PlatformsFaststream SiliconFaststream RadioFaststream VisionConnected EdgeFaststream SecureMobility & Rail
ProductsSemiconductor IPWireless & RANEdge & GatewaysTracking & IdentificationSoftware & FrameworksConnected Systems
Technology5G protocol stackWireless and RF architectureBaseband and low PHYForward error correctionControl and data planeHigh-speed interfacesFirmware and bootSilicon root of trustSoftware-defined vehicleAutomotive OTAFunctional safety
AIAI Engineering ServicesEdge AI & Embedded MLComputer Vision EngineeringSensor Fusion & PerceptionAI Silicon & AccelerationMLOps for DevicesAI Visual InspectionPredictive MaintenanceDriver MonitoringVideo Analytics & Safety
SolutionsSemiconductorIndustrial AIConnected ProductsAsset TrackingBluetooth AoA RTLSWearable TrackingAutomotive & MobilitySmart InfrastructureSecure IdentityWireless & SatellitePrivate 5GSmart WashroomsFuel ManagementSmart BuildingsWorker SafetyEnergy MonitoringSmart AgricultureSmart CityAutonomous PlatformsAssembly AutomationLiDAR Rail SafetyHardware Wallet
IndustriesSemiconductorTelecommunicationsIndustrial & ManufacturingAutomotive & MobilityTransportation & RailAerospace & DefenceHealthcare & MedicalEnergy & UtilitiesOil & GasRetailConsumer ElectronicsMedia & EntertainmentSmart Infrastructure & IoT
ServicesSystem Integration overviewASIC & SoC DesignRTL to GDSIIVerification methodologyDFT and silicon testLow-power designMixed-signal integrationDesign enablementFPGA DesignFPGA-to-ASIC ConversionAnalog, Mixed-Signal & RFHardware & High-Speed PCBEmbedded SoftwareCloud, OTA & Device ManagementManufacturing TransitionHow we engage
CompanyAbout FaststreamEngineering ExcellenceLeadership & OrganisationHow We EngageQuality & ComplianceStandards & EcosystemPartners & EcosystemTrust CentreLocations & DeliveryNewsroom & MediaCareersCase StudiesKnowledge CenterWhite PapersGlossaryNewsletterResources & Support
ContactStart a projectHow we engage
Talk to us
SENSOR VALIDATION

Validating a 2,000-frame-per-second image sensor with an FPGA.

A high-frame-rate image sensor cannot be evaluated by pointing a host at it. At around two thousand frames a second the data arrives faster than software can reliably accept, and the people who need the images — the analog designers of the read-out circuit — need them captured deterministically, not best-effort.

DomainHigh-speed sensor interface & characterisation
PlatformsFaststream Silicon · Vision
ScopeFPGA design, acquisition and host tooling
Binding constraint~2,000 fps, deterministic capture
DisclosureProperty level; customer not named
CONTEXT

Where this started.

The DROIC under test integrated the analog signal path with only minimal digital logic: an LVDS interface for image data and a serial programming interface. Everything else — acquisition timing, framing, and getting the images off the board fast enough to be useful — had to live in an FPGA sitting between the sensor and the host.

The sensor targets industrial and military use and images in both the infrared and visible bands, so the acquisition had to be exact rather than approximate: the characterisation is only as trustworthy as the capture beneath it.

CHALLENGES

4 problems, named.

Stated before they had answers.

01

The sensor outruns the host

At ~2,000 fps the image stream exceeds what a host can capture directly, so acquisition and buffering had to move into the FPGA.

02

The DROIC is almost all analog

With only minimal digital logic on the sensor, the FPGA had to provide LVDS capture, framing and the serial programming interface itself.

03

Getting frames off the board

A single link could not carry the rate, so the images had to be split across two high-speed paths to the host without loss or reordering.

04

Capture has to be exact to be useful

Characterisation feeds the analog design team, so any artefact introduced in acquisition would be mistaken for a sensor property.

ARCHITECTURE

How the system was built.

DROIC → FPGA ACQUISITION → HOST CHARACTERISATIONSENSORDROIC sensorAnalog + LVDS, IR/visibleLVDS captureGTX/GTH deserialiseSerial configProgramming interfaceFPGA — XILINX KINTEX ULTRASCALEAcquisition FSMFrame timing & controlFrame buffering4 GB DDR4USB 3.0 controller2× FX3 SuperSpeedHOSTDual USB 3.0 link≈5.2 Gbps combinedOffline processingCharacterise DROICFeedbackTo analog design team

The FPGA sits between an almost-all-analog sensor and a host that cannot keep up. A framing state machine captures the LVDS image stream into DDR4, and two FX3 SuperSpeed controllers carry it to the host at roughly 5.2 Gbps combined, where the images are processed offline to characterise the DROIC and feed the sensor manufacturer's analog teams.

CONTRIBUTION

What Faststream did.

The specific scope, rather than a capability list.

WHAT WAS HARD

The parts that consumed the schedule.

Written out because a reader facing the same programme gets more from this than from a summary of what went well.

01

The sensor outruns the host

Acquisition and buffering were moved into the FPGA so frames are captured deterministically rather than best-effort.

02

Splitting the stream without loss

Dividing the image stream across two USB 3.0 controllers meant getting ordering and flow control right so the host reassembles frames intact.

03

Exact capture for trustworthy characterisation

Because the images feed analog design decisions, the datapath was designed to add nothing of its own to the sensor's output.

04

IR and visible in one path

Supporting both bands meant the acquisition could not assume a single data format or range.

OUTCOME

What resulted.

A deterministic ~2,000 fps capture path

LVDS acquisition, framing and buffering on the FPGA, capturing the sensor at full rate.

Images on the host at ≈5.2 Gbps

Dual USB 3.0 SuperSpeed links delivering acquired frames fast enough to be characterised in bulk.

Characterisation feedback to the sensor maker

Offline processing that turned captured images into feedback for the analog design teams.

A reusable sensor bring-up rig

An FPGA-based interface and acquisition pattern reusable for other high-rate sensors.

Confidentiality

Customer projects are presented at property, capability, outcome and integration level. Customer names, internal architecture and confidential deliverables are not disclosed. Where a figure would identify a customer or a design, it is omitted rather than approximated. More detail is available under a non-disclosure agreement, within the limits each customer has agreed.

CAPABILITY USED

What this was built from.

Every item links to its own page.

WHERE THIS APPLIES

Industries this serves.

COMMON QUESTIONS

Questions this programme gets asked.

01

What is a DROIC?

A Digital Read-Out Integrated Circuit — the read-out chip bonded to an imaging array. Here it integrated the analog signal chain with only minimal digital logic, exposing LVDS image data and a serial programming interface.

02

Why use an FPGA to characterise a sensor?

Because the sensor runs faster than a host can capture directly, and the analog team needs deterministic high-rate acquisition an FPGA provides through LVDS capture, a framing state machine and a high-throughput link.

03

How is 5.2 Gbps to a host achieved?

Two USB 3.0 SuperSpeed (FX3) controllers run in parallel, each carrying part of the image stream, so frames reach the host at roughly 5.2 Gbps combined.

04

What does characterisation feed back to?

Images are processed offline on the host to characterise the DROIC and feed the sensor manufacturer's analog design teams — closing the loop between silicon and system.

KEEP READING

Related work.

BUILD WITH FASTSTREAM

Bring us the difficult part.

Tell us the sensor, the frame rate and the interface. High-rate acquisition and characterisation on FPGA fabric is core Faststream work.