Multi-Spectral Daylight Sensor

The Problem with Conventional Daylight Sensing Technology

Responsive daylight control, in which shading is automatically adjusted to maintain a desired level of glare-free natural illumination, is the key to unlocking the full energy-saving, ergonomic, health, and wellness benefits of automated shading.

Unfortunately, conventional responsive daylight control technology just isn’t practical for mainstream use—and a major reason is that conventional light sensors don’t respond to daylight in the same way that humans do.

Responsive daylight control is the most valuable form of automated shading, but also virtually non-existent in today’s buildings because it’s so hard to implement with conventional technology.

Conventional Sensor-Based Systems

Many automated shading product include provisions for a light sensor aimed at providing some form of responsive daylight control.

All such products we’ve tested provide what is referred to in control theory as bang-bang control : the shading setting is alternated between only two positions (a fully-closed position and a user-selected position), depending on the sensor output relative to a user-set threshold:

If daylight always caused only either severe glare or no glare, then bang-bang control would be the perfect form of responsive daylight control.  Of course, that’s not the case, so the shading is often either too open or too closed for the prevailing conditions.

Still, a bang-bang control system that reliably blocks daylight glare would be extremely useful.  Unfortunately, that’s not possible with the type of system shown in Figure 1.

That’s because a conventional light sensor is a scalar sensor that measures just a single quantity, i.e. illuminance in the case of a visible-wavelength sensor.  However, humans perceive not just the intensity of light, but also the angular distribution of that intensity.  Daylight glare can be due to either high illuminance (which can be sensed by a scalar sensor) or high contrast in the visual field (which can’t be sensed by a scalar sensor).

The limitations of conventional sensors are especially evident when sensing direct sunlight from the solar disc, which is a prime cause of daylight glare. Direct sunlight is easy to sense on a cloudless day when the sun is high in the sky, because it results in an extremely high irradiance and temperature.  But that’s not the case when the sun is partly obscured by clouds or is low on the horizon.

As a result, a system like the one shown in Figure 1 will often close the shading when there is no risk of glare, or open it when there isn’t—so users often just turn off the automatic capability because it’s more annoying than useful.

Conventional Systems Based on Calculated Solar Geometry

Modern smart products have substantial on-board processing capability and/or are connected to a home-automation hub, voice-assistant, building-automation system, or the cloud.  Is there a way to harness those resources to improve the operation of a system like that shown in Figure 1?

The answer is “yes”—but only to a limited degree.

Specifically, the reliability of inferring the presence of direct sunlight can be increased by augmenting the output of the sensor with the azimuth and elevation of the solar disc.  Here’s how it works:

• The azimuth and elevation of the solar disc is calculated on the basis of the time of day, day of year, and window latitude and longitude.
• If the sun’s calculated position is more than 90 degrees from normal to the window surface, the system infers that direct sunlight is absent because it would be geometrically impossible.
• On the other hand, if the sun’s calculated position is less than 90 degrees from normal to the window surface, the system infers that direct sunlight is present when the output of the irradiance sensor exceeds a threshold which varies as a function of the solar position.  This variable threshold significantly increases the reliability of the inference relative to a fixed threshold.

The simplest such systems fully close a shading device when direct sunlight is inferred.  This type of bang-bang control imposes no special requirements on the shading device, which enables it to be implemented via a software or firmware upgrade to a system like the one shown in Figure 1.

In more advanced systems, a venetian blind or a roller shade is automatically closed only enough to block light coming from the direction of the calculated solar position.  This type of selective sun-blocking is far less obtrusive than bang-bang control and results in a higher average level of glare-free natural illumination—but also comes at a significant increase in complexity:

Actually, a selective sun-blocking system like the own shown in Figure 2 is possible even without substantial processing resources or internet connectivity.  In fact, we patented such a system back in the 1990’s.  However, we let that patent lapse because we came up with something much, much better…but more on that later.

One key disadvantage of solar-position-based systems like the one of Figure 2 is that they must “know” the window’s latitude, longitude, and compass orientation to calculate the relative solar position, which adds a step to the set-up process.

An additional disadvantage is that such systems need to “know” the setting of the shading device (e.g. the slat-tilt angle or roller-shade height) in absolute terms relative to the solar geometry.  This requires either an absolute-position sensor or careful calibration of the shading device, both of which increase cost and installation complexity. Backlash in the shading device is also an issue if an absolute position sensor isn’t used, and must be accounted for in the shading-control algorithm.

Finally, all solar-position-based systems have the disadvantage that the irradiance sensor must “see” the sun from the same perspective as the room occupants in order to ensure fully reliable operation.  Unfortunately, this generally isn’t practical, so the sensor will often be shaded from direct sunlight which reaches the occupants, and vice-versa.  As with the simpler system of Figure 1, this can prompt users to simply turn-off automatic operation because it’s more trouble that it’s worth.

Conventional Camera-Based Systems

Actually, is possible to implement a highly effective daylight control capability with conventional technology.  The solution is to use a wide-angle digital image sensor (i.e. a camera) and specialized image-processing.  Glare can then be inferred on the basis of the image contrast across the camera’s field-of-view, and the shading device can be adjusted to keep the inferred glare below a desired threshold:

However, as with a sensor-based system, the camera must be positioned and oriented to “see” the shading device from the same perspective as the occupants; if that’s not possible (and it typically isn’t), the camera might not provide any glare-sensing advantage over a simple scalar sensor.

Further, the complexity of the camera and required image processing can be prohibitive for cost-sensitive products intended for the mainstream market.

Those aren’t the only reasons that camera-based responsive daylight-control systems have failed to gain a foothold in the market, but they’re enough: as far as we’re aware, no camera-based system has been successfully commercialized (at least in the U.S.).

A Much Better Way to Sense Daylight: Our Multi-Spectral Glare Sensor

Conventional attempts to sense the glare-inducing properties of daylight are based on luminance and contrast.

However, through long-term testing of daylight control systems, we’ve discovered that subjective perceptions of daylight are also highly correlated with the spectral composition of the daylight.  Specifically, the risk of glare increases not just with the total irradiance (as would be expected), but also with the proportion of irradiance at longer wavelengths.

The reason is that direct sunlight is the primary cause of daylight glare, especially when the sun is low on the horizon—and the spectral composition of daylight indicates not just the presence of direct sunlight, but also the approximate solar elevation.  That’s due to two phenomena related to the scattering of light in the atmosphere:

• Rayleigh scattering is the isotropic scattering of light by air molecules by an amount that varies inversely with the fourth power of the wavelength.  This is what causes a cloud-free sky to appear blue, an overhead sun to appear yellow, and  a setting sun to appear reddish orange.
• Mie scattering is the mostly forward scattering of light by larger particles such as water droplets and smoke.  Unlike Rayleigh scattering, it has only a weak dependence on wavelength, and is what causes clouds to appear achromatic (i.e. white or gray).

As a result of these scattering phenomena, the spectral composition of daylight reveals much more about sun-sky conditions that just the total irradiance as sensed by conventional sensors.

In fact, when combined with the total irradiance, the spectral composition of daylight tells us everything we need to know to reliably sense the risk of glare.  Even better, we don’t need to fully characterize the daylight spectral composition; all we really need to know is the ratio of irradiances in two different spectral bands in the daylight spectrum—and we can do that with a pair of inexpensive scalar sensors.

To better illustrate why the multi-spectral sensor works so well, the following chart summarizes the results of some of our testing of responsive daylight control systems in occupied spaces.  It plots various sun-sky conditions on a two-dimensional map: the Y-axis represents the square root of the total irradiance (consistent with Stevens’ Power Law for apparent brightness) , while the X-axis represents the ratio of irradiance in the Near-Infra-Red (NIR) band to the irradiance in the Near-Ultra-Violet (NUV) band.  Also shown is the subjective threshold between acceptable and unacceptable glare (based on occupant adjustments of the daylight set-point for responsive daylight control).

As the chart shows, irradiance alone isn’t a good indicator of daylight glare—but augmenting it with the ratio of long-to-short wavelength irradiance can yield an excellent indicator of daylight glare under all conditions. And that’s the essence of our multi-spectral sensing approach.

A good test of a glare sensor is to see how it responds to low-angle sunlight.  To that end, the following chart compares the output of one of our multi-spectral glare sensors with the output of a conventional irradiance sensor, over the course of an clear sunny afternoon through sunset:

As the sun descends toward the horizon, the irradiance drops due to the increased atmospheric path length traveled by the sunlight.  However, while the irradiance is dropping, the risk of glare is actually increasing, because low-angle direct sunlight can penetrate deeper into a room and the solar disc is directly within the occupants’ field-of-view.

If the output of the conventional sensor were used for responsive daylight control under these conditions, then either the shading would open prematurely in the afternoon (admitting glare)—or else the daylight-control setpoint would have to be decreased so much that the shading would remain excessively closed even when there is no risk of glare.

On the other hand, our multi-spectral sensor correctly senses the increasing risk of glare as the sun descends toward the horizon.  Using this sensor for responsive daylight control greatly reduces the risk of glare while maximizing glare-free natural illumination.

There’s also another key advantage of this approach: while the spectral composition of daylight is affected somewhat by reflection from room surfaces, the irradiance ratio still provides a reliable indication of glare from low-angle direct sunlight even when the direct sunlight doesn’t directly impinge the sensor.

So, unlike conventional daylight sensors, our multi-spectral sensor doesn’t need to “see” the sky from the same perspective as the building occupants in order to reliably sense glare under the most stressing conditions.

Further, the two spectral bands used to obtain the irradiance ratio (and the total irradiance) don’t actually have to be in the visible spectrum, as long as they’re close to it.  For example, the two irradiances can be sensed in the Near-Ultra-Violet (NUV) and Near-Infra-Red (NIR) bands which overlap the daylight spectrum, but are just outside the visible wavelengths produced by fluorescent and LED lamps:

The NUV and NIR spectral responses shown in Figure 6 are typical of inexpensive NUV and NIR photodiodes with visible-light-blocking filters.   Because of the relatively large difference in wavelength, using the NUV and NIR bands provides greater sensitivity to glare than using, e.g., the blue and red bands in the visible spectrum.

However, an even more significant benefit of using the NUV and NIR bands is that the photodiodes don’t have to be shielded from artificial illumination.  In addition to simplifying packaging, this also provides wide flexibility in locating and orienting the multi-spectral sensor, and in particular enables closed-loop as well as open-loop daylight control:

Our multi-spectral sensor offers two key advantages for open-loop control:

• It provides a far more reliable indication of direct sunlight—especially when the sensor is partially shaded—than conventional sensors.
• Unlike systems that rely on the calculated solar geometry, it needs no specialized commissioning process for highly reliable bang-bang glare control (commissioning would still be needed for selective sun-blocking with a venetian blind or roller-shade).

However, while our multi-spectral sensor finally makes open-loop glare control cost-effective, it’s even more advantageous for closed-loop control.

Closed-loop daylight control offers compelling advantages, including the potential for selective sun-blocking without need for specialized commissioning or an absolute position feedback sensor.  However, as might be expected, closed-loop control presents its own unique challenges:

• A closed-loop daylight sensor must respond only to daylight while ignoring any artificial illumination.  Otherwise, any switching or dimming the lamps will cause a spurious shading adjustment.
• As with open-loop control, a closed-loop sensor must respond to the daylight (and, in particular, the angular distribution of the daylight) in the same way a human would.

Our multi-spectral sensor overcomes both challenges: inexpensive NUV and NIR sensors provide highly reliable sensing of daylight glare, and their immunity to artificial illumination allows them to be located on the room-side of the shading device and physically integrated into a wide variety of automated shading products.

But whether used in an open-loop or closed-loop configuration, our multi-spectral sensor greatly simplifies implementation of a responsive daylight control system, as shown in the following chart:

In addition to enabling a simplified system configuration, the multi-spectral sensor itself is also simple and inexpensive to implement:

• It requires no optics, which minimizes cost and volume.
• The processing overhead for the multi-spectral glare algorithm is only a tiny fraction of the capabilities of the microcontroller in a typical smart-shading product, so a dedicated microcontroller is unnecessary.
• Finally, our sensor can directly leverage commodity ambient-light-sensing chips developed for the huge market for personal electronic devices.  Such chips integrate multiple photodiodes, amplifiers, and A/D converters along with power-management circuitry and an I2C interface in a single tiny package.

As a result, a multi-spectral sensor for open-loop applications can be implemented with as little as just a single ambient light-sensing chip (along with a lightweight algorithm hosted on an existing microcontroller).

Closed-loop control requires both NIR and NUV responsivity, which doesn’t currently exist in any single ambient light sensing chip currently on the market.  Therefore, a multi-spectral sensor for closed-loop control currently requires two chips (occupying a volume of less than 10 mm³), but a single-chip solution is coming soon.

In both of these configurations, the two multi-spectral bands are close enough to the visible band that ordinary acrylic plastic can be used as a window to admit light to the photodiodes when they are enclosed in a housing. Alternatively, because the sensor chips are so small, a small hole can be used instead of an acrylic window.

Complementary Technologies

While our patented multi-spectral sensing technology provides a highly cost-effective means of sensing daylight glare, it’s even more advantageous when combined with three of our other innovations:

• Our fluctuation-mitigation technology enables a responsive daylight-control system to respond quickly to isolated changes in the daylight level while ignoring sustained high-amplitude fluctuations due to moving clouds.
• Our closed-loop daylight-control algorithm for horizontal venetian blinds mitigates spurious daylight components which would otherwise confuse the output of a closed-loop sensor.
• When integrated into a device that is mounted near or on a window, our multi-spectral sensor technology  works synergistically with our patented thermal-gradient sensing technology to enable minimization of HVAC energy consumption as well as high-performance daylighting.

See our IntelliBlind™ reference design for an example of how all of these technologies can be advantageously leveraged in a single product.

Applications

Our multi-spectral sensing technology’s low-cost, small size, absence of the need for optics, minimal processing overhead, and immunity to artificial illumination enable it to be integrated into virtually any device that could benefit from an advanced, cost-effective daylight-sensing capability.

For example, our IntelliBlind™ Smart Miniblind Actuator and IntelliLux™ Smart Headrail Sensor incorporate an inward-facing multi-spectral sensor to enable closed-loop daylight control, while our IntelliLux™ Smart Window Sensor incorporates an outward-facing multi-spectral sensor to enable open-loop daylight control: