A digital frequency domain near-infrared brain function acquisition circuit

Abstract

A digital frequency-domain near-infrared brain function acquisition circuit comprises: a pre-mixing analog front-end circuit for downconverting photocurrent and converting it into a voltage signal; an amplitude-to-digital conversion circuit, whose input is connected to the output of the analog front-end circuit and is used to digitize the amplitude information of the analog front-end output signal; and a phase-to-digital conversion circuit, whose input is connected to the output of the analog front-end circuit and is used to digitize the phase information of the analog front-end output signal. The power-saving pre-mixing analog front-end circuit has a gain of up to 141dBΩ, uses a Σ-Δ architecture phase detector to achieve a phase quantization noise of 0.0096°, and has a total power consumption of only 6.8 milliwatts, realizing an FD-NIRS readout chip that can operate at a low carrier frequency of 10.01024.

Description

Technical Field

[0001] The present invention belongs to the technical field of wearable brain-computer interfaces, and in particular relates to a digital frequency-domain near-infrared brain function acquisition circuit. Background Art

[0002] Near-infrared spectroscopy (NIRS) optical imaging is a non-invasive, non-ionizing neuroimaging technique for monitoring human cerebral cortical hemodynamics and has been widely used in functional brain imaging and brain-computer interfaces. Although many imaging modalities can be used, such as positron emission tomography (PET), electroencephalography (EEG), magnetoencephalography (MEG), magnetic resonance imaging (MRI), and ultrasound, NIRS-based optical imaging can be used to provide functional information of the cortex without the use of additional harmful radiation, radioactive materials, or bulky instruments. Therefore, NIRS can achieve brain imaging without any environmental restrictions, which is crucial in many wearable applications.

[0003] Continuous wave (CW), frequency domain (FD), and time domain (TD) are three different forms of near-infrared spectroscopy. CW-NIRS is very commonly used due to its simple implementation: constant-intensity light is injected into the brain, and changes in the amplitude of the emitted light are detected to obtain relative changes in the absorption coefficient. Because the light source does not require modulation and the readout circuit bandwidth only needs to cover ultra-low frequencies, CW-NIRS consumes only microwatts of power. However, CW-NIRS only measures the attenuation of light amplitude in tissue, making it insensitive to information about neural activity reflected by the time of flight of photons in the brain. Consequently, CW-NIRS cannot separate absorption from scattering effects or detect the absolute value of the absorption coefficient or the attenuation-scattering coefficient, which limits the ability to quantitatively compare measurements across time or between subjects. Furthermore, many neuroimaging applications focus on baseline and absolute changes in hemoglobin concentration. From this perspective, FD-NIRS is an upgraded version of CW-NIRS. FD-NIRS modulates the light source frequency to megahertz and detects both amplitude and phase, allowing it to measure the absolute values of the absorption coefficient and the attenuation-scattering coefficient. Since the FD-NIRS light source is modulated at high frequency, low-frequency ambient light can be filtered by the circuit, which is a great advantage compared to CW-NIRS or TD-NIRS technology because it makes FD-NIRS suitable for wearable applications in non-light-proof environments. In addition, compared with CW-NIRS, which only has amplitude information, FD-NIRS with phase information has higher deep sensitivity and spatial resolution. The third NIRS method is TD-NIRS, which uses a pulsed light source to emit light pulses with a duration of tens of picoseconds, and the time resolution of the detection instrument is in the sub-nanosecond level. Although TD-NIRS can measure information from deeper tissues of the brain than FD-NIRS and CW-NIRS, TD-NIRS is not wearable due to the bulky picosecond laser driver. Therefore, the present invention adopts the FD-NIRS technical route.

[0004] However, for today's advanced FD-NIRS chips, there are problems such as high power consumption of the front-end chip, low energy efficiency of the off-chip phase digital converter, high power consumption of the off-chip amplitude digital converter, high driving power consumption due to the high carrier frequency, and low integration.

[0005] 1. High power consumption of the front-end chip: The analog front-end used in Y. Miao and V. J. Koomson, “A CMOS-Based Bidirectional Brain Machine Interface System With Integrated fdNIRS and tDCS for Closed-Loop Brain Stimulation,” IEEE Transactions on Biomedical Circuits and Systems, vol. 12, no. 3, pp. 554–563, 2018, suffers from high power consumption. This is because the high-frequency photocurrent is first amplified by a transimpedance amplifier and converted into a voltage signal, then filtered by a filter. The frequency is then heterodyned by a mixer to reduce the frequency and amplify the delay. The high-bandwidth transimpedance amplifier and filter consume a significant amount of power, approximately 30 milliwatts.

[0006] 2. Low energy efficiency of off-chip phase-to-digital converters: Y. Miao and V. J. Koomson, “A CMOS-Based Bidirectional Brain Machine Interface System With Integrated fdNIRS and tDCS for Closed-Loop Brain Stimulation,” IEEE Transactions on Biomedical Circuits and Systems, vol. 12, no. 3, pp. 554–563, 2018. and C. C. Sthalekar, Y. Miao, and V. J. Koomson, “Optical Characterization of Tissue Phantoms Using a Silicon Integrated fdNIRS System on Chip,” IEEE Transactions on Biomedical Circuits and Systems, vol. 11, no. 2, pp. 279–286, 2017. In both papers, the analog front-end output is directly fed to a high-performance analog-to-digital converter, which then performs an off-chip Fourier transform to calculate the phase. This wastes power and reduces accuracy because the high-performance analog-to-digital converter requires a large oversampling clock and high bandwidth.

[0007] 3. The problem of high power consumption of off-chip amplitude-to-digital converters: Y. Miao and V. J. Koomson, “A CMOS-Based Bidirectional Brain Machine Interface System With Integrated fdNIRS and tDCS for Closed-Loop Brain Stimulation,” IEEE Transactions on Biomedical Circuits and Systems, vol. 12, no. 3, pp. 554–563, 2018. and C. C. Sthalekar, Y. Miao, and V. J. Koomson, “Optical Characterization of Tissue Phantoms Using a Silicon Integrated fdNIRS System on Chip,” IEEE Transactions on Biomedical Circuits and Systems, vol. 11, no. 2, pp. 279–286, 2017. In both articles, the front-end output is directly fed to a high-performance analog-to-digital converter, which then demodulates the amplitude off-chip. This wastes power because the high-performance analog-to-digital converter requires a large oversampling clock and high bandwidth.

[0008] 4. The problem of high driving power consumption caused by high carrier frequency: Due to the insufficient resolution of the phase digital converter, it is necessary to increase the carrier frequency to increase the input phase signal power, but this causes the bandwidth of the light source driver to increase, which significantly increases the driving power consumption.

[0009] 5. Low integration: Y.Miao and VJKoomson, “A CMOS-Based Bidirectional Brain Machine Interface System With Integrated fdNIRS and tDCS for Closed-Loop Brain Stimulation,” IEEE Transactions on Biomedical Circuits and Systems, vol.12, no.3, pp.554–563, 2018. and C.C.Sthalekar, Y.Miao, and VJKoomson, “Optical Characterization of Tissue Phantoms Using a Silicon Integrated fdNIRS System on Chip,” IEEE Transactions on Biomedical Circuits and Systems, vol.11, no.2, pp.279–286, 2017. The digital part in both articles is an off-chip high-performance ADC, which reduces the integration and is not conducive to the implementation of wearable systems. Summary of the Invention

[0010] The technical problems solved by the present invention are:

[0011] The current FD-NIRS chip has problems such as high power consumption of the front-end chip, low energy efficiency of the off-chip phase digital converter, high power consumption of the off-chip amplitude digital converter, high carrier frequency resulting in high driving power consumption and low integration.

[0012] The technical solution of the present invention is:

[0013] A digital frequency domain near-infrared brain function acquisition circuit is characterized by including:

[0014] The analog front-end circuit before the mixer is used to down-convert the photocurrent and convert it into a voltage signal;

[0015] an amplitude-to-digital conversion circuit, whose input terminal is connected to the output terminal of the analog front-end circuit and is used to digitize the amplitude information of the analog front-end output signal; and

[0016] The phase-to-digital conversion circuit has an input end connected to the output end of the analog front-end circuit and is used to digitize the phase information of the analog front-end output signal.

[0017] First, the 10.01024MHz photocurrent is filtered through a DC-blocking capacitor to remove the DC component before being demodulated by a mixer with a 10MHz local oscillator frequency. A bandpass transimpedance amplifier then performs current-to-voltage conversion and extracts the lower sideband frequency. This leaves only the 10.24kHz lower sideband signal. The mixed lower sideband signal retains the amplitude and phase information of the original signal and is sent to the subsequent amplitude-to-digital converter and phase-to-digital converter for digitization.

[0018] The pre-mixing analog front-end circuit shifts the transimpedance amplifier's operating bandwidth from the high-band (10MHz-100MHz) to the low-band (10.24kHz) via the pre-mixer. This reduces the required operating bandwidth of the subsequent transimpedance amplifier and lowers the power consumption of the entire analog front-end. This pre-mixing analog front-end architecture not only increases the transimpedance amplifier's gain (5MΩ to 25MΩ) but also significantly reduces the power consumption required to maintain the transimpedance amplifier's bandwidth. This solves the problem of high front-end chip power consumption.

[0019] The phase digitizer utilizes a phase-domain Σ-Δ structure with first-order noise shaping. This leverages the naturally high oversampling ratio between the intermediate frequency (10.24kHz) and the desired phase signal bandwidth (10Hz) to shape quantization noise into a high-frequency band, reducing in-band phase quantization noise. Furthermore, this phase digitizer boasts a simple structure and operates at only 10.24kHz, eliminating the need for an additional high-frequency clock and further reducing power consumption in the FD-NIRS readout circuit. The on-chip phase digitizer also enhances integration.

[0020] The amplitude-to-digital converter described above can classify FD-NIRS signals into two categories based on their frequency band: slowly varying DC signals representing fundamental tissue characteristics, and AC signals such as heart rate (~1-4 Hz), respiration (~0.2 Hz), and neuro-oximetry (~0.1 Hz). Typically, the amplitude of an AC signal is approximately 1% to 2% of that of a DC signal. In this chip, the 10.24 kHz intermediate frequency signal undergoes amplitude envelope extraction, then is separated and amplified independently according to the AC and DC signal frequency bands. Two identical time-domain analog-to-digital converters (ADCs) with enhanced linearity are then used for digital-to-analog conversion to improve detection of weak AC components. The on-chip amplitude-to-digital converter with envelope demodulation saves power and increases integration.

[0021] The beneficial effects of the present invention compared with the prior art are:

[0022] 1) Using a delta-sigma phase-to-digital converter provides high phase detection accuracy, which can reduce the carrier frequency to save drive bandwidth and reduce drive power consumption. The carrier frequency can be reduced to 10.01024MHz to save power.

[0023] 2) A FD-NIRS readout chip capable of operating at a low carrier frequency of 10.01024 kHz was developed. The low-power mixer analog front-end circuit achieved a gain of up to 141 dBΩ, and a Σ-Δ architecture phase detector achieved a phase quantization noise of 0.0096°, while consuming only 6.8 mW of power. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a block diagram of the digital frequency domain near-infrared brain function acquisition circuit architecture of the present invention;

[0025] Figure 2 This is a block diagram of the analog front-end circuit of the mixer proposed by the present invention;

[0026] Figure 3 This is a block diagram of the amplitude-to-digital conversion circuit proposed by the present invention;

[0027] Figure 4 This is a block diagram of the phase-to-digital conversion circuit proposed by the present invention;

[0028] Figure 5 This is a working timing diagram of the phase-to-digital conversion circuit proposed in the present invention. DETAILED DESCRIPTION

[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments, but the scope of protection of the present invention should not be limited thereto.

[0030] like Figure 1 As shown in the figure, the digital frequency-domain near-infrared brain function acquisition circuit architecture proposed by the present invention consists of three major modules: the analog front-end circuit before the frequency mixing, the amplitude digital conversion circuit, and the phase digital conversion circuit. The photocurrent is first down-converted and converted into a voltage signal by the analog front-end circuit before the frequency mixing. The input of the amplitude digital conversion circuit is connected to the output of the analog front-end for digitizing the amplitude signal. The input of the phase digital conversion circuit is connected to the output of the analog front-end for digitizing the phase signal.

[0031] like Figure 2 As shown in the figure, the analog front-end circuit of the mixer is mainly composed of a mixer, a transimpedance amplifier, and an instrumentation amplifier.

[0032] The photodiode on the left receives a photocurrent with a frequency of 10.01024 MHz, and the photodiode is connected to a DC blocking capacitor.

[0033] One end of the mixer is connected to a DC-blocking capacitor, and the other end is connected to a transimpedance amplifier. The mixer consists of four switches in a complementary configuration. The mixer multiplies the photocurrent input frequency of 10.01024 MHz by the 10 MHz local oscillator signal fb. The mixer outputs currents at frequencies of 10.24 kHz and 20.01024 MHz.

[0034] One end of the transimpedance amplifier is connected to the mixer output, and the other end is connected to the instrumentation amplifier input. The transimpedance amplifier consists of two stages: the first stage is a differentiator, and the second stage is an integrator. These two stages form a bandpass characteristic, used to filter out the 20.01024MHz signal and convert the 10.24kHz photocurrent signal into a voltage signal, which is then amplified and filtered.

[0035] One end of the instrumentation amplifier is connected to the output of the transimpedance amplifier and the other end is connected to the amplitude-to-digital converter and phase-to-digital converter. The instrumentation amplifier is used to convert the input differential signal into a single-ended signal and amplify the voltage signal. The output of the instrumentation amplifier includes an amplitude signal and a phase signal.

[0036] like Figure 3 As shown in the figure, the amplitude-to-digital converter digitizes the amplitude of the analog front-end output signal. This amplitude information reflects the intensity of the emitted light. The amplitude-to-digital converter consists of the following components: envelope detection, Gm-C filter 1, Gm-C filter 2, instrumentation amplifier, buffer, and time-domain ADC.

[0037] One end of the envelope detector is connected to the output of the analog front end, and the other end is connected to the positive input end of the Gm-C filter 1. The envelope detector is used to detect the envelope of the intermediate frequency signal output by the analog front end.

[0038] One end of Gm-C filter 1 is connected to the output of the envelope detector, and the other end is connected to the input of the instrumentation amplifier and Gm-C filter 2. Gm-C filter 1 is used to filter out glitches in the envelope detector output. Gm-C filter 1 is a two-stage Gm-C filter with a passband of 0-25 Hz. Therefore, the output of Gm-C filter 1 reflects the amplitude of the AC signal in the human brain.

[0039] One end of Gm-C filter 2 is connected to the output of Gm-C filter 1, and the other end is connected to the input of the buffer and the input of the instrumentation amplifier. Gm-C filter 2 is also a two-stage Gm-C filter, but its passband is 0-0.05Hz. Gm-C filter 2 is used to extract the DC level of the output of Gm-C filter 1. Therefore, the output of Gm-C filter 2 reflects the amplitude of the DC signal in the human brain.

[0040] The inputs of the instrumentation amplifier are connected to the outputs of Gm-C filter 1 and Gm-C filter 2, respectively. The instrumentation amplifier is used to subtract the AC signal output by Gm-C filter 1 from the DC signal output by Gm-C filter 2, and then amplify the difference. The output of the instrumentation amplifier is the amplified human brain AC signal. It should be noted that amplifying the AC signal helps reduce the performance requirements of the analog-to-digital converter and saves power.

[0041] The input of the buffer is connected to the output of the Gm-C filter 2, and the output is connected to the time domain ADC. Its function is to act as a buffer to drive the next stage time domain ADC.

[0042] The input terminals of the two time domain ADCs are connected to the output of the instrumentation amplifier and the output of the buffer, respectively, for digitizing the DC and AC amplitude signals.

[0043] like Figure 4 As shown in the figure, the phase-to-digital converter (PDC) digitizes the phase information of the analog front-end output signal. This phase information reflects the phase delay between the outgoing and incoming light, i.e., the delay of RF light in the human brain. The PDC primarily consists of the following components: a comparator, a phase detector, a charge pump, a loop filter, a quantizer, a multiphase generator, an edge detector, and a selector.

[0044] The comparator input at the input of the phase-to-digital converter is connected to the output of the analog front end, and the output is connected to one of the inputs of the phase detector. Since the phase detector is rising-edge triggered, the comparator's function is to convert the output of the analog front end into a square wave.

[0045] The two inputs of the phase detector are connected to the comparator output and the selector output, respectively. The output of the phase detector is connected to the charge pump input. The phase detector compares the phase of the comparator output waveform with the phase of the selector output waveform. It then subtracts the phases of the comparator output waveform from the selector output waveform to generate a pulse signal whose duration is proportional to the phase difference. The UP and DN outputs represent the lead and lag relationship between the two input signals, respectively. If the reference phase leads the input phase, the pulse width is output from the UP terminal; otherwise, the pulse width is output from the DN terminal.

[0046] The charge pump's input is connected to the phase detector's output, and its output is connected to the loop filter's input. The charge pump converts the pulse's time signal into a current signal, which is then integrated by the subsequent loop filter.

[0047] The loop filter's input is connected to the charge pump's output, and its output is connected to the quantizer's input. The loop filter consists of a single capacitor, which acts as the loop filter for the entire loop. When UP is high, the upper plate of the capacitor is charged and the lower plate is discharged. When DN is high, the upper plate of the capacitor is discharged and the lower plate is charged. The charge accumulated on the capacitor is proportional to the cumulative time difference between charging and discharging, thereby integrating the phase of the analog front-end output.

[0048] The quantizer's input is connected to the loop filter's output, and its output is connected to the selector. The quantizer performs a 1-bit quantization of the voltage difference across the capacitor. Controlled by a zero-phase clock, CLK, the quantizer performs a quantization comparison on each rising edge pulse. If the voltage on the top plate of the capacitor is greater than the voltage on the bottom plate, a digital code 1 is output; otherwise, a digital code 0 is output. The average value of the 1-bit digital code stream output by the quantizer can be used to calculate the phase of the input signal relative to two reference phases.

[0049] The input of the 10kHz multiphase generator is connected to a 10MHz crystal oscillator, which generates 10MHz from the crystal. The output is connected to an edge detector. This module generates four phase signals with different phases as reference signals. These four phases are 0 degrees, 90 degrees, 180 degrees, and 270 degrees. To improve phase detection accuracy, the phase difference between the reference signals can be reduced. A 90-degree difference between the two reference phases was selected, but this reduces the phase detection range. Because the phase changes caused by human brain activity are small, manually adjusting the interval between the two reference phases during range selection can solve the problem of insufficient phase detection range.

[0050] The edge detector's input is connected to a 10kHz multiphase generator, and its output is connected to a selector. Its function is to detect the rising edge of the input multiphase clock signal and convert it into a short-duration pulse to avoid a high-impedance state in the phase detector.

[0051] The input of the selector is connected to the edge detector, and the output is connected to the phase detector. The function of the selector is to select one of the two reference phase signals for output based on whether the output code stream of the quantizer is 0 or 1.

[0052] like Figure 5 As shown, the quantizer output signal is φ s , the selector selects a reference signal φ according to the output level of the quantizer Ref Output, the phase detector compares the phase difference between the quantizer output signal and the reference signal, converts it into pulse time signals UP and DN, and then the charge pump converts the pulse time signals UP and DN into current signals to charge and discharge the capacitor, so the integrated voltage on the capacitor is V Σ , then the quantizer clock control signal CLK is compared with V Σ The output of 1-bit code stream after the bias voltage is adjusted controls which reference signal φ is selected by the next cycle selector. Ref Output.

[0053] The phase-to-digital converter has an input signal frequency of 10.24kHz, a sampling frequency of 10.24kHz, a bandwidth of 10Hz, and an oversampling ratio of 512. Because the integrating capacitor acts as an integrator, the quantization noise is shaped by first-order noise. Test results show that the phase-to-digital converter has an in-band integrated noise of only 0.0096°.

[0054] The contents not described in detail in the specification of the present invention belong to the common knowledge of professionals in this field.

Claims

1. A digital frequency domain near-infrared brain function acquisition circuit, characterized in that: include: The analog front-end circuit before the mixer is used to down-convert the photocurrent and convert it into a voltage signal; an amplitude-to-digital conversion circuit, whose input end is connected to the output end of the analog front-end circuit, for digitizing the amplitude information of the analog front-end output signal; and a phase-to-digital conversion circuit, whose input end is connected to the output end of the analog front-end circuit, for digitizing the phase information of the analog front-end output signal; The amplitude-to-digital conversion circuit includes an envelope detector, a Gm-C filter 1, a Gm-C filter 2, an instrumentation amplifier, a buffer, a first time-domain ADC, and a second time-domain ADC; The input end of the envelope detector is connected to the output end of the analog front-end circuit, and the output end of the envelope detector is connected to the positive input end of the Gm-C filter 1; The negative input terminal of the Gm-C filter 1 is connected to the negative input terminal of the instrumentation amplifier and the input terminal of the Gm-C filter 2 respectively; One end of the Gm-C filter 2 is connected to the output of the Gm-C filter 1, and the other end is connected to the input end of the buffer and the positive input end of the instrumentation amplifier; The input end of the instrumentation amplifier is connected to the output end of the Gm-C filter 1 and the output end of the Gm-C filter 2 respectively, and the output end is connected to the input end of the first time domain ADC; The input end of the buffer is connected to the output end of the Gm-C filter 2, and the output end is connected to the second time domain ADC; The input terminals of the first time domain ADC and the second time domain ADC are connected to the output of the instrumentation amplifier and the output of the buffer respectively; The phase-to-digital conversion circuit includes a comparator, a phase detector, a charge pump, a loop filter, a quantizer, a multi-phase generator, an edge detector and a selector; The input end of the comparator is connected to the output end of the analog front-end circuit, and the output end is connected to the first input end of the phase detector; The second input terminal of the phase detector is connected to the output terminal of the selector, and the UP output terminal and the DN output terminal of the phase detector are respectively connected to the input terminal of the charge pump; The output end of the charge pump is connected to the input end of the loop filter; The output end of the loop filter is connected to the input end of the quantizer; The output end of the quantizer is connected to the input end of the selector; The input end of the multi-phase generator is connected to a 10 MHz crystal oscillator, and the output end is connected to the edge detector; The output end of the edge detection is connected to the input end of the selector; The selector is used to determine which reference phase signal to output from the two reference phase signals according to the output code stream of the quantizer.

2. The digital frequency domain near-infrared brain function acquisition circuit according to claim 1, characterized in that: The analog front-end circuit includes: DC blocking capacitor, used to filter out the DC component of the input photocurrent; A mixer, whose input end is connected to the output end of the DC blocking capacitor, for changing the frequency of the output photocurrent; a transimpedance amplifier, whose input end is connected to the output end of the mixer, for converting the photocurrent signal into a voltage signal and extracting the lower sideband frequency; and an instrumentation amplifier, whose input end is connected to the output end of the transimpedance amplifier, for converting the input differential signal into a single-ended signal and amplifying the voltage signal, which includes amplitude information and phase information.

3. The digital frequency domain near-infrared brain function acquisition circuit according to claim 2, characterized in that: The mixer package is composed of four complementary switches, which multiplies the input photocurrent by the local oscillator signal fb and outputs two frequency-converted photocurrent signals.

4. The digital frequency domain near-infrared brain function acquisition circuit according to claim 2, characterized in that: The resistance amplifier is composed of a first-stage differentiator and a second-stage integrator to form a bandpass characteristic, which is used to filter out high-frequency photocurrent signals and convert low-frequency photocurrent signals into voltage signals.

Images