I bet you’ve heard multiple times that “lidar-on-a-chip” is the ultimate goal and you may have even heard announcements of other LiDAR companies saying they’ve got such a chip.
This is something I partially agreed with, as for me it was always clear that using chips to miniaturize, reduce the cost and power consumption, and improve the performance was very desirable and a truism; but I remember thinking “which chip are they talking about?”.
A LiDAR has two jobs to perform: measure distance (ranging) and steering a light (scanning), and it’s a complex system that integrates electronics, optics, integrated photonics, advanced software algorithms for signal processing with the rigor of industrial quality and automotive cost and scale.
Thus, my question “which chip are they talking about?” refers to the fact that the multiple sub-systems in a LiDAR would necessarily integrate to completely different classes of chips and using different processes; for example: Silicon chips using CMOS processes typically used for electronics, or Silica glass chips used to miniaturize optics like lenses and prisms used to steer light, or photonic chips using Silicon, Germanium, Indium-Phosphide, or Gallium-Arsenide to combine or modulate light, convert electrical currents into light (Tx) or viceversa (Rx). So, for a solution to be truly powerful you’d need to take to the chip level all of these sub-systems. Baraja is the only company that does, and this is the impact of the work we’ve conducted for 8 years to have access to all the components of Tx, Rx, Amplification, Processing, Steering on heterogeneous chips.
In today’s post we’re announcing the availability of all these chips that combine to make the world-first Doppler RMCW Spectrum-Scan™ LiDAR. The beauty of the solution is that the technologies developed to get to these chips self-reinforce and make possible further integration in the future.
Light amplification – the SOA
The Semiconductor Optical Amplifier (SOA) integrates the amplifier chip, together with a custom-developed high-efficiency Thermo-electric cooler into a hermetically sealed and automotive-grade package and it’s able to amplify the laser light to achieve the >200m range requirements, while being mass-produceable (chip process) and 50-80x smaller and lower power consumption compared to fiber amplifiers used by legacy 1550nm LiDAR, including our own Off-Road first generation product. We decided to package it separately to make the power dissipation easier to achieve when packaging the SOA into the final LiDAR module.
Light emission and reception – the BOSA
The Bidirectional Optical Sub-Assembly (BOSA) integrates no less than 3 types of chips in an automotive-grade, hermetically sealed and temperature controlled enclosure (3x3.5cm). These chips are: the laser chip (InP), the homodyne receiver chip (silicon photonics) and the TIA chips (silicon). These, plus micro optics and automotive-grade Thermo-electric coolers deliver an end-to-end solution to produce and receive the wavelength-tuneable laser light that is the hallmark of Spectrum-Scan™, so in a sense the BOSA also partially integrates the “steering” job of the LiDAR.
Steering
In Spectrum-Scan™ Baraja’s LiDAR uses the changing wavelength of our laser paired with dispersive optics (e.g. prisms) to steer the light onto different angles based on the wavelength of the light. The same principle enables the optics to capture the light returning from the environment and couple it back into the fiber optic receiver. Thus, Spectrum-Scan™ is used for both transmit and receive, and it inherently builds a spatial and wavelength optical filter that prevents interference from sunlight and other LiDAR sources. At Baraja, we’ve integrated (not shown in the pictures) several of the optical functions of lenses and prisms into optical chips built using inexpensive Silica glass. Our Spectrum HD 2025 product incorporates micro-optics and a silica chip for dispersion and steering; and beyond we’ve developed and tested a waveguide chip that further integrates and miniaturizes most of the optical components into a single chip, this will be available for our next generation.
Processing – not one, but a family of SoC
This is the most “traditional” of our chips, being silicon CMOS chip used for controlling and processing the LiDAR. While we initially spun-up the internal development of an ASIC, we abandoned this in favour of a new SoC family of chips that have recently hit the market from a top chip manufacturer in automotive and other industries. This comes with the added benefit that the automotive qualification is already complete and for us and our customers it gives flexibility to select from the family of SoCs the one most appropriate for the type of LiDAR being designed, i.e. we chose a more powerful SoC for a higher performance (longer range, more points per second) LiDAR, and a lower-grade SoC from the same family for a mid-range low power consumption LiDAR. All these can be built using the same chips that Baraja has developed. Finally, this chip supports over-the-air upgrades to the core processing algorithms much more flexibly than what an ASIC ever could.
Development process – from the forest to the summit
Developing optical sub-assemblies and “chips”, if done well is about a 36-month process. Compared to “simple” electronic silicon chips this process is much more involved and the planning and design for qualification testing must be central to the ethos of the designers and the partners selected (the fabs, suppliers and the assembly manufacturers).
I’m pleased to share that now with our A-Samples available, we’re right on track to launching Spectrum HD 2025 C-samples/General Availability in 2025, and if you want to have access to this truly transformational technology then now is the time to work together: info@baraja.com
