Current Domain Analog Front End for Intensity-Modulated Direct Time-of-Flight LIDAR

Abstract

A current-domain analog front end for intensity-modulated direct time-of-flight (TOF) LIDAR. A circuit for filtering a signal corresponding to the time of flight (TOF) of light from a laser reflected from an object to a photodetector includes a preamplifier, a DC cancellation loop, and an AC cancellation loop. The preamplifier can be configured to receive a signal from the photodetector corresponding to the output of the laser reflected from an object distant from the laser and the photodetector. The DC cancellation loop includes a current-feedback DC servo loop. The AC cancellation loop includes a feedback network driven by a floating class AB output stage, and a preamplifier configured to drive the floating class AB output stage, wherein the preamplifier is driven by an error signal from the feedback network and creates an AC signal path through the feedback network and the floating class AB output stage.

Description

Technical Field

[0001] The present invention generally relates to a current domain analog front-end circuit for intensity modulated direct time-of-flight photodetection and ranging. Background Art

[0002] There are various categories of transmit and receive schemes for light detection and ranging (LIDAR) systems, ranging from continuous wave frequency or intensity modulation systems that measure distance with the aid of phase or frequency measurements, to direct time-of-flight systems where the travel time of the transmitted laser pulse is measured. State-of-the-art long-range automotive light detection and ranging often incorporates a direct time-of-flight single-pulse scheme that complies with eye-safe laser operation that requires limiting the transmitted laser energy.

[0003] Automotive light detection and ranging systems with long-range requirements (>200 meters) at low object reflectivity (~10% to 20%) need to be eye-safe. The power of the reflected optical signal decreases quadratically with distance and linearly with the object's reflectivity, which easily translates to a dynamic range on the order of 90 dB. To maximize range, the emitted laser pulses need to have very high peak optical powers, on the order of tens of watts. Eye-safe operation requires that the average power be maintained at the standard's maximum allowable level, which necessitates reducing the optical signal's pulse width to a few nanoseconds. This provides the best performance in terms of achieved range, but imposes system challenges on the light detection and ranging hardware implementation, contributing to cost and complexity. The laser light source's driver needs to deliver large amounts of power to the laser in very short periods of time with fast switching times, often necessitating the development of specialized high-speed power semiconductor components, such as those using gallium nitride technology. Furthermore, the receive chain for time-of-flight (TOF) measurements needs to handle the large dynamic range imposed by the long distance and range of object reflectivity. Summary of the Invention

[0004] A current-domain analog front-end (AFE) circuit for a LIDAR system includes a photodetector configured to provide current, a current-feedback DC servo loop configured to cancel a DC component of the current, a feedback network configured to cancel an AC component of the current, a floating class AB output stage driving the feedback network, and a preamplifier. The preamplifier can be configured to drive the floating class AB output stage, wherein the preamplifier is driven by an error signal from the feedback network and creates an AC signal path including the feedback network and the floating class AB output stage.

[0005] A method of operating a LIDAR system includes providing current via a photodetector, canceling a DC component of the current via a current feedback DC servo loop, canceling an AC component of the current via a feedback network, driving the feedback network via a floating class AB output stage; and driving the floating class AB output stage via a preamplifier, wherein the preamplifier is driven by an error signal from the feedback network.

[0006] A circuit for filtering a signal corresponding to the time of flight (TOF) of light from a laser reflected from an object to a photodetector includes a preamplifier, a DC cancellation loop, and an AC cancellation loop. The preamplifier can be configured to receive a signal from the photodetector corresponding to the output of the laser reflected from an object distant from the laser and the photodetector. The DC cancellation loop includes a current-feedback DC servo loop. The AC cancellation loop includes a feedback network driven by a floating class AB output stage and a preamplifier configured to drive the floating class AB output stage. The preamplifier is driven by an error signal from the feedback network, creating an AC signal path through the feedback network and the floating class AB output stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 is a block diagram illustrating the direct time-of-flight (dTOF) signal path.

[0008] Figure 2 is a block diagram illustrating an alternative direct time-of-flight (dTOF) signal path.

[0009] Figure 3 yes Figure 1 or Figure 2 A graphical representation of the relationship between the amplitude of a signal and time.

[0010] Figure 4 is a block diagram illustrating a circuit of a direct time-of-flight (dTOF) signal path including a transimpedance amplifier (TIA), an intermediate amplifier, an analog-to-digital converter (ADC), and a digital signal processor.

[0011] Figure 5 is an open loop TIA amplifier and has A graphical representation of the feedback factor of a TIA amplifier's amplitude versus frequency.

[0012] Figure 6 Figure 1 is a schematic diagram of a common-gate input buffer and TIA.

[0013] Figure 7 Figure 1 is a schematic diagram of regulating the common gate input buffer and TIA.

[0014] Figure 8 is a schematic diagram of a capacitive feedback transimpedance amplifier (TIA).

[0015] Figure 9 is a block diagram illustrating an intensity modulated phase coded LIDAR receive path of an intermediate signal.

[0016] Figure 10 Figure 1 is a block diagram of the current domain analog front end (AFE) in the pulse-coded LIDAR receive path.

[0017] Figure 11 Figure 1 is a schematic diagram of a current domain analog front end (AFE).

[0018] Figure 12 Figure 1 is a block diagram of a current domain analog front end (AFE).

[0019] Figure 13 is a schematic diagram of the capacitive / resistive feedback network associated with a floating signal.

[0020] Figure 14 is a schematic diagram of a class AB output stage with a source follower bias network.

[0021] Figure 15 is a schematic diagram of a dynamically degraded high-speed current mirror.

[0022] Figure 16 The figure is a schematic diagram of a current-domain analog front-end (AFE) preamplifier, including bias and common-mode adjustment circuits.

[0023] Figure 17 Schematic diagram of a DC servo loop with a current feedback class AB current conveyor. DETAILED DESCRIPTION

[0024] As required, detailed embodiments of the present invention are disclosed herein; however, it should be understood that the disclosed embodiments are merely exemplary of the invention that may be embodied in various and alternative forms. The figures are not necessarily to scale; some features may be exaggerated or minimized to illustrate details of particular components. Therefore, the specific structural and functional details disclosed herein should not be interpreted as limiting, but merely as a representative basis for teaching one skilled in the art to variously employ the present invention.

[0025] The term "substantially" may be used herein to describe embodiments disclosed or claimed. The term "substantially" may modify a value or relative property disclosed or claimed in the present disclosure. In such instances, "substantially" may indicate that the value or relative property it modifies is within the range of the value or relative property. 0%, 0.1%, 0.5%, 1%, 2%, 3%, 4%, 5% or within 10%.

[0026] Lidar is a sensor used for ranging in applications such as autonomous driving and robotics. It measures distance optically, for example, by emitting laser pulses and measuring the time it takes for the light to bounce back from a target (the direct time-of-flight, or dTOF, method). One of the challenges of these optical detection and ranging systems lies in their receive signal path electronics, which must detect light reflected from distant objects (i.e., >200 meters) with limited reflectivity (dark objects). This imposes several requirements on the receive signal path's noise performance, dynamic range, and bandwidth. The current state-of-the-art in dTOFLIDAR typically passes the output current of its photodetector (optical transducer) to a transimpedance amplifier (TIA), whose output voltage is then processed by a high-speed, high-resolution analog-to-digital converter (ADC). TIA design presents a quadrature performance tradeoff between bandwidth and stability, on the one hand, and noise performance, on the other. If high bandwidth is desired, this often makes the TIA the dominant noise source in the lidar receiver signal path, limiting the overall link signal-to-noise ratio (SNR) and, consequently, its range performance. Furthermore, TIAs have limitations in terms of fast saturation recovery (e.g. when processing signals from the near to mid range), which limits their use to alternative modulation techniques, besides the frequently used single-pulse transmission scheme (e.g., with phase-coded pulse trains similar to those disclosed in PCT / EP2019 / 052094).

[0027] This application discloses a current-domain analog front-end circuit (current AFE) for automotive LiDAR that addresses the aforementioned challenges of TIA-based systems. It allows for alternative trade-offs beyond the TIA design space (i.e., decoupling of input noise and bandwidth), optimizing noise and recovery speed for the high dynamic range and high speed requirements of automotive LiDAR. Furthermore, the current-domain signal path enables the use of a current-domain ADC, which is better optimized for high oversampling ratios, making it particularly well-suited for alternative modulation schemes such as intensity-modulated phase-coded pulse trains.

[0028] One of the challenges of automotive Lidar sensors (laser rangefinders) is to achieve the required range and resolution performance at permissible eye-safe laser energy levels. Among the possible Lidar modulation technologies, direct time of flight (dTOF) is one approach that can achieve these requirements. Currently, the most advanced dTOF systems ( Figure 1 and Figure 2) emits a single laser pulse with high power (~ >100 watts) and narrow width (~ <5 nanoseconds) and measures the time it takes for the light to bounce back from the target, which is proportional to the range through which the speed of light passes. The detection path often used involves: a photodetector, such as an avalanche photodiode (APD), which produces an output current; a transimpedance amplifier (TIA), which converts the photodiode current to an output voltage; and a time-to-digital converter or analog-to-digital converter (TDC or ADC), which digitizes the signal for further processing.

[0029] Figure 1 is a block diagram illustrating a direct time-of-flight (dTOF) signal path 100. A laser 102 transmits an optical signal onto an object 104, which is then received by a photodetector such as a photodiode. The signal is then amplified by a transimpedance amplifier (TIA) 206 and converted to a digital signal via an analog-to-digital converter (ADC) 108. The digital signal can then be manipulated by a processor 110 (e.g., a digital signal processor (DSP)) that outputs a time-of-flight (TOF) signal.

[0030] Figure 2 2 is a block diagram illustrating an alternative direct time-of-flight (dTOF) signal path 200. A laser 202 transmits a light signal onto an object 204, which is then received by a photodetector such as a photodiode. The signal is then amplified by a transimpedance amplifier (TIA) 206 and forwarded to a constant fractional discriminator (CFD) 208, and then to a comparator 210 and a time-to-digital converter (TDC) 212, which then outputs a time-of-flight (TOF) signal.

[0031] Figure 3 yes Figure 1 or Figure 2 3. A graphical representation 300 of a transmitted amplitude 302 and a received amplitude 304 of a signal of FIG. 3 is shown with respect to time 306.

[0032] For this approach, improved lidar performance is achieved by minimizing input-referred noise and maximizing TIA bandwidth. Unfortunately, this is difficult to achieve with TIAs because their input noise and bandwidth are fundamentally coupled—that is, they are interdependent and proportional. In fact, to increase the TIA bandwidth, its input noise must be increased, and conversely, to increase the TIA input noise, its bandwidth must be increased. This dependency / relationship between TIA bandwidth and input noise is exactly the opposite of what one would expect to maximize the performance of a lidar analog front-end, and it often makes the TIA one of the dominant performance limiters in lidar systems. Furthermore, TIA circuits have difficulty recovering from saturation for large inputs, and as such, their outputs are typically clipped for near- to mid-range reflections. For a single pulse transmit scheme, the arrival time can be resolved as long as the rising edge of the signal can be detected. However, the saturation problem may be less important. This limitation prohibits the system designer from making higher-level compromises, such as by choosing a lower peak transmit power and regaining system performance in the back end through matched filters and cross-correlation techniques similar to the cross-correlation technique disclosed in PCT / EP2019 / 052094.

[0033] Figure 4 4 is a block diagram illustrating a transimpedance amplifier (TIA) circuit of a direct time-of-flight (dTOF) signal path 400. A signal from a photodetector (such as a photodiode) is amplified by a transimpedance amplifier (TIA) 406, which is then amplified by an amplifier 410 and converted to a digital signal via an analog-to-digital converter (ADC) 408. The signal can then be processed by a processor 412 (e.g., a digital signal processor (DSP)) which cross-correlates the digital signals and outputs a time-of-flight (TOF) signal.

[0034] Figure 4 The diagram shows the signal path for a TIA-based dTOF Lidar used in current state-of-the-art implementations. An intermediate gain stage is typically required to boost the signal at the far end of the range. The performance and design criteria of the TIA are affected by several contributing terms: parasitic input capacitance C D is one of the terms (e.g. caused by the photodiode and the amplifier) ​​that has a negative impact on the stability of the amplifier. To compensate for this, a feedback capacitor C is added. F This stability requirement boils down to a design criterion where the 3dB bandwidth of the signal path, f BW , the unity gain frequency f of the TIA amplifier U , transimpedance factor R TIA and capacitors will be related, as Figure 5 The frequency response curve is shown in the graph. Figure 5 The figure shows the amplifier open-loop gain and noise gain (reverse feedback factor or The difference between the two forms a feedback network ( ), where the stability criterion requires a 20dB / dec crossing slope when passing through the unity gain frequency. The noise gain after adding the compensation capacitor can be written as:

[0035] .

[0036] like Figure 5 As shown in , this characteristic has a zero due to parasitic capacitance (denoted by f Z ), and the poles (denoted by f BW Marked), we select the stabilizing capacitor C F The value of f BW Marking):

[0037] .

[0038] For a Butterworth response (Q = 0.7), one can choose to have a value that would be f BW The closed-loop bandwidth, which is the unity-gain frequency and is determined by C D The geometric mean of zeros introduced.

[0039]

[0040] Figure 5 is the amplifier open loop response magnitude 502 and the reverse feedback magnitude 504 relative to the open loop TIA amplifier and has A graphical representation of the feedback factor 510 versus the frequency 506 of a TIA amplifier.

[0041] Therefore, the TIA design tradeoff can be in one of two paths: bandwidth sets the noise (eg, SNR) or noise (eg, SNR) sets the bandwidth.

[0042] Where bandwidth sets noise (SNR), applications such as dTOF Lidar transmit pulse width, determined by the available eye-safe laser energy and the required range specifications, fix the required bandwidth f. BW , and the photodiode has a fixed parasitic capacitance C D , and the Lidar link SNR budget determines the feedback resistor R based on the noise requirement TIA (Assuming a balanced TIA design, R TIA is the dominant noise source). This fixes the amplifier's required unity-gain bandwidth, f U choice, which becomes a technical limitation.

[0043] In the case of noise (e.g., SNR) that sets the bandwidth, the link SNR fixes the noise (and hence R TIA ), and the ability of this technique to fix the unity-gain bandwidth of the amplifier f U , and together with a given photodiode parasitic capacitance, this will result in a maximum achievable bandwidth. To some extent, this will translate back to a limitation in Lidar range, since for a fixed laser energy, range performance must now be compromised by reducing the laser peak transmit power.

[0044] As can be observed from the above trade-offs, larger R TIA The current noise is improved, but this comes at the expense of reduced BW, which may be limited by the speed of the available technology. Finally, it is also worth noting that regarding the final noise performance of the resistive TIA, we also have the amplifier noise Vn (in addition to the noise caused by R TIA The input current noise is represented by the input parasitic capacitance and the compensation capacitance, which appears at the input as input referred current noise:

[0045] .

[0046] This shows that in a resistive TIA, the parasitic capacitance not only TIA The CMOS value (for stable operation) not only places a limit on the noise performance, but also creates a noise gain by which the voltage noise of the amplifier is converted into input referred current noise.

[0047] Another challenge associated with current state-of-the-art dTOF is the need for very high-speed and high-resolution ADCs, which need to produce a small number of samples over the duration in which the narrow laser pulse reflection can be observed. This is primarily because there are only a few pulse samples (2 to 5 nanoseconds) available for each shot, and therefore ADC quantization noise has a negative impact on the accuracy with which the pulse arrival time can be estimated. This can be seen as an oversampling ratio limitation, where the bandwidth is determined by the pulse width (larger bandwidth for narrower pulses). If the noise bandwidth can be reduced, and potentially the sampling rate can be increased, the oversampling ratio and quantization noise can be traded off, allowing for a looser ADC resolution.

[0048] An improvement to the above approach can be hypothesized to differ from the single narrow transmit pulse scheme used in current methods. However, given a fixed available eye-safe energy, simply reducing the pulse amplitude and extending it in time (to maintain energy but reduce the noise bandwidth and help increase the oversampling ratio) would result in a substantial loss of signal-to-noise ratio (SNR) due to a degradation of the pulse autocorrelation peak slew rate. This can potentially be addressed by using specific phase encoding within the pulse train (pulse train) to achieve an autocorrelation function with a maximized peak slew rate, similar to that disclosed in PCT / EP2019 / 052094. For example, this could be imagined as a sequence of 10 pulses, each at a 10-fold lower peak amplitude and the same pulse width as the single pulse, but with specific phase-shift keying implemented within them. This could potentially enable the use of even a single-bit ADC for time-of-flight estimation, similar to that disclosed in PCT / EP2019 / 052094. Using such an encoding scheme with reduced peak transmit power is particularly attractive because it reduces system cost and complexity.

[0049] A problem associated with TIA-based front-ends is that their transimpedance is typically optimized for long-range SNR requirements, meaning the TIA will saturate and clip for short- to medium-range signals (assuming a dynamic range of four orders of magnitude for automotive lidar). This has several limitations. One is that the TIA's recovery time from saturation is typically much slower than the modulation period suitable for phase-coded lidar transmissions. This destroys the phase encoding and degrades the pulse train's autocorrelation function. Furthermore, the clipping TIA does not allow for measurement of reflected signal energy, which is desirable for object classification.

[0050] In addition to the TIA structures described above, there are also TIAs introduced in the prior art that incorporate a common gate input current buffer or a regulated common gate input current buffer before the resistive TIA. These are intended to decouple the photodiode capacitance and the TIA feedback resistor (e.g., Figure 6 and Figure 7 ), which has D The benefits of stability and noise gain. However, these structures have the disadvantage of additive noise due to the channel noise of the regulation loop and the common-gate transistors. In addition, the clipping and saturation issues discussed earlier remain unsolved with these methods.

[0051] Figure 6 is a schematic diagram of a common gate input buffer 600 including a transimpedance amplifier (TIA) 606 . Figure 7 is a schematic diagram of a regulated common gate input buffer including a transimpedance amplifier (TIA) 706 .

[0052] A promising architecture is Figure 8 The TIA with a capacitive feedback network shown in effectively attempts to break the coupled stability and BW / noise tradeoff of the resistive feedback TIA. Figure 8 FIG is a schematic diagram of a capacitive feedback transimpedance amplifier (TIA) 800 including a transimpedance amplifier (TIA) 806. In this topology, the output buffer transistor M out Driver shunt feedback capacitor C SH , to generate a voltage V across its two ends x . V x The feedback capacitor C F Sensed and converted into feedback current i FB , the feedback current i FB Eliminate input current i by closed-loop operation in . C SH and C F The ratio between them leads to the out and i FB (and therefore i in ) generates a gain between the currents:

[0053] .

[0054] The prior art then outputs the buffer M out The resulting output current is fed directly to the resistor R TIA , to create the output voltage:

[0055] .

[0056] This topology has several advantages over resistive TIAs. The use of a capacitive gain network means there is no thermal noise penalty like with resistive feedback TIAs. Another advantage is that the parasitic capacitance at the input (C D ) does not cause instability, but reduces the in-band loop gain and affects the bandwidth. Finally, at a value greater than (1+C SH / C F ) in the case of a current gain, the transimpedance resistor R TIA The current noise of the amplifier will be reduced by the same gain factor when it is returned to the input. In this architecture, we still need to consider the noise contribution of the amplifier through the input capacitance, and there is also the output buffer M. out It also directly contributes to current noise.

[0057] The capacitive TIA architecture is promising for dTOF lidar because it enables a better trade-off between noise and bandwidth; however, it still has some drawbacks considering the specific needs of dTOF links:

[0058] ● The architecture shown in the prior art remains challenging in handling 4 orders of magnitude of signal dynamic range (over >200m range and 10% to 100% reflectivity variation).

[0059] • If the signal should not be limited, the voltage output will suffer from limited headroom.

[0060] ● Saturation recovery is particularly important for intensity modulated transmit coding schemes, but remains a challenge, especially for capacitive feedback networks (the integrating characteristic of capacitors).

[0061] ● Handling a large dynamic range requires efficient quiescent operation because current noise requires low quiescent current (a disadvantage of Class A operation).

[0062] • For example, if two functions are applied to the Lidar received signal in the current domain, namely one function optimized for TOF estimation and another function optimized for received signal energy estimation, such as reflectivity estimation, these architectures have low flexibility. DETAILED DESCRIPTION

[0064] The main vision of the Current Domain Analog Front End proposal for Intensity Modulated Direct Time of Flight (dTOF) Lidar Applications is to create a flexible front end capability that enables:

[0065] ● Remains relatively linear over the nearly 4 orders of magnitude dynamic range expected for automotive LiDAR systems that rely on linear photodetectors such as avalanche photodiodes

[0066] ● Achieve the required bandwidth and noise performance

[0067] ● Sufficient flexibility to split the signal in such a way that functions optimized for TOF and intensity measurements can be applied to the received signal

[0068] ● The DC current caused by the backlight can be processed in the photodetector.

[0069] This should help enable a signal path that overcomes the limitations of the resistive and capacitive TIA topologies described earlier.

[0070] A key application of such a circuit is to use it in conjunction with circuits similar to those disclosed in PCT / EP2019 / 052094 and in Figure 9 The state-of-the-art single-pulse transmission scheme is shown compared when using an intensity-modulated pulse train with embedded coding (such as a phase-shift keying pulse train).

[0071] Figure 9Figure 9 is a block diagram illustrating an intensity-modulated phase-coded LIDAR receive path 900 for an intermediate signal. Based on input from a phase-keyed pulse train pattern generator 914, a laser 902 transmits an optical signal onto an object 904, which is then received by a photodetector such as a photodiode. This signal is then amplified by a current AFE 906 and forwarded to a sign comparator 908, which then forwards it to a single-bit cross-correlation block 910, which also receives input from the phase-keyed pulse train pattern generator 914 and to a peak detection circuit 912, which then outputs a time-of-flight (TOF) signal. The intermediate signal includes transmitted and received amplitudes relative to time 916, transmitted key and received signs relative to time 918, and amplitudes with peak detection relative to time 920.

[0072] Figure 9 The system shown in is a potential use case where the system can utilize such a current domain analog front end (current AFE) to extract phase coding information from the received signal without clipping and thus destroying the phase information. Such modulation allows the use of, for example, highly oversampled single-bit ADCs (e.g., symbol comparators) to extract TOF by applying the transmitted phase coding pattern to the output bit stream (single-bit cross-correlator, e.g., application of a matched filter). This is an example where phase coding enables the use of very low resolution ADCs to accurately estimate TOF. A high-level block diagram of the proposed current domain analog front end in the potential phase coding direct TOF architecture is shown in Figure 10 As shown in .

[0073] Figure 10 100 is a block diagram of a current domain analog front end (AFE) in a pulse coded LIDAR receive path 1000. The current domain analog front end (AFE) receives a signal from a photodetector and outputs a time-of-flight (TOF) signal to a time-of-flight measurement circuit 1004 and an intensity signal to an intensity measurement circuit 1006.

[0074] like Figure 10 As shown in FIG, the proposed Lidar current AFE does not utilize a conventional TIA followed by a multi-bit high-speed ADC signal path. The current AFE receives the APD current, which in some cases may have an amplitude ranging from 1uA to 20mA, which is superimposed on the backlight-induced DC current, which in some cases may be on the order of 50uA. The received signal may involve a binary phase encoding embedded in an intensity modulated pulse train. The current AFE frees the design space from the trade-offs in resistive TIAs, where the APD parasitic capacitance C DThe modulation period of the transmit signal together with the fixed unity gain bandwidth of the amplifier places a box around the choice of the signal path BW, and therefore around the feedback resistor, ultimately resulting in a trade-off for noise and range.

[0075] The key architectural improvements of the proposed current AFE over current resistive or capacitive TIAs are as follows:

[0076] The current domain signal path enables processing of the required signal range of four orders of magnitude without amplitude limiting, thereby enabling modulation that cannot tolerate saturation.

[0077] Depending on whether the signal is coming from the far end of the range (low noise operation) or the near end of the range (high dynamic range operation), the signal path noise and dynamic range can be automatically adjusted for low noise or high dynamic range accordingly.

[0078] This is enabled by using nonlinear capacitive and resistive feedback networks (rather than pure resistors or capacitors). This allows low-noise operation when operation in capacitive feedback network mode is required (for mid-range to long-range input signal levels) and by switching the sensing network to a resistive feedback network when the signal is above a certain amplitude (in which range noise is not a concern).

[0079] There is an amplitude sensitive mechanism for enabling dynamic degradation of the current handling circuitry to have low noise operation for long range (small signals), and for short range large signals to bypass the degradation circuitry to prevent clipping.

[0080] It utilizes a floating Class AB output stage to drive a shunt feedback network, which then mirrors its output current to the TOF and intensity measurement channels. Class AB operation enables very low quiescent current operation, which is suitable for low current noise when the signal is from mid-range to long-range (and therefore small), and automatic adjustment of the quiescent current when the signal is large.

[0081] The backlight-induced DC current of the APD, if passed to the AFE, results in a loss of dynamic range and increased noise due to the increased quiescent current. The proposed current AFE utilizes a current-feedback DC servo circuit to compensate for the expected level of the backlight-induced DC current. This servo loop then creates a high-pass frequency response within the overall AFE frequency response, whose corner frequency can be simply adjusted within the servo loop to scan the LiDAR system's needs. This can be accomplished without an additional noise penalty and has the advantage of freeing the design from the use of external components such as DC block capacitors. This paves the way for multi-channel scaling of the system.

[0082] APD parasitic capacitance affects the corner frequency rather than making the signal path less stable.

[0083] The current from the floating class AB output stage can be circulated through a current mirror to create parallel signal paths for applying different functions to the received signal. For example, as shown previously, this could involve two copies of the TOF and intensity measurement.

[0084] The following sections focus in more detail on the five main differentiating aspects of the proposed current AFE. The overall current-domain AFE architectural benefits include: a signal-dependent capacitive / resistive feedback network, a floating Class AB output stage with a signal-dependent degeneration current mirror, a high-speed capacitive-coupled preamplifier with a bias voltage regulation loop that enables a floating Class AB output stage with capacitive feedback (bias regulation), and a current-feedback Class AB DC servo loop for backlight compensation.

[0085] Overall Current Domain AFE Architectural Benefits

[0086] A more detailed block diagram of the current domain AFE is given in Figure 11 The various blocks of the architecture are discussed in detail in the following subsections.

[0087] Figure 11 Figure 1 is a schematic diagram of a current-domain analog front-end (AFE) 1100, 1102. The current AFE consists of an AC signal path consisting of a preamplifier (input stage) driving a floating Class AB output stage consisting of a source-follower-based bias circuit and a DC servo loop. This structure enables controlled setting of the output DC level, which allows the use of a capacitive feedback network.

[0088] The output current of the output stage drives a capacitive / resistive feedback network, which is primarily capacitive for low-noise operation when processing medium- to long-range signals, and primarily resistive for short-range operation when dynamic range is the primary consideration. Feedback is closed in the current domain at the input to eliminate current from the APD photodetector. The feedback network senses the output swing and automatically transitions from a capacitive to a resistive network. Because the cross-correlation applied to phase-coded modulation (or single pulse) is robust to such non-idealities, potential glitches or distortion caused by this crossover are minimally problematic for lidar applications.

[0089] The feedback current generated by the floating class AB output stage is sensed and replicated with the aid of a dynamically degraded output current mirror, creating a current signal replica that is used by various functions in the rest of the signal path, such as time-of-flight (TOF) or intensity measurement. Dynamic degradation enables low-noise and wideband operation for medium- to long-range signals at low quiescent current levels.

[0090] One path to scaling the AFE is to enable interfaces to multiple photodetectors, for example, to scale the spatial resolution of a dTOF lidar through partial flash operation. This requires a higher level of integration, leading to the development of a DC servo loop to compensate for the backlight-induced APD current. Due to potential requirements associated with lidar scan rates, backlight current compensation may need to provide fast settling capability as the lidar scans from very reflective pixels to non-reflective pixels. The DC servo loop features a current buffer input stage that drives a class AB output stage using unity current domain feedback. Its low-pass frequency response creates the overall current AFE's bandpass frequency response by determining its high-pass corner frequency. The class AB low quiescent current current feedback architecture enables low-noise operation. The bandwidth can be adjusted according to system requirements to cover the high-pass corner frequency, allowing for fast settling of the backlight current compensation.

[0091] Ignoring most minor effects, the small signal frequency response of the current domain AFE can be described by considering Figure 12 The simplified block diagram shows the contributing circuit parameters to be designed as first order. This includes:

[0092] ● AC signal path parameters include preamplifier in-band gain: A AC , floating output stage effective gm: g mo，AC , capacitive feedback network value: C SH and C FB , APD parasitic capacitance: C APD

[0093] ● DC servo loop parameters, including DC servo loop forward path gain: A DC , the input transimpedance gm of its current domain input stage i，DC .

[0094] Figure 12 is a block diagram of a small signal current domain analog front end (AFE) 1200 including a main signal path circuit 1202 and a DC servo loop 1204. The corner frequency of the closed loop transfer function will be proportional to:

[0095]

[0096] .

[0097] The closed-loop in-band current transfer gain will be For example, in some specific implementations, this can be achieved by using to set the gain to ~1.

[0098] AFE's high-pass corner frequency f HPDetermined by the following two elements:

[0099] ● The input impedance of the DC servo loop divided by its loop gain, e.g. input DC resistance

[0100] ● The feedback capacitor of the AC loop is multiplied by the AC loop gain, i.e. the Miller effect is applied to C due to the gain FB , the gain is dominated by the load capacitance of the DC servo loop.

[0101] AFE low-pass corner frequency f LP Determined by the following two elements:

[0102] ● Transconductance of the class AB output driver of the AC loop (g mo，AC ) divided by the AC loop gain (which determines the output impedance of the closed-loop amplifier)

[0103] ● Shunt feedback capacitor C SH Add C FB and C APD The series connection of the P .

[0104] It should be noted that in order not to introduce non-dominant poles, the BW of the preamplifier in the main AC loop needs to be at least 2xf LP .

[0105] For bandwidth accuracy, the parasitic C APD Still should be kept to a minimum, i.e. , so that C SH It will mainly determine the LP To increase bandwidth, it is better to keep C SH Small, but now C SH with C APD compared to.

[0106] An overview of the overall benefits of current domain AFEs, which enable various benefits within the LiDAR system architecture:

[0107] ● High dynamic range operation.

[0108] Dynamic feedback networks include capacitive feedback for low noise for medium to long range signals, and switching to low resistance (or current buffers) for large signals, which can help prevent clipping.

[0109] ○ Dynamic transistor source degeneration enables low noise and infinite amplitude current mirrors.

[0110] ○ Floating Class AB output stage enables dynamic behavior, where low quiescent current enables low noise, and high drive capability enables large-signal behavior.

[0111] ● Backlight compensation.

[0112] ○ No external components blocking DC enable scaling.

[0113] Class AB current feedback servo loop enables low-noise operation.

[0114] ○ High-pass frequency response can be adjusted for Lidar scan rate to enable fast response to backlight sensing amplitude Scan from dark pixels to light pixels.

[0115] ● Parasitic input capacitance does not interfere with stability (only affects corner frequency, and there is still a noise penalty through the preamplifier noise performance.

[0116] ● Split signal paths for different functions.

[0117] ○ Separate TOF and intensity signal paths are each optimized for their own requirements (frequency response, speed, dynamic range).

[0118] • Both sink and source capabilities of the entire current AFE (bipolar APD connection).

[0119] ○ Signal dependent capacitive / resistive feedback network.

[0120] In the case of a single-pulse dTOF Lidar system using a resistive or capacitive TIA, a large input signal saturates the front end, and thus there is a need for a reset to resume the next measurement transmission.

[0121] This application discloses a signal-dependent feedback network for use with current-domain signal paths. This network automatically switches between a capacitive network designed for low-noise operation (without a noise penalty) for small signals (over distances from 20m to 200m), and a resistor-based feedback network to prevent saturation for large signals from 2m to 20m (for which noise is negligible). Figure 13 One embodiment of such a network for signal-dependent control of the feedback network is shown. This is accomplished by sensing the voltage at the output of the floating class AB output stage (the input to the feedback network) and using a level-shifted amplifier to drive a parallel pair of PMOS and NMOS transistors. N,RES and M P,RES The gate of the transistor is realized so that when a certain voltage swing is sensed at the amplifier output (by the level shift parameter and the M N,RES and M P,RES When the threshold voltage is designed), they bypass the feedback capacitor C FB .

[0122] Figure 13 is a schematic diagram of a signal dependent capacitive / resistive feedback network 1300 including a main signal path circuit 1302 and a capacitive / resistive feedback network 1304. Figure 13 In FIG. 1 , an alternative capacitive / resistive feedback network 1306 is illustrated using a specific implementation of a level-shifting amplifier providing feedback type 1308 .

[0123] When the signal is small, the level shifter, together with the DC voltage regulation of the feedback network input and output nodes (from the amplifier internal nodes), ensures that the bypass transistor M N,RES and M P,RES is turned off (in which case each transistor is effectively supplied with a negative |V GS |). In this case, the feedback network is capacitive. For the large Lidar return signal, during the rising edge of the first pulse, the capacitive network results in an integrated voltage at the amplifier output. This results in a voltage across the feedback capacitor C FB The net voltage across CFB is determined by the sink or source nature of the input current signal. When the voltage difference across CFB becomes greater than two threshold voltages, the M N,res or M P,res When the voltage difference across CFB is two NMOS threshold voltages, the bypass path is enabled. N,B Enable M N,res path, and when the voltage difference is negative two PMOS threshold voltages, then through the source follower M P,B Enable M P,res Path. Figure 10 Either of these scenarios shown allows for a predetermined signal swing for which the feedback network is capacitive and for swings greater than (towards positive or negative values) the capacitor is replaced by M N,res or M P,res The resistive network formed is shunted. There will be a gain difference between the two feedback networks, however this is of little significance for TOF evaluation and calibration may be required for intensity (energy of the reflected signal).

[0124] One potential application of the current AFE is to be used as a current buffer and current replica (i.e., in C SH Value and C FB values ​​differ by a factor of 10), in which case the gain variation is irrelevant since one can simply make M N,RES and M P,RES Channel resistance ratio C SH The impedance of the MOSFET is much smaller to maintain the unity-gain transfer characteristic between the capacitive mode and the resistive mode.

[0125] Floating Class AB Output Stage with Signal-Dependent Degeneration Current Mirror

[0126] The output stage of the current AFE performs the following functions:

[0127] Create a source-follower-based Class AB driver that can: 1) self-regulate the DC bias at the output node and drive feedback current into the capacitive / resistive feedback network. DC regulation is important in both operating modes of the feedback network, and 2) enable four orders of magnitude more current handling capability at low quiescent current (IQ) levels when in Class AB operation, which is important for low-noise operation with mid-range to long-range receive signals.

[0128] ● Sink / source output current capability allows both sink / source connections of photodetectors.

[0129] ● The sinking / sourcing feedback current injected by the class AB output stage (sinking from the supply rail or sourcing into the ground rail) is recovered by the dynamic degeneration current mirror to create the desired number of output current copies for various signal paths, such as for TOF and intensity detection paths.

[0130] Figure 14 Figure 1400 is a schematic diagram of a Class AB output stage with a source-follower bias network, illustrating the source-follower-based Class AB biasing and output transistors. The source-follower configuration (as opposed to the common-source output device configuration) has the disadvantage of limited voltage swing, however, it allows for adjustment of the output DC bias, which is required when using a capacitive feedback network. The voltage swing limitation is not an issue here because the signal path is in the current domain, and the swing is kept limited by the signal-dependent capacitive / resistive feedback network to prevent saturation. This configuration enables a high-speed and low-noise solution, rather than using common-mode regulation techniques at the output stage. To bias the Class AB driver, the preamplifier of the current AFE is incorporated into the output DC bias regulation (discussed later).

[0131] Figure 15 is a schematic diagram of a dynamically degraded high-speed current mirror 1500, which illustrates a dynamically degraded low noise and high dynamic range current mirror that recycles and replicates the current injected into the floating class AB output stage of the capacitive / resistive feedback network. Here, the mirror transistor M NCS and M PCS Degraded resistor R deg This degradation has various benefits for the accuracy and noise performance of the current mirror, however it comes at the expense of overvoltage margin requirements for large signals, which quickly becomes impractical when the current mirror must handle a wide dynamic range.

[0132] For the dTOF lidar signal path, mid-range to long-range received signals have the most stringent requirements on noise and offset, especially when considering alternative modulation schemes such as phase encoding in intensity-modulated pulse trains. These requirements do not hold for short-range to mid-range signals, where the signal is greater than the noise level. Within those ranges of the input signal, it is crucial that the circuit does not clip, so that the phase information embedded in the pulse train can be retrieved.

[0133] Based on the above considerations, one can consider adapting the circuit to the needs of the signal. This means that we can consider a dynamically degraded current mirror, where the benefits of the degeneration resistor are present for small signals. In those ranges, we are concerned with low offset and low noise operation. Once the signal becomes large, the mirror point V mirrorP and V mirrorN The voltage built on the source follower becomes large enough to make the level shifter buffer M based on the source follower NSF and M PSF Can turn on the bypass transistor M Nb and M Pb , and therefore bypass the degeneration resistor R deg In this way, a bypass mechanism for degradation can be designed to approximately apply the degradation for signals from 20m to 200m, and bypass it for signals from distances less than 20m.

[0134] In addition, to further enhance the accuracy of mid-range to long-range signal levels, the amplifier has an adjustable P and A N Cascode transistor M PCAS and M NCAS Ensure that the drain-source voltage of the mirror transistors MNCS and MPCS remains the same (equal to V mirrorP and V mirrorN ), thereby significantly increasing their output impedance.

[0135] To summarize the current mirror: it recycles the output current of the floating class AB output stage (combined low noise + high dynamic range capabilities) and mirrors it with precision (low noise and low offset) for mid-range to long-range signals, and with dynamic range (unlimited amplitude) for short-range signals.

[0136] High-speed capacitive-coupled preamplifier with bias voltage regulation loop enabling floating class-AB output stage with capacitive feedback (bias regulation)

[0137] The current AFE's single-transistor preamplifier drives a floating Class AB output stage. It is optimized for speed and noise. Its bandwidth should be several times (at least twice) that of the current AFE so as not to affect the stability of the front-end. Its noise performance should also be designed based on the expected input noise requirements of the current AFE, taking into account the expected parasitic capacitance of the photodetector (APD).

[0138] Closed loop regulation of its output bias point allows the output stage of the bias front end to be DC biased, which is required due to the use of a capacitive feedback network. Figure 16 Shown in. Figure 16 is a schematic diagram of a preamplifier of a current domain analog front end (AFE) including bias and common-mode adjustment circuit 1600.

[0139] The preamplifier consists of transistor M NA As a common source amplifier and load transistor M NL Formed, they have transconductance g m，A and g m，L The input-to-output gain of the amplifier is determined by:

[0140] .

[0141] The amplifier transistors are connected from the bias branch on the left (transistor M bNA ) bias, which also biases the right-hand replica branch (by M bNA and M bNL The replica branch uses transistors with the same current density and aspect ratio as the transistors in the amplifier branch (and bias branch). The feedback loop regulates the gate of the load transistor (M NL and M bNL ), so that the source of the load transistor in the replica branch is regulated to the desired V CM，ref (This will be the DC level at the input, and hence the output, of the floating class AB output stage.) Since the transistors are matched, the source of the load transistor in the amplifier leg (and hence V out DC level) will be equal to the source of the replica branch, which is regulated to V CM，ref The variation due to mismatch between the branches is not critical since this slightly shifts the bias point which, given the current-in, current-out topology of the current AFE, does not cause headroom and swing issues that are typical problems of TIA circuits.

[0142] The input is connected through capacitor C C is capacitively coupled, capacitance C C Through the resistor R Cis designed to have a crossover frequency much lower than the high-pass corner frequency of the overall current amplifier so as not to interfere with its frequency response. To achieve this, the resistor R C Very big.

[0143] Current Feedback Class AB DC Servo Loop for Backlight Compensation

[0144] The DC servo loop is a current feedback current input and current output circuit that has several key characteristics for the dTOF lidar system and overall current AFE operation:

[0145] ● Considering the integrated solution, it is more suitable for scaling with more detectors (rather than using off-chip DC block capacitors).

[0146] ● Absorbs the backlight-induced DC current generated by the photodetector (APD) and thus increases the dynamic range of the signal path, reducing excess noise in the current AFE (current noise increases when quiescent current increases, e.g. due to processing DC backlight-induced current)

[0147] ● Its low-pass frequency response creates the overall bandpass frequency response for the current AFE. The high corner frequency achieved by this feature can be designed so that the DC servo loop meets the maximum expected backlight-induced DC current in the short time period available between two adjacent pixels scanned by the Lidar (the target reflectivity may cause the full dynamic range of the DC backlight to shift)

[0148] ● Provide DC bias voltage for the front end.

[0149] The circuit architecture is based on the previously proposed current conveyor, however, this embodiment has some new features that are particularly critical to the needs of the Lidar signal path:

[0150] ● A combined current input stage that drives a class AB biased output stage to drive a unity feedback path (also self-regulating the DC bias at the output).

[0151] • Low quiescent current operation allowing for combined low noise and high dynamic range operation capable of rejecting a wide range of DC currents with minimal noise impact on the current AFE.

[0152] Figure 17 FIG. 1 is a schematic diagram of a DC servo loop with a current feedback class AB current conveyor 1700. The input stage is composed of PMOS and NMOS transistors M NI and M PIA mesh structure is formed in which the input and feedback currents are summed at a point on the right branch. The left branch, connected to the reference voltage, regulates the DC bias point of the input / output terminals (and ultimately also the DC bias of the output transistor via the unit current feedback). The output sink / source current of the input stage is passed to the output stage, which drives the class AB biased output stage transistor M. PO and M NO The feedback is unity and is therefore performed by short-circuiting from the output to the input, and the feedback current i Feedback Compensates for the input DC current generated by the backlight. The frequency response of this stage determines the high-pass corner frequency of the overall signal path, which is discussed in the overall current AFE discussion. The low-pass corner frequency formed by it defines the high-pass corner frequency of the current AFE.

[0153] Program code embodying the algorithms and / or methodological techniques described herein can be distributed separately or collectively as a program product in a variety of different forms. The program code can be distributed using a computer-readable storage medium having computer-readable program instructions thereon for causing a processor to perform aspects of one or more embodiments. Computer-readable storage media, which are inherently non-transitory, may include volatile and non-volatile, removable and non-removable tangible media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media may further include RAM, ROM, erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other solid-state memory technology, portable compact disk read-only memory (CD-ROM) or other optical storage devices, cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and is readable by a computer. The computer-readable program instructions can be downloaded from the computer-readable storage medium to a computer, another type of programmable data processing apparatus, or another device, or downloaded via a network to an external computer or external storage device.

[0154] The computer-readable program instructions stored in a computer-readable medium can be used to direct a computer, other types of programmable data processing devices, or other devices to operate in a specific manner so that the instructions stored in the computer-readable medium produce an article of manufacture including instructions for implementing the functions, actions, and / or operations specified in the flowchart or diagram. In certain alternative embodiments, the functions, actions, and / or operations specified in the flowchart and diagram can be reordered, processed serially, and / or processed in parallel consistent with one or more embodiments. In addition, any flowchart and / or diagram may include more or fewer nodes or blocks than those illustrated consistent with one or more embodiments.

[0155] While the present invention has been described in detail through various embodiments, and while these embodiments have been described in considerable detail, it is not the intention of the applicant to restrict or in any way limit the scope of the appended claims to such detail. Additional advantages and modifications will readily become apparent to those skilled in the art. Therefore, the invention in its broader aspects is not limited to the specific details, representative apparatus and methods, and illustrative examples shown and described. Accordingly, departures from such details may be made without departing from the spirit or scope of the overall inventive concept.

Claims

1. A current domain analog front end (AFE) circuit for a LIDAR system, comprising: a photodetector configured to provide a current; a current feedback DC servo loop configured to eliminate a DC component of the current; a feedback network configured to eliminate an AC component of the current; A floating class AB output stage driving the feedback network; and A preamplifier is configured to drive a floating class AB output stage, wherein the preamplifier is driven by an error signal from a feedback network and creates an AC signal path with the feedback network and the floating class AB output stage, wherein a current feedback DC servo loop sets a frequency response of the AC signal path at a high-pass frequency corner based on an input impedance of the current feedback DC servo loop with a low-pass frequency corner based on a transconductance of the floating class AB output stage.

2. The circuit of claim 1 , wherein the floating class AB output stage comprises a dynamically degenerated current mirror.

3. The circuit of claim 1 , wherein the feedback network comprises a switch coupled between an input and an output of the floating class AB output stage.

4. The circuit according to claim 3, wherein In response to an absolute value of the output of the floating class AB output stage exceeding twice a threshold voltage of a level shift switch of the feedback network, the feedback network is configured to transition from capacitive operation to resistive operation based on a resistive path of a switch controlled by the level shift switch.

5. The circuit of claim 1, wherein the preamplifier includes a bias voltage regulation loop that sets an input bias voltage for the floating class AB output stage. The circuit of claim 1 , wherein the photodetector is an avalanche photodiode.

7. The circuit of claim 6 , wherein the error signal is based on an output current (I APD )The output current of the feedback network (I FB，AC )reduce.

8. A method of operating a LIDAR system, comprising: providing a current via a photodetector; Eliminate the DC component of the current via a current feedback DC servo loop; Eliminate the AC component of the current via the feedback network; Drives the feedback network via a floating class AB output stage; driving a floating class AB output stage via a preamplifier, wherein the preamplifier is driven by an error signal from a feedback network; and The frequency response of the AC signal path is set by a current feedback DC servo loop at a high pass frequency corner based on the input impedance of the current feedback DC servo loop with a low pass frequency corner based on the transconductance of the floating class AB output stage.

9. The method of claim 8, wherein the feedback network comprises a switch coupled between an input and an output of the floating class AB output stage.

10. The method of claim 9 , further comprising configuring the feedback network to transition from capacitive operation to resistive operation based on a resistive path of a switch controlled by the level-shift switch in response to an absolute value of the output of the floating class AB output stage exceeding twice a threshold voltage of a level-shift switch of the feedback network.

11. The method of claim 8, wherein the preamplifier includes a bias voltage regulation loop that sets an input bias voltage for the floating class AB output stage.

12. The method of claim 8, wherein the floating class AB output stage comprises a dynamically degenerated current mirror.

13. The method of claim 8, wherein the photodetector is an avalanche photodiode.

14. The method according to claim 13, wherein the error signal is based on the output current (I APD )The output current of the feedback network (I FB，AC )reduce.

15. A circuit for filtering a signal corresponding to the time of flight (TOF) of light from a laser reflected from an object to a photodetector, the circuit comprising: a preamplifier configured to receive a signal from the photodetector, the signal corresponding to an output of the laser reflected from an object remote from the laser and the photodetector; a DC cancellation loop comprising a current feedback DC servo loop; and an AC cancellation loop including a feedback network driven by a floating class AB output stage, and a preamplifier configured to drive the floating class AB output stage, wherein the preamplifier is driven by an error signal from a feedback network and creates an AC signal path together with the feedback network and a floating class AB output stage, wherein a current feedback DC servo loop sets a frequency response of the AC signal path at a high-pass frequency corner based on the input impedance of the current feedback DC servo loop with a low-pass frequency corner based on the transconductance of the floating class AB output stage.

16. The circuit of claim 15, wherein the feedback network comprises a switch coupled between the input and the output of the floating class AB output stage.

17. The circuit of claim 16, wherein: In response to an absolute value of the output of the floating class AB output stage exceeding twice a threshold voltage of a level shift switch of the feedback network, the feedback network is configured to transition from capacitive operation to resistive operation based on a resistive path of a switch controlled by the level shift switch.

18. The circuit of claim 15, wherein the preamplifier includes a bias voltage regulation loop that sets an input bias voltage for the floating class AB output stage.

Images