias circuit technology, specifically relating to a multi-channel noise shaping data converter based on code division multiplexing. Background Technology

[0002] An analog-to-digital converter (ADC) is an electronic device that converts analog signals into digital signals. Since most physical signals in nature exist in the form of analog signals, and computer systems are good at processing digital signals efficiently, ADCs are indispensable for efficient processing of physical signals.

[0003] To improve the efficiency of analog-to-digital converters (ADCs) and reduce hardware overhead, interface circuits typically employ multi-channel multiplexed ADCs, where a single ADC simultaneously converts analog information acquired from multiple channels. This results in various multiplexing methods, such as frequency division multiplexing (FDM), time division multiplexing (TDM), and code division multiplexing (CDM). Benefiting from the CDM scheme's ability to utilize the same carrier wave and bandwidth for simultaneous transmission and reception, it is widely used in multi-channel data converters to improve the utilization efficiency of both the time and frequency domains. The basic idea of ​​CDM is to modulate signals from different channels using different, mutually orthogonal address codes. The final result is that the signals are modulated to different frequency points or into orthogonal phase states at the same frequency point. At the receiving end, to demodulate information from a specific channel, only the address code of that channel needs to be used for demodulation to obtain information unaffected by interference from other channels. For example, the capacitance-to-digital converter system based on code division multiplexing proposed in the literature [Luo Y, Heng CH. An 8.2w 0.14mm 2 16-Channel CDMA-Like Period Modulation Capacitance-to-Digital Converter with Reduced Data Throughput[C] / / 2018IEEE Symposium on VLSICircuits.IEEE,2018] adopts a code division multiplexing scheme to realize the simultaneous conversion of 16 channels of capacitance information by a single interface circuit, thereby improving the efficiency of the interface circuit and reducing hardware overhead; however, this scheme does not process in-band noise, resulting in limited accuracy of the final converted digital information.

[0004] To significantly improve the conversion accuracy of analog-to-digital converters (ADCs), noise-shaped ADCs have been proposed. Noise-shaped ADCs typically possess two characteristics: oversampling and noise shaping. First, oversampling occurs when the ADC's sampling frequency exceeds the Nyquist sampling rate. The sampling circuitry in the ADC oversamples the input signal to extend the bandwidth. Second, noise shaping utilizes a control loop to process the input signal and quantization noise. A loop filter provides gain to signals in specific frequency bands, shifting noise introduced at the quantizer outside the effective bandwidth and improving the signal-to-noise ratio (SNR) within that specific frequency band during ADC conversion. For example, the fourth-order noise-shaping successive approximation analog-to-digital converter proposed in the literature [Wang T, Xie T, Liu Z, et al. An 84dB-SNDR Low-OSR 4th-Order Noise-Shaping SAR with an FIA-Assisted EF-CRFF Structure and Noise-Mitigated Push-Pull Buffer-in-Loop Technique[C] / / 2022IEEE International Solid-State Circuits Conference(ISSCC).IEEE,2022,65:418-420] can effectively quantize the in-band noise introduced by the quantizer, thereby improving the signal-to-noise ratio of the in-band signal; however, this scheme is single-channel, that is, one interface circuit can only process one channel of analog signal, which is inefficient and increases hardware overhead.

[0005] Based on the above two considerations, in order to simultaneously improve the conversion accuracy and multi-channel multiplexing characteristics of the data converter, a noise-shaping data converter based on a code division multiplexing multi-channel scheme can be adopted. However, since the demodulation process uses a square wave to multiply with the digital signal after analog-to-digital conversion, the higher harmonics of the square wave will fold the noise-shaped quantization noise back into the band. Simply performing multi-channel modulation can easily cause the noise suppressed in the conversion process to be shifted back into the band during demodulation, thereby affecting the in-band signal-to-noise ratio.

[0006] In summary, current multi-channel noise-shaping data transcoders with code division multiplexing schemes struggle to handle the relationship between noise shaping and code division multiplexed square wave demodulation, resulting in performance loss. Noting the strong correlation between the encoding frequency and oversampling rate of the quadrature encoder, a code division multiplexer for noise-shaping data transcoders is proposed. This multiplexer uses correlation sampling to ensure high-precision demodulation during multi-channel analog-to-digital conversion. Summary of the Invention

[0007] In view of the above, the present invention provides a multi-channel noise shaping data converter based on code division multiplexing, which has the characteristics of adapting to multi-channel noise shaping and the advantages of high precision, high frequency bandwidth utilization and multi-channel transmission.

[0008] A multi-channel noise-shaping data converter based on code division multiplexing includes an encoder, a chopper switch array, an adder, a data converter, and a decoder, wherein:

[0009] The encoder is used to convert the pre-assigned address code and generate control code signals corresponding to each channel to provide to the chopper switch array;

[0010] The chopper switch array controls the switching of each channel according to the control code signal, thereby modulating the input signal of each channel;

[0011] The adder is used to superimpose the multi-channel signals modulated by the chopper switch array, so that the superimposed signal becomes a single-channel analog signal.

[0012] The data converter synchronously converts a single-channel analog signal into a corresponding digital signal based on the principles of oversampling and noise shaping.

[0013] The decoder is used to convert the pre-assigned address code to generate demodulation code signals corresponding to each channel. Then, the demodulation code signals are used to demodulate the digital signal and complete subsequent downsampling and low-pass filtering to restore the data information of each channel.

[0014] Furthermore, the address code is based on orthogonal coding. The characteristics of orthogonal coding ensure that different address codes will not be correlated, that is, signals from different channels will not interfere with each other. Orthogonal coding also ensures that the modulation frequency bands of all channel signals are multiples of each other, that is, the signal will be modulated to f. c ,2f c , 4f c … Considering this characteristic, the sampling frequency set according to the above scheme will enable the data converter of the present invention to have coherent sampling. The advantage of coherent sampling is that the energy is maintained at the sampling frequency point and no spectral leakage occurs. Therefore, high-frequency noise will not be excessively reflected back into the band during sampling, thereby causing a decrease in the in-band signal-to-noise ratio.

[0015] Furthermore, the data converter includes a sampling circuit, a noise-shaping analog-to-digital converter, and a control logic circuit, wherein the sampling circuit is used to coherently sample the modulated and superimposed single-channel analog signal at a sampling frequency of 2. N f c,max N is the number of channels and N is a positive integer greater than 1, f c,maxThis represents the highest signal frequency among all channel input signals. The noise-shaping analog-to-digital converter circuit is used to synchronously convert the sampled single-channel analog signal into the corresponding digital signal, and it has noise-shaping characteristics. This scheme can effectively ensure that the multi-channel signals do not interfere with each other, and can avoid the noise being shaped to a high frequency and folded back into the effective bandwidth during the demodulation stage, thereby enabling the noise-shaping data converter to achieve high precision while supporting multiple channels.

[0016] Furthermore, the encoder generates a set of control code signals related to the address code based on the control signals provided by the control logic circuit in the data converter (which determine its working start time) and the current state. These signals are used to control the on / off state of each channel switch in the chopper switch array, thereby determining the current input state of each channel and thus realizing channel encoding.

[0017] Furthermore, the decoder includes a decoder logic module, a logic gate array, and a downsampling low-pass filter. The decoder logic circuit generates a set of demodulation code signals related to the address code based on the control signals (which determine its start time) provided by the control logic circuit in the data converter and the current state. These signals are then input to the logic gate array to control its demodulation of the digital signals. The downsampling low-pass filter is used to downsample and low-pass filter the demodulated digital signals of each channel, filtering out high-frequency noise and restoring the information of each channel.

[0018] Furthermore, the chopper switch array periodically changes the state of the channel switches according to the control code signal provided by the encoder, thereby performing chopping processing on the channel input signal according to the address code. The specific chopping processing methods include the following two categories:

[0019] ② Connect the ground or channel input signal to the circuit according to the address code "0" or "1";

[0020] ② For differential signals, the positive or negative channel input signal is connected to the circuit according to the address code "0" or "1".

[0021] Furthermore, the control logic circuit in the data converter adopts a Millitype state machine, which controls the operation of the sampling circuit and the noise shaping analog-to-digital converter on the one hand, and outputs corresponding control signals to the encoder and decoder on the other hand to control their operation.

[0022] Furthermore, the logic gate array is an XOR gate array.

[0023] Furthermore, the digital signal output by the data converter can be in bit stream form (PDM modulation) or in parallel multi-bit digital signal form (PCM modulation).

[0024] This invention modulates multi-channel input signals using an encoder, enabling analog multi-channel input signals to be converted into data by a single-channel analog-to-digital converter (ADC). Through the coherent oversampling strategy proposed in this invention, high-precision ADC conversion and demodulation of each channel signal can be achieved. The data converter in this invention can achieve noise shaping characteristics, suppressing noise within the bandwidth. To achieve noise shaping, the data converter oversamples the analog signal, suppressing noise within the effective bandwidth and shaping it to other frequency bands. Simultaneously, the oversampling rate (OSR) of the control circuit is 2. N This configuration can effectively avoid aliasing between channels and interference from out-of-band noise, enabling the system to be applied to multi-channel code division multiplexing noise shaping data converters.

[0025] This invention enables simultaneous data conversion of multi-channel analog information on a noise shaping data converter. The set oversampling rate allows high-frequency harmonic energy to be reflected back to the observation frequency band, ensuring no crosstalk between channels. Furthermore, the oversampling rate ensures coherent sampling, avoiding spectral leakage. This reduces the reflection of high-frequency shaping noise during demodulation, preventing out-of-band noise reflection from significantly impacting the in-band signal-to-noise ratio and maintaining a high in-band signal-to-noise ratio. Thus, it combines multi-channel multiplexing with noise shaping data conversion technology. Attached Figure Description

[0026] Figure 1 This is a system structure block diagram of the multi-channel noise shaping data converter of the present invention.

[0027] Figure 2 This is a schematic diagram of chopping operation for a single channel signal.

[0028] Figure 3 This is a schematic diagram illustrating a specific implementation of the data converter of the present invention.

[0029] Figure 4 This is a schematic diagram of the address encoding used in an embodiment of the data converter of the present invention.

[0030] Figure 5 (a) is a spectrum diagram of the digital signal portion converted by the data converter in an embodiment of the present invention.

[0031] Figure 5 (b) is a schematic diagram of the process by which out-of-band energy is returned to the observation frequency band after sampling at a frequency of fs in an embodiment of the data converter of the present invention.

[0032] Figure 6 This is a schematic diagram of the sampling process in the time domain of an embodiment of the data converter of the present invention.

[0033] Figure 7(a) is a schematic diagram of the demodulation results after adopting the sampling frequency scheme of the present invention.

[0034] Figure 7(b) is a schematic diagram of the noise return caused by not using the sampling frequency scheme of the present invention.

[0035] Figure 7(c) is a schematic diagram of the channel crosstalk caused by not using the sampling frequency scheme of the present invention. Detailed Implementation

[0036] To describe the present invention in more detail, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0037] Figure 1 The diagram shows the system structure of the multi-channel noise shaping data converter of the present invention. Signals from N channels are input respectively. The control logic circuit in the data converter outputs an encoding control signal CS to the encoder. After receiving the encoding control signal CS, the encoder outputs a set of switch control signals. These switch control signals are related to the pre-assigned address codes and control the on / off state of the chopper switches of the N channels respectively. The chopping is specifically manifested by multiplying the input signal with the chopping frequency, so that the input signal is modulated to a high frequency.

[0038] The chopping-processed signals from each channel are summed and then sent to a noise-shaping analog-to-digital converter (ADC) via a sampling circuit. The sampling circuit synchronously samples the data from multiple channels, and the on / off frequency of the sampling switch is the sampling frequency of the data converter. In this invention, the sampling frequency is set to 2 times the highest data rate of all channels. N The sampling switch is controlled by the sampling control signal fs output by the control logic circuit. The sampled signal is converted from analog to digital by a noise-shaping analog-to-digital converter. The output digital signal can be in multi-bit parallel form or bitstream form. The output digital signal includes information from N channels. To obtain the information of a specific channel, demodulation is required. The demodulation process is similar to the modulation process. Since the signal being processed is digital, an XOR gate processing scheme is used instead of chopping to achieve frequency multiplication. The decoder logic module outputs control logic to the XOR gate array, causing the signal to be multiplied with the orthogonal code of that channel, thereby recovering the signal of that channel. Due to oversampling and noise shaping, in order to obtain the digital signal corresponding to the initial analog signal, it is necessary to pass downsampling filtering and low-pass filtering to restore the signal frequency to the initial state and filter out high-frequency shaping noise. The final output digital signal is the result corresponding to that channel.

[0039] The address codes for each channel are designed during the encoder design process and will not be changed subsequently. There are multiple possible configurations for the address codes, but the basic requirement is orthogonal encoding. Orthogonal encoding ensures that different address codes are not correlated, meaning there will be no crosstalk between signals from different channels. Orthogonal encoding also ensures that the modulation frequency bands of all channel signals are multiples of each other, meaning the signal will be modulated to f... c ,2f c , 4f c …Considering this characteristic, the sampling frequency set according to the above scheme will enable the scheme to have coherent sampling. The advantage of coherent sampling is that the energy is maintained at the sampling frequency point and no spectral leakage occurs. Therefore, high-frequency noise will not be excessively reflected back into the band during sampling, thus causing a decrease in the in-band signal-to-noise ratio.

[0040] Figure 2 The diagram illustrates the chopping operation of a single channel signal. The two input switch control signals are non-overlapping, and the two codewords correspond to two different switch topologies, as shown in the diagram. Figure 2 As shown in the lower middle section, these correspond to in-phase and out-of-phase inputs, respectively. When the control code word is 1, the output Vout is IN, and when the control code word is -1, the output Vout is -IN.

[0041] Example

[0042] like Figure 3 As shown, this embodiment is a four-channel code division multiplexing noise shaping data converter. The two inputs of the chopper switch array are analog signal and ground, respectively. When the system starts working, the data converter logic circuit first sends a CS valid signal to the encoder. The encoder outputs a set of codewords to the chopper switch array according to the current state, controlling the chopper switch connection of the four channels respectively. This example uses the simplest orthogonal coding, where the frequency of each set of codewords is twice that of the previous set. The data converter logic circuit also sends an fs valid signal simultaneously to control the sampling switch to close, so that the four channel inputs are sampled and held in the data converter. Then, under the control of the data converter logic circuit, the data converter converts the sampled analog signal. After the conversion is completed, the resulting digital signal enters the decoder. The decoder logic module performs an AND operation between the codeword of the corresponding frequency transmitted by the channel and the converted digital signal, modulating the information of that channel back to a low frequency. Then, after downsampling and low-pass filtering, the digital result of the corresponding channel is obtained.

[0043] like Figure 4As shown, this example uses Walsh code to encode different channels. Other encoding methods can also be used, such as Goledn code, pseudo-random code, etc. According to Walsh code, the encoder generates a four-bit codeword at each clock cycle. For example, at t1, it generates [1, 1, 1, 1], at t2, it generates [1, -1, 1, -1], and so on. Based on the four-bit codeword, the chopper switches of the four channels are controlled respectively, thereby realizing the encoding of the channel signals.

[0044] like Figure 5 (a) shows a partial spectrum of the digital signal converted by the data converter in this example. Since chopping is equivalent to multiplying the signal with a square wave in the time domain, the frequency domain is represented by the convolution of the signal spectrum and the square wave spectrum. The maximum energy exists at the fundamental frequency fc of the square wave, while the energy decreases by the square law at the odd harmonics of the fundamental frequency. Figure 5 (b) illustrates the process by which out-of-band energy is folded back into the observation band after sampling at frequency fs, since fs is 2 times the maximum channel data rate. 4 Therefore, fs is 16f. c4 From the properties of the orthogonal codes used, we can know f c4 =2f c3 =4f c2 =8f c1 Consider the harmonics that will bounce back to the fundamental frequency, i.e. Figure 5 (b) The four harmonics shown on the right have a difference of f from the sampling frequency point fs, respectively. c1 f c2 f c3 f c4 Therefore, when the harmonics return to their respective channels, they will return to the fundamental frequency of their respective channels. That is, the harmonics of channel one will only return to the fundamental frequency of channel one, the harmonics of channel two will only return to the fundamental frequency of channel two, and the situation for channels three and four is similar. This achieves non-crosstalk between channels.

[0045] like Figure 6 As shown, taking the input signals of channels three and four as examples, the white circles indicate that the data converter will sample the input signal at this time. Since the sampling frequency is a multiple of the signal frequency and is 16 times the highest channel data rate, according to the characteristics of orthogonal coding, the sampling frequency is 128 times the lowest channel data rate. Therefore, by controlling the sampling to 128 points each time, the signal sample value of the complete cycle can be obtained on each channel, and there will be no truncated signal that is not the number of cycles of the signal. Figure 6The input signal frequency of channel three is 1 / 32 of the sampling frequency, so sampling 128 points can complete exactly 4 cycles. Similarly, the input signal frequency of channel four is 1 / 16 of the sampling frequency, so sampling 128 points can complete exactly 8 cycles. The same situation occurs in all channels of this example, thus ensuring that the sampling process is coherent and does not cause spectral leakage. Therefore, it does not reflect out-of-band high-frequency noise back into the band, ensuring a high signal-to-noise ratio within the band.

[0046] Figures 7(a) to 7(c) The three spectrograms shown are simulation results of the same noise shaping model, the only difference being the sampling frequency in each system. The noise shaping model has four input channels, with each channel's bandwidth set to 1–1 kHz. To verify the effect of crosstalk, the sine wave frequencies transmitted through each channel are 400 Hz, 500 Hz, 600 Hz, and 700 Hz, respectively, and the chopping frequencies for each channel are 10 kHz, 20 kHz, 40 kHz, and 80 kHz, respectively. Therefore, according to the sampling frequency setting proposed in this invention, it should be 80 kHz × 2. 4 =1280kHz; Figure 7(a) shows the spectrum of the system using this sampling frequency after demodulating the data of channel one. It can be seen that within the bandwidth, the effective number of bits of the corresponding demodulated signal of channel one can reach 11.57 bits; The simulation system corresponding to Figure 7(b) has a system sampling frequency of 1278.752kHz, which also meets the Nyquist sampling requirements. It can be seen from the spectrum that the demodulated signal is still not affected by the signals of other channels. However, due to the failure to meet the coherent sampling requirements, there is a large amount of noise from the high-frequency band within the bandwidth, which causes the signal-to-noise ratio of the signal to drop sharply, with the effective number of bits only 8.93 bits; The simulation system corresponding to Figure 7(c) has a system sampling frequency of 1290kHz, which meets the Nyquist sampling requirements. Since this sampling rate meets the coherent sampling requirements for the chopping frequency of channel one, there is no obvious noise within the bandwidth. However, because the signals of other channels are folded back into the band at this time, signal crosstalk occurs, and the signal-to-noise ratio of the signal is still affected, with the effective number of bits dropping to 5.88 bits.

[0047] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. Those skilled in the art can readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.

Claims

1. A code division multiplex based multichannel noise shaper data converter, characterized by, It includes an encoder, a chopper switch array, an adder, a data converter, and a decoder, among which: The encoder is used to convert the pre-assigned address code and generate control code signals corresponding to each channel to provide to the chopper switch array; The chopper switch array controls the switching of each channel according to the control code signal, thereby modulating the input signal of each channel; The adder is used to superimpose the multi-channel signals modulated by the chopper switch array, so that the superimposed signal becomes a single-channel analog signal. The data converter synchronously converts a single-channel analog signal into a corresponding digital signal based on oversampling and noise shaping principles. The data converter includes a sampling circuit, a noise-shaped analog-to-digital converter, and a control logic circuit. The sampling circuit performs coherent sampling on the modulated and superimposed single-channel analog signal at a sampling frequency of [frequency missing]. , N The number of channels and N It is a positive integer greater than 1. This represents the highest signal frequency among all channel input signals; the noise shaping analog-to-digital converter is used to synchronously convert the sampled single-channel analog signal into the corresponding digital signal, and has noise shaping characteristics; The decoder is used to convert the pre-assigned address codes to generate demodulated code signals corresponding to each channel. These demodulated code signals are then used to demodulate the digital signal and perform subsequent downsampling and low-pass filtering to restore the data information of each channel. The address codes are based on orthogonal coding, which ensures that the modulation frequency bands of all channel signals are in a multiple relationship, meaning the signal will be modulated to... , This is the fundamental frequency.

2. The multi-channel noise shaping data converter according to claim 1, characterized in that: The encoder generates a set of control code signals related to the address code based on the control signals provided by the control logic circuit in the data converter and the current state. These signals are used to control the on / off state of each channel switch in the chopper switch array, thereby determining the current input state of each channel and thus realizing channel encoding.

3. The multi-channel noise shaping data converter according to claim 1, characterized in that: The decoder includes a decoder logic module, a logic gate array, and a downsampling low-pass filter. The decoder logic circuit generates a set of demodulation code signals related to the address code based on the control signals provided by the control logic circuit in the data converter and the current state. These signals are then input to the logic gate array to control the demodulation of the digital signals. The downsampling low-pass filter is used to downsample and low-pass filter the digital signals of each channel after demodulation, filtering out high-frequency noise and restoring the information of each channel.

4. The multi-channel noise shaping data converter according to claim 1, characterized in that: The chopper switch array periodically changes the state of the channel switches according to the control code signal provided by the encoder, thereby performing chopping processing on the channel input signal according to the address code. The specific chopping processing methods include the following two categories: ① Connect the ground or channel input signal to the circuit according to the address code "0" or "1"; ② For differential signals, the positive or negative channel input signal is connected to the circuit according to the address code "0" or "1".

5. The multi-channel noise shaping data converter according to claim 1, characterized in that: The control logic circuit in the data converter adopts a Milli-type state machine, which controls the operation of the sampling circuit and the noise shaping analog-to-digital converter on the one hand, and outputs corresponding control signals to the encoder and decoder on the other hand to control their operation.

6. The multi-channel noise shaping data converter according to claim 3, characterized in that: The logic gate array is an XOR gate array.

7. The multi-channel noise shaping data converter according to claim 1, characterized in that: The digital signal output by the data converter can be in bitstream form or in parallel multi-bit digital signal form.

Images