Stanford EE259 I Lidar range and direction of arrival estimation I 2023 I Lecture 18

03 Feb 2024 (9 months ago)

Photodiodes

Pin photodiodes are used to detect light in lidars.
The intrinsic material in a pin photodiode generates photocurrent when photons are absorbed.
The total current through a pin photodiode is the sum of the dark current and the photocurrent.
Reverse bias is applied to eliminate the dark current and make the total current proportional to the photocurrent.
The voltage-current characteristic of a pin photodiode can be divided into three regions: forward bias, reverse bias, and breakdown.
In the reverse bias region, the pin photodiode can operate in three modes: photoconductive, non-linear, and Geiger mode.
The photoconductive mode offers a linear response with a responsibility of around 1 A/W.
The non-linear mode, also known as avalanche photodiode (APD) mode, provides high responsivity but with a non-linear response.
The Geiger mode, achieved by operating beyond breakdown voltage, offers infinite responsibility but has an on-off switching behavior.

Photodiodes can be biased in three different modes: photovoltaic, photoconductive, and Geiger mode.
Reverse biasing a photodiode with a negative V bias results in a linear response.
A transimpedance amplifier (TIA) is used to achieve a constant negative bias voltage and convert the current generated in the photodiode to a voltage.
The output of a TIA is a voltage type output, which is compatible with ADCs for digitization.
The frequency response of a TIA is a low pass filter with a cutoff frequency Omega zero.
The output voltage of a TIA is the inverse Fourier transform of the frequency response multiplied by the Fourier transform of the incident optical power.

Direct detection lidar (TOF lidar) uses a single photodetector and detects the received optical power.
Direct detection lidar is simpler in hardware and algorithms and is more commonly used in autonomy.
State-of-the-art direct detection lidar can see up to 400 meters, with an average range of 150-200 meters.
Range resolution is typically a few centimeters, and the field of view can be up to 360 degrees horizontally and 50-70 degrees vertically.
DOA resolution can be as good as 0.05 degrees, and the measurement rate is typically 10-30 Hertz.
Direct detection lidar systems typically consume a few tens of watts.
Direct detection lidars cannot measure velocity because the Doppler shift is lost in the detection process.
Different pulse shapes can be used in direct detection lidars, but typically, short pulses are used to achieve high resolution.

Coherent detection lidar detects both the amplitude and phase of light using a coherent optical receiver.
Coherent lidar achieves better SNR than direct detection lidar because the signal is proportional to 1/R for coherent lidar and 1/R^2 for direct detection lidar.
Coherent lidar systems send very short pulses (3-10 nanoseconds) and wait for echoes, similar to pulse radar.
Coherent lidars use phase or frequency modulation instead of amplitude modulation like direct detection lidars.
Coherent lidars are usually monostatic, using the same optics for transmission and reception.
A beam splitter or circulator is used to separate the transmitted and received signals.
The coherent receiver mixes the received signal with a local oscillator to obtain a complex electrical signal proportional to the full electric field.
Coherent detection uses multiple photodiodes to extract the real part of the complex electric field.
Combining the outputs of these two balanced photodetectors provides a full IQ receiver for coherent lidar.
The 90-degree optical hybrid is a passive device that generates the four outputs required for the balanced photodetectors.
Coherent lidar can transmit and receive complex signals, which enables the use of the same waveforms and DSP techniques as in radar.
FMCW and PMCW lidars can be built using coherent lidar technology.
Coherent lidars continuously transmit periodic waveforms and can use frequency chirps (FMCW) or ranging codes (PMCW) for range and Doppler estimation.

Lidar and radar signal processing algorithms are similar because they use the same signal model and estimate the same parameters.
Lidar beams are much narrower than radar beams, making lidar DSP easier as it can assume there is only one target in the signal.
Doppler shifts in lidar are much higher than in radar due to the shorter wavelengths of lidar.
The high Doppler shifts in lidar make it difficult to ignore Doppler and only use phase shifts for estimation, unlike in radar DSP.