Total Measured Optic Response: A New Approach to UV LED Measurement
By Jim Raymont, Joe May, Mark Lawrence and Paul Mills
EIT Instrument Markets
The Fundamental Paradox of Measurement
Measurement is important to set process parameters and to identify if the process is running within those parameters. When things change in the process, measurement provides valuable clues that help locate the source of the problem(s). Without measurement, we cannot optimize the process for greatest profit and efficiency. Measurement allows communication with precision within a facility or between customers and their supply chain partners.
However, the act of measurement frequently presents its own problems. For instance, suppose you wish to measure the air pressure in your cars spare tire (Figure 1). How can you measure the pressure within the tire without changing the pressure itself? Opening the tires valve opens a can of worms that form the fundamental paradox of measurement. Scientists call the problem of changing somethings properties through the act of measuring it the observer effect. The engineers challenge is to devise an instrument that causes the least distortion of whatever we intend to measure.
The Challenges of Measuring UV
Measuring the output of a UV LED source presents similar challenges. To measure UV energy, we must first capture the sources output and then transform the optical energy into an electronic signal with a value that can be stored and displayed. Figure 2 illustrates a simplified, generic chain of components needed to perform this task.
In this example, UV energy is collected by the optical components (or optical stack) of the radiometer. Usually, a high-quality optical window is used to the condition the incoming energy and protect the sensitive components beneath. To ensure that all the energy from the source is captured and presented evenly, the UV passes through a series of diffusers and apertures. Since the detector also is prone to measure wavelengths other than UV, such as visible or infrared, this unwanted energy is removed using filters. Conditioned (uniform and filtered) UV energy is presented to the surface of a sensitive electronic photodiode detector. The photodiode converts the UV energy into an electronic potential with a magnitude that depends on the UV energy intensity. This signal is processed, stored and displayed. But, the accuracy of the UV measurement depends on the spectral response of each of the components that come between the original signal and the display. And, the radiometer designer must engineer the system to be affordable, easy to use and robust to the harsh conditions frequently encountered in industrial UV curing processes.
Over several decades of development, UV radiometers have evolved to meet these technical and commercial objectives. Instruments are available to measure the output of a wide range of conventional UV sources. Some devices try to accomplish this using a single set of optical and electronic components to capture the entire (100nm to 400nm) range of UV wavelengths. However, the spectral response of the components across such a large spectrum may not always be flat. This could cause errors in measurement. An alternative approach uses separate optical and electronic devices that each measure smaller portions of the broad spectrum. With broad band sources, measuring in different spectral areas allows the user to better understand the efficiency of the UV source. Concerned about adhesion or penetration through opaque coatings? Measuring the longer wavelengths (UVA, UVV) may be useful. If you are concerned about surface cure properties, measuring the short wave UVC or UVB will be more important. Comparing the values in multiple UV bands allows users to identify bulb types, spectral changes in the bulb and when maintenance is required. For example, by comparing readings, a user can detect short wave UV values dropping while long wave UV values remain the same. With a little experience, multiband radiometers also allow users to confirm the correct bulb type has been installed in a lamp housing.
As a bonus, the optics used in narrower spectral bands generally are more uniform. Figure 3 illustrates data gathered using this multi-band approach from a radiometer designed to measure UV output that has been segmented into the four distinct bands shown in Table 1. These designations were originally derived by CIE, the International Commission on Illumination, based on the observed health effects of various wavelengths. In practice, radiometer designers may use different definitions of these bands in their own instrument design that correspond more closely to the spectral output of commercial UV sources.
This approach is popular because the engineer can select devices with a predictable, repeatable, uniform spectral response within each band. This method solves the problem of trying to design a one size fits all solution that requires a set of components that must perform over a wide spectrum. This solution has served the industry well for decades of UV curing by measuring UV energy coming from arc and microwave UV sources.
LED Light Sources are Different
In the last ten years, the landscape for UV curing has been changing as new UV LED sources have been adopted for commercial applications. The output of a UV LED source has substantially different spectral characteristics than that of broad band mercury-based lamps. Figure 4 compares the output of a traditional mercury light source with the output of a UV LED array. While the output of mercury lamps appears as a continuous spectrum with many sharp peaks, UV LEDs produce a single peak, centered at a specific wavelength. Today, although development is occurring at other wavelengths, most UV LEDs used for curing have output ranging from 365nm to around 405nm.
Industrial UV LED sources are not a single tiny UV LED diode, but are arrays that contain hundreds or thousands of individual diodes (or dies) arranged in geometric columns and rows to provide uniform irradiance over the cure surface.
To assemble these arrays, the UV LED source manufacturer typically purchases individual diodes in large quantities. These are specified to have a given desired output wavelength. For example, to build a commercial 6″ x 6″ 395 nm UV array the manufacturer must purchase and mount hundreds of individual 395 nm diodes to the array. In reality, however, the output of any of these individual diodes may differ from another one depending on how precisely the semiconductor chemistry and manufacturing is controlled. In practice, UV LED diodes are created from a complex sandwich of semiconductors and other materials layered together in a delicate architecture whose exact size and geometry can vary the spectral output when the wafer is sliced into tiny diodes. Thus, a 395 nm array also will contain diodes with a central peak at 390 nm, 400 nm or any of a wide range of other possible wavelengths. Diode suppliers often price their components by how tight the tolerance of individual diodes in a batch may be (through a process known as binning). The tighter the binning specification, the higher the price, since the diode manufacturer must reject more non-conforming diodes to meet the guaranteed tolerance. A 395 nm source is sold with the Center Wave Length (CWL) typically specified as +/- 5 nm. In a competitive market, low-cost, low-quality manufacturers may be tempted to loosen up binning standards and assemble arrays made of a wide range of individual diodes. They may still refer to these diodes as though they are a single wavelength, but the measured output of a 395 nm array could vary considerably from the expected CWL of +/- 5 nm.
This variation in diode wavelengths presents an additional concern for UV radiometer design, since the spectral response must be kept flat over a wider range of wavelengths. As shown in Figure 5 a diode with output at a central wavelength, say 395 nm, actually emits energy over a substantially wider range. With good binning, the variation in the CWL from 390 to 400 nm on a 395 nm source, the output of the energy may be as low as 370 nm and as high as 425 nm (Figure 6). The effects of poor LED binning may have the energy even further outside of the bandwidth shown.
The bands proposed in Table 2 incorporate and capture almost all but a very tiny percentage of the energy from different LEDs. Keeping the bands narrow has several optics advantages, as described in the next section. It also would let end users know when a claimed 395 nm LED is closer to having energy output closer to 405 nm or more.
Total Measured Optic Response: A Different Approach
What if a radiometer design incorporated all the optical optic components shown in Figure 3 in the instrument response instead of just the filter? This would require incorporating all components in the optic stack in the instrument response.
A Total Measured Optic Response requires that the overall response of the instrument take into account the contribution of each component (e.g., the optical window, diffuser, aperture, spacers, filters and detector), each of which has its own response characteristics, as shown in Figure 7. This approach results in a flat overall instrument response.
While this integrated approach seems conceptually straightforward, until now radiometer design did not consider the Total Measured Optic Response of the sum of these individual components. One reason is that the output spectra of conventional mercury lamps is very predictable, due to the natural structure of the mercury atom. Since all mercury atoms are identical to all others, there is no need for binning atoms used in mercury lamps. Second, as shown in Figure 4, the mercury spectra is very broad, and the consequences of cutting off a few nanometers of output are not as great as with the sharp spectral output that characterizes the output of LED light sources. As can be seen in Figure 6, a filter that cuts off at 410 nm instead of 405 nm would have substantial impact on the measured UV output, since much of the energy would be cut off.
To solve this problem, a new approach considers the Total Measured Optic Response of the instrument. To assure uniformity within the range that commercial LEDs may vary due to differences from diode to diode in the array, the commercial UV LED spectrum is segmented into a series of LED or L bands. Each L band is constructed by considering a central peak wavelength that might vary by ±5nm. Thus, each band is 50 nm wide, as shown in Table 2. Thus, an L395 radiometer is equipped to measure diodes whose center wavelength vary between 390 nm and 400 nm using filters that capture all the energy emitted between 370 nm and 422 nm at the 50 percent power responsivity point. The widths of the L bands were chosen to balance flatness with performance at an affordable cost. Figure 8 shows the overall measured response achieved for the popular L395 band. Notice that the achieved response is exceptionally uniform over the desired region.
By considering the total optical response, not only are readings highly accurate within each designated L-band, but there is vast improvement in the correspondence of measurements made from instrument to instrument. This means that a process measured in the lab with one instrument may be replicated in the field with little error.
Figure 9 illustrates the accuracy and reliability of two production radiometers1. The pair of radiometers made 20 successive laboratory measurements of a single 395 nm UV LED array. The variation in absolute UV output from run to run, which averages substantially less than one percent, may even be due in part to small intertemporal fluctuations in the source itself. The difference between instruments, approximately 0.2 percent overall, is extremely narrow and significantly better than results with traditional, filter-only designs. Table 3 illustrates how closely absolute measurements of a 395 nm LED array match a national, traceable, primary standard at various working distances (data courtesy Excelitas Lumen Dynamics.)
The performance benefits of this new UV LED radiometer design are already being proven in the field. For example, Phoseon Technology found good spectral response in testing done using 385 nm, 395 nm and 405 nm light sources with an L-395 band instrument. The lamps were calibrated using a third-party meter with a known (measured) spectral response. When exposed to a 365 nm UV LED lamp, the L-395 radiometer measured very little UV energy, indicating that L395 spectral response has a sharp response in the L-395 band. The radiometer also demonstrated consistent peak irradiance and energy density measurements at various scan speeds varying from 1.2 to 6.0 meters/min. Finally, the unit had strong correlation with a NIST traceable meter from another manufacturer. These results have been replicated by others.2
This robust evidence points to the success of re-engineering the radiometer to meet the reality of man-made semiconductor light sources. Although LED proponents are quick to point to the longevity and stability of the arrays, irradiance, until now the effects of diode variation have been somewhat ignored, though every array is a mixture of diodes with different spectral emission. By creating L-bands across which the instrument response may be very flat, and by considering the individual contributions of each optical and electronic component on the radiometers overall response, a new generation of LED radiometers has been designed to help characterize the UV curing processes of the future.
- EIT LEDCure L395 radiometers were utilized.
- Integration Technology Ltd has tested the EIT LEDCure L395 and found it produces consistent readings on a variety of sources, with the profiler function being an extremely useful feature for research and product development.