Virtual Photodetectors: Building Your Own Detector

Andy Whitehead, Southern Vision Systems Inc.

Photodetectors are available in multiple configurations, with point, quadrant, linear and 2-D arrays being among the better known. These devices offer optimal performance for dedicated tasks, and they require processing and digitizing electronics specific to each detector. The profusion of commercially available products can be intimidating when one sets out to solve a detection task. Factors such as spectral response, bandwidth, signal-to-noise ratio, amplifier noise, detector area and sensitivity frequently are traded off one against the other.

Virtual photodetectors replace multiple pieces of dedicated equipment and leverage the digital nature of the control and output signals.

Virtual photodetectors offer users an element of flexibility: the ability to select the best compromise of these parameters in a package that also can be reprogrammed to do the same for the next photodetection task. Reconfigurable detectors that provide digitally sampled signals at rates fast enough for most applications, they are enabled by the marriage of high-speed large-area imagers and field-programmable gate arrays. The devices offer the advantages of lower cost, a faster time to market and lower risk than custom photodetectors.

Applications for virtual photodetectors span multiple fields. For example, when Advanced Optical Systems Inc. of Huntsville, Ala., was designing a wide-angle lens for a seeker head, it needed to demonstrate tracking ability to the customer as well as to verify the laser spot profile. Faced with tedious realignment for multiple tasks and a short project time line, the company opted for a virtual photodetector.

“Our sampling requirements were several hundred hertz for position-sensitive information as well as careful characterization of the imaged spot to validate the collection optics,” explained Keith B. Farr, the company’s president, chief operating officer and principal scientist. Switching between functions was as simple as uploading a software mask to the virtual photodetector.

Similarly, a large test-and-measurement company developing commercial metrology equipment had to sample up to 100 regions spaced several hundred microns apart at rates of several kilohertz. It found that virtual photodetector technology was an attractive alternative to a custom photodetector run at a silicon foundry.

Virtual principles

Start with a 0.5 × 0.5-in. piece of silicon, subdivide it into more than 1 million micron-size photodiodes, overlay a mask with specific geometric patterns, sample at up to 500 kHz, digitize, present the data over a common digital interface, and one has the essence of the virtual photodetector. The device combines the features of an analog photodiode, amplifier, digitizer and interface in one package.

As with the virtual photodetector itself, components are used for multiple functions depending on the needs of a specific detection task. Memory is used for functions such as mask storage and data buffering. The serial processor controls communications and aids in data processing. Similarly, the parallel processor controls the imager, applies the geometric mask and does the bulk of the sensor processing.

The major components of the virtual photodetector are a high-speed, directly addressable CMOS imager and adjacent parallel and serial processors. Sufficient memory is available to store a mask and to buffer data for transmission to a remote processor.

Parallel and serial processors perform specific functions depending on the nature of the optical signal processing tasks. Masking, thresholding or binning is most efficient on the parallel processor, while the serial processor is better at normalizing, averaging or performing tasks that involve floating point operations.

The SpecterView virtual photodetector offers standard prepackaged modes that enable it to act as a single photodiode, a quadrant detector, a linear array or a 2-D array with arbitrary windowing.

Southern Vision Systems Inc.’s SpecterView product line is built around this technology. Cost advantages result from the use of off-the-shelf technology for all components and from a design with the goal of flexibility in application. The SpecterView package, measuring 3.5 × 4 × 3 in., includes BNC connectors for external triggering and analog video output. RS-170 video circuitry aids optical alignment.

A mask is created using the accompanying graphical user interface or third-party software that defines up to 256 detection areas. Each area can be as small as one pixel or as large as the entire imager array, and it can consist of nonadjacent pixels. This mask is then resident in the device’s memory, and every acquisition event samples it to determine how to interpret the optical signal.

In application

The software recognizes that the smaller the sampling area of the mask, the higher the sampling bandwidth, and it adjusts accordingly. Each sampled area can be plotted independently or relative to another region using ratioing or differential amplification. The digital nature of input and output signals facilitates control-loop interaction with the host computer as arbitrator.

The USB 2.0 interface is attractive for this purpose because of its ubiquitous presence, but its nondeterministic nature requires some enhancements for time-critical measurements. The digital data is buffered inside the camera and is sent to the host via the USB 2.0 interface in a pipeline fashion, so processing occurs in parallel with data transmission with no lost packets. Without buffering, USB 2.0 relies on the host computer’s operating system to ask for more data when it is ready. If the system does not make such a request quickly enough, data is dropped. Buffering allows the operating system to accommodate other bus requests without losing data.

Virtual photodetectors do not have the low noise and high gain of photomultiplier tubes, the high sensitivity and high bandwidth of avalanche photodiodes, or the infrared spectral response of germanium detectors. But as generic, multipurpose photodetectors, they provide significant savings in terms of cost and time. Masks for the most common photodetector tasks (e.g., photodiode, quad detector or line-scan) are preloaded to provide out-of-the-box functionality.

Imagine that one is building a free-space optical switch with multiple input/output channels and wishes to measure the crosstalk and dynamic range at kilohertz rates. The output channels may be too close together for an array of individual PIN photodiodes. A fiber bundle coincident with the output plane to capture all output channels and to reroute them to each PIN photodiode is expensive and time-consuming to implement for something that will not be in the final product. A focal plane array would be nice but would not be nearly fast enough for this application.

The operation of a virtual photodetector sensor involves a trade-off among sample rate, detector dimension and dynamic range. In general, the sample rate decreases and the dynamic range increases with expanding detection area. The maximum sample rate generally is determined by the sensor hardware, with the sustained data rate limited by the digital interface to the host computer.

Using a virtual photodetector, however, one can capture an image of the output plane at an appropriate exposure, build a mask in software defining each output channel, upload the mask to memory inside the photodetector, start the detection sequence and thereby produce an X-Y plot in time of each output.

Consider a different application. Imagine that one has a high-resolution monochromator with curved output slits and would like to sample large spectral regions rapidly after a pulsed event. All of the light coming through the slit is required, but focusing on a linear array with a cylindrical lens destroys the resolution.

The virtual photodetector solves this problem. In only a matter of minutes, the user can create a mask with up to 512 curved “lines,” externally trigger the photodetector and then plot the spectra in real time on the host computer.

From prototyping to competitor

Virtual photodetectors offer real advantages for optical prototyping and product development. They may not be the final detector embedded in the system, but they can rapidly help the designer arrive at the photodetection requirements.

Nevertheless, as advances in CMOS imager technology increase sampling speed, and reduce noise and cost, virtual photodetectors may be expected to compete head-to-head with dedicated detectors.

Meet the author

Andy Whitehead is president of Southern Vision Systems Inc. in Madison, Ala.; e-mail: awhitehead@southernvisionsystems.com.