Bench Talk

(Source: yallowww/stock.adobe.com)

How does a self-driving car recognize a stop sign? It starts with image sensors capturing video of the scene and software that identifies a stop sign’s characteristic shape and lettering. The same basic capability lies within many applications—from video conferencing and mobile devices to machine-vision systems that inspect products on the factory line.

Today, image sensors are designed for all types of specialized applications. While these applications vary widely, the image sensors encounter a common challenge when it comes to their hardware: they are unforgiving when it comes to power. Noise, voltage drift, poor sequencing, or thermal stress can degrade image quality and introduce errors that cannot be corrected with software. As sensors push toward higher resolutions and faster frame rates, their power demands become more exacting.

Image sensors straddle the boundary between optical and electrical engineering and contain sensitive precision electronics. As such, they require multiple separate power inputs for their subsystems, each with its own voltage tolerance and sequencing requirements. With these challenges in mind, this blog covers design considerations for building a power tree that meets the voltage tolerance and sequencing requirements for image sensors. We’ll look at the tradeoffs between buck converters and low-dropout (LDO) linear voltage regulators, and how choosing the right topology impacts efficiency, cost, sensor performance, thermal management, and long-term reliability.

Sensor Basics

Image sensors convert an image formed by a lens into digital signals. This capability can be used to facilitate a video conference, support self-driving cars, or check that every bottle of ketchup on a filling line has a cap. Naturally, each application has unique demands, and the image sensor has to be mated to its specific needs. A video conferencing camera needs to show people clearly against various backgrounds, whereas machine-vision cameras must quickly capture objects passing by on a production line. Manufacturers have created all kinds of sensors with readout speeds, dynamic range, and low-light performance tailored to the needs of particular applications.

Voltage Control Strategies

Converting light into digital data requires precision throughout the entire system, starting with the power supply. Image sensors typically require a power tree built from discrete components or a power management integrated circuit (PMIC), supplying each voltage rail with appropriate power levels. Normally, input power to the power tree or PMIC starts with a higher voltage, which is then reduced as necessary for each branch. PMICs are a frequent choice, but their level of integration can limit flexibility, so an optimal match to a specific sensor may not always be available. More commonly, this down-conversion can be accomplished by building a discrete power tree with LDOs or buck converters. This provides more design flexibility and is less expensive. Choosing between LDOs and buck converters depends on the application and the characteristic differences and tradeoffs between the two (Table 1).

Generally, key considerations when choosing between LDOs and buck converters include:

• LDOs tend to be a lower-cost choice, and can operate efficiently with smaller voltage drops. However, efficiency decreases as the voltage drop increases. The decreased efficiency leads to higher operating temperatures and may affect the practicality of enclosed modules.

• Buck converters are more efficient, especially with large voltage drops and demanding loads. Their intrinsic efficiency allows them to run cooler, but they are more expensive and tend to be bulkier, requiring more printed circuit board (PCB) space for both the converter and an external inductor. Buck converters also have more ripple in their outputs than LDOs which can interfere with sensitive image sensor inputs.

Table 1: The practical differences between LDOs and buck regulators are clearly defined, assuming greater than or equal to 500mV step down.^[1] (Source: onsemi)

In many cases, an LDO is the preferred starting point unless specific constraints rule it out. A practical way to evaluate regulator options is to work through a short series of design questions:

• Can a less expensive LDO do the job? If the voltage drop is relatively low, an LDO is usually the first choice, but there are additional considerations.
• Can an LDO handle the temperature? Designers typically estimate power dissipation (PD) and resulting junction temperature to ensure the regulator stays within its specified thermal limits. If the expected thermal rise approaches the device limits, a switching regulator, such as a buck converter, may be preferred to improve efficiency and reduce heat.
• Is enough power available to compensate for the LDO’s low efficiency? A unit with mains power won’t be a problem, but a battery-powered unit will operate better with a buck.

Power Rail Conventions

Image sensors have unique architectures and pinouts depending on the manufacturer and application, but generally, they all need to power at least a few internal subsystems: digital I/O, image processing, and analog image sensing. A power tree design for an image sensor is typically based on with the following power input pins:

• VDDIO: powers the digital I/O communication interface
• VDD: powers the digital core that handles the internal logic and image processing
• VAA: powers the analog image sensing pixel array and analog-to-digital converters (ADCs)

Manufacturers use different designations for these rails, but their functions are similar. When designing a power tree, it is critical to study the sensor specs to ensure a full understanding of the requirements and tolerances for each, so they can be applied to the solution. Moreover, in operation, the power-up and power-down rails must occur in the correct sequence.

A Typical Power Tree in Practice

Consider a representative power tree arrangement for an onsemi Hyperlux^™ Image Sensor working from a 5V input rail (Figure 1). Rather than dropping all the way in a single stage, the design uses multiple reduction steps to avoid excessive dissipation at any one point. Where the first drop is modest, that intermediate node may directly feed the highest‑voltage sensor input, reducing the total number of devices. In other cases, four stages are appropriate to distribute thermal and efficiency considerations. In a four-stage design, the first LDO has a maximum current rating equal to the combined maximum current rating for the subsequent three inputs.

It should be noted that each input has a different voltage and tolerance band, as well as more subtle requirements. VDD usually has the smallest rail voltage tolerance, as it does in the diagram below. This calls for an LDO or buck converter with good transient handling capabilities. The VAA, which powers the image-sensing pixel array, is the most sensitive to noise. Noisy power inputs cause the least significant bits to fluctuate, reducing image quality. The ideal input for VAA is an LDO with a high power supply rejection ratio (PSRR)—greater than 85dB—and very low noise—less than 10mV. On the board, the LDO should be as close as possible to the ball grid array of the VAA input to mitigate trace noise.

Figure 1: This example of a power tree indicates typical power characteristics necessary to support an image sensor. (Source: onsemi)

Evolution of Next-Generation Image Sensors

Leveraging deep expertise in image sensor design and implementation, onsemi Hyperlux Image Sensor Power Trees provide verified combinations of buck converters, low-noise LDOs, and PMICs to take the guesswork out of power tree design for Hyperlux portfolio of image sensors. onsemi power trees provide flexible input options of 5V, 5V-18V, and 48V, simplifying integration into applications ranging from battery-powered devices to automotive camera systems.

Each component within a power supports the specific needs of each image sensor input. For example, in Figure 1 above, the onsemi T30LMPSR165 Ultra-Fast Linear Regulator was chosen for the VDD input due to its fast transient response time (1μs typical), low noise, and high PSRR, making it an excellent LDO for the tight tolerance of the VDD input.

Taking advantage of verified designs helps new designs reach the market faster, especially when designing around new image sensors that offer high image quality at low power, such as the onsemi AR0544 and AR0830 Hyperlux LP image sensors. Designed for ultra-low power consumption, these image sensors benefit from precise power delivery for their sleep and wake functions. The AR0544 and AR0830 deliver advanced image capture at the edge and offer programmable operating modes that allow designers to balance resolution, frame rate, and bandwidth against power consumption.

Conclusion

Achieving a successful vision system requires a carefully selected image sensor that is supported by an optimized power tree. Designing a tree using discrete components provides the greatest flexibility to meet the tight voltage tolerances, noise constraints, and sequencing requirements across many applications. The distinctions between LDOs and buck converters directly shape system reliability and image quality. Likewise, understanding the unique needs of each rail, from the noise‑sensitive VAA input to the tightly regulated VDD core, ensures stable operation. As imaging continues to move into more applications, like autonomous systems, the quiet decisions made in power architecture will determine how clearly machines perceive the world and how reliably they act on what they see.

Author

Peter Welander, now semiretired, has been working as a freelance writer and editor for more than 10 years, following seven years as a senior editor and content manager for Control Engineering magazine. During this time he has written heavily about industrial automation, primarily in process industries. Responsibilities have also included audio and video production, podcasts, and blogging.

Moving into the publishing world followed many years working in sales and marketing management, engineering, and operations for a variety of industrial manufacturers. One capability that he has had throughout his career is the ability to grasp technical concepts and explain complex ideas clearly. His personal interests include photography, a machine shop, woodworking, and pipe organ building.

[1] https://www.youtube.com/watch?v=e4vnTTCoe1U