Solid State Automotive LiDAR: Physics Principles, Design ...

🄫 Hamamatsu Photonics K.K. and its affiliates. All Rights Reserved. 1

Solid State Automotive LiDAR: Physics Principles,

Design Challenges, and New Developments

Slawomir Piatek

New Jersey Institute of Technology &

Hamamatsu Photonics, Bridgewater, NJ

06.2. 2020

2🄫 Hamamatsu Photonics K.K. and its affiliates. All Rights Reserved.

■ LiDAR Concepts: ToF & FMCW

■ Light Sources & Beam Steering

■ Solid State (Flash) LiDAR

Index


LiDAR Concepts


Basic Layout of Time of Flight LiDAR

start pulse

stop pulse

laser

target

photodetector

timer

beam steering

system

Distance d = Δ𝑡 ∙ 𝑐/2

Measure the time of flight Δ𝑡


FMCW LiDAR: Heterodyne Mixing

tunable

laser

BS

M

receiving optics

collimator

PD

returned light

M

M

M

M

to target

optical mixing

occurs on the

detector

frequency

shifter

electronics

Legend:

𝑓𝐿𝑂 – Local oscillator frequency

𝑓𝑜𝑓𝑓𝑠𝑒𝑡 – Offset frequency added by the frequency shifter, ~10 – 100 MHz

𝑓𝑃𝑂 – Power oscillator frequency of transmitted light

𝑓𝑎 – Frequency of returned light; Δf due to distance and Doppler’s effect

These are instantaneous values

𝑓𝐿𝑂

𝑓𝑃𝑂 = 𝑓𝐿𝑂 + 𝑓𝑜𝑓𝑓𝑠𝑒𝑡

𝑓𝑎 = 𝑓𝐿𝑂 + 𝑓𝑜𝑓𝑓𝑠𝑒𝑡 + Δ𝑓


Example of Frequency Modulation: Double Linear Ramp

time

frequency

T ~ 10’s μs – 1 ms, B ~ 100’s MHz – 10’s GHz

0time

frequency

𝐵𝑓0 𝑓0𝑓𝑚𝑎𝑥 𝑓𝑚𝑎𝑥

0 𝑇 2𝑇

𝑓0

𝑓𝑚𝑎𝑥

𝑇 2𝑇


Frequency Shift in FMCW LiDAR

time

local oscillator

received wave

𝑑 =𝑐 𝑓𝐵1 + 𝑓𝐵2 𝑇

8𝐵𝑣𝑟 =

𝑐 𝑓𝐵2 − 𝑓𝐵14𝑓0

𝛿𝑑 =𝑐

2𝐵𝛿𝑣𝑟 =

𝑐

𝑓0𝑇

Δ𝑡

𝑓𝐵1

𝑓𝐷

𝑓𝐵2

𝑇

𝐵

𝑓0

𝑓(𝑡)

𝑓𝑚𝑎𝑥


Heterodyne Optical Mixing

PDBS

fa

fLO

measure thisamplification!

𝐸𝑡𝑜𝑡2 = 𝐸𝑎 + 𝐸𝐿𝑂

2 = 𝐴𝑎 cos 2𝜋𝑓𝑎𝑡 + 𝜑𝑎 + 𝐴𝐿𝑂 cos 2𝜋𝑓𝐿𝑂 + 𝜑𝐿𝑂2

𝐸𝑡𝑜𝑡2 = 𝐸𝑎

2 + 𝐸𝐿𝑂2 + 𝐴𝑎𝐴𝐿𝑂 cos 2𝜋 𝑓𝑎 − 𝑓𝐿𝑂 𝑡 + 𝜑𝑎 − 𝜑𝐿𝑂

𝑃𝑠𝑖𝑔 = 𝑃𝑎 + 𝑃𝐿𝑂 + 2 𝑃𝑎𝑃𝐿𝑂 cos 2𝜋 𝑓𝑎 − 𝑓𝐿𝑂 𝑡 + 𝜑𝑎 − 𝜑𝐿𝑂

𝑖𝑠𝑖𝑔 =𝜂𝑒𝑃𝑠𝑖𝑔

ℎ𝑓= 𝑖𝑎 + 𝑖𝐿𝑂 + 2 𝑖𝑎𝑖𝐿𝑂 cos 2𝜋 𝑓𝑎 − 𝑓𝐿𝑂 𝑡 + 𝜑𝑎 − 𝜑𝐿𝑂

Δ𝑓 = 𝑓𝑎 − 𝑓𝐿𝑂 − 𝑓𝑜𝑓𝑓𝑠𝑒𝑡

Δ𝑓 gives 𝑑 and 𝑣𝑟


Comparison Between ToF & FMCW Concepts

ToF FMCW

Pros

Easy optical layout

Easy distance calculation

Any wavelength of light can be used Pros

Optical amplification of the returned signal

Photon shot noise detection possible

Immunity to background and interference

Cons

Large detection bandwidth →increased noise

Weak returned signal

Susceptible to background and

interference Cons

Complex optical layout

Expensive tunable laser with a long

coherence length needed

Complex distance and velocity calculation

Gives both distance and radial velocity

of the target


Light Sources & Beam Steering


Light Sources

ToF FMCW

High peak power: ~ 100 W

Short-duration pulses: ~ few ns

Repetition period: ~ ms - μs

Wavelength: NIR, e.g., 905 nm, 1550 nm

Tunable output frequency

Coherence length 𝐿 > 2𝑑𝑚𝑎𝑥

Stable phase

some light source offerings by Hamamatsu


905 or 1550?

905 nm 1550 nm

Figure from Matson et al. 1983

Solar irradiance at sea level

𝑃𝐵 @ 905 nm > 𝑃𝐵 @ 1550 nm S

pect

ral i

rrad

ianc

e μ

W c

m-2

nm-1

Wavelength [nm]

Solar background at 905 nm is higher than at 1550 nm


905 or 1550?

H2O absorption @ 1550 nm > (100×) @ 905 nm

From Wojtanowski et al. 2014

Wavelength [nm]

Ab

so

rptio

n c

oe

ffic

ien

t o

f w

ate

r [c

m-1

]

Reference: “Comparison of 905 nm and 1550 nm semiconductor laser rangefinders’ performance deterioration due to adverse

environmental conditions,” Wojtanowski et al. 2014


905 or 1550?

1550 nm

- Requires IR (non-silicon) photodetectors

+ Best eye safety

905 nm

+ Lower background

+ Better transmission in atmosphere

+ Silicon-based photodetector

+ Coherence length 𝐿 ∝𝜆2

𝐵


Mechanical Beam Steering: Rotating Platform

𝜔

Each laser is matched with its own detector Scan direction

Horizontal sweep

Vertical sweep


Mechanical Beam Steering: Galvo Mirrors

ω

ω

Laser

𝑥 − 𝑦 scan

Rotating Galvo mirrors are another example of a mechanical beam steering.


Beam Steering: Optical Phase Array

Laser

Amplitude & phase

control unit

Beam in the far field

Amplitude and phase of the light emitted by

each pixel (emitter) can be controlled

electronically.

Optical phased array is an example of a solid state or ”no moving parts” beam steering.


OPA Emitter and Receiver

Light feed for

emission

Local oscillator

light for optical

amplification

emitted beam detected beam

Control unit

Current output


LiDAR Based on OPA Emitter and Receiver

First prototype of a FMCW LiDAR that uses

optical phased arrays for beam steering and

light detection.

The figure is from “Coherent solid-state LIDAR with silicon

photonic optical phased arrays” by Poulton et al. 2017


Beam Steering: Flash and Structured Light

Array of lasers such as

VCSELs

Laser and pulse

expander

Divergent pulse of light

■ Wider the angle, smaller the surface brightness

■ For a Gaussian beam, the surface brightness is

not uniform

■ Lateral resolution limited by the 2D sensor

■ Beams have the same intensity

■ Lateral resolution limited by the angular

separation between the beams


Solid State (Flash) LiDAR


Basic Layout of ToF Flash LiDAR

projector

to the target𝜃𝑒

𝜃𝑑

focal plane array

control unit

𝑓

The emission FOV, 𝜃𝑒, should be matched with the detection FOV, 𝜃𝑑.

2D detector


Focal Plane Distance Measurement

Focal plane

image of the scene

detector array

A single “pixel” in the 2D detector determines distance to a single element of the scene.


Distance Resolution

𝛿𝑑 =𝑐

2𝐵

𝛿𝑑

𝐵~1

𝑇𝑇 − pulse duration

For 𝑇 = 5 ns, 𝛿𝑑 ≈ 0.75 m → 𝐵 ≈ 200 MHz

(distance resolution)

Better distance resolution requires pulses of even shorter duration. The limitation is in the laser

technology (parasitic inductance).


Distance Uncertainty

𝜎𝑑2~

𝛿𝑑 2

𝑆/𝑁=

𝑐2

4𝐵2𝑆/𝑁

𝑑

𝜎𝑑

𝑑

(distance uncertainty)

1. The larger the 𝑆

𝑁, the smaller the uncertainty, all else being the same

2. Photodetector and electronic time jitters also contribute to the uncertainty.


Photon Budget: Single Photodetector

TX

𝑃 𝑑 = 𝑃0𝜌𝐴0𝜋𝑑2

𝜂0𝑒−2𝛾𝑑

𝑃 𝑑 − Peak power received

𝑃0 − Peak power transmitted

𝜌 − Target reflectivity

𝐴0 − Aperture area of the receiver

𝜂0 − Receiving optics transmission

𝛾 − Atmospheric extinction coefficient

Lambertian reflection

Laser spot smaller than the target

Normal incidence

𝛾 is constant

Assumptions:


Photon Budget: Single Photodetector

𝑃 𝑑 = 𝑃0𝜌𝐴0𝜋𝑑2

𝜂0𝑒−2𝛾𝑑

Example: 𝑃0 = 100 W, 𝜌 = 0.1, 𝐴0 = 3.14 × 10−4 m2, 𝜂0 = 0.5, 𝛾 = 0.5 km-1

for 𝒅 = 𝟏𝟎𝟎 m, 𝑷 = 𝟒𝟓 nW

Reference: “Comparison of 905 nm and 1550 nm semiconductor laser rangefinders’ performance deterioration due to adverse

environmental conditions,” Wojtanowski et al. 2014

1. Atmospheric extinction 𝛾 depends on weather conditions and wavelength. Its value can range

from about 4 km-1 to about 0.1 km-1 at 905 nm.

2. For a square 5-ns pulse (𝜆 = 905 nm), the number of emitted photons is ~2.3 × 1012 and the

number of received is ~1 × 103.


Photon Budget: Flash

TX

RX

There is a tradeoff between spatial resolution and 𝑆

𝑁.

𝐴


Photon Budget: Flash

𝑃 𝑑 = 𝑃0𝜌𝜃𝑝2

Θ2𝐴0𝜋𝑑2

𝜂0𝑒−2𝑑𝛾

(received peak power per pixel)

𝜃𝑝 − angular subtense (field of view) of a pixel

Θ − angular field of view of the pulse projector (pulse divergence)

Reference: “Comparison of flash lidar detector options,” McManamon et al. 2017

Note that 𝜃𝑝 ≪ Θ, so if the pulse peak power is the same, the amount of light received by a single

pixel is proportional to 𝜃𝑝2

Θ2for a given distance 𝑑.


Photodetection (Single Element)

TIA PD

Trigger

circuit

Pulse of light

Optical bandpass filter

Focal plane

To timerFOV

■ Active area of the photodetector, focal length of the lens, and placement of the optical bandpass

filter determine the photodetector’s field of view.

■ Avalanche photodiode or silicon photomultiplier are commonly used photodetectors.


Noise in Transimpedance Amplifier

−

+

𝑣𝑛

𝑅𝐹

𝐶𝑓

𝑖𝐽

𝐶𝑜𝑝𝑖𝑛

𝑒𝑛

𝑖𝑑𝑖𝑝ℎ

𝑅 = 𝑅𝑆𝑖𝑃𝑀||𝑅𝑜𝑝

𝐶𝑡 + 𝐶𝑆𝑅

𝑣𝑛 = 𝑒𝑛2 1 +

𝑅𝐹𝑅

2

+4𝜋2

3Δ𝑓 2𝐶𝑇

2𝑅𝑓2 + 𝑅𝑓

2𝑖𝑇2 + 4𝑘𝑇𝑅𝑓

1/2

Δ𝑓 1/2

𝐶𝑇 = 𝐶𝑡 + 𝐶𝑓 + 𝐶𝑜𝑝 + 𝐶𝑠

𝑖𝑇 = 𝑖𝑛2 + 𝑖𝐽

2 + 𝑖𝑑2 + 𝑖𝑝ℎ

2

All else being equal, noise increases with terminal capacitance of the photodetector.

(total capacitance)

noise output


Bandwidth and Stability of Transimpedance Amplifier

−

+

𝑣𝑜𝑢𝑡

𝑅𝐹

𝐶𝑗

Simplified equivalent circuit of a photodetector

connected to an uncompensated TIA

𝐼

𝑣𝑜𝑢𝑡 = 𝐼−𝑅𝑓

1 +1

𝐴𝑜𝑙𝛽

𝐴𝑜𝑙 − Open loop gain of TIA

𝛽 𝑗𝜔 =1

1 + 𝑗𝜔𝑅𝐹𝐶𝑖

where 𝐶𝑖 = 𝐶𝑗 + 𝐶𝑜𝑝

1

𝛽(𝑗𝜔)= 1 + 𝑗𝜔𝑅𝐹𝐶𝑖

𝑓𝐹 𝑓𝑖 𝑓𝐺𝐵𝑊𝑃

Gain peaking and

oscillations occur around

this frequency

𝑓𝐺𝐵𝑊𝑃 − Unity gain bandwidth of the op-amp


Bandwidth and Stability of Transimpedance Amplifier

Simplified equivalent circuit of a compensated

photodiode connected to an uncompensated TIA

𝐶𝐹

𝑓𝐹 𝑓𝑖 𝑓𝐺𝐵𝑊𝑃

𝑓𝐹 =1

2𝜋𝑅𝐹 𝐶𝑖 + 𝐶𝐹𝑓𝑖 =

1

2𝜋𝑅𝐹𝐶𝐹

𝐶𝐹 =1

4𝜋𝑅𝐹𝑓𝐺𝐵𝑊𝑃1 + 1 + 8𝜋𝑅𝐹𝐶𝑖𝑓𝐺𝐵𝑊𝑃

(Optimal value of the compensating capacitor)

−

+

𝑣𝑜𝑢𝑡

𝑅𝐹

𝐶𝑗𝐼

𝛽 𝑗𝜔 =1 + 𝑗𝜔𝑅𝐹𝐶𝐹

1 + 𝑗𝜔𝑅𝐹(𝐶𝑖+𝐶𝐹)


Importance of Noise and Bandwidth

time

High bandwidth → higher noise but high fidelity

trigger level

trigger time

time

Low bandwidth → lower noise and lower fidelitytrigger level

trigger time

flat top


Importance of Excess Noise (F)

time

Sig

nal

Fixed trigger level

Constant fraction trigger

Waveform 1

Waveform 2

Fixed trigger level gives different round-trip-times, Δ𝑡1 ≠ Δ𝑡2 ✖

Constant-fraction trigger gives the same round-trip-times, Δ𝑡1 = Δ𝑡2 ✔

Δ𝑡1 = Δ𝑡2

Δ𝑡1 Δ𝑡2


Takeaway

Analog photodetection in ToF LiDAR, especially in flash LiDAR, is very challenging.

Is there anything else we can do?

What about a statistical measurement using SPAD?


Single-Photon Avalanche Photodiode (SPAD)

steady state

transient

quench

recharge

𝑅𝑄 must be large enough to ensure quenching

quasi-stable “ready”

state

(Quench resistor)

(Load resistor)

𝑅𝑄

𝑉𝐵𝐼𝐴𝑆

𝑅𝑙

𝑆𝑃𝐴𝐷

𝑅𝑄 ≫ 𝑅𝑙

𝐼𝑆𝑃𝐴𝐷

𝑉𝐵𝐼𝐴𝑆𝑅𝑄

𝑉𝐵𝐷 𝑉𝐵𝐼𝐴𝑆 𝑉𝑆𝑃𝐴𝐷

𝑣𝑜


Measuring Distance with a Single SPAD

t = 0

time

timer

targetSPAD

laser

Puls

e e

mis

sio

n

𝑡 = 𝑇

Δ𝑡 =2𝑑

𝑐


Measuring Distance with a Single SPAD

time

𝑡 = 0 𝑡 = 𝑇Δ𝑡 =

2𝑑

𝑐

Multiple pulse illumination provides distance information to the target. The information comes

from a histogram of trigger times.

Histogram of trigger times


SPAD Arrays

Micro-bumps

ASIC

(application-specific integrated circuit)

Photodetector array


Detection Techniques with a SPAD Array

Time gating:

𝑡 = 0 𝑡 = 𝑇

Time window

Only the events in the pre-defined time window are counted. The choice of the time window

depends on the expected knowledge of the target distance.

Pixel 1

Pixel 2

Pixel 3



Temporal and spatial correlation:

𝑡 = 0 𝑡 = 𝑇𝑡 = 𝑡1

𝑡 = 𝑡1

𝑡 = 𝑡1

Event not counted: temporal

correlation but no spatial

correlation.

Pixel 1

Pixel 2




𝑡 = 0 𝑡 = 𝑇𝑡 = 𝑡1

𝑡 = 𝑡1

𝑡 = 𝑡2

Event not counted: spatial correlation but

no temporal correlation.

𝑡 = 𝑡2

Pixel 1

Pixel 2




𝑡 = 0 𝑡 = 𝑇𝑡 = 𝑡1

𝑡 = 𝑡1

𝑡 = 𝑡1

Event counted: spatial correlation but

no temporal correlation.

Pixel 1

Pixel 2


SPAD Pixel for Correlated Detection


SPAD Pixel for Correlated Detection

Pulse

coincidence

detection

Signal photon

Background photon


Comments

1. This technique can be used both in a scanning and flash LiDAR.

2. In a scanning LiDAR, an OPA array is well-suited for beam steering.

3. The greatest advantage is a reduced sensitivity to the background light.

4. Additional advantages: less affected by gain variations, sensitivity in IR (using non-silicon

structures), compatible with CMOS-based ASICs.

5. Challenges: SPAD arrays can exhibit crosstalk and high dark count rates.


Hamamatsu Assists LiDAR Companies

Photodetectors (Silicon or InGaAs, PIN, APD, SiPM, SPAD and more) for all LiDAR concepts

Light sources (PLDs or VCSELs) for selected LiDAR concepts

Custom integrated optical assemblies, from front-end electronics to complete ASICs

Support automotive grade qualifications (AEC, ISO and more)

Full customization of photodetectors, light sources, and optical assemblies

Because of our wide offering of optical components, Hamamatsu is unbiased

when recommending the correct detector and/or light source to each unique

LiDAR concept (customer) in the market.


Photodetectors for LiDARS (850 nm – 940 nm)

Si PIN photodiode Si APD

High Photosensitivity;

Internal Gain = 1


Internal Gain ~ 100

SPPC (or SPAD)

Low Photosensitivity;

Internal Gain ~ 105 to 106

MPPC (or SiPM)

Low Photosensitivity;

Internal Gain ~ 105 to 106


Photodetectors for LiDARS (1550 nm)

InGaAs APD

Wavelength [nm]

Photo

sensitiv

ity [A

/W]


Internal Gain ~10-20

Wavelength [nm]

Photo

sensitiv

ity [A

/W]

InGaAs PIN PD


Internal Gain – 1


Closing Remarks

1. Two distinct LiDAR systems, ToF and FMCW, are actively researched

2. Each system presents a unique set of engineering challenges

3. Beam steering and photodetection are the two most outstanding challenges

4. Flash LiDAR together with a SPAD-based statistical detection is a new avenue of research


Thank you

Thank you for listening

Contact information:

[email protected]


Join Us for 10 Weeks of FREE Photonics Webinars (17 Topics)

Week # Weekly Topics # of Talks Talk #1 Date Talk #2 Date

1 Introduction to Photodetectors 2 26-May-20 28-May-20

2 Emerging Applications - LiDAR & Flow Cytometry 2 2-Jun-20 4-Jun-20

3 Understanding Spectrometer 2 9-Jun-20 11-Jun-20

1 Weeks Break

4 Specialty Products – Introduction to Light Sources & X-Ray 2 23-Jun-20 25-Jun-20

5 Introduction to Image Sensors 2 30-Jun-20 02-Jul-20

1 Weeks Break

6 Specialty Products – Laser Driven Light Sources 2 14-Jul-20 16-Jul-20

7 Image Sensor Circuits and Scientific Camera 2 21-Jul-20 23-Jul-20

8 Mid-Infrared (MIR) Technologies & Applications 2 28-Jul-20 30-Jul-20

1 Weeks Break

9 Photon Counting Detectors – SiPM and SPAD 1 11-Aug-20

10 Using SNR Simulation to Select a Photodetector 1 18-Aug-20

To register and attend other webinar series, please visit link below:

https://www.hamamatsu.com/us/en/news/event/2020/20200526220000.html

Solid State Automotive LiDAR: Physics Principles, Design ...

Documents

Transcript of Solid State Automotive LiDAR: Physics Principles, Design ...