Better Filters Yield Better Sensor Performance

HANK HOGAN, CONTRIBUTING EDITOR, hank.hogan@photonics.com

From weather satellites to lidar for self-driving cars, precision optical filters play a vital role in making sensors work. Today, the ability to interrogate more and more sections of spectrum promises enhanced capabilities for these and other applications. On the horizon are materials and techniques that could make precision optical filters tunable. However, performance, cost, and other issues must be addressed.

Thanks to precision optical filters, today’s state-of-the-art weather satellites cover a broad spectral range with a number of high-performance bands and can collect imaging data at four times the resolution of their predecessors. The latest satellite has filters from Materion Precision Optics. Courtesy of Materion/muratart/Shutterstock.com.

An illustration of today’s state-of-the-art capabilities can be found in the next-generation weather satellite GOES-16, which began operating in late 2017. The satellite collects three times the data at four times the resolution of its predecessors, thanks to precision optical filters that span the spectrum, from blue at 470 nm to LWIR at 13.3 µm.

“It does cover a broad spectral range with a number of high-performance bands, especially in the longwave and midwave [IR] that allows some additional details,” said Dave Harrison, business development manager for image and sensing at Materion Corp.’s Precision Optics, based in Westford, Mass. The company supplied the filters used in the satellite.

A next-generation weather satellite, GOES-16, on the launchpad. Thanks, in part, to advances in precision optical filters, the satellite’s imager, which is now operational, has three times the spectral bands and four times the resolution of previous-generation instruments. Courtesy of NASA and NOAA.

The satellite’s Advanced Baseline Imager has 16 spectral bands, compared to five for previous GOES imagers. The increase is part of an overall trend of more spectral bands in instruments.

Increasingly, precision optical filters must be multiband (a) or have very high transmission and blocking with steep slopes (b). OD: optical density. Courtesy of Alluxa.

For every band, the goal is to maximize transmission of a spectral region and block everything else, with a sharp slope in the transition region. This is accomplished by putting down multiple layers of coatings on a substrate, with perhaps 100 or more of the right thickness and composition needed to create the desired spectral response.

Band size challenges

An increase in the number of bands has resulted in some challenges. Typically, more bands mean fewer pixel rows on the sensor for each. This translates to a decrease in filter size, down from a typical millimeter width to as narrow as 20 µm, Harrison said.

Other challenges involve improving protection against scatter and stray light, thereby decreasing crosstalk between sensor channels. There also is a need to enhance the large angle performance because the pixels in a spectral band must collect data over a wider span. The decrease in the number of pixels per band also mandates fewer defects in filter films, as each defect can kill a bigger percentage of the available pixels.

Besides more bands, the future should see increasing transmission, with this moving from previously 95 percent or so to perhaps as much as 99 percent, according to Harrison. Volumes are also increasing, pushed up by growing demand for imagery and data.

There is a desire among customers to deposit the filter directly atop the sensor because this eliminates manufacturing steps and cost. Direct deposition has been done with some simple RGB detector wafers. However, the cost and risk of this approach for expensive sensors is too great for current technology, given the chance of snafus during the manufacturing process.

“If you’re doing 20 deposition runs on somebody’s detector wafer, there will be something that happens,” said Kevin Downing, director of business development for space and defense at Precision Optics.

The ability to interrogate more and more sections of spectrum promises enhanced capabilities for a range of applications.

In addition to satellites, lidar for self-driving cars and mobile applications are putting new demands on filters. For lidar, the requirement is for narrow filters with flat tops and the highest possible transmission, according to Peter Egerton, chief commercial officer at Alluxa in Santa Rosa, Calif. Across all fields, there is a need for more and more bands with the highest possible transmission or to cut costs, he said.

Alluxa addresses both the higher performance and lower cost angles through proprietary manufacturing technology based on an advanced plasma deposition process. The company created its own production equipment, developing hardware, software, and control solutions. Controlling the deposition of layers quickly, accurately, and repeatedly has been key to Alluxa’s success, Egerton said, as has been its ability to adapt the technology to keep up with changing requirements.

Complex demands

A clear trend in filters is that the complexity of requests has been increasing. This may be in the form of more spectral bands, with steeper slopes, higher transmission, and deeper blocking. Consequently, both the design of the filter and its manufacturing are more challenging.

“Some of our coating designs are multiple thousands of layers,” Egerton said. “We had one that was over 10,000 layers. That’s our record.”

Volumes are also growing, with an initial run of a few filters being followed by a ramp-up to a thousand or more. This comes without any loosening of performance requirements, which places a premium on manufacturing capabilities.

Regarding other uses, fluorescence applications in the life sciences and elsewhere benefit significantly from enhanced performance of precision optical filters, said Ken Pihl, senior product marketing manager for thin-film coatings, filters, and mirrors at the Franklin, Mass.-based Light & Motion division of MKS Instruments.

These performance improvements include lower loss, reduced size, single-surface construction, tighter surface quality specifications, and maximized signal-to-noise ratio. The last is critical.

“The highest performing filters in terms of signal-to-noise enable detection of very weak fluorescence emission signals,” Pihl said.

He added that achieving the maximum performance demands precise control of the deposition of refractory metal oxide coatings. Thus, the manufacturing equipment (which consists of chambers where the deposition is done under a vacuum) and the environment where the coating is done require stringent controls. There also must be sophisticated software capable of managing a deposition process that involves hundreds of layers, some of which are extremely thin.

Predictions

For the future, Pihl predicted smaller filters with exacting surface quality and scatter requirements. He noted that spectral performance is approaching a maximum, with greater than 95 percent transmission on the blocked bandpass filters now often produced. Squeezing out the remaining few percent possible will be a challenge, he said.

In addition to the current commercial approach to making precision optical filters, research is underway into totally new methods. For instance, Howard Lee, assistant professor of physics at Baylor University in Waco, Texas, is investigating (among other things) metasurfaces. These consist of planar arrays of artificial structures that are smaller than a wavelength of interest, with that arrangement giving the metasurface properties not found in natural materials.

A man-made metasurface composed of subwavelength structures can radically change its absorption with a voltage change. Such tunable optical properties could lead to tunable filters, consisting of gold (Au), hafnium dioxide (HfO₂), and indium tin oxide (ITO). The wavelength (λ) at which absorptance spikes (or the reflectance [R] dips) moves with applied voltage. ENZ is the epsilon near-zero point of metamaterial, which is useful in antennas. Courtesy of Howard Lee, Baylor University.

The structures can be very refined, only 10 to 20 nm thick, and composed of a transparent conductor, such as indium tin oxide (ITO). When a voltage is applied, the electron density in the conductor changes, and that alters its optical properties. Reflectance or absorption can be affected radically.

“For example, we can absorb most of the light, 99.9 percent of the light,” Lee said. “So, it’s really efficient. We can design the film for a particular wavelength. It’ll absorb only for one wavelength. Or we can combine a couple of films to get a broadband absorber.”

Transmission is currently around 70 percent. The points at which transmission happens is tunable over 150 or so nanometers. That range, as well as the efficiency, could go up as the research continues, Lee said, with the tuning perhaps spanning 200 nm or more and with transmissivity as high as 90 percent.

The metasurface responds rapidly to changes in voltage, making it possible for switching speeds to be 10 GHz or more. An application for this filter technology would be in a color CMOS imaging sensor. Today, sensor pixels are dedicated to red, green, and blue. With a tunable filter, each pixel could be set, in turn, for each of the colors, effectively tripling the number of pixels in the sensor.

To do that, though, the metasurface filters must get out of the lab and be integrated into an actual device. There is commercial interest in the research, with large consumer electronics and internet companies supporting it. The most likely applications are in areas where the need is for thin and smart devices, perhaps in displays for virtual or augmented reality, Lee said.

Another approach to precision optical filtering comes from work done at the National Institute of Standards and Technology (NIST) in Boulder, Colo. Researchers there, in conjunction with teams from the University of California, Santa Barbara and the California Institute of Technology, combined two chip-size frequency combs with a programmable semiconductor laser chip. The result was as much as 100,000 times more stable than traditional compact lasers.

Researchers combined a pair of frequency combs, miniature lasers, and other components to replicate the capabilities of a tabletop-size optical frequency synthesizer on microchips, shrinking the system significantly. The technology could be used for precision optical filtering. Artist’s rendering of the system (a). Setup 1 m on a side, and now 1 cm (b). Courtesy of DARPA (Figure a) and University of California, Santa Barbara (Figure b).

It shows hertz-level frequency stability over hours of time, said Daryl Spencer, a physicist on the research team. Spencer was lead author of the Nature article titled “An optical-frequency synthesizer using integrated photonics,” which reported on the project in April 2018.

Comparable stability today is achieved by calibrating a laser source against a gas cell or Fabry-Pérot cavity, Spencer said. That is more expensive, more power hungry, and larger than what can be done with the new approach.

Ultrastable lasers can be used for an optical spectrum analyzer and filtering, according to Spencer. This could be done by sweeping the tunable laser through a spectral region, and, by the resulting interaction with any incoming light, it can create a much lower frequency output. Standard electronic means can extract that output. Knowing the exact wavelength of the source makes it feasible to precisely pinpoint the incoming light’s spectral location, thereby enabling both filtering and analysis.

The new technology could be widely useful because of its characteristics. “By utilizing chip-scale components,” Spencer said, “there is great promise for efficient, portable systems to be realized.”