The photobiological safety of optical radiation emitted by image projectors, particularly of laser illuminated projectors (LIP), is addressed by the recently published product safety standard IEC 62471-5 [*Photobiological Safety of Lamps and Lamp Systems—Part 5: Image Projectors* (IEC, 2015)]. According to IEC 62471-5, the accessible emission is determined at a distance of 1 m from the projection lens. A classification framework is used to categorize projectors into risk groups (RG), indicating the degree of risk from potential optical radiation hazards to the eye and skin, ranging from the exempt risk group (RG0) to risk group 3 (RG3). According to IEC 62471-5, the highest classification permitted for consumer products is RG2. In this paper, a risk analysis for exposure to the emission of LIP classified as RG2, at distances less than 1 m is provided. The analysis shows that the risk for retinal injury associated with RG2 LIP or conventional projectors at distances less than 1 m can be considered as very low to negligible.

## I. INTRODUCTION

### A. Photobiological safety standard

Driven by innovations in the field of solid state light sources, laser illuminated projectors (LIP) represent a trend for increased optical output. Consequently, the risk for ocular injury needs to be characterized and the emission for consumer products limited to acceptable levels. Since LIP have image formation and optical systems that are equivalent to projectors with conventional light sources, the international laser product safety standard^{1} IEC 60825-1 was amended to the effect that the emission of the LIP can be classified under the lamp safety standard series. The International Electrotechnical Commission, IEC, developed a safety standard specifically for image projectors,^{2} which was published in June 2015 as IEC 62471-5 in the standard series “Photobiological Safety of Lamps and Lamp Systems.” The range of projectors includes small data projectors, projectors for home-cinema, and high power cinema projectors both lamp and laser driven. The scope of the standard excludes scanning beam projectors, but includes LIP that fulfill the requirements specified in subclause 4.4 of^{2} IEC 60825-1:2014. In IEC 62471-5, the basic scheme for risk groups (RG) as defined in^{3} IEC 62471:2006 was adopted, but while the risk group classification in IEC 62471:2006 is defined to apply to the bare lamp (such as the xenon arc lamp), IEC 62471-5 applies to the complete projector. A classification framework is defined to categorize projectors into RG, indicating the degree of risk from potential optical radiation hazards, ranging from exempt (RG0) to risk group 3 (RG3).

As a general concept, risk group classification of a product is based on determining the emission level (also referred to as the “accessible emission,” AE) at a standardized reference distance. The emission level is then compared to the “accessible emission limit” (AEL) for the different hazards and risk groups (RG0, RG1, RG2). The IEC determined that the two reference distances used in IEC 62471 for lamps (the 500 lux distance for lighting application and 20 cm in all other instances) are not appropriate for image projectors, as the 500 lux distance is too far away and the 20 cm distance is too close to appropriately reflect a realistic exposure scenario. For projectors in the scope of IEC 62471-5, a reference distance of 1 m appeared as the most appropriate to determine the risk group.

IEC 62471-5 gives consideration to the photobiological safety of image projectors with respect to five hazards to the skin and eye and defines AEL for each: two ultraviolet hazards, two retinal hazards, and one infrared hazard. For each hazard-AEL, the difference between the risk groups is the “time base”—the defined duration for which the underlying time dependent limit is not exceeded. The lower the risk group, the longer the time base. For RG0, the underlying time dependent limit is not exceeded even for the extreme case of intentional long term (the whole day) exposure at the reference distance. RG2 is considered safe due to normal aversion response reflexes to bright light and heat. RG1 has time base values intermediate between RG0 and RG2. For RG3, within the hazard area, the underlying limit is exceeded for exposure durations shorter than as assumed for aversion responses to take place.

The critical emission limit for image projectors is the retinal thermal limit, i.e., the limit to protect against thermally induced retinal injury,^{4} where the assumed accidental exposure duration (and therefore time base for classification) is 0.25 s, based on aversion responses to bright light.^{5} For the remainder of the paper, when there is reference to a limit, the retinal thermal limit is meant. For longer exposure durations, photochemical retinal injury is the dominating injury mechanism (refer to the Appendix for an analysis based on the photochemical exposure limits (EL)) and is known to occur from deliberate staring into the sun at noon or staring into welding arcs for a considerable amount of time.^{6}

While RG3 projectors are intended to be sold only for professional applications where the proper installation (for instance, with adequate beam height above the audience) ensures safety, RG2 projectors are also appropriate for consumer products. The worst case (maximum) emission level for an RG2 projector under IEC 62471-5 is achieved when the accessible emission is equal to the AEL for RG2 at a distance of 1 m from the projector. The retinal thermal AEL for RG2 is numerically equal to the retinal thermal EL for 0.25 s exposure duration (as promulgated^{7} by ICNIRP). This means that for projectors where at 1 m distance the emission is equal to the AEL for RG2, the EL for the retinal thermal hazard for accidental exposure will be exceeded for exposure distances less than 1 m. A validation of the requirements of IEC 62471-5 has to demonstrate that exposure at distances less than 1 m represents an acceptably low level of risk. While the technology that has driven the development of IEC 62471-5 was LIP, the following discussion applies also to projectors with conventional light sources, provided that they can emit levels of radiation that approach the RG2 limit. Since the optical system and the image formation of LIP are equivalent to those of projectors with conventional light sources, in the following we refer to projectors in a general way.

### B. Retinal thermal hazard

The time base of the retinal thermal hazard for RG2 is 0.25 s, as also used for laser products for Class 2 classification according to Center for Devices and Radiological Health (CDRH as part of the US Food and Drug Administration (FDA)) and IEC product safety standards.^{2,8} The duration of 0.25 s is based on aversion responses to bright light and is also the internationally used “default” exposure duration for unintentional exposure to bright light. Longer exposure durations can occur when a person overcomes the natural aversion response and intentionally stares into the bright light. Retinal thermal injury is extremely rare for nonlaser sources, as the “brightness” (more accurately “radiance”) has to be so high that detrimental temperature increases are induced in the retina before the aversion response sets in. Besides collimated laser beams, retinal thermal injury is known only from nuclear blasts, intentional exposure to xenon arc lamps (used for retinal surgery before the advent of the laser^{5}) and looking into the sun with telescopes. The retinal thermal exposure limit for an exposure duration of 0.25 s is based on a 7 mm eye pupil diameter. This is overly restrictive for most exposure scenarios, as is specifically noted in the 2013 ICNIRP guidelines.^{7} The assumed pupil size of 7 mm and the correspondingly low exposure/emission limit when expressed as radiance, for instance, result in the sun at moderate and high elevation angles to exceed the retinal thermal limit for 0.25 s by a factor^{9} close to two. The sun, when classified at the reference distance of the earth, would be classified RG3 based on the retinal thermal limit, although in reality no cases of thermally induced retinal injury from short-time sun exposure have been reported in the literature. Solar retinopathy is known only for prolonged staring into the sun such as during eclipses or from sun-gazers.^{10} Besides the assumed pupil diameter of the eye, there is a significant safety margin between the injury threshold and the emission limit, and exceeding the emission limit (or the respective exposure limit) by some factor does not necessarily represent a real risk for injury. This is well known from laser products, where the exposure limit for 0.25 s for collimated laser beams in the visible wavelength range equals 1 mW. However, experience shows that power levels of 5 mW and more have not needed negligible risk for thermal injury.^{11} This power range of up to 5 mW was also given a dedicated laser classification group, Class 3 R under IEC 60825-1 and Class IIIa under CDRH regulations.^{8} Thus exceeding the exposure limit by some factor (factor 5 for laser products) is not “new” in the field of radiation safety, and Class IIIa lasers in the USA are also acceptable to be the marketed as consumer products.

Many years of field experience have shown that lower to medium power projectors of a few thousand lumen light output—which have been available as consumer products for many years, and where exposure at close distance (<1 m) surely has occurred—did not produce any known detrimental effect on the retina.^{4} Even the highest power cinema projectors in use for several decades did not produce any known injury from momentary viewing, but would be hazardous for intentional staring at close distance.^{4} High power projectors are not used in home or other uncontrolled environment due to cost and minimum screen size requirements to avoid uncomfortably bright images. Cinema projectors, or image projectors in general, as far as the authors are aware, are accredited only one documented^{12} case of retinal injury published in 1936 when a Russian projector operator during adjustment stared into a carbon-arc lamp where the filter glass in the projector was broken.

RG2 devices are low to moderately high power projectors (compared to professional projectors for large venues), and IEC 62471-5 defines RG2 projectors to be fit to be placed on the market as consumer devices. For a risk analysis, it is important that projectors with emission levels at the RG2 limit are very expensive high-end systems with luminous power levels of at least 7000 lm, that would only rarely be purchased for home-cinema use (small screen cinema projectors have luminous power levels of 3000 lm).

The risk analysis detailed in this paper is based on two main parameters:

the diameter of the pupil of the eye, and

the safety margin inherent in the AEL applicable to image projectors, where the wavelength distribution and apparent source size are well defined for relevant types of projectors.

In order to have a robust risk analysis, conservative to worst-case scenarios are chosen for all relevant parameters.

## II. EXPOSURE LIMIT ANALYSIS

### A. Retinal thermal EL

As mentioned above, this risk analysis is based on the worst-case emission level for RG2 where at a distance of 1 m the AE is just below (in effect equal to) the AEL of RG2. The AEL of RG2 in^{1} IEC 62471-5 is equal to the “retinal thermal” exposure limit (EL) from the^{7} ICNIRP 2013 guidelines for 0.25 s exposure duration and equals

where α is in units of rad and is the angular subtense of the apparent source with a maximum value of 0.1 rad for continuous emission.

The exposure level that is compared against the EL is defined as the radiance of the projector over the spectral range 380–1400 nm, weighted with *R*(*λ*) and averaged over a field of view of 11 mrad (continuous emission) or 5 mrad (pulsed emission). Since the wavelength weighting function *R*(*λ*) is equal to 1 in the relevant visible wavelength range, the unweighted radiance for the maximum emission (white image) can be used for the risk analysis.

The apparent source for image projectors (producing the most restrictive combination of value of *α* and radiance) is the exit pupil of the projection lens system.^{1,4} The exit pupil can also be approximated as the virtual beam waist and the origin of the beam when the borders of the emitted beam are extrapolated back into the projector (see Fig. 1). The horizontal opening angle of the beam is characterized for projectors by the throw ratio *TR*, defined as the ratio of distance to the projector exit pupil over the width of the beam at that distance *W*. The divergence of the beam in horizontal direction, measured in radian, is the inverse of the *TR*.

As schematically shown in Fig. 1, the exit pupil is positioned at a specific distance Δ*L* within the outer surface of the outer projection lens, *L* is the distance to the projection lens, and *D*_{EP} is the diameter of the exit pupil. Then for a top-hat profile of the exit pupil, the angular subtense of the apparent source *α* is equal to

Radiance, as exposure level, can be calculated by dividing the total beam power by the area of the beam at distance *L* from the lens (equals beam irradiance) and further division by the solid angle that is subtended by the exit pupil as seen from distance *L* + Δ*L*. This quantity, as is generally the case for radiance,^{9} does not depend on distance. What does depend on distance is the angular subtense of the apparent source and therefore the EL, since the EL depends inversely on *α*; the angular subtense of the apparent source increases by the same factor that the distance *L* + Δ*L* decreases. For example, if at the distance of *L* + Δ*L* the exposure is equal to the EL and that distance is halved, the angular subtense of the apparent source is doubled so that the EL is exceeded by a factor of 2. The dependence of the factor that the EL is exceeded as a function of the distance to the projector is “dampened” by the distance Δ*L* (halving *L* does not decrease the EL by a factor of 2). For a conservative analysis, a rather small value of Δ*L* = 0.1 m is used. The type of projector under discussion, which approaches the RG2 limit at 1 m distance, is in the higher output range of at least 7000 lm but typically exceeding 10 000 lm. These projection systems will require a correspondingly large projection lens and an exit pupil location significantly within the lens system. Considering Δ*L* = 0.1 m, the EL is exceeded by a factor of 1.9 for *L* = 0.5 m and by a factor of 5.5 for *L* = 0.1 m from the outer surface of the lens. These values do not depend on the angular subtense of the source *α* as long it is not larger than 0.1 rad. For a distance of *L + ΔL* = 0.2, this independency is satisfied as long as the exit pupil diameter is smaller than 20 mm, which for the projection optics is usually the case, as shown below. For exit pupils larger than 0.1 rad, the EL no longer decreases with decreasing distance, which is less restrictive.

### B. Accounting for eye pupil size

While the ambient illumination and other factors such as age “determine” the diameter of the pupil of the eye, studies show that for the same exposure condition individual variability of the pupil diameter is considerable. The variability is in the range of a factor of up to two spread for a given lighting scenario (see references cited by Slamovits *et al*.^{13}), where not all factors that influence the pupil diameter were necessarily controlled. The different factors that influence the pupil diameter for a given target luminance were reviewed by Watson and Yellott^{14} who derived a unified formula for light-adapted pupil size. The plots Watson and Yellott redrew from data by Winn *et al*.^{15} demonstrate the relatively large variability of pupil diameters, but also show the trend for reduced pupil diameter as a function of luminance; the target subtended 10°. The maximum luminance used was 4400 cd m^{−2} where the maximum pupil diameter was about 5.8 mm and the average about 4 mm for the youngest eyes. We note that for looking into a 11 000 lm projector set to project a black image, the luminance is at least ten times 4400 cd m^{−2}, even with an excellent contrast ratio of 10 000:1 (even when a “black” image is projected, the light source in the projector is on and due to stray light there is a significant amount of light emitted by the projector, see details on this calculation below). In the unified formula developed by Watson and Yellott^{14} the target luminance, the angular subtense of the target, age and the number of eyes exposed (only one or both eyes) were considered. A parameter variation for the unified formula reveals that an age of 20 years is associated with the largest average pupil diameters, and the pupil diameter decreases with increasing angular subtense of the target. The distance to the target (and therefore the accommodation state of the eye), however, was not specifically considered; likely, the studies that form the basis of the unified formula used a target at some distance.

The pupil size, however, does not only depend (with some variability) on light levels, age, and target size as considered in the unified formula, but also on the eye's accommodation condition. To image a close-by object, not only the lens power changes to optically image the target (accommodation) and the axis of the eyes center on the target (convergence) but also the pupils constrict (miosis), referred to as the “near triad of accommodation”^{16} (Fig. 2).

The near triad of accommodation is a reflex, controlled by the Edinger–Westphal nucleus in the midbrain.^{17} The response has some individual variability, and the literature^{18–22} is not fully consistent.

For a projector showing a fully white image, the exit pupil can be considered as the relevant target, being the location within the (virtual) beam that is associated with the maximum radiance. If the eye accommodates to a point further behind the exit pupil, then the near triad of accommodation would not be induced, or less pronounced. This means that the pupil might be larger (but see the results of unified formula, in which close accommodation is not considered), but then the retinal image of the exit pupil would be blurred representing a lower retinal irradiance compared to accommodation to the exit pupil (for the same pupil diameter of the eye).

The EL as given in Eq. (1) is based on the assumption that the pupil of the eye has a diameter of 7 mm. The critical parameter for retinal hazards is the retinal irradiance, which for a given corneal irradiance or radiance is directly proportional to pupil area.^{5,9} The ICNIRP 2013 guidelines^{7} note that the EL being based on a 7 mm pupil is conservative for a normal, reactive pupil, and can be increased when smaller pupil diameters are assured. The factor to adjust the EL is the ratio of the areas of a 7 mm pupil to the area of the actual pupil *D*_{P}, and therefore 7^{2}/*D*_{P}^{2}. Thus for a risk analysis, the diameter of the eye's pupil for the case that exposure occurs at close distance to the projector is the central factor. For instance, if the pupil diameter for an actual exposure is 3.5 mm rather than 7 mm, only 25% of the light enters the eye. When the 7 mm-EL is exceeded by a factor of 3.7 at 20 cm distance from the projector lens, the EL adjusted for a pupil diameter of 3.5 mm (factor 4 in area) is not exceeded. In Table I, the eye pupil diameter that results in an exposure equal to the adjusted EL is shown for a projector with Δ*L* = 0.1 m; note that this analysis does not depend on the exit pupil diameter of the projector.

Distance to lens (m) . | Distance to exit pupil of projector (m) . | Factor exceeding EL with 7 mm eye pupil . | Eye's pupil diameter to equal EL (mm) . | Predicted average pupil diameter, neglecting near triad of accommodation^{a} (mm)
. |
---|---|---|---|---|

0.1 | 0.2 | 5.5 | 3.0 | 2.7 |

0.2 | 0.3 | 3.7 | 3.7 | 3.0 |

0.3 | 0.4 | 2.8 | 4.2 | 3.2 |

0.4 | 0.5 | 2.2 | 4.7 | 3.4 |

0.5 | 0.6 | 1.8 | 5.2 | 3.6 |

0.6 | 0.7 | 1.6 | 5.6 | 3.7 |

0.7 | 0.8 | 1.4 | 6.0 | 3.8 |

0.8 | 0.9 | 1.2 | 6.3 | 3.9 |

0.9 | 1.0 | 1.1 | 6.7 | — |

1.0 | 1.1 | 1.0 | 7.0 | — |

Distance to lens (m) . | Distance to exit pupil of projector (m) . | Factor exceeding EL with 7 mm eye pupil . | Eye's pupil diameter to equal EL (mm) . | Predicted average pupil diameter, neglecting near triad of accommodation^{a} (mm)
. |
---|---|---|---|---|

0.1 | 0.2 | 5.5 | 3.0 | 2.7 |

0.2 | 0.3 | 3.7 | 3.7 | 3.0 |

0.3 | 0.4 | 2.8 | 4.2 | 3.2 |

0.4 | 0.5 | 2.2 | 4.7 | 3.4 |

0.5 | 0.6 | 1.8 | 5.2 | 3.6 |

0.6 | 0.7 | 1.6 | 5.6 | 3.7 |

0.7 | 0.8 | 1.4 | 6.0 | 3.8 |

0.8 | 0.9 | 1.2 | 6.3 | 3.9 |

0.9 | 1.0 | 1.1 | 6.7 | — |

1.0 | 1.1 | 1.0 | 7.0 | — |

^{a}

From the NASA unified formula;^{14} minimum target angle in the NASA unified formula is 1°, therefore no predictions for distances further than 0.8 m.

Table I in the last column also shows the predicted average pupil diameters from the unified formula with the following input parameters: an exit pupil diameter of the projector of 15 mm (representing the target, see discussion on minimum exit pupil diameters in Sec. III B) located 100 mm behind the projector lens surface; an age of 20 years resulting in the largest pupil diameters in terms of age dependence; a target luminance of 40 000 cd m^{−2}, the minimum luminance for projecting a “black image” (see derivation of this value in Sec. II C). We see that the predicted pupil diameter becomes smaller with smaller distance due to the larger angular subtense of the target. The assumed projector's exit pupil diameter of 15 mm is a conservative choice for projectors that can achieve RG2 limits at 1 m distance (see discussion in Sec. III B). The dependence on target angular subtense, is, however, not great. For a 6 mm exit pupil diameter, the predicted average eye's pupil diameter for a distance of 0.2 m from the lens equals 3.7 mm compared to 3.0 mm for a 15 mm exit pupil. The NASA unified formula only accounts for target subtended angle and does not account for the accommodation distance and near triad of accommodation which would also result in smaller pupil diameters for close distances.

### C. Conclusions for EL analysis

For a risk analysis, one needs to consider what scenarios can lead to an exposure at close distance (<1 m from the projection lens). The worst case scenario should be: a dark room or very dimly lit; the projector is switched off or is set to a black image; a person happens to be in the path of the beam and looks into the projector lens from a distance less than 1 m at the moment when the projector is switched on or to emit a “white” image; and the human eye is focused on the exit pupil of the projection lens. This situation appears as highly unlikely considering that these higher power projectors are usually either mounted on the ceiling or in a projection booth, but it cannot be excluded. IEC 62471-5 for RG2 and RG3 projectors requires (subclause 6.3) a “soft-start,” i.e., the full emission level is only permitted to be reached 1 s after the projector light source has been switched on. Due to the soft-start, there is sufficient time for aversion responses to bright light, i.e., reduction of the pupil diameter to take effect (see Fig. 4 in ICNIRP guidelines^{7}) or for a person to look away so that the above scenario is not relevant for the case of the projector being switched on.

The only scenario where an instantaneous emission can occur is the projector being on but a black image is projected, which is then switched to white at the time a person is at close distance and looks into the projector. As noted before, a projected black image does not mean there is no light coming out of the projection lens. Due to the limited contrast ratio, there is a significant amount of light leakage, which for a 11 000 lm projector (see Fig. 3) is quite bright so that for the case of intrabeam exposure, the pupil would constrict within a short period of time. For example, for a high (and therefore conservative) contrast ratio of 10 000:1, a 11 000 lm projector with an exit pupil diameter of 15 mm and a throw ratio of 2, for a black image has a luminance of about 40 000 cd m^{−2}. This is roughly ten times the luminance of the clear blue sky and four times the luminance of a fluorescent lamp, resulting in a correspondingly smaller pupil when looking into the light source.

Even for a position outside of the beam (Figs. 3(b) and 3(c)), there is stray light visible that exits the projection lens. This stray light also makes a person aware of the projector and the beam, as well as the light results in some pupil constriction when one approaches the beam. The amount of stray light increases with increasing optical output of the device.

Given the above analysis, it is possible to conclude for normally reacting pupils that the eye pupil at half a meter distance from the projector is not larger than 5 mm (the unified formula, see Table I, predicts an average of 3.6 mm for a 15 mm exit pupil and 4 mm for a 6 mm exit pupil) and at 20 cm from the projector is not larger than 3 mm (the unified formula predicts an average of 3 mm, but neglects the near triad of accommodation). Compared with the fourth column in Table I, listing pupil values that produce an exposure equal to the adjusted EL, it can be concluded that for normally reacting pupils (i.e., not medically dilated pupils), the adjusted EL is not exceeded for a projector where the exposure is equal to the 7 mm-EL at 1 m distance (maximum emission for RG2). For a complete and conservative risk analysis, the case that the pupils are larger than the above values need also be considered, which is discussed in the second part of the risk analysis, where the safety margin between the injury threshold and the EL is taken into account.

## III. INJURY THRESHOLDS ANALYSIS

Although the predicted pupil diameters for a black projected image are smaller than those necessary to remain below the exposure limit (Table I), the risk analysis can be further strengthened by considering injury thresholds.

### A. Prediction of injury threshold

The Seibersdorf Laboratories computer model was used to predict retinal thermal injury thresholds for the specific exposure parameters applicable to projectors. The model, described in detail by Jean and Schulmeister,^{23,24} is validated against all applicable experimental injury threshold data obtained with rhesus monkeys. The computer model was adjusted to conservatively predict injury thresholds for humans by assuming a smaller minimum lesion size but still being based on the absorption properties and the injury thresholds of the rhesus monkey model. Stuck^{25} noted that the injury threshold for humans, where available, was consistently higher than for the nonhuman primates; a factor of 2 for regions outside of the macula and a factor of at least 1.3 for the macular region for the case of a broadband white light source with a retinal image diameter of 1 mm. However, direct comparisons (for comparable exposure conditions and endpoints) of thresholds from human volunteers with nonhuman primate experiments are very scarce. The model is optimized to predict experimental ED_{50} levels for a given wavelength, irradiance profile, and exposure duration. The ED_{50} level is the total intraocular energy required to induce a minimum visible lesion to the retina with a probability^{26} of 50%. The model was optimized so that it predicts the mean of the experimental ED_{50} data for a given parameter set of exposure duration, spot size, and wavelength. This means that there are experimental data that have higher as well as lower reported ED_{50} than the predictions. When taking the set of threshold data applicable for the macula (where the thresholds are lowest due to highest pigmentation) and exposure durations less than 10 s, the maximum factor of the computer model being higher than the nonhuman primate data is 1.3. This value of 1.3 has to be applied as a factor to scale the model prediction to the lower edge of experimentally found ED_{50} values and is referred to here as “uncertainty reduction factor.” Additionally, a reduction factor needs to be applied to scale the ED_{50} (which is in the center of the dose-response curve, where 50% of the exposures resulted in an effect, which means that due to individual variability also lower exposure levels can produce an effect, see Sliney *et al*.^{26}) to a value where the probability of injury is in a range which can be considered to be “negligible”; this additional reduction factor is referred to as the “risk reduction factor”. The reduction of the predicted ED_{50} to account for uncertainty in the experimental ED_{50} data is, however, correlated with the necessary risk reduction factor: the more conservative the ED_{50} is reduced to account for the uncertainty in the experimental data (i.e., the larger the uncertainty reduction factor is), the steeper the dose-response curve can be assumed and therefore the smaller the risk reduction factor can be.^{26}

As an extreme case, when the uncertainty reduction results in a value equal to the lowest threshold that is found in a given population, then there is no further risk reduction necessary. A steeper dose-response curve means that the need for further reduction to transform the 50% probability into a negligible probability is not as strong as if the dose-response curve were of a shallower type. A dose-response curve slope of *S* = 1.1 and less (closer to 1.0, *S* being defined as ratio of the exposure level resulting in 84% probability for observing an injury over the ED_{50}, see, for instance, Sliney *et al*.^{26}) is regularly reported for higher quality experimental studies. When applying an uncertainty reduction factor of 1.3 to the model predictions, then the assumption of a slope of *S = *1.1 for the dose-response curve appears justified. A risk reduction factor of 1.5 being applied to the ED_{50} of a dose-response curve with slope *S = *1.1 results in a probability value for a log-normal distribution of 10^{−5}. Overall these two reduction factors result in a factor of 1.3 × 1.5 = 1.95, which can be rounded to 2.0. If a larger uncertainty reduction factor of, for instance, 1.5 were applied (i.e., to a level below the observed experimental data range), then this would be an even stronger argument for a very steep (close to a step-function) dose-response curve, of, for instance, a slope of *S = *1.05. Applying a risk reduction factor of 1.3 to the ED_{50} of a dose-response curve with *S = *1.05 results in a probability value for a log-normal distribution of 4 × 10^{−8}. Overall these two reduction factors again result in a factor of 1.95, but in this treatment the ED_{50} was reduced more strongly (uncertainty reduction) in turn supporting the assumption of a steeper dose-response curve and to apply a somewhat smaller risk reduction factor. The overall effect is in both cases a negligible probability for injury, with the same overall reduction factor of 1.95 which is rounded here to a factor of 2.0. For the interpretation of the above probability values, it is relevant that, as discussed by Sliney *et al*.,^{26} the dose-response curve based on a log-normal distribution is not applicable for very small exposure levels and over-predicts the probability for observing an injury. Therefore, for reduction below the lowest threshold expected for a given population, the risk can be considered as negligible; as anecdotal evidence, the risk for becoming blind from a fever that results in a certain temperature increase in the retina is also negligible. An overall reduction factor of 2 can even be considered to be somewhat conservative; an uncertainty reduction factor of 1.3 together with a risk reduction factor of 1.3, resulting in an overall reduction factor of 1.7 should also result in a level that is associated with a very small risk. This also fits with the observed minimum factor between the unreduced model predictions and the EL for top-hat retinal irradiance profiles which for 530 nm (the wavelength with the lowest injury thresholds), 5 ms exposure duration and 3 mrad angular subtense equals a factor of only 1.5 (i.e., less than the reduction factor applied here).

An overall reduction factor of 2 was derived as a reliable basis to classify the resulting level as “safe,” i.e., negligible risk for injury (for *α* of about 10 mrad, the reduced model data is almost equal to the EL, providing support of the conservative nature of the reduction). This level (the mean ED_{50} reduced by a factor 2) is here also referred to as the “Highest Safe Level” (HSE) and can be understood as level where the risk for retinal injury is considered negligible, because the actual injury threshold is with good reliability higher than the HSE. Thereby, an uncertainty and variability analysis that is otherwise necessary for a quantitative risk analysis is accounted for in a simplified way by a conservative choice of reduction factors. It needs to be kept in mind that the model data were optimized for the thresholds for nonhuman primates. All available comparative studies with human volunteers exhibited higher injury thresholds even for highly pigmented humans.^{25} The EL and HSE for 530 nm and 0.25 s exposure duration are shown in Fig. 4 for a disk shaped (top-hat profile) retinal image as a function of angular subtense of the image as seen from the pupil of the eye, which is equivalent to the angular subtense of the source.

The injury thresholds were also calculated for typical wavelength distributions for a white image of projectors with a xenon lamp, a laser pumped phosphor projector, and red-green-blue laser illuminated projectors. As expected,^{11} all of these spectra had a higher predicted injury threshold as compared to 530 nm monochromatic emission. As a worst case smallest difference found for one of the laser illuminated projectors, the conservative value of 1.1 ratio between the threshold for the projector spectrum and the threshold for 530 nm is used in the following. The ratio of HSE over EL for the worst case wavelength distribution to produce white (data of Fig. 4 corrected for wavelength distribution) is plotted in Fig. 5.

The variation of the ratio of HSE over EL with source size is considerable: it is minimum (close to 1) at 10 mrad (noting again that the ratio is not based on the predicted injury threshold but on a reduced level which can be associated with negligible risk for injury), but increases up to 2.8 for 100 mrad. Thus, the margin between the HSE and the EL increases for closer distances to the projector. This is relevant, as also the degree by which the 7 mm EL is exceeded increases with closer distances, so that the larger “safety margin” for closer distances has a compensating effect (additionally to smaller pupils).

### B. Consideration of projector parameters

While the conclusions of Sec. II, being based on EL adjusted for pupil size, do not depend on the exit pupil diameter, for the injury threshold analysis, projector optics properties become relevant. As shown in Sec. III A, the safety margin is largest for large retinal images. A conservative risk analysis therefore needs to be based on the smallest exit pupil that is associated with projectors which approach the AEL of RG2 at 1 m distance. Restricting this discussion to the worst-case level of emission (where the radiance at 1 m distance is equal to the AEL of RG2) at the same time results in a certain minimum size of the imager chip and optics (and therefore exit pupil diameter) to be able to “handle” this level of power. As a first step the luminous flux that is permitted for RG2 is determined. The annex of IEC 62471-5 gives examples of how to perform such an analysis for a given exit pupil diameter.

In order to base the discussion on more generic projector properties such as imager size and f-number, we derive the following relationship. The diameter of the projection lens' exit pupil is directly related to the size of the imager chip, which is the part of the projector which in terms of pixels is imaged onto the screen and is either^{27,28} a digital micromirror device (DMD), a liquid crystal on silicon (based on reflection of light), or a liquid-crystal display (based on transmission of light). The relationship can be derived as follows: the projection lens with focal length *f* images the chip onto the screen. Further, the location of the chip is approximated to be located in or very close to the focal plane of the lens, which is justified as the screen is at least several meters away. It follows that the ratio of the focal length *f* to the width of the chip *w* is the same as the throw ratio *TR* (the ratio of distance to the screen over the width of the projected image). Therefore, *f* = *TR* · *w*. The *f*-number is the main parameter characterizing the projection lens and is equal to the ratio of *f* to the exit pupil diameter *D*_{EP}. Thus, the exit pupil diameter is equal to *D*_{EP} = *TR* · *w* · *f*-number^{−1}. As a general relationship,^{27} the *f*-number is larger for larger chips, and typical values are *f*/1.6 for a 0.45 in. chip, *f*/2 for a 0.67 in. chip, and an *f*/2.5 for a 0.98 in. chip. The dependence of the exit pupil diameter is shown in Fig. 6 for two chip sizes, as a function of the throw ratio *TR*.

For larger imager chip sizes, (for a given throw ratio and *f*-number) the exit pupil is larger. For a conservative risk analysis, the smallest applicable exit pupil is relevant, and therefore also the smallest imager chip that can be utilized in a projector that can reach the RG2 AEL at a distance of 1 m.

The critical and limiting factor regarding minimum chip size is the maximum permitted temperature of the imager, for example, the DMD chip,^{29} which due to cooling reasons limits the maximum optical power. Some of the optical power that is incident on the imager chip is absorbed; for DMD devices mainly due to the nonperfect reflectance of the aluminum mirror material^{27} of about 88% as well as because of the gaps between the micromirrors which in relation to the mirror size is larger for the smaller chips, (smaller mirrors). Since the imager chips are cooled from the back, the maximum optical power that is permitted to be incident on the chip is directly proportional to the area of the chip, resulting in a permitted irradiance on the chip which is approximately constant for the different chip-sizes. Correspondingly, the thermal resistance of the chip, defined between front and back and relevant for not exceeding the maximum permitted temperature at the front of the imager, is smaller for larger chips. For the 0.45 in. chip, 2.0 °C/W is given^{30,31} and for the 0.67 in. chip, 0.7 °C/W based on data from the manufacturer, respectively. When the thermal resistance for these two chips is multiplied with the imager area, a constant value is found. Due to the limitations of cooling, including the noise of the cooling, and limitations on the light through-put (étendue), the maximum luminous flux for a single 0.67 in. imager chip design is approximately 7000 lm for a white image which is typically associated with a luminous efficacy of about 300 lm per optical (unweighted) Watt. With higher noise levels or improved cooling technology, up to 10 000 lm can be reached, but this can be considered as an upper limit for a single 0.67 in. chip. For an imager chip size of 0.45 in., the maximum permissible luminous flux can be assumed to be not more than 5000 lm, even with some future improved cooling technology. Higher luminous flux levels are achievable only with three chip designs (one imager chip for each color), and/or with larger imager chips. Due to the high cost^{32} of the imaging chip, three-chip designs are employed only in professional projectors and typically the imager chip size is a minimum of 0.96 in. Products with three chips and 0.67 in. are available as small movie projectors but are rare and expensive.

The above technological limits on maximum optical output can be compared against the maximum emission values that are calculated for an RG2 projector for a given exit pupil diameter and luminous efficacy of 300 lm/W. The typical *f*-number for a 0.45 in. imager is *f*/1.6 resulting in an exit pupil diameter of 12 mm and about 13 000 lm being permissible for RG2 (*TR* = 2), which is well above the technological limit. For a more restrictive choice of parameters (producing smaller permitted luminous flux values for RG2), a *TR* = 3 and *f*/1.6 results in 8700 lm permitted luminous flux; additionally choosing a conservative *f*-number *f*/2 results in 7000 lm permitted luminous flux. Consequently, the luminous flux levels necessary to approach the RG2 AELs at 1 m for nonprofessional products are significantly higher than the levels that are technically achievable for a 0.45 in. chip device, i.e., for a 0.45 in. chip size, the RG2 limits cannot be reached due to technical limitations. In terms of risk, these are consequently less critical than devices that reach up to the RG2 limit (a lower exposure level produces a correspondingly smaller temperature and thermal injury is extremely nonlinear with temperature^{33}).

From the above, we can conclude that the smallest imager size to be considered for this analysis is 0.67 in. Even for this chip-size, the calculated maximum emission level permissible for RG2 is, for the typical case of *TR* = 2 and *f*/2 (being associated with an exit pupil diameter of 14.4 mm), with 16 000 lm significantly higher than what is possible technologically for a single chip device. For a 0.67 in. chip, only larger throw ratios together with larger *f*-numbers result in low-enough RG2 emission limits (such as 8700 lm for *TR* = 3 and *f*/2.5, producing an exit pupil of 17.3 mm) that are technologically achievable. For these projector parameters, however, the exit pupil is larger, (less restrictive for the risk analysis) as compared to the more typical values. Therefore, a projector that can reach RG2 limits will either be a three chip 0.67 in. device, or a device with a larger imager chip, in either case it will certainly be too expensive to be a consumer product. The high luminance levels necessary to achieve RG2 limits also dictate a correspondingly large screen to achieve comfortable brightness, so that projectors in this regime are cinema projectors, mounted in a projection booth (also because of the noise), and not consumer products where exposure at less than 1 m is of concern. Basing the risk analysis for exposure at distances less than 1 m on a 0.67 in. chip device that could be a consumer product with emission levels equal to the RG2 AEL at 1 m is currently not realistic from a technological standpoint of view, but it is a prudent worst-case assumption in order to have a solid risk analysis with a “buffer” for future developments.

Having decided on a conservative chip size of 0.67 in., the smallest applicable exit pupil needs to be defined for risk assessment purposes. First, a conservatively large *f*-number of *f*/2.5 is assumed (the typical value is *f*/2). For a *TR* = 2, the diameter of the exit pupil equals 12 mm and the associated luminous flux to approach RG2 equals 13 000 lm. Although a smaller *TR* results in a smaller exit pupil, this would be associated with RG2 luminous flux values which would well exceed the maximum thermal load for a 0.67 in. chip, as well as be unrealistic as consumer product due to required minimum screen size. Therefore, the risk analysis is based on the conservative assumption of an exit pupil diameter of 12 mm. A more typical device that could approach RG2 limits would be a device with a 0.96 in. chip-set (usually three chips) which for *TR* = 2 and *f*/2.5 is associated with an exit pupil diameter of 17 mm, being larger and less conservative for the risk analysis.

### C. Risk analysis including injury thresholds

In Sec. III B, a projector's exit pupil diameter of 12 mm was identified as the smallest applicable diameter, even considering possible future developments. This means that at 1 m distance, and the conservative assumption of the exit pupil being located 100 mm behind the outer lens surface, the angular subtense of the apparent source *α* equals 11 mrad. With closer distances, the angular subtense of the apparent source increases, being associated with an increasing ratio of the HSE to the EL. With this ratio, Table II lists pupil diameters which are associated with a negligible risk level.

Distance to lens (m) . | α (mrad)
. | Min. safety margin . | “Negligible risk” eye pupil diam. (mm) . | For 8 mm pupil: factor above HSE . |
---|---|---|---|---|

0.1 | 60 | 1.89 | 4.1 | 3.8 |

0.2 | 40 | 1.46 | 4.4 | 3.3 |

0.3 | 30 | 1.30 | 4.8 | 2.8 |

0.4 | 24 | 1.21 | 5.2 | 2.4 |

0.5 | 20 | 1.16 | 5.6 | 2.1 |

0.6 | 17 | 1.14 | 5.9 | 1.8 |

0.7 | 15 | 1.13 | 6.3 | 1.6 |

0.8 | 13 | 1.12 | 6.7 | 1.4 |

0.9 | 12 | 1.11 | 7.0 | 1.3 |

1.0 | 11 | 1.11 | 7.4 | 1.2 |

Distance to lens (m) . | α (mrad)
. | Min. safety margin . | “Negligible risk” eye pupil diam. (mm) . | For 8 mm pupil: factor above HSE . |
---|---|---|---|---|

0.1 | 60 | 1.89 | 4.1 | 3.8 |

0.2 | 40 | 1.46 | 4.4 | 3.3 |

0.3 | 30 | 1.30 | 4.8 | 2.8 |

0.4 | 24 | 1.21 | 5.2 | 2.4 |

0.5 | 20 | 1.16 | 5.6 | 2.1 |

0.6 | 17 | 1.14 | 5.9 | 1.8 |

0.7 | 15 | 1.13 | 6.3 | 1.6 |

0.8 | 13 | 1.12 | 6.7 | 1.4 |

0.9 | 12 | 1.11 | 7.0 | 1.3 |

1.0 | 11 | 1.11 | 7.4 | 1.2 |

With respect to the calculated eye pupil diameters see discussion in Sec. II where it was concluded that for normally reacting pupils the adjusted EL is not exceeded. In this section, we additionally consider the minimum safety margin of the EL. With this minimum safety margin (“minimum” because the HSE is scaled to lie below injury thresholds), the pupil diameter that is associated with negligible risk levels is correspondingly larger as compared to the pupil diameter that is based on not exceeding the EL. Consequently, the conclusion that pupils for reasonably foreseeable situations can be assumed as not to be larger than the values listed (Table II column 4) is further strengthened and supported.

The absolute worst-case scenario of a dilated pupil that does not react to light stimulus is considered. The results are shown in the last column in Table II. A pupil diameter of 8 mm is used here even for very small exposure distances. This scenario is possible for medically dilated pupils which do not constrict upon exposure to light; but in combination with no spherical nor chromatic aberrations which would lead to increased image size is very rare. Up to a distance of 50 cm, the factor exceeding the HSE is less than 2, and since the difference between injury thresholds of human retinas and nonhuman primates can also be estimated to be at least in that range, the risk for injury for short-time exposure for these distances can still be assumed to be relatively small. For distances less than 50 cm, the factor is larger than 2, up to almost 4 for 10 cm. From this factor alone, exceeding the injury threshold for short-time exposure cannot be precluded. It should, however, be considered that in addition to low probability that a person is exposed within this short distance, these are relatively large projectors usually mounted on the ceiling or in a restricted projection booth. In addition, there will always be light leakage present, a soft-start is a required safety feature and besides pupil constriction there are also other aversion responses to bright light such as squinting and looking away. These all reduce the exposure and therefore the risk, even if the pupil does not constrict.

That exposure with medically dilated pupils occurs within 50 cm from a higher power projector, in combination with good visual acuity and accommodation to the exit pupil can be assumed to occur with extremely low probability. Due to aversion responses to bright light, a potentially critical exposure can be considered as intentionally looking into the bright light. Such intentional staring into bright light for extended periods of time also bears a certain risk for photochemically induced injury, (see Appendix for an analysis) which is known from staring into welding arcs and the sun, where the “time to injury” is shorter for dilated pupils. Considering that for RG2 projectors, a corresponding warning label not to stare into the light source is mandatory in IEC 62471-5, this type of risk from bright sources (as long as they are not child appealing products, which is the case for a projector unless, for example, the projector is designed in a Mickey Mouse style) is usually considered an “acceptable” risk, at least under European^{34} product safety legislation. Thus although the risk for these absolute worst-case scenarios is not “zero,” i.e., not negligible, it should still be considered as acceptable.

## IV. CONCLUSIONS AND SUMMARY

The safety standard IEC 62471-5 defines the criteria for determining risk groups for LIP as well as conventional image projectors. Risk group 2 (RG2) products are understood—and intended to be treated—to be safe as a consumer product. For projectors which approach the emission limit of RG2 determined at a distance of 1 m from the projector lens, exposures at closer distances with eye pupils of 7 mm diameter will exceed the exposure limit. Due to the dependency of the retinal thermal exposure limit (EL) on source size and therefore exposure distance, the maximum pupil diameter for which the EL is not exceeded can be calculated. Considering that RG2 projectors are (1) required to have a soft-start, (2) there is significant light emission even for black images, (3) light emission is also visible from the side of the projected beam, and (4) for accommodation to the apparent source at close distance the “near-triad of accommodation” leads to pupil constriction, it can be concluded that for normally reacting pupils, the EL will not be exceeded when adjusted for a realistic pupil size, even at very close distance to the projector. Thus based on an EL analysis, it is possible to conclude that there is negligible risk of injury for normally reacting pupils for unintentional exposure to RG2 projectors. This conclusion holds independently of the specific optical properties of the projector, such as exit pupil diameter. The conclusion can be strengthened further by extending the risk analysis to include the minimum safety margin between the injury threshold and the EL, where even larger eye pupils can be associated with negligible risk for injury for unintentional exposure. This conclusion can be extended to intentional exposures, considering that for a projector at the RG2 limit, the NASA unified formula predicts a pupil of 2 mm diameter, being reached after about a 2 s exposure.^{35} Additionally, the reduction of the retinal thermal injury threshold^{11,33}—when expressed as radiance and not as radiance dose value—as a function of exposure duration is small, an effect that is further enhanced by eye movements. Consequently, if there is negligible risk for retinal thermal injury for unintentional exposure of the order of up to 0.25 s, then retinal thermal injury is also not expected for intentionally staring into the beam. These factors were the very basis for setting the EL as a constant radiance value for exposure durations exceeding 0.25 s. The risk for photochemically induced retinal injury is discussed in the Appendix.

The absolute worst-case is a fully dilated pupil that is nonresponsive due to medical conditions or medication and exposure very close to the projector (less than 50 cm from the projection lens), combined with intentionally looking into the projector and suppressing aversion responses to bright light. In this case, retinal injury can potentially occur for high power RG2 projectors. The probability, and the corresponding risk, for this to occur is very small, and is usually accepted in terms of product safety to be managed by warning labels.

This analysis supports the product classification scheme and test conditions defined in IEC 62471-5, including the delineation of the risk groups, where projectors up to RG2 are considered sufficiently safe to be marketed as consumer products.

## ACKNOWLEDGMENTS

This work was supported by LIPA, the Laser Illuminated Projector Association. The authors gratefully acknowledge comments and a critical review of the manuscript by Leslie Lyons.

### APPENDIX: ANALYSIS OF PHOTOCHEMICAL LIMIT

The body of the paper refers to the retinal thermal exposure limit. It is known from general experience and quantitative analysis that for short (unintended) exposures, the retinal thermal limit is more restrictive than the blue light hazard (BLH) limit (the limit that protects against photochemically induced retinal injury). For the case that injury occurs following unintended, i.e., short exposure, it is thermally and not photochemically induced. Photochemically induced injury becomes more restrictive for prolonged intentional staring and luminance levels that are below those necessary to induce thermal injury after a short exposure duration, as can also be shown for image projectors^{36} and is summarized here briefly. The spectral distribution of typical projectors (LIP and conventional) was used to obtain a scaling factor between *R*(*λ*)-weighted and *B*(*λ*)-weighted spectra, as applicable for the BLH. The scaling factor is applied to an exposure level assumed to be equal to the retinal thermal limit for 0.25 s exposure duration. The resulting BLH effective exposure level, for the worst case of an apparent source angular subtense *α* = 11 mrad is a factor of at least (depending on spectral distribution) 6.3 below the BLH EL for 0.25 s exposure duration. For larger sources, the factor increases, for instance, to 8.6 for *α* = 15 mrad, which is the typical angular subtense at 1 m from the image projector. Since the underlying BLH limit is a constant dose limit, this factor can also be applied to the exposure duration for reaching the BLH limit: for *α* = 11 mrad, when the exposure is equal to the retinal thermal EL, the BLH limit (being based on pupil diameters of about 3.5 mm) will be reached after 1.6 s or longer staring duration (depending on spectral distribution) or at a minimum of 2.1 s for *α* = 15 mrad. Considering (a) that the high brightness would induce pupil constriction to about 2 mm as well as aversion responses within that time, and (b) the safety margin of roughly^{37} 20 inherent in the BLH EL for a 3 mm pupil of the eye, the risk for photochemically induced retinal injury is negligible for staring durations of up to at least 20 s. While intentionally looking into a bright projector for longer than 0.25 s cannot be excluded, for a product safety analysis (even a very conservative one) it should not be necessary to consider staring durations into an uncomfortably bright source of longer than a few seconds. Consequently, the risk for photochemically induced injury for reasonably foreseeable intentional staring durations is negligible.

## References

^{®}Series-600 DMD, DLPA053 (

^{®}Technology and Products Texas Instruments Q2 (