1stVision, Author at 1stVision Inc. - Machine Vision Articles

July 23, 2025July 23, 2025

Alvium Frame – better than a bare board

OEM vision system builders often favor bare board cameras. They are typically very small and compact, attractively priced, and easy to fit into small spaces. Ideal for embedding into systems of your own design.

Alvium housed camera (red) vs. Alvium Frame in metallic aluminum – Courtesy Allied Vision

But sensor alignment can be challenging – enter the Alvium Frame

With a traditional housed camera, the sensor is typically aligned within the camera, to very precise tolerances. That insures that when the lens is mounted, the field of view is optimally transferred through the optics and onto the sensor, without introducing tip/tilt/focus losses.

Conversely, a board level camera is wonderfully compact, “just” a sensor and other electronics on a printed circuit board (PCB). But the integrator, often an OEM systems developer, is then left to his own skills aligning the PCB optical sensor and a corresponding lens. Even the most skilled mechanical engineers and systems builders can find optical alignment challenging. The tolerances are very fine, and besides optics knowledge and experience, it often takes special instrumentation and tooling to set, test, and tune the alignment. And then to insure the positioning remains stable relative to future vibrations once deployed in the target system.

Alvium Frame – sized like a board, aligned like a camera

Allied Vision’s innovation with the Alvium Frame was to recognize a market for something almost as small as a board level camera, yet sensor-aligned in a frame. And with helpful heat dissipation characteristics besides.

Click to contact — Give us some brief idea of your application and we will contact you to discuss camera options.

What happens if the sensor isn’t well-aligned?

Image quality can be negatively affected, by getting the geometry wrong. Optics is all about mapping the real world target through a lens and onto the 2D array of pixels in the sensor. So true orientation without rotation in the X or Y axis is preferred, as is proper depth in the Z axis, and avoidance of tip or tilt.

Consider the following illustration of sensor rotation off by just 1 degree. While the camera body and PCB board being imaged were squared to each other, for this illustration the sensor was rotated by 1 degree around the Z axis. It would not pass the alignment process in manufacturing production and quality control like this! Notice how the white line slopes down to the right, for this misaligned sensor.

Whether your application is PCB inspection, optometry, lasers, or any precision work, optical alignment matters.

Misaligned sensor causes horizontal white line to appear sloped – Courtesy Allied Vision

Sensor tip/tilt impacts Modulation Transfer Function (MTF)

In our blogs, newsletters, Tech Briefs, and Knowledge Base, we periodically talk about the Modulation Transfer Function (MTF). It’s an important measure of lens performance. One typically seeks a lens that’s more than good enough for the task at hand, just as one’s camera, sensor, lighting, and data rates each have to be up to the job – a system is no better than its weakest link.

But a lens’ reported MTF is based on testing instances of the lens mounted on cameras with precisely aligned sensors. Any tip or tilt of the sensor away from the orthogonal plane might not greatly impact the central area of the image. But even a little tip or tilt, with the top leaning forward and the bottom backward, or the other way around, may lead to out-of-focus conditions away from the center of the sensor.

So even if the lens’ MTF is very good, an insufficiently aligned sensor can drag down imaging performance. Which is why it’s important to know the engineering tolerances of every optical component – or to trust your provider publishes and warranties their tolerances.

Alvium camera family – more than 200 variants

First came the “original” Alvium housed camera, of which there are more than 10 members in each of 1GigE and 5GigE interfaces. These compact cameras are attractively priced, with a range of sensors, and are feature-rich.

Alvium G1 with lens – Courtesy Allied Vision

Variations are available through the Alvium Flex concept, with options for Open Housing, Flex Frame, and Bare Board. Designed for space-constrained applications, the Alvium Flex cameras offer USB and MIPI CSI-2 interfaces. Still feature-rich and with many sensor options, pricing in single or small quantities, or OEM volumes are available.

Alvium Flex options – Courtesy Allied Vision

In another nod to OEM customers, the USB3 version of the Alvium Flex Frame has a choice of 2 interface positions, a rear exit (1800) or a side exit (90), as shown below:

Interface options helpful for tight spaces – Courtesy Allied Vision

So the Flex Frame offering fills a niche between housed vs. board level cameras. It’s close in size to board level, but with the alignment benefits of a housed camera.

So what are the alignment tolerances?

Let’s use the diagram and definition of terms below as a framework:

Sensor shift and rotation framework – Courtesy Allied Vision

For housed-sensor Alvium models, Allied Vision asserts the following manufacturing accuracy for sensor positioning:

Sensor positioning accuracy for housed-sensor Alvium models – Courtesy Allied Vision

For Alvium Frame models, the table below shows sensor positioning tolerances:

Sensor positioning accuracy for Alvium Frame models – Courtesy Allied Vision

Did I read that right? Frame model tolerances tighter than housed?

Exactly. Sensor shift in the x/y axis is +/- 90 µm in for Alvium Frame models fully 60 µm tighter tolerance than the +/- 150 µm for sensors in the housed Alvium models.

Likwise in the z axis, the optical back focal length is within 0 to -50 µm for the Alvium Frame models, while the Alvium housed models are factory calibrated to within 0 to -100 µm.

Other features of the Alvium Frame

In addition to the optically important sensor alignment, the frame helps to dissipate heat via the metallic surface area, a benefit over a bare-board approach. The frame has 4 M2 screw holes for easy mounting. And there are two options to help you align the Alvium Frame within your own system:

Two options to align Alvium Frame units – Courtesy Allied Vision

1st Vision’s sales engineers have over 100 years of combined experience to assist in your camera and components selection. With a large portfolio of cameras, lenses, cables, NIC cards and industrial computers, we can provide a full vision solution!

About you: We want to hear from you! We’ve built our brand on our know-how and like to educate the marketplace on imaging technology topics… What would you like to hear about?… Drop a line to info@1stvision.com with what topics you’d like to know more about.

June 30, 2025June 30, 2025

What can you see with a 67MP camera?

Remember when machine vision pioneers got stuff done with VGA sensors at 0.3MP? And the industry got really excited with 1MP sensors? Moore’s law keeps driving capacities and performance up, and relative costs down. With the Teledyne e2v Emerald 67MP sensor, cameras like the Genie Nano-10GigE-8200 open up new possibilities.

12MP sensor image – Courtesy Teledyne DALSA

67MP sensor image – Courtesy Teledyne DALSA

So what? 67MP view above right doesn’t appear massively compelling…

Well at this view, without zooming in, we’d agree….

But at 400% zoom, below, look at the pixelation differences:

Both images below show the same target region, with the same lighting and lens, and each zoomed (with Gimp) to 400%. There is so much pixelation in the 12MP image to raise doubts about effective edge detection on either the identifying digits (33) or for the metal-rimmed holes. Whereas the 67MP image has far less pixelation, thereby passing a readily usable image to the host for processing. How much resolution does your application require?

Important “aside”: Sensor format and lens quality also important

Sensor format refers to the physical size of the sensor and the pixel shape and pixel density. Of course the lens must physically mount to the camera body (e.g. S, C, M42, etc.), but it must also create an image circle that appropriately covers the sensor’s pixel array. The Genie Nano-10Gige-8200 uses the Teledyne e2V Emerald 67M packs just over 67 million pixels, each square pixel just 2.5 µm wide and high, onto a CMOS sensor measuring only 59mm x 59mm.

Consider other good quality cameras and sensors, with pixel sizes in the 4 – 5 µm range, which leads to EITHER fewer pixels overall in the same size sensor array; OR to a much larger sensor to accommodate more pixels. The former may limit what can be accomplished with a single camera. The latter would necessarily make the camera body larger, the lens mount larger, and the lens more expensive to manufacture.

The lens quality, typically expressed via the Modulation Transfer Function (MTF), is also important. Not all lenses are created equal! A “good” quality lens may be enough for certain applications. For more demanding applications, one would be wasting a large format sensor if the lens’ performance fails below the sensor’s capabilities.

Two different lenses were used to take the above images, both fitting the sensor size. However the right image was taken with a lens designed for smaller pixels versus the left. – Courtesy Teledyne DALSA

The high-level views of the test chart above tease at the point we’re making, but it really pops if we zoom in. Look at the difference in contrast in the two images below!

Lens nominally a fit for the sensor format and mount type, but NOT designed for 2.5 µm pixels.

The takeaway point of this segment is lensing matters! The machine vision field benefits users tremendously with segmented sensor, camera, lensing, and lighting suppliers. Even within the same supplier’s lineup, there are often sensors or lenses pitched at differing performance requirements. Consider our Knowledge Base guide on Lens Quality Considerations. Or call us at 978-474-0044.

Another example:

Below see the same concentric rings of a test chart, under the same lighting. The left imaged was obtained with a good 12MP sensor and good quality lens matched to the sensor format and pixel size. The right imaged used the 67MP sensor in the Genie-Nano-10GigE-8200, also with a well-matched lens.

If you need a single-camera solution for a large target, with high levels of detail, there’s no way around it – one needs enough pixels. Together with a well-suited lens.

Teledyne DALSA 10GigE Genie Nano — Genie Nano 10GigE 8200 – Courtesy Teledyne DALSA

The Genie Nano 10GigE 8200, in both monochrome and color versions, is more affordable than you might think.

Once more with feeling…

Which of the following images will lead to the more effective outcomes? Choose your sensor, camera, lens, and lighting accordingly. Call us at 978-474-0044. Our sales engineers love to create solutions for our customers.

June 25, 2025July 16, 2025

Prophesee event-based vision – a new paradigm

We don’t use terms like paradigm-shift lightly. Event-based vision (EBV) is shaking things up just as much as the arrival of frame-based and line-scan imaging once did . It’s that different. And why we are excited to offer Prophesee sensors and cameras.

Applications examples… and just enough about concepts

This informational blog skews towards applications examples, with just enough about concepts and EBV technology to lend balance. Event-based vision is so new and so different than previous vision technologies. We believe our readers may appreciate an examples-driven approach to understanding this radically new branch of machine vision.

Unless you’re an old hand at event-based vision (EBV) …

…and as of this writing in Summer 2025 few could be, let’s show a couple of short teaser videos before we explain the concepts or go deeper on applications.

Example 1: High-speed multiple-object counting and sizing

The following shows a use of event-based imaging to both count particles or objects and to estimate their size. All in a high-speed multiple concurrent target environment.

Courtesy Prophesee

Example 2: Eye-tracking – only need to track pupil changes

Now consider eye-tracking: in the video below, we see synchronized side-by-side views of the same scene. The left side was obtained with a frame-based sensor. The right side used a Prophesee event-based sensor. Why “waste” bandwidth and processing resources separating a pupil from an iris, eyelids, and eyebrows, when the goal is eye-tracking?

Courtesy Prophesee

Radically increased speed; massively less data

Prophesee Metavision‘s sensor tracks “just” what changes in the field of view, instead of the frame-based approach of reading out whole sensor arrays, transmitting voluminous data from the camera to the host, and algorithmically looking for edges, blobs, or features.

Example 3: Object tracking

Not unlike the eye-tracking example above, but this time with multiple moving targets – vehicular traffic. With frame-based vision, bandwidth and processing has to repeatedly process static pavement, light poles, guard rails, and other “clutter” comprising 80% of the field of view. But with event-based vision the sensor only detects and transmits the moving vehicles.

Whether counting traffic for congestion planning/reporting, or collision avoidance abord a given vehicle, we don’t care what the vehicle looks like – only that it’s there and moving along a certain trajectory and at a detected speed.

Example 4: Surveillance and security

Prophesee named this video “Optical Flow Crowd”, which is best understood in the context of security and surveillance, we suggest. It’s not unlike the vehicular flow example above, except that cars and trucks mostly stay in-lane. Whereas pedestrians move at diverse angles. And their arm, leg, head, and torso movements also convey information.

The (computed) vector overlays indicate speed and directional changes, important to reveal potentially dangerous actions taken or likely to emerge. For example, does a raised arm indicate a handgun or knife being readied, or is it just a prelude to scratching an itchy nose? Is there a pursuer turning towards another pedestrian, to which a nearby policeman should be alerted?

Example 5: Vibration monitoring and preventative maintenance

Motorized equipment is typically service either on a preventative maintenance schedule and/or a break-fix approach, depending on costs, legal liabilities and risks, etc. What if one could inexpensively identify vibration patterns that reveal wearing belts or bearings before a breakdown and before there is preventable collateral damage to other components?

Courtesy Prophesee

Enough with the examples – how can I get event-based sensors?

1stVision is pleased to represent Prophesee with a wide range of sensors, evaluation kits, board-level and housed cameras, and an SDK designed for event-based vision applications.

Built on neuromorphic engineering principles inspired by the brain’s neural networks and human vision, Prophesee products are surprisingly affordable, enabling new fields of event-based vision. Or improving on frame-based vision applications that may be done better, faster, or less-expensively using an event-based approach.

June 25, 2025June 25, 2025

T2IR – Trigger to Image Reliability

Machine vision systems are widely accepted for their value-add in so many fields. Whether in manufacturing, pick-and-place, quality control, medicine, or other areas, vision systems either save money and/or create competitive advantage.

BUT deploying new systems can be tricky

For experienced machine vision system builders, sometimes the next new system is either “simple” in complexity, or a minor variant on previously built complex systems in which one has good confidence. Deployment, systems acceptance testing and validation may be straightforward, with little doubt about correctness or reliability. Congratulations if that’s the case.

For new undertakings, even for experienced vision system builders, there’s often some uncertainty in the “black box” nature of a new system. Can I trust the results as accurate? Am I double counting? Am I missing any frames? If the system overloads, how can I track which inspections were skipped while the system recovers? How to troubleshoot and debug? And rising QA standards are pushing us to verify performance – yikes!

T2IR: Trigger to Image Reliability can open up the black box

Take the mystery out of system deployment, tuning, and commissioning. T2IR’s powerful capabilities essentially open up the black box. And for users of many Teledyne DALSA cameras and framegrabber products, T2IR is free!

Two minute overview of T2IR: Trigger to Image Reliability – Courtesy Teledyne DALSA

Bundled with many Teledyne DALSA products

The T2IR features are available at no cost for customers who purchase (or already own) any of:

Cameras: Genie Nano, Linea, Piranha, and others

Frame Grabbers: Xtium and Xcelera

Software: Sapera LT SDK (for both Windows and Linux)

T2IR in a nutshell

T2IR, Trigger to Image Reliability, is a set of hardware and software features that help improve system reliability. It’s offered as a free benefit to Teledyne DALSA customers for many of their cameras and framegrabbers. It helps you get inside your system to audit and debug image flow.

Example One: Coping with unexpected triggers

Consider the following diagram:

To the left of the illustration we see the expected normal trigger frequency. Each trigger, perhaps from a mechanical event and an actuator sending an electronic pulse, triggers the next camera image, readout, and processing, at intervals known to be sufficient for complete processing. No lost or partial images. No overtriggering. All as expected.

But in the middle of the illustration we see invalid triggers interspersed with valid ones. What would happen in your system if you didn’t plan for this possibility? Lost or partial images? Risk of shipping bad product to customers due to missed inspections? Damage to your system? What’s the source of the extra triggers? How to debug the problem if found during design and commissioning? How to recover if system is already in production?

With T2IR, the Sapera API, may be programmed to detect the invalid triggers that are out of the expected timing intervals. An event can be generated for the application to manage. The application might be programmed to do something like “stop the system”, “tag the suspect images”, “note the timestamps”, or otherwise make a controlled recovery.

Key takeway: instead of the system just blundering along with the invalid triggers undetected, with possible grave consequences, T2IR can help create a more robust system that recovers gracefully.

Example Two: Tracking and Tracing Images – Tagging with timestamps

Consider the challenges of high-speed materials handling, let’s say 3000 parts per minute. Some such systems require multiple cameras, lets say 4 per object, to make an accept, reject, or re-inspect decision on each component. For this hypothetical system, that would be 12,000 images per minute to coordinate!

Correlating up to thousands of images per minute – Courtesy Teledyne DALSA

The illustration above shows the timestamps one might assign with T2IR, based on ticks in microseconds, according to known or believed belt motion speed. But what if results seem corrupted using that method, or if there are doubts about whether speed is constant or variable?

With the Xtium frame grabber, one can use T2IR to choose among alternate timestamp bases beside the “hoped for” microsecond calculations:

Multiple timestamp tagging options – Courtesy Teledyne DALSA

So one might prefer to tag the images with and external trigger or a shaft encoder, which are physically coupled to the system’s movements, and hence more precise than calculated assumptions.

Sounds promising, how to learn more about T2IR?

Pursue any/all of the following:

Detailed primer: T2IR Whitepaper

Read further in this blog: Additional high-level illustrations and another example below

Speak with 1stVision: Sales engineers happy to speak with you – call 978-474-0044.

Discuss T2IR with your 1st Vision sales engineer

Diagnostic tool: a brief overview

Consider the following screenshot, in the context of an Xtium framegrabber together with T2IR:

Diagnostic T2IR window for Xtium framegrabber – Courtesy Teledyne DALSA

The point of the above annotated screenshot is to give a sense of the level of detail provided even at this top level window – with drill ins to inspect an control diverse parameters, to generate reports, etc. This is “just” a teaser blog meant to whet the appetite – read the tutorial, speak with us, or see the manual, for a deeper dive.

T2IR Elements and Benefits – in more detail

Conclusion

Every major machine vision camera and software provider offers an SDK and at least some configuration and control features. But Trigger to Image Reliability – T2IR – takes traceability and control to another level. And for users of so many Teledyne DALSA cameras, framegrabbers, and software, it’s bundled in at no cost!