Sony STARVIS Camera Portfolio: What STARVIS and STARVIS 2 Actually Mean
STARVIS is not a sensor. It is a Sony specification that a sensor either meets or does not, and the threshold is a published number: sensitivity of at least 2000 mV per square micrometer, measured in color at F5.6 with a 706 cd per square meter light source over a one-second exposure. A conventional back-illuminated sensor of the same era reaches roughly 700. That is the whole basis of the name, and it is why a Sony STARVIS camera holds a usable image in conditions where an ordinary module returns noise. This page explains how that number is achieved, what changed in STARVIS 2, what the technology does not do, and which of the thirteen Vadzo modules built on it suits a given interface. Each is supplied as a board-level OEM camera module rather than a finished camera.
What Is Sony STARVIS?
Sony STARVIS is a back-illuminated pixel technology and qualification for image sensors used in security and surveillance cameras. It applies to CMOS sensors in Sony's Exmor R family that achieve a minimum sensitivity of 2000 mV or more per square micrometer, in color at F5.6 with a 706 cd per square meter light source over a one-second exposure. Sony applies the name only to sensors meeting that threshold, so it is a qualification rather than a product line.
That distinction matters when reading specifications. Not every Sony low-light sensor is a STARVIS image sensor, and not every sensor described as STARVIS in third-party listings carries the designation. Sony treats STARVIS, STARVIS 2, and STARVIS 3 as three specific generations of pixel technology, each with its own criteria. A STARVIS camera module is therefore a module built on a sensor that Sony has qualified, which is a verifiable claim rather than a marketing adjective.
Every sensor carrying the name shares two properties. The pixel is back-illuminated, and the readout chain is designed for very high conversion gain. Everything else about the family, including resolution, pixel size, and frame rate, varies by part. A Sony STARVIS embedded camera is best understood as a module built for one job, which produces a clean image when there is very little light to work with.
How Sony STARVIS Sensor Technology Works
A back-illuminated image sensor inverts the conventional pixel stack. In a front-illuminated design, the metal wiring that reads the pixel sits between the lens and the photodiode, so some of the incoming light is blocked or scattered before it reaches the silicon. Back illumination moves that wiring behind the photodiode, leaving a clear optical path. The photodiode receives light that a front-illuminated pixel of identical size would have lost.
The second half of Sony STARVIS technology is a conversion gain. Once a photon has been absorbed and a charge generated, that charge must be converted to a voltage the readout circuit can measure. The efficiency of that conversion is set largely by the capacitance of the node where it happens, and a lower capacitance produces a larger voltage swing for the same charge. A Sony STARVIS sensor is built with a very low-capacitance conversion node, which amplifies each captured photon before any electronic noise is added downstream.
Those two mechanisms together produce the 2000 mV per square micrometer figure. Set against roughly 700 for a standard back-illuminated part; the gain is close to three times, which corresponds to about one and a half stops of exposure. In practice, that means a low-light image sensor of this class can use a shutter three times shorter for the same image brightness or hold the same shutter and return an image with substantially less noise.
The third element is the photodiode itself. Sony uses a deep photodiode structure to increase the volume of silicon available to absorb light, which matters most for long wavelengths, since red and near-infrared photons penetrate further into silicon before being absorbed. A shallow photodiode lets those photons pass through without being captured. A deep one collects them.
STARVIS vs STARVIS 2: What Changed
STARVIS 2 keeps the same sensitivity threshold and improves the photodiode, increasing charge capacity and near-infrared absorption. The headline sensitivity number did not move because it was already the qualifying criterion. What changed is how much signal the pixel can hold before it saturates, and how well it responds beyond the visible range.
The mechanism is geometric. A STARVIS 2 photodiode is deeper, and its surface is deliberately uneven rather than flat. The extra depth adds absorbing volume. The uneven surface refracts incoming light at an angle instead of letting it pass straight through, which lengthens the path a photon travels inside the silicon and raises the chance it is absorbed rather than lost. Both effects are strongest at long wavelengths, which is why the near-infrared improvement is the most visible difference between the generations.
A greater charge capacity has a second consequence. A pixel that can hold more electrons before saturating has more room between its noise floor and its clipping point, which widens the usable range within a single exposure. A Sony STARVIS 2 camera therefore handles a scene containing both a bright region and a dark one better than a first-generation part, without depending on any multi-exposure technique.
Sony has since introduced a third generation using a lateral overflow capacitor structure to extend a single exposure range further. No module in the Vadzo range uses it, so it is noted here only so the generation numbering makes sense.
The practical selection rule between the first two generations is simple. If the application involves infrared illumination or a scene with strong brightness variation inside one frame, the second generation is the better engineering choice. If the requirement is simply a clean image in low visible light, a first-generation part meets it, and pixel size will matter more to the result than the generation.
Does STARVIS Work in Complete Darkness?
No. No image sensor of any kind produces an image in complete darkness, because there are no photons to detect. Sensitivity determines how few photons are enough, not whether zero will do. Any claim that a camera sees in total darkness is describing infrared illumination, which means the scene is lit, just not visibly.
What STARVIS delivers is a lower floor. The term starlight camera module is used for sensors in this class because they can return a recognizable image under illumination on the order of starlight, roughly a thousandth of a lux, given a fast lens and a long enough exposure. That is a real capability, and it is why the technology exists.
Two conditions are attached, and both are easy to overlook. Exposure time has to be long enough to collect the photons, so very low light performance and high frame rate are opposed requirements. A sensor delivering a clean image at a thousandth of a lux is not delivering it at 60 frames per second. And the lens matters as much as the sensor, since an F1.4 lens passes roughly four times the light of an F2.8, which is a larger difference than one sensor generation.
The honest framing for a low-light camera module is therefore that it lowers the illumination at which the application still works. It does not remove the need for light. If the scene will genuinely have none, the answer is an infrared illuminator, and the sensor's question becomes a question about near-infrared response.
Near Infrared Response
Silicon responds to light out to roughly 1100 nm, well past the 700 nm limit of human vision, and the deep photodiode structure used across the family is designed to capture that long wavelength response rather than let it pass through. This is what makes a NIR-sensitive camera practical using an ordinary silicon sensor.
The two wavelengths that matter are 850 nm and 940 nm, because that is where inexpensive illuminators are manufactured. At 850 nm, the sensor retains a useful fraction of its peak response, and the illuminator emits a faint red glow that a person looking directly at it can see. At 940 nm, the glow is effectively invisible, and the response is lower, so the same scene needs more illumination of power or a longer exposure.
Choosing between them trades visibility against a light budget. The rule of thumb is short: if covert operation matters more than power budget, choose 940 nm, and if range and efficiency matter more than remaining unseen, choose 850 nm.
A near-infrared camera sensor used with its own illuminator becomes independent of ambient conditions. The scene is lit by the illuminator, the exposure is set for it, and changes in daylight or room lighting stop affecting the image. For anything that has to behave identically at every hour, that predictability is often worth more than the sensitivity gain.
One point applies to the optics rather than the sensor. Most lenses are corrected for visible wavelengths, so a lens focused under white light will not be exactly in focus at 850 nm. The shift is small but real, and a fixed focus assembly should be set under the illumination the application will use.
Every STARVIS Sensor Is Rolling Shutter
There is no global shutter for STARVIS sensors. The entire family is a rolling shutter, which means the sensor exposes and reads one row at a time rather than capturing the whole frame at a single instant. This is a hard constraint, and it decides whether the technology suits an application before any other specification is considered.
A rolling shutter camera module records the top of the frame slightly earlier than the bottom. If nothing in the scene moves during that interval, the result is indistinguishable from a simultaneous exposure. If something does move, it is recorded at a different position on each row, so a straight vertical edge is rendered as a slanted one and a rotating object appears bent.
The reason the family is built this way is that global shutters and low-light performance pull against each other. A global shutter pixel needs in-pixel storage to hold charge while the frame is read out, and that storage occupies area which would otherwise be photodiode. Every square micrometer given to the storage node is taken from light collection, which is precisely what this family optimizes for.
So, the rules are straightforward. If the subject or the camera moves during exposure, this family is the wrong choice, and a global shutter sensor is required instead. If the scene is fixed or slow, rolling shutter costs nothing, and the sensitivity is available in full.
Vadzo Sony STARVIS Camera Portfolio
Model | Sensor | Generation | Resolution | Interface | Buy Now |
|---|---|---|---|---|---|
Merlin-290CRS | IMX290 | STARVIS | 2.1MP / 1080p | USB 2.0 | |
Merlin-291CRS | IMX291 | STARVIS | 2.1MP / 1080p | USB 2.0 | |
Bolt-291CRS | IMX291 | STARVIS | 2.1MP / 1080p | MIPI CSI-2 | |
Merlin-415CRS | IMX415 | STARVIS | 8.46MP / 4K | USB 2.0 | |
Falcon-415CRS | IMX415 | STARVIS | 8.46MP / 4K | USB 3.2 Gen 1 | |
Bolt-415CRS | IMX415 | STARVIS | 8.46MP / 4K | MIPI CSI-2 | |
Innova-715CRS | IMX715 | STARVIS 2 | 8.46MP / 4K | GigE | |
Merlin-662CRS | IMX662 | STARVIS 2 | 2.1MP / 1080p | USB 2.0 | |
MerlinPlus-662CRS | IMX662 | STARVIS 2 | 2.1MP / 1080p | USB 2.0 | |
Innova-662CRE | IMX662 | STARVIS 2 | 2.1MP / 1080p | GigE | |
Wave-662CRE | IMX662 | STARVIS 2 | 2.1MP / 1080p | Wi-Fi | |
Innova-678CRS | IMX678 | STARVIS 2 | 8.29MP / 4K | GigE | |
Wave-678CRE | IMX678 | STARVIS 2 | 8.29MP / 4K | Wi-Fi |
The 2.9 um parts collect roughly four times the light per pixel of the 1.45 um parts, so those are the modules to specify when illumination is the binding constraint. A Sony IMX290 camera and a Sony IMX291 camera sit there at 1080p on first-generation STARVIS, and a Sony IMX662 camera covers the same tier on STARVIS 2 with better infrared response and more charge capacity.
At the other end, the smaller pixels trade per-pixel signal for four times the pixel count. A Sony IMX415 camera delivers 8.46MP at 1.45 um and is the part offered on the widest choice of interfaces. A Sony IMX678 camera uses a 2.0 um pixel, which sits between the two extremes and is the most balanced option in the 4K group. A Sony IMX715 camera returns to 1.45 um at the same resolution. Neither end of the range is better than the other. They answer different questions, and pixel size answers the low-light question more directly than generation does.
One Sensor Across Five Interfaces
The most useful property of this portfolio has nothing to do with low light. The IMX662 is available on USB 2.0, GigE, and Wi-Fi. The IMX415 is available on USB 2.0, USB 3.2 Gen 1, and MIPI CSI-2. That means connectivity can change without the image changing, because the sensor, the optics, and the tuning underneath stay the same.
For a product line shipping in more than one configuration, this removes a whole category of work. A camera qualified once does not need requalifying when a variant moves from a cabled interface to a network one. Exposure behavior, noise characteristics, color response, and field of view are properties of the sensor and lens, not of the connector, so validation carries across.
Each interface answers a different constraint. A STARVIS USB camera on USB 2.0 suits a host within a few meters and keeps the bill of materials low, with compression making the higher resolutions fit the available bandwidth. USB 3.2 Gen 1 removes that bandwidth pressure. A STARVIS MIPI camera integrates at board level with no intermediate bus and no protocol overhead, in exchange for a sensor-specific driver and a device tree description on the host. A STARVIS GigE camera reaches a hundred meters over ordinary network cable and can draw power from the same cable. A STARVIS Wi-Fi camera removes the cable entirely, at the cost of a compressed stream and latency in the low hundreds of milliseconds.
Prototyping on one and shipping on another is therefore a practical route. Development on USB is quick because the module needs no driver work, and moving to MIPI or GigE for production changes the transport without disturbing the image.
Applications
-
Surveillance and monitoring is the first group and the one the technology was designed for. Scenes are usually fixed, so rolling shutter is not a limitation, and illumination is often poor and outside anyone's control. The 1080p modules with 2.9 um pixels suit corridors, doorways, and enclosed spaces, while the 4K modules suit wide areas where the field of view has to be covered in detail.
-
Remote inspection and monitoring of equipment is the second. A low-light embedded camera placed inside machinery, in a service void, or inside an enclosure works where adding lighting is impractical, and pairing it with an infrared illuminator makes the image independent of whatever light happens to be available.
-
Traffic and access monitoring is the third. These are mixed lighting environments with strong brightness variation inside a single frame, which is where the greater charge capacity of the second-generation parts earns its place, and they frequently run under infrared illumination at night.
-
The case against is as clear as the case for. Anything where the subject or the camera moves during the exposure needs a global shutter sensor, and no part of this family provides one. Sensitivity does not compensate for geometric distortion, because the distortion is in the timing of the readout rather than in the amount of light collected.
Quick Samples Dispatch
Find the Right Camera for Your Project Today.
Samples procured typically ship within 5-7 business days of order confirmation, with tracked international shipping. Compressed development timeline? You can be capturing frames from a Vadzo camera the same week you shortlist it.
RoHS 3 · REACH · NDAA (Selected Models)
No MOQ (Minimum Order Quantity)
Ships Worldwide
Frequently Asked Questions
1. What is a Sony STARVIS camera?
A Sony STARVIS camera is a camera built on an image sensor that meets Sony's STARVIS specification, which requires sensitivity of at least 2000 mV per square micrometer, measured in color at F5.6 with a 706 cd per square meter light source over a one-second exposure. A conventional back-illuminated sensor of the same generation reaches roughly 700, so the qualifying threshold is close to three times higher. STARVIS is a qualification applied to specific sensors rather than a product family, so not every Sony low-light sensor carries it. A STARVIS camera module is a board-level module built on a qualified sensor.
2.What is the difference between STARVIS and STARVIS 2?
STARVIS 2 keeps the same sensitivity threshold and improves the photodiode, adding charge capacity and near-infrared response. The photodiode is deeper, which increases the volume of silicon available to absorb light, and its surface is deliberately uneven rather than flat, which refracts incoming light at an angle and lengthens the path a photon travels inside the silicon. Both effects are strongest at long wavelengths, so infrared response is where the generations differ from most. The extra charge capacity also widens the range a single exposure can cover. Both generations are back-illuminated image sensors, and both are rolling shutter.
3.Does STARVIS work in complete darkness?
No. No sensor produces an image without photons, so sensitivity sets how few are sufficient rather than whether none will do. Cameras described as seeing in total darkness are using infrared illumination, which lights the scene invisibly rather than not at all. What this class of sensor does deliver is a usable image at illumination on the order of a thousandth of a lux, given a fast lens and a long enough exposure. Both conditions matter: low light performance and high frame rate are opposed, and an F1.4 lens passes about four times the light of an F2.8, which is a bigger difference than a sensor generation.
4.How well do these sensors respond to infrared?
Silicon responds out to roughly 1100 nm, and the deep photodiode structure used across the family is designed to capture that long-wavelength response, which is what makes a NIR-sensitive camera practical with an ordinary silicon sensor. The two working wavelengths are 850 nm, where response is stronger but the illuminator shows a faint red glow, and 940 nm, where the glow is effectively invisible and more illumination power is needed. Choose 940 nm when covert operation outweighs power budget, and 850 nm when range and efficiency outweigh remaining unseen. A near-infrared camera sensor paired with its own illuminator becomes independent of ambient light, so the image stays consistent at any hour.
5.Is there a global shutter STARVIS sensor?
No. The entire family is rolling shutter, which exposes and reads the sensor one row at a time rather than capturing the whole frame at once. The reason is a direct trade-off: a global shutter pixel needs in-pixel storage to hold charge during readout, and that storage takes area away from the photodiode, which is exactly what this family optimizes. A rolling shutter camera module is fine for a fixed or slow scene and wrong when the subject or the camera moves during exposure, because the distortion comes from readout timing and is not corrected by sensitivity or resolution.
6. Which interface should be chosen?
It depends on distance to the host and how much integration work is acceptable. A STARVIS USB camera needs no driver and suits a host within a few meters. A STARVIS MIPI camera integrates at board level with no bus overhead, in exchange for a sensor-specific driver and a device tree description. A STARVIS GigE camera reaches a hundred meters over ordinary network cable and can take power from the same cable. A STARVIS Wi-Fi camera removes cabling entirely, at the cost of compression and latency in the low hundreds of milliseconds. Because the IMX662 is offered on three interfaces and the IMX415 on three, the choice can change without the image changing.