Scope and source note
This note covers digital CMOS lidar: direct time-of-flight ranging systems that use CMOS single-photon detectors and in-pixel or per-pixel time-to-digital conversion to digitize photon arrival times. The strongest technical/performance/IP claims below are supported by granted patents and peer-reviewed journal papers; adoption and business claims are supported by company filings and engineering documentation, which are labeled as self-reported. No SEO content farms, anonymous benchmark sites, or AI summaries are used.
The principal evidence base:
- Granted patent: US10222475B2, assigned to Ouster, Inc.
- Peer-reviewed review: Bruschini et al., Light: Science & Applications 2019, DOI 10.1038/s41377-019-0191-5.
- Peer-reviewed prototype paper: Niclass et al., IEEE Journal of Solid-State Circuits 2013, DOI 10.1109/JSSC.2012.2227607.
- Company regulatory filing: Ouster, Inc. 2023 Form 10-K, SEC EDGAR, CIK 0001816581.
- Company technical documentation: Ouster technology page, self-reported.
Access date for all cited sources: 2026-05-07.
What “digital CMOS lidar” means at the device level
Digital CMOS lidar is usually distinguished from analog lidar by where the signal becomes a number. In an analog lidar receiver, a photodetector such as an avalanche photodiode produces a current or voltage pulse whose amplitude is processed by an analog front-end and later digitized by an ADC. In a digital CMOS lidar receiver, the detector is typically a single-photon avalanche diode (SPAD) array fabricated in a CMOS process, and the detection event is a digital avalanche that is quenched and time-stamped by a time-to-digital converter (TDC) on or next to the pixel.
The Bruschini et al. review describes CMOS SPAD arrays as capable of detecting individual photons with picosecond-class timing and of integrating time-correlated single-photon counting circuits directly into the sensor Bruschini 2019, DOI 10.1038/s41377-019-0191-5. This is a peer-reviewed, DOI-verified source and is Tier 1 for technical claims. The review reports SPAD arrays, in-pixel TDCs, histogramming, and direct time-of-flight ranging as a major application. It does not establish that any specific commercial product achieves those figures; it establishes that the architecture is real and has been demonstrated across multiple research groups.
Niclass et al. provide an independent peer-reviewed prototype demonstration: a CMOS time-of-flight depth sensor implemented in 0.18 μm CMOS with 340×96 pixels, 10 frames/s, and a reported 100 m range Niclass 2013, DOI 10.1109/JSSC.2012.2227607. This is a different institution and different lineage from the Ouster patent and marketing materials; it helps cross-verify the feasibility claim that CMOS direct time-of-flight ranging can operate at useful ranges, though it is a research prototype rather than a commercial automotive lidar.
Ouster’s granted patent US10222475B2 is titled “Optical system for collecting distance information within a field” and is assigned to Ouster, Inc. US10222475B2, granted 2019-03-05. The patent generally claims an optical system with a light source, a detector array, and optical elements arranged to collect distance information from a field. The patent is a recorded claim, not a performance measurement. It supports the IP/moat claim only insofar as it shows Ouster holds a granted patent in the digital lidar/optical collection area. It does not prove the system works, passes automotive qualification, or has better range/accuracy than competing lidar.
Core technical building blocks
2.1 CMOS SPAD arrays
A SPAD is a p-n junction biased above its breakdown voltage in “Geiger mode.” A single absorbed photon can trigger a self-sustaining avalanche, producing a digital pulse. A quenching circuit stops the avalanche and resets the detector. This gives single-photon sensitivity but introduces dead time and afterpulsing Bruschini 2019, DOI 10.1038/s41377-019-0191-5. In CMOS, the detector and readout circuitry can be integrated monolithically, which is the basis for the commercial pitch that digital lidar is more scalable than exotic-CMOS or discrete receiver designs.
The Bruschini review is a synthesis of many SPAD imagers; it reports typical SPAD timing jitter in the tens of picoseconds, TDC resolution in the tens to hundreds of picoseconds, and architectures that include in-pixel histogramming or counter-based accumulation. The review also reports that high background photon flux can saturate SPADs and cause pile-up; this is a major constraint for outdoor lidar. Because the review covers multiple devices, exact jitter/binning values must be regarded as representative, not a single product spec.
2.2 Time-to-digital conversion and histogramming
Distance is calculated from the photon round-trip time:
\[
R = \frac{c \Delta t}{2}
\]
where \(c\) is the speed of light and \(\Delta t\) is the time delay from laser pulse to detected photon. A TDC quantizes \(\Delta t\) into bins. Multiple photon detections are accumulated into a histogram; the peak of the histogram estimates the true target time, suppressing random background photon timestamps Bruschini 2019, DOI 10.1038/s41377-019-0191-5; Niclass 2013, DOI 10.1109/JSSC.2012.2227607.
The quantitative consequence is worked through in Section 5 below.
2.3 Optical architecture and patent claims
US10222475B2 describes an optical arrangement for collecting distance information. It is cited here only as a granted US patent; no claim is made that the patent’s described system was independently benchmarked or that the commercial Ouster sensor exactly implements the claimed optical system. Under the source-tier rules, this patent is Tier 1 for the existence and assignment of recorded claims, but not for performance or commercial viability US10222475B2.
Pending patent applications, and patents not yet granted, are not used here to support moat/feasibility claims. The distinction matters: an application is a claim, not an enforceable right. A grant is still not proof of performance.
Peer-reviewed evidence and independent validation
3.1 Bruschini 2019 review
Citation: Bruschini, C., Homulle, H., Antolovic, I.M., Burri, S., Charbon, E., “Single-photon avalanche diode imagers in biophotonics: review and outlook,” Light: Science & Applications, vol. 8, article 87, 2019. DOI: 10.1038/s41377-019-0191-5. URL: [doi.org . Granted/peer-reviewed: peer-reviewed journal article. Access date: 2026-05-07.
This is not a company source. It reviews the field of CMOS SPAD image sensors, including time-resolved single-photon imaging and direct time-of-flight. It supports these load-bearing technical claims:
- CMOS SPAD arrays can detect single photons and can integrate timing circuits.
- SPAD sensors can operate in high-background environments but suffer saturation/dead-time limits.
- TDC and histogramming architectures are standard in time-resolved SPAD arrays.
These claims are cross-verified in the sense that Bruschini et al. is an independent academic review; but specific commercial values should be sourced to the vendor or a third-party benchmark.
3.2 Niclass 2013 JSSC paper
Citation: Niclass, C., Soga, M., Matsubara, H., Kato, S., Kagami, M., “A 100-m Range 10-Frame/s 340×96-Pixel Time-of-Flight Depth Sensor in 0.18-μm CMOS,” IEEE Journal of Solid-State Circuits, vol. 48, no. 2, pp. 559–572, 2013. DOI: 10.1109/JSSC.2012.2227607. URL: [doi.org . Access date: 2026-05-07.
This paper is peer-reviewed and reports a CMOS direct ToF sensor with:
- 0.18 μm CMOS implementation;
- 340×96 pixel array;
- 10 frame/s;
- 100 m reported range.
Those numbers are specific to the research prototype and are not an automotive production device. They still help disconnect the architecture’s feasibility from purely vendor self-reporting. This is a different institution from Ouster and from the Bruschini review, so it provides independent corroboration that CMOS direct ToF is capable of 100 m class ranging in controlled or favorable conditions.
Company filings and marketing cross-reference
4.1 Ouster 10-K
Ouster, Inc. 2023 Form 10-K was filed with the SEC and is available through EDGAR under CIK 0001816581. URL: [sec.gov . Access date: 2026-05-07.
Ouster’s 10-K is used here as a company regulatory filing, which is appropriate for adoption/deployment context and risk disclosures. It is not an independent technology benchmark. The filing discloses a history of losses, ongoing operating cash use, and uncertainty about profitability. It also describes customer concentration, supply-chain dependence, competition, and long sales cycles. Those are standard risk factors, but they matter for the business part of the deep dive.
Cross-reference with marketing: Ouster’s technology page presents digital lidar as a CMOS-based, scalable architecture Ouster technology page, self-reported. The 10-K does not contradict the existence of the technology; it does caution that the company has a history of losses, faces intense competition, and may not achieve or maintain profitability. Therefore a marketing claim of inherent CMOS cost advantage is not independently confirmed by the filing. The appropriate wording is: the company markets custom CMOS silicon as a cost/scalability advantage, while the filed risk factors warn that this may not translate into positive margins or commercial success.
Second cross-reference: The granted patent US10222475B2 supports the existence of Ouster IP, but the 10-K also discloses that competitors may hold or obtain their own IP, and that Ouster’s patents may not prevent competition. Thus the patent is a moat signal, but not a proven moat.
No financial figures are restated from memory in this note. Any exact revenue, gross-margin, or net-loss figure must be read directly from the 10-K and should be marked needs primary source validation if quoted outside that filing.
Worked quantitative calculation
5.1 Range precision from TDC quantization and photon accumulation
For a direct ToF SPAD lidar, the TDC bin width sets a quantization limit on the measured photon time delay. If the TDC bin width is \(T_{\text{bin}}\), a single photon timestamp has quantization standard deviation:
\[
\sigma_t^{(1)} = \frac{T_{\text{bin}}}{\sqrt{12}}
\]
If \(N\) independent photon timestamps from the same target are averaged, the timing precision improves:
\[
\sigma_t = \frac{T_{\text{bin}}}{\sqrt{12 N}}
\]
The one-way range standard deviation is:
\[
\sigma_R = \frac{c}{2}\sigma_t
= \frac{c T_{\text{bin}}}{2 \sqrt{12 N}}
\]
Inputs:
- \(c = 299,792,458\) m/s (physical constant).
- \(T_{\text{bin}} = 100\) ps \(= 1 \times 10^{-10}\) s. This is a representative TDC bin width for CMOS SPAD arrays; source literature reports TDC bins in the tens to hundreds of picoseconds Bruschini 2019, DOI 10.1038/s41377-019-0191-5. It is an assumed example, not a product spec.
- \(N = 100\) photons per pixel per frame. This is an assumed photon count chosen to illustrate accumulation; it is not a measured value from any product.
Single-photon case \(N=1\):
\[
\sigma_t^{(1)} = \frac{100\text{ ps}}{\sqrt{12}}
= 28.87\text{ ps}
\]
\[
\sigma_R^{(1)} = \frac{299,792,458 \times 28.87 \times 10^{-12}}{2}
\]
\[
\sigma_R^{(1)} \approx 4.33\text{ mm}
\]
So with a 100 ps TDC bin, a single detected photon gives about 4.33 mm of one-standard-deviation range quantization in this simplified model.
Accumulated case \(N=100\):
\[
\sigma_t = \frac{28.87\text{ ps}}{\sqrt{100}}
= 2.887\text{ ps}
\]
\[
\sigma_R = \frac{299,792,458 \times 2.887 \times 10^{-12}}{2}
\]
\[
\sigma_R \approx 0.433\text{ mm}
\]
Sensitivity:
- If \(T_{\text{bin}} = 200\) ps, \(N=1\) gives \(8.66\) mm; \(N=100\) gives \(0.866\) mm.
- If \(T_{\text{bin}} = 50\) ps, \(N=1\) gives \(2.16\) mm; \(N=100\) gives \(0.216\) mm.
This calculation is derived from first principles with inputs marked as assumed where journal/product-specific measured values are not available. It is not a claim that Ouster or any product achieves the 100-photon case outdoors. The photon count \(N=100\) is especially sensitive: if only 10 photons are collected, the accumulated precision degrades to \(1.37\) mm for a 100 ps TDC; if only 1 photon is collected, it is \(4.33\) mm. Therefore, precision claims based on histogram accumulation depend heavily on the signal photon count actually reaching the pixel.
5.2 Max unambiguous range vs. pulse repetition rate
A repeating laser pulse train produces an ambiguity because a photon from an earlier pulse can be associated with a later gate. The maximum unambiguous range for a fixed pulse repetition rate \(f_{\text{pulse}}\) is:
\[
R_{\text{unambig}} = \frac{c}{2 f_{\text{pulse}}}
\]
For \(f_{\text{pulse}} = 1\) MHz:
\[
R_{\text{unambig}} = \frac{299,792,458}{2 \times 1 \times 10^6}
\approx 149.9\text{ m}
\]
For \(f_{\text{pulse}} = 10\) MHz:
\[
R_{\text{unambig}} \approx 14.99\text{ m}
\]
This is derived from the ToF equation and is not sourced to a product. It illustrates why long-range lidar systems must either use low repetition rates, multiple pulse codes, or multi-modal disambiguation.
Parameter table
| Parameter | Value | Source (URL + ID + date) | Tier | Measured/derived/assumed/inferred | Cross-verified? |
|---|---|---|---|---|---|
| Ouster granted patent number | US10222475B2, granted 2019-03-05 | [patents.google.com | Tier 1 | Patent record | Yes, by patent grant number |
| CMOS SPAD single-photon sensitivity | Single-photon avalanche detection demonstrated | Bruschini 2019, DOI 10.1038/s41377-019-0191-5 | Tier 1 | Peer-reviewed review | Yes, multiple cited works within review |
| TDC bin width | 100 ps example | Bruschini 2019, DOI 10.1038/s41377-019-0191-5 | Tier 1 | Assumed/derived | Single-source |
| Accumulated photons per pixel | 100 photons example | Assigned in calculation, not measured | Derived | Assumed | No |
| Single-photon range quantization at 100 ps TDC | 4.33 mm | Derived from \(cT_{\text{bin}}/(2\sqrt{12})\) | Derived | Derived | Calculation only |
| 100-photon range quantization at 100 ps TDC | 0.433 mm | Derived from \(cT_{\text{bin}}/(2\sqrt{12N})\) | Derived | Derived | Calculation only |
| CMOS ToF prototype range | 100 m reported | Niclass 2013, DOI 10.1109/JSSC.2012.2227607 | Tier 1 | Peer-reviewed measured prototype | Single-source |
| Ouster profitability | History of losses; no assurance of profitability | Ouster 2023 Form 10-K, SEC EDGAR | Tier 2 | Filed regulatory disclosure | Single-source |
| Ouster marketing CMOS scalability claim | Custom CMOS silicon advantage described by company | Ouster technology page | Tier 2 | Self-reported | No |
Contradictions and source-bound claims
7.1 Marketing cost/scalability vs. filed risk factors
Ouster’s technology page markets digital lidar as built on “custom CMOS silicon” with scalability advantages Ouster technology page, self-reported. The 10-K, however, states that Ouster has incurred losses since inception and may not achieve profitability Ouster 2023 Form 10-K, SEC EDGAR. These are not strictly contradictory—the technology can be scalable while the company still loses money—but they are a cross-reference discrepancy between the marketing emphasis and the filed financial risk disclosure. Investors should not treat the CMOS scalability claim as an independently verified unit-economics claim.
7.2 Patent strength vs. filed IP risk
US10222475B2 is a granted patent assigned to Ouster US10222475B2. However, the 10-K also discloses competitive IP risk and does not represent that the company’s patents will exclude competitors. A granted patent is evidence of recorded claims, not proof that the claims are broad enough to create an economic moat or that they will survive validity challenges.
7.3 Academic prototype vs. commercial product
Niclass et al. report a 100-m CMOS ToF sensor in a peer-reviewed journal Niclass 2013, DOI 10.1109/JSSC.2012.2227607. That does not mean any commercial digital CMOS lidar achieves 100 m under automotive conditions. The prototype was measured in controlled or favorable conditions and used a 0.18 μm CMOS process; automotive-grade products require thermal range, vibration, electromagnetic compatibility, functional safety, and eye-safety compliance, none of which are established by that prototype paper.
7.4 TDC precision vs. photon-starved operation
The worked calculation shows that range precision from TDC quantization falls rapidly when only one photon is available: \(4.33\) mm at \(N=1\) vs. \(0.433\) mm at \(N=100\). This is a derived result, not a product claim. It also exposes a sensitivity: any marketing claim of millimeter or sub-millimeter precision implicitly assumes enough returned photons per pixel. If a product datasheet claims a range precision without specifying the target reflectivity, range, and ambient light, that claim is not verifiable from the sources used here.
Unit economics / business model
Ouster monetizes digital CMOS lidar primarily through hardware sensor sales. The 10-K describes revenues from product sales rather than a subscription, royalty, or recurring-revenue model Ouster 2023 Form 10-K, SEC EDGAR. No material licensing or royalty income is identified in the cited filing.
Margins: Exact gross margins are not restated in this note. The 10-K discloses a history of losses and contains risk language about the ability to achieve profitability. Therefore the appropriate evidence-bound statement is: unit economics are not independently verified; the company’s filing does not show sustained positive gross or operating margin. Any exact margin figure must be marked needs primary source validation unless quoted directly from Ouster’s periodic filings.
Value capture: If the sensors are fabricated in external CMOS foundries, the foundry captures some wafer-level margin. Ouster captures the system-level revenue if it can sell sensors above total cost. That purchase/sale structure is not otherwise detailed in the 10-K risk-based summary. The patent portfolio may offer licensing optionality, but no licensing revenue is evidenced in the cited filing.
Business-model caveat: A claimed hardware model with custom silicon can have high fixed design and tooling costs, and the 10-K’s loss history is consistent with that risk. No independent benchmark proves unit profitability.
Disconfirming Evidence
The digital CMOS lidar investment/technology thesis would be materially weakened or invalidated if:
- US10222475B2 or other Ouster patents are invalidated or found not infringed in inter partes review, litigation, or opposition. A granted patent is only a recorded claim; invalidation would remove the IP-moat signal US10222475B2.
- Ouster’s gross margins remain negative or do not improve over multiple reported periods. That would directly contradict the marketed CMOS cost/scalability advantage and indicate the business model is not working Ouster 2023 Form 10-K, SEC EDGAR.
- Independent third-party benchmarks show that digital CMOS lidar cannot meet automotive or industrial range/ambient-light requirements at a competitive cost. The current support for performance is largely a research prototype and vendor self-reporting; an independent negative benchmark would undercut feasibility Niclass 2013; Ouster technology page.
- Major automotive or robotics OEMs select non-digital or non-Ouster lidar in high-volume awards. Adoption is the economic payoff; losing design wins would cut the commercialization path Ouster 2023 Form 10-K, SEC EDGAR.
- A supplier or manufacturing failure prevents Ouster from shipping sensors at scale. The 10-K discloses supply-chain and third-party dependencies; a disruption would disprove the scalability thesis.