Physical and Visual Evaluation of Filters for Direct Observation of the Sun and the International Standard ISO 12312-2:2015

We investigated the compliance of 43 commercially available solar filters (eclipse glasses) with the ISO 12312-2:2015 standard by measuring their spectral transmittances (280–2000 nm) and calculating their luminous, solar ultraviolet A, ultraviolet B, and infrared (IR) transmittances. We also evaluated the filters for usability by observing the full midday Sun and rating the view on a seven-point balanced scale, from “far too dark, details seen only with great difficulty” to “far too light, uncomfortable to view the Sun.” The mean ratings of two observers, one experienced and one inexperienced in solar observing, differed by 0.28 (95% confidence interval of the mean = 0.26). The inexperienced observer tended to be less accepting of high transmittances. All 43 solar filters complied with the UV and IR requirements. Eighteen filters passed the luminous transmittance requirements, and 24 were borderline too light or too dark. Seven of the 15 solar filters with a luminous transmittance darker than the requirement were rated as acceptable. One filter that passed and another that was borderline too light were rated as too light or far too light. The ISO 12312-2 limits derive from welding filter standards and do not represent an appropriate evidence base for direct solar viewing. This work provides the evidence base for a maximum 0.0012% and a minimum 0.00004% luminous transmittance for solar filters. The results of this study also support the use of welding filters between shades 12 and 16. Lighter welding filters are more acceptable than solar filters of the same luminous transmittance.


Introduction
Around solar noon, the solar disk has a luminance of about 1 × 10 9 cd m −2 (Sliney & Worlbarsht 1980). For comfort, it is necessary to reduce the luminance to 1 × 10 4 cd m −2 , requiring a filter with a luminous transmittance of about 0.001%, or optical density (OD) of 5 (Sliney & Worlbarsht 1980). As a consequence, prolonged unfiltered viewing of the Sunʼs disk is uncomfortable and the view is unbearably bright.
During the partial phases of a solar eclipse, the area of the Sun is diminished, and the view may become bearable, but the luminance of the exposed portion remains the same. The mechanism of retinal damage in eclipse blindness was once thought to be thermal but is now known to be biochemical. Given that the cornea and crystalline lens of the eye protect the retina from much of the ultraviolet spectrum, the sensitivity to damage is limited to short visible wavelengths and is known as the blue-light hazard (International Commission on Non-Ionizing Radiation Protection 2013).
The solar blue-light radiance ranges up to 1.71 × 10 6 W m −2 sr −1 with a median of 1.31 × 10 6 W m −2 sr −1 ; these represent maximum safe exposure durations of 0.82 s and 1.07 s, respectively (Okuno 2008). With a gray (neutral density) filter, the luminous transmittance of 0.001% that provides comfort extends the safe viewing times to 82,000 s and 107,000 s, respectively. As a general rule, white light sources that do not contain excessive amounts of blue and are comfortable to view are not a blue-light hazard (Sliney & Worlbarsht 1980). As a consequence, there is not a pressing need to have a specific blue-light hazard requirement in addition to the luminous transmittance requirement for solar filters. The current ISO 12312-2 upper limit of luminous transmittance, 0.0032%, provides maximum exposure durations of over 25,000 s and 33,000 s, respectively.
One of the continuing controversies raised in advance of each solar eclipse is the question of how to view an eclipse safely. In the mid-20th century, some solar observers began to advocate the use of dark welding filters (shade number 13 or 14) as an inexpensive means to view the uneclipsed or partially eclipsed Sun safely. During the total phase of an eclipse no filters are needed, as the solar corona is similar in brightness to the full Moon. Glass or polycarbonate welding filters provided a direct, unmagnified view of the solar photosphere with a predictable and consistent reduction of its brightness. Among eclipse aficionados, welding filters replaced potentially hazardous viewing materials such as glass with a deposit of soot, overexposed and developed black-and-white film negatives, and others.
By the 1980s a variety of solar filter materials suitable for both direct viewing of the Sun and filtering of camera lenses, binoculars, and telescopes had been developed. A polyester film with a coating of aluminum (SolarSkreen ® ) gained popularity among solar observers. Several competing products soon became available. The atomic aluminum coating and extremely thin filter substrate resulted in a sharp blue-white solar image, and these products soon became used in both solar viewers and as filters for solar imaging through lenses and telescopes. Pinholes in the metallic coating, artifacts of the manufacturing process, raised some concerns about safety, but in practice they proved not to be a hazard.
In the 1990s an alternative material was developed: black polymer incorporated carbon particles imbedded in a resin substrate and was similar in function to soot-covered glass. The thicker filter material rendered a yellow to orange solar image, and the filter was more resistant to physical abuse. (Although aluminum-coated films would show marring as a result of rough handling, they were shown to be very puncture resistant, and the surface blemishes did not materially affect the transmittance of solar radiation.) Black polymer filter material has now come to dominate the market for what are often referred to as solar eclipse glasses. More recently, metal-coated black polymer and a stiffer form of black polymer mounted in plastic sunglass frames have been introduced.
As a result of the eye safety controversy that occurred in advance of the 1999 August total solar eclipse over Europe, the European Committee for Standardization (CEN) developed a standard for filters for direct observation of the Sun that specified a range of acceptable levels of luminous transmittance and transmittance of ultraviolet (UV) and infrared (IR) radiation (CEN 2005).
An agreement between CEN and the International Organization for Standardization (ISO) resulted in the replacement of EN 1836:2005 with a global standard ISO 12312-2 (ISO 2015). This is also published as the European Standard EN ISO 12312-2:2015 and the Australian Standard AS ISO 12312-2:2020.
In 2016 and 2017, the American Astronomical Society (AAS) and NASA publicized the use of solar viewers that complied with ISO 12312-2:2015 for observing the partial phases of the total solar eclipse of 2017 August 21 over the United States of America. This approach was also adopted by the American Academy of Optometry and the American Academy of Ophthalmology in their public outreach. The AAS invited manufacturers, distributors, and retailers of solar eclipse viewers to provide samples and compliance reports of their products for the AAS list of reputable vendors (American Astronomical Society 2017). The AAS also received samples of solar viewers that were not in compliance with ISO 12312-2 from vendors wishing to be included in the online list, as well as from the general public.
The purpose of this study was to investigate the compliance of commercially available solar eclipse filters with the ISO 12312-2 standard. The results were also used to assess the appropriateness of the provisions of ISO 12312-2.

Samples
Samples of eclipse glasses were obtained online through ebay.com.au by one of the authors (S.J.D.). Other samples were submitted to author R.T.F. at the AAS for inclusion in the online list of vendors or provided by another author (B.R.C.) from a personal collection of eclipse viewers and glasses.
A total of 22 samples were obtained. They comprised 14 with a thin cardboard spectacle-like frame, 4 with a thin cardboard frame intended to be handheld, 3 with plastic spectacle frames, and 1 filter-only sample.

Transmittance Measurements
The spectral transmittances were measured in the range 280-2000 nm using a Varian Cary 5000 UV-VIS-NIR spectrophotometer without an integrating sphere (Varian Australia Pty. Ltd., Melbourne). The Cary 5000 is a dual beam, double monochromator instrument with photomultiplier and lead sulfide detectors and quartz halide and deuterium sources. The detectors are changed at 900 nm, the grating at 850 nm, and the sources at 350 nm. The wavelength accuracy is ±0.2 nm (visible) and ±0.4 nm (infrared) assessed with deuterium and mercury discharge sources. The spectral halfbandwidth is 5 nm (visible) and 20 nm (infrared).
The uncertainty of measurement has been assessed using a set of gray filters of transmittance from 73.2% to 9.5%. These filters were calibrated according to the methods for which the University of New South Wales's Optics & Radiometry Laboratory (ORLAB) is ISO 17025 accredited as a calibration facility by the National Association of Testing Authorities, Australia. This methodology allows the uncertainty of measurement to be estimated for very low transmittances without having to make repeated measurements of each sample, which can be inordinately time consuming. The results translate to uncertainties of measurement as listed in Table 1. ORLAB is accredited, under its Performance and Approvals accreditation, to test solar filters to ISO 12312-2 and welding filters, which have transmittances of the same order, to EN 169 (CEN 2002), ANSI Z87.1 (ANSI/ISEA 2015), and AS/NZS 1338.1 (AS/NZS 2012).
Baselines of 100% and 0% were recorded before each measurement session. To achieve measurements of the filters' low transmittances, the method involves the technique known as reference beam substitution. Three perforated sheet metal gauzes of nominal optical densities (OD) 0.5, 1.1, and 1.5 (total OD 3.1 = 0.08% transmittance) were placed in the reference beam and the transmittance was measured and recorded by the instrument. The sample was then placed in the test location with the boxed center of the filter at the measurement point. The values were extracted at intervals of 5 nm (visible) and 10 nm (infrared).

Calculations
The luminous transmittance to International Commission on Illumination standard illuminant D65, τ v, D65 was calculated. ISO 12312-2 (ISO 2015) does not specify the reference source, but ISO 12312-1 (ISO 2013b) specifies D65. As a check, and since it is, in principle, more correct, the luminous transmittance to direct solar radiation (Moon 1940), τ v, S , was also calculated. The solar ultraviolet B transmittance, τ SUVB , was calculated. The solar ultraviolet A transmittance, τ SUVA , was calculated, according to ISO 12311 (ISO 2013a) and ISO 4007 (ISO 2018), as a summation from 280 to 380 nm. However, ISO 4007:2018 now specifies two measures of solar ultraviolet A transmittance, τ SUVA380 and τ SUVA400 . In the latter the summation is extended to 400 nm. Accordingly, both measures have been calculated. The infrared solar transmittance, τ SIR , was also calculated.
In addition, the solar blue-light transmittance, τ SB , was calculated in the range of 380-500 nm. The uniformity of luminous transmittance (ISO 12312-2:2015 Clause 4.1.2) was not measured. Knowledge of the products shows that repeated measurements are tedious; the specified 10% tolerance is rarely achieved and is not necessary. However, the difference in transmittance between the two filters of a pair and of the two test points on a single filter covering both eyes was calculated.

Dimensions
The physical dimensions of the filters were checked using a template constructed with the dimensions specified in ISO 12312-2:2015 Clause 4.3.2.

Labeling
The labeling of the filter frames or their packaging was examined for claims of compliance with ISO 12312-2. Where this was present, compliance with ISO 12312-2:2015 Clause 5 was assessed. Other claims of compliance were documented.

Evaluation of the Solar Image
The eclipse glasses were evaluated by two colocated observers (one experienced in solar eclipse viewing and one not) on a seven-point balanced scale: 1. Far too dark, details seen only with great difficulty. 2. Too dark, details can be seen and the filter is usable, but more light preferable. 3. Dark but within the acceptable range. 4. Ideal for viewing detail and comfort. 5. Light but within the acceptable range. 6. Too light, details can be seen and the filter is usable, but less light preferable. 7. Far too light, uncomfortable to view the Sun.
The observers independently viewed a full Sun around noon (Australian Eastern Standard Time) in Kensington, New South Wales, Australia (latitude 31°.9S) in 2018 September. The Sun was at an altitude of about 50°. 4 There was no visible cloud, except low on the horizon (< 1 8 sky coverage), and the sky was deep blue. The observers used the scale values to rate the appearance of the Sun through the filters.
For reference, the two observers also evaluated the view through six welding filters of known shades between 9.4 (luminous transmittance 0.025119%) and 14.4 (0.000181%). Table 2 is a list of the test samples used in this study with a description of each viewer, whether it met the criteria specified in ISO 12312-2 for dimensions and labeling, and the details of standards identified in the labeling (see the discussion section below). We report our transmittance measurements and calculations in Table 3, including luminous transmittance, solar ultraviolet A and B transmittance, solar infrared transmittance, and solar blue-light transmittance calculated according to ISO 12312-2. Also indicated is whether each filter met the transmittance criteria specified in ISO 12312-2. Table 4 lists the performance of the right and left filters of a pair in rendering the solar image, based on subjective evaluation by an experienced observer (B.R.C.) and an inexperienced observer (S.J.D.). Also tabulated is the measured luminous transmittance as well as compliance with the luminous transmittance requirements of ISO 12312-2.

Results
All results are subject to the uncertainty of measurement listed in Table 1. ISO 12312-2 defines a maximum uncertainty of measurement for the measurement of spectral transmittance (25%), and ISO 12311 defines how this is to be interpreted with respect to compliance. If a result lies within the uncertainty of measurement of the requirement, it is deemed to have failed the requirement. As a consequence, the effective transmittance requirements differ from the stated requirements by the uncertainty of measurement of the testing laboratory and up to the maximum required in the standard. This means that different laboratories may report different passes and fails depending on their uncertainties of measurement. This is an incentive to testing laboratories to minimize their uncertainties of measurement. Acceptable variation between laboratories is As a consequence, the results listed as "borderline" here fail the requirements of ISO 12312-2 on the basis of the data reported here, but it is possible (however remotely) that another laboratory, particularly with smaller uncertainties of measurement, might report some as passing. Figure 1 shows the correlation between the evaluations of the filters by the two observers. There is good agreement between them in determining what is an acceptable brightness of the solar image; the mean ratings of the two observers differed by 0.28 (95% confidence interval of the mean = 0.26). The inexperienced observer tended to be less tolerant of high transmittances.

Discussion
The calculated transmittance values for ultraviolet A, ultraviolet B, and infrared were well within the limits specified in the ISO 12312-2 standard. There is negligible risk of eye injury due to exposure to these wave bands when viewing the Sun through the filters in accordance with the manufacturerʼs instructions.
The luminous transmittance findings are of concern because some filters rendered a very dim solar image. If the Sun cannot be seen easily through the filter, the user is more likely to stop using it and to steal brief views of the uneclipsed or partially eclipsed Sun and be more at risk of eye injury as a result. Excessively dark filters should be excluded, but the present minimum luminous transmittance specified in ISO 12312-2, 0.000061%, also excludes a number of filters that are not rated as too dark.
There were 15 filters with values of τ v, D65 below the standardʼs lower limit of luminous transmittance. Of those just below the limit, eight were rated as "too dark" or "far too dark," and it is appropriate to exclude them. The remaining seven below the minimum were rated as acceptable but would be excluded by the standard. A modification to the minimum luminous transmittance, lowering it to 0.00004%, would exclude the eight filters rated as "too dark" or "far too dark" and would allow the inclusion of the filters that are just below the present luminous transmittance limit but are rated as acceptably dark. It would also exclude two filters rated as "dark" by both observers. Note that if the Sun is viewed at an altitude below about 10°, atmospheric extinction could dim the view so much that a compliant filter near the new recommended minimum luminous transmittance would nevertheless be too dark to be usable.
Only two samples and one welding filter were rated as allowing an unacceptably bright solar image. Although the luminous transmittances of the solar filters are within currently acceptable limits, the image is too bright for extended comfortable viewing of the solar photosphere and could possibly impair the observerʼs ability to perceive the much fainter solar corona upon removing the filter during totality.
While the maximum luminous transmittance specified in ISO 12312-2 is consistent with the earlier recommendation of at least shade 12 that were based on welding filter shade numbers (Chou 1998), it allows filters that permit excessively bright solar images. The lighter welding filters are rated better than the solar filters, given the same luminous transmittance. This may be due to the fact that they are green and reduce the blue transmittance more than luminous transmittance. The bluer wavelengths have been identified as being particularly involved in the causation of discomfort glare, or photophobia (Kardon 2012). Lowering the maximum allowable luminous transmittance to 0.0012% (the dividing line between shades 12 and 13) would screen out such products.
These proposals for amendment of the maximum and minimum luminous transmittances are illustrated in Figure 2.
The minimum dimensions specified in Clause 4.3.2 of the ISO 12312-2 standard were based on spectacle-shaped solar eclipse viewers (eclipse glasses) made with cardboard mountings (frames and sides). Such viewers are now available in different sizes for adults and children. Solar filters are also available in the form of handheld rectangular cardboard viewers that cover both eyes, as clip-on viewers meant to be attached to the brim of a baseball cap, and in plastic sunglass frames. The latter design fails the specified dimensions of the cutaway area that accommodates the bridge of the nose. However, this more robust viewer fits closer to the eyes than a cardboard viewer and sits more securely on the face. The ISO standard should be revised to allow for the new more durable spectacle type of viewer ( Figure 3) that typically has a narrower nasal gap but covers the eyes better than the cardboard viewer, and it should allow for other widely available formats. Several samples failed the requirements for labeling (Table 1). This may have been due to the samples being manufactured