搞定车载激光雷达VCSEL驱动!思瑞浦发布高压大电流高速驱动技术白皮书Solved Automotive LiDAR VCSEL Driving! Silergy Publishes High-Voltage, High-Current, High-Speed Drive Technology White Paper


AI free token/AI免费积分

激光雷达需在纳秒量级内向VCSEL注入数十安培的精准电流,既要保证探测距离,又不能在任何情况下灼伤人眼——难点大多集中在发射端这颗驱动芯片上。本文从电压预算、驱动架构、人眼安全到选型,拆解它为何需要60V–80V耐压、20A–50A峰值、ns级边沿。

 

图片1.png

图1 车载LiDAR扫描架构分类与演进——机械式/半固态/纯固态,趋势向纯固态电子扫描收敛

 

车载LiDAR正转向纯固态2-D可寻址VCSEL阵列,对发射端多通道集成要求更高。dToF测回波飞行时间,发射驱动须在精确时刻注入高峰值、窄脉宽、快边沿脉冲。

图片2.png

图2 LiDAR系统框图——涵盖发射Tx、接收Rx、电源转换、系统监控与接口,其中蓝色为3PEAK产品覆盖范围

 

一、为什么发射驱动需要60V–80V?

图片3.png

图3 dToF测距信号链——TX发射、目标回波、RX接收到信号处理

 

VL = L × ( di / dt )

核心约束可用一个公式概括:电流变化越快,回路寄生电感产生的感性压降越大——每nH电感都按V=L·di/dt占去一截电压裕量,同时拖慢脉冲边沿。而边沿每增加1ns,测距误差约增大15cm,约一本书的宽度。

 

 

  • 驱动电压预算:六结、20A、1ns上升时间基线

2.png

各项压降逐一叠加(见上表),六结即达四十余伏;随多结VCSEL向八结、十结演进、电流加大,电压预算很快逼近80V。约束由此明确:高电压、低电感、快开关。

 

二、几十安培,怎么一瞬间放出来?

tu4.jpeg

图4 2-D Flash LiDAR TX 驱动系统架构——Boost升压、TPM8909Q高边充电、TPM8918Q低边脉冲与VCSEL阵列

采用“高边充电加低边脉冲”架构,其原理类似相机闪光灯:由高边电路预先将能量储入本地电容,再由低边开关在发射瞬间释放,从而把慢速充电与快速脉冲解耦,停止充电即切断能量源。整条链路含四个模块:Boost升压、高边充电IC、储能电容阵列、低边脉冲IC(见上图)。

 

  • 充电拓扑与封装

充电拓扑有恒流与谐振两种方式:恒流方案电路成熟、时序确定、EMC表现好;谐振方案借LC谐振转移能量、效率更高,适合大电容、高帧率,代价是电路更复杂。封装则在物理上决定回路寄生电感——集成度越高、键合线越少,寄生越小,而在大电流、快边沿下寄生会直接抬高关断过冲,故低寄生封装对高压快脉冲尤为关键。

 

  • 2-D 阵列驱动

阵列驱动是2-D固态路线的核心。传统1-D线阵加机械扫描存在三处短板:任意时刻只有一列发光、光子利用率低,热耗集中在同一组发射器、限制峰值功率,外加扫描光学的损耗与可靠性风险。2-D可寻址 VCSEL面阵改以电子方式逐区点亮,把光能集中到目标区域,单个发射器每帧只承担一小部分脉冲、热负载下降,并省去运动光学。

由此带来的一项关键能力是高反膨胀抑制:车牌、路牌等强反射目标的回波会使接收端饱和、向邻近像素溢出,模糊甚至淹没周边目标,而 2-D逐区可寻址可对这些已知高反区域单独调低驱动电流或发射器数量,从发射端做区域级功率控制、从源头抑制,又不牺牲其它区域的探测能力。

 

实现上有两处关键设计:行列寻址让阵列按“行+列”寻址、控制线数大幅减少,多颗芯片级联即可覆盖整片阵列;双寄存器组的乒乓机制在当前行发光期间预写下一行参数、切换即时生效,几乎不占用时序。系统再以大小波交替兼顾远近——远距用大波保证回波信噪比,近距用小波避免接收端饱和。

 

  • 关得太快,会出什么问题?

tu5 (2).png

图5 TPM8915Q实测光脉冲——脉宽约 6.1ns、上升 / 下降约 1.8ns,关断后可见振铃引起的二次发光

关断越快脉宽越窄,但回路电感对负向di/dt同样按V=L·di/dt,带来两类风险。一是二次发光:振铃让VCSEL再次正偏,主脉冲后冒出小光脉冲、形成虚假回波,即上图衰减振铃。

tu6.png

图6 TPM8915Q关断电压振铃实测——约56.5A、回路寄生约3nH下节点对地过冲约110V

二是过电压:回路电感像急刹车时的惯性一样抗拒电流骤变——电流被快速切断时,它为维持电流而在两端激起反向电压尖峰,实测可达约110V,既冲击器件耐压、又反偏VCSEL。对策是降低回路寄生、将过冲控制在80V以内,并辅以可编程斜率与续流钳位。

 

三、芯片坏了,会闪伤眼睛吗?

图7.jpeg

图7 正常脉冲模式与低边短路故障模式的对比——后者脉冲退化为连续发光

VCSEL的单脉冲能量本就超出IEC 60825-1 Class1的MPE一到两个数量级,正常工作全靠不到0.1%的占空比,把平均功率压在安全线内。风险在于故障:一旦低边开关短路,激光便从脉冲退化为连续发光(CW),占空比跳至100%、平均功率约等于峰值功率,相对安全平均限值可超标上千倍。因此标准要求:任何单一故障下都不得超过 MPE,并达到车规功能安全ASIL B以上。

 

人眼安全先从识别链路失效模式及其危害入手:

1.png

双芯片架构在此提供了一道独立兜底:即便低边IC彻底失效,高边仍能停止充电并主动放电,待储能电容能量耗尽后激光自然熄灭——这是将全部功能集成于单颗芯片的纯低边方案难以具备的。

 

四、按场景怎么选?

图片8.png

图8 TPM8915Q 集成功率级驱动系统——WLCSP封装内集成 80V/50A功率级,封装寄生小于0.1nH

 

  • Flash固态多通道:高边加低边

高边充电IC TPM8909Q与TPM8909AQ给16通道80V储能与 CVD/RVD诊断,低边脉冲IC TPM8918Q与TPM8918BQ给8通道 20A脉冲,级联覆盖2-D阵列。

 

  • 扫描式/MEMS:GaN驱动

外挂第三方GaN FET配GaN驱动IC:车规TPM1025Q、TPM2025Q,工规TPM1020、TPM1025、TPM2025,用于单或少通道EEL、1-D VCSEL。

 

  • 单通道高性能:集成功率级

TPM8915Q在WLCSP 3.35×1.65mm内集成80V/50A功率级,封装寄生<0.1nH,脉宽可窄至1ns。

全系列覆盖高边充电、低边脉冲、GaN驱动到集成功率级,均有车规与工规版。低成本是规模标配门槛:集成方案把BOM从50+压到20–30颗,下一代充电IC向24通道以上、更高电流、SPI取代并行接口。

 

完整版《激光雷达驱动白皮书》

本文为精简版;完整白皮书含电压预算、架构与封装、大小波与2-D阵列、关断振铃与人眼安全及全线规格。

 

Laser radar requires precise current injection in the nanosecond range, which is critical for both detection range and eye safety. The primary challenge lies in the drive chip's ability to manage high voltage, high current, and fast edges without compromising performance or safety.

图片1.png

Figure 1: Classification and Evolution of Automotive LiDAR Scanning Architectures — Mechanical, Semi-Solid, and Solid-State. The trend is toward pure solid-state electronic scanning, where the ability to address individual emitters is key.

Automotive LiDAR is transitioning to pure solid-state 2-D addressable VCSEL arrays, which demand higher levels of integration and precise timing at the transmitter end. The drive pulses must be sharp, narrow, and occur at exact moments to enable accurate distance measurement.

图片2.png

Figure 2: LiDAR System Block Diagram — Transmitter (TX), Receiver (RX), Power Conversion, System Monitoring, and Interface. The blue areas highlight the components covered by 3PEAK products.

Why Do Transmitter Drivers Require 60V–80V?

图片3.png

Figure 3: dToF Ranging Signal Chain — from TX Emission to Signal Processing. The voltage required is directly proportional to the rate of change in current (di/dt) and the inductance of the circuit.

V_L = L × (di/dt)

This formula indicates that faster switching times and larger current steps lead to higher inductive voltage drops, which must be accommodated within the driver's supply voltage to prevent errors.

Drive Voltage Budget: A Baseline of Six Jumps, 20A, and a 1ns Rise Time

2.png

The voltage budget accumulates from the forward voltage of the VCSEL to the inductive drops in the transmission line. As the number of junctions (from 6 to 10) and the current amplitude increase, the required driving voltage quickly exceeds 60V, necessitating a robust 80V design.

How Are Dozens of Amps Released Instantly?

tu4.jpeg

Figure 4: 2-D Flash LiDAR TX Drive System Architecture — featuring a Boost converter, High-Side charging (TPM8909Q), Low-Side pulsing (TPM8918Q), and the VCSEL array.

The system employs a "High-Side Charge, Low-Side Pulse" architecture, akin to a camera flash. Energy is stored in a capacitor bank by the High-Side circuit, then released by the Low-Side switch at the exact moment of transmission.

Charging Topology and Packaging

There are two main charging topographies: Constant Current (CC) and Resonant. CC is standard and well-understood, while Resonant charging can transfer energy faster due to its sinusoidal nature, though it requires a more complex circuit. The packaging must minimize parasitic inductance to allow for these high-speed transitions.

2-D Array Driving

The shift from 1-D lines (where only one row lights up at a time) to 2-D arrays allows for parallel operation and better energy distribution. This improves the Signal-to-Noise Ratio (SNR) and allows for dynamic power control (e.g., reducing intensity on bright surfaces to prevent saturation).

Row and Column Addressing

The 2-D array is addressed using a grid system where each point corresponds to an emitter. This allows for the precise direction of the laser pulse and the ability to adjust power levels dynamically based on the specific target characteristics.

What Happens if the Signal Turns Off Too Fast?

tu5 (2).png

Figure 5: Measured Optical Pulse of the TPM8915Q — showing a clean 6.1ns pulse with distinct rise and fall times.

Turning the signal off too quickly can lead to several issues, most notably "ringing" or oscillations in the voltage.

tu6.png

Figure 6: Shutdown Voltage Ringing of the TPM8915Q — illustrating the overvoltage spike caused by the rapid change in current interacting with the circuit's inductance.

This ringing can cause the laser to "turn back on" momentarily (secondary emission) or push the junction voltage into dangerous reverse-breakdown territory. To mitigate this, the circuit must be designed with low inductance and may incorporate programmable slope control or active clamping.

Would a Broken Chip Blind Someone?

图7.jpeg

Figure 7: Comparison of Normal Pulse Mode vs. Low-Side Short Circuit Fault — demonstrating the difference between a controlled pulse and a continuous wave (CW) fault state.

The primary risk is not the normal operation, which is well within safety limits, but rather a hardware failure. A shorted transistor on the low-side could cause the laser to remain on continuously (CW), resulting in a massive increase in average power that would likely exceed safety thresholds.

To ensure safety, the system utilizes a dual-chip architecture. Even if the low-side switch fails, the high-side control will detect the error and prevent further charging, ensuring the laser does not remain active indefinitely.

1.png

图片8.png

Figure 8: TPM8915Q Integrated Power Stage Drive System — highlighting the compact WLCSP package and the integrated power stage.

图片8.png

For Scanning/MEMS Systems: Pair discrete GaN FETs with specific gate drivers (TPM1025, TPM2025) for higher voltages and simpler single-channel setups.

For Single-Channel High Performance: The TPM8915Q offers an integrated solution with a power stage and very low parasitic inductance, suitable for precise, high-current pulses.

Single-channel high-performance: integrated power stage

扫描式/MEMS:GaN驱动

Standalone third-party GaN FET paired with GaN driver IC: automotive grade TPM1025Q, TPM2025Q; industrial grade TPM1020, TPM1025, TPM2025; for single or few-channel EEL, 1-D VCSEL applications.

Full lineup covers high-side charging, low-side pulsing, GaN driving to integrated power stages, all with automotive and industrial grades. Low cost is the scalability threshold: integrated solutions reduce BOM from 50+ to 20–30 components; next-gen charging ICs aim for 24+ channels, higher currents, and SPI replacing parallel interfaces.

TPM8915Q integrates 80V/50A power stage in WLCSP 3.35×1.65mm, with package parasitics <0.1nH and pulse width down to 1ns.

Full lineup covers high-side charging, low-side pulsing, GaN drivers to integrated power stages, all with automotive and industrial versions. Low cost is the threshold for mass deployment: integrated solutions reduce BOM from 50+ to 20–30 parts. Next-gen charging ICs target 24+ channels, higher current, and SPI over parallel interfaces.

Complete “Lidar Driver White Paper”

This is a condensed version; the full white paper includes voltage budgeting, architecture and packaging, large- and small-wave arrays, turn-off ringing and eye safety, and full specifications.

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry