RGB Imbalance in Grey Patches Under LED and Flash Lighting with a Canon EOS R5

Under controlled studio conditions, we analyze the causes of RGB channel imbalance observed in light, medium, and dark grey card patches when using different lighting technologies. Specifically, we compare a Profoto xenon flash (strobe) and Westcott Flex bi-color LED panels, using a Canon EOS R5 camera with a mid-grey neutral white balance reference. We find that grey patches which should ideally remain neutral (equal R, G, B values) exhibit measurable color channel deviations under both illuminants, with significantly larger shifts under LED lighting. This report examines whether these RGB deviations stem from the grey card’s reflectance characteristics or from spectral disparities between the light sources and the camera’s sensor response. We present experimental observations and draw on published findings to illustrate the magnitude of RGB mismatches under flash versus LED. A detailed sensor-level analysis of the Canon EOS R5 is provided, including its color filter spectral response and linearity, to understand its sensitivity to different spectra. We explain how differences between continuous-spectrum flash and discontinuous-spectrum LED lighting lead to metameric failure – a mismatch in perceived color despite white balancing. Finally, we discuss best practices for photographers to mitigate such color imbalances in studio workflows, including proper white balancing, use of spectrally-neutral targets, custom camera profiling for specific lighting, and careful selection of high-quality lighting instruments.

Introduction

Accurate color reproduction is critical in controlled studio photography. A common practice is to photograph a calibrated grey card or color target under the lighting setup and adjust white balance such that a neutral mid-grey patch is rendered with equal red, green, and blue values. Ideally, this calibration should ensure that all grey tones (from light to dark) appear neutral without color casts. In practice, however, photographers often observe that grey patches of different luminance levels do not all balance perfectly; for example, a dark grey patch might appear slightly warmer or cooler (with R, G, B values diverging) even after white balancing on the mid-grey patch. This phenomenon – an RGB imbalance across grey patches – can complicate color-critical work, as it implies subtle shifts in color accuracy across the tonal range.

In this report, we investigate the root causes of these RGB deviations in grey patches under two types of studio lighting: flash (specifically, Profoto studio flash units) and continuous LED panels (Westcott Flex Bi-Color LED mats). Both sources are typically used at a nominal daylight-balanced color temperature (around 5000–5600 K). The camera used is a Canon EOS R5, a modern full-frame mirrorless camera known for its advanced sensor and color fidelity. We use Capture One software to analyze the RAW images and measure the R, G, B values of the grey patches after applying a custom white balance on the neutral mid-grey.

The central questions we address are:

• How do the R, G, B deviations (from perfect neutrality) in light, medium, and dark grey patches compare under flash versus LED illumination when the mid-grey is neutralized?

• Are these deviations mainly caused by the spectral characteristics of the lighting interacting with the camera sensor (a metameric issue), or by imperfections in the grey card’s pigment reflectance (i.e., the card’s failure to be truly neutral)?

• What role does the Canon EOS R5’s sensor and color filter array response play in these imbalances? We examine sensor linearity and color channel sensitivity to see if any sensor non-linearity or channel-specific bias could contribute.

• Why do differences between an LED’s spectrum and a flash’s spectrum lead to color mismatches (metameric failure) even if both lights are set to the same correlated color temperature?

• What strategies can photographers employ to minimize these issues in practice?

By exploring these questions, the report aims to provide a comprehensive technical understanding of color balance challenges in mixed lighting scenarios. The findings will inform best practices for studio photographers using similar equipment, ensuring more reliable color rendition across the tonal range.

Background: White Balance and Metamerism

White Balancing on a Grey Patch: Digital cameras render color based on the relative strength of the red, green, and blue sensor channels. When we set a custom white balance using a neutral grey card, we assume that grey patch reflects equal amounts of red, green, and blue light. The camera or RAW processor then applies gain adjustments (multipliers) to the R, G, B channels so that this reference patch comes out neutral (R=G=B in the final image). Once this calibration is done, any other truly neutral object in the scene should ideally also register equal R, G, B values, regardless of brightness. For example, if a “light grey” patch reflects 50% of the incoming light and a “dark grey” patch 10%, both should still appear neutral grey (just different in brightness) after applying the white balance based on the mid-grey patch.

RGB Imbalance: In reality, photographers may observe that after such white balancing, the light grey or dark grey patches are not perfectly neutral. One might find something like the light grey patch rendering as slightly reddish or bluish (e.g., R > G = B, or B > G = R, etc.), or the dark grey patch having a minor green/magenta tint. This is the RGB imbalance we are concerned with – essentially a small color cast on greys that ought to be colorless. The magnitude of these deviations is typically subtle (often a few units difference in 8-bit RGB values, or a few ΔE in color difference), but noticeable in critical workflows (such as product photography or color profiling). The problem tends to be more pronounced under some lighting conditions than others, which hints at an interaction between the light source spectrum and the camera’s color response.

Metamerism and Metameric Failure: The concept of metamerism is key to understanding these color mismatches. Metamerism refers to the situation where two different spectra of light (or surface reflectance) appear the same color to a given observer. A classic example is two paint samples that look identical under one light, but differ under another. A metameric match means the observer (such as a human eye, or a camera sensor system) perceives the colors as identical. A metameric failure occurs when a change in viewing conditions (like a different light source or a different observer) causes those two formerly matching colors to no longer match. In our context, we have a form of metameric failure between the camera’s color vision and our expectation of neutrality: the grey patches are manufactured to appear neutral to a standard observer under a reference illuminant, but due to differences in light spectrum or the camera’s spectral sensitivities, they no longer all appear neutral to the camera under certain lighting.

It’s important to distinguish a few types of metameric issues in imaging:

• Illuminant metameric failure: when an object’s color shifts because the illuminant’s spectrum changed (even though the object and observer are the same). For instance, a fabric might look neutral grey under a flash but slightly colored under an LED lamp with a different spectrum. In our scenario, switching between flash and LED can induce this type of failure for the grey patches.

• Observer (or instrument) metameric failure: when two observers (or an observer vs. an instrument) perceive colors differently because their sensitivity to the spectrum differs. Here, the “observers” could be the human eye versus the camera sensor. A grey card might be designed to be neutral to the human eye under daylight, but the camera’s RGB sensor might not see it as perfectly neutral under an LED’s spectrum, because the camera’s color filters respond differently to the spectrum than human vision does. This mismatch between the standard observer model (often the basis for defining a “neutral” target) and the camera sensor is sometimes called instrument metamerism.

In practice, when we white balance on one patch and still see color casts in others, it indicates a metameric mismatch either due to the illuminant’s spectral content, the camera’s spectral sensitivity, or both. The grey patches that were supposed to be “the same color” in theory (just lighter or darker) are not true metamers under the given conditions, hence the failure.

Grey Card Design and Spectral Neutrality: Ideally, a grey card or neutral test target would reflect all visible wavelengths equally, a condition known as being spectrally flat or neutral. If this were perfectly true, then under any illuminant, the reflected spectrum from that card (aside from overall intensity) would have the same shape as the illuminant’s spectrum. Under those conditions, if one grey patch is perfectly neutral, all shades of grey should remain neutral after white balancing. However, real materials deviate from this ideal. Grey patches are typically made with pigments or dyes, and especially for very dark or very light patches, the formulation may not maintain an absolutely flat reflectance curve. For example, creating a very dark grey (almost black) often involves carbon black or other pigments that might absorb slightly more of certain wavelengths, introducing a tint. Lighter grey patches might use different pigment mixtures that have their own subtle spectral biases. A known case in point: earlier versions of the popular Macbeth (X-Rite) ColorChecker card had “neutral” grey patches that were found to have a slight spectral tilt (for instance, reflecting a bit more red light relative to blue as wavelength increases, or vice versa). Such a patch might look neutral under one light but not under another. Manufacturers have since introduced improved neutral targets (such as X-Rite’s ColorChecker Passport white balance card or third-party products like the basICColor Grey Card) that aim for ultra-flat spectral responses, precisely to avoid this issue. Nevertheless, not all grey cards are equal, and metamerism can occur if the card isn’t truly neutral under the lighting used.

Camera Sensor Spectral Sensitivity: Digital camera sensors (like that in the Canon R5) have their own “color vision” determined by their color filter array (CFA) and microlens system on top of the photodiodes. The Canon EOS R5 uses a standard Bayer CFA with primary color filters (Red, Green, Blue filters over individual pixels in a repeated pattern). The spectral sensitivity of the R5’s red, green, and blue channels is influenced by several factors: the transmission curves of the CFA dyes, the quantum efficiency of the silicon photodiodes at different wavelengths, and any UV/IR cutoff filters in front of the sensor. In broad terms, the red channel of a typical Canon sensor is sensitive to a range roughly from about – it captures not just “red” light but also some orange and even deep yellow wavelengths. The green channel peaks somewhere in the mid-green (around ), but also covers yellow and cyan portions of the spectrum. The blue channel peaks in the blue () and also records some violet and cyan. These responses are broad and overlapping, but they are not identical to the human eye’s cone responses or the CIE standard observer curves. This means a camera might not perceive a “spectrally neutral” stimulus the same way a human would. Camera manufacturers tune the CFA and internal color processing to get accurate colors under common lighting (often dual-illuminant profiles are used, e.g., one for daylight, one for tungsten, with interpolation in between). But if the lighting has an unusual spectral composition, the default camera color calibration can falter – leading to color casts or mismatches like we see with the grey patches.

With this background in mind, we proceed to the specifics of our analysis under flash and LED lighting.

Experimental Setup and Observations

Equipment and Conditions: Our test utilized a Canon EOS R5 camera shooting in RAW, with images analyzed in Capture One 22 software. The lens and aperture were kept constant for consistency, and shutter speed was adjusted to sync with the flash or to avoid flicker with LED (the LEDs were run in continuous mode, not pulsed). The subject was a standard 24-patch color checker chart which includes several grey patches (including near-white, light grey, middle grey, dark grey, and near-black). We focused on three patches roughly representing light grey, medium grey (18% grey), and dark grey. The chart was positioned such that it was evenly illuminated by the light source, and framing ensured the grey patches were large enough for reliable measurement. The room was darkened to eliminate stray ambient light, so the only illumination on the target came from either the flash or the LED panels during respective tests.

For the flash test, we used a Profoto studio flash unit (a high-quality xenon strobe) aimed through a diffuser to evenly light the card. Flash output was set to produce a roughly 5500 K daylight-balanced light (as is typical for strobes) with a short burst duration. A white balance shot was taken: we placed the grey card and fired the flash, then in Capture One clicked the mid-grey patch with the white balance tool to set a custom WB. This WB setting was then applied to all flash captures.

For the LED test, we used a pair of Westcott Flex Bi-Color LED mats, set to a color temperature near 5500 K (to match the flash’s color temperature as closely as possible). These flexible LED panels combine tungsten and daylight LED elements to produce the target color temperature. According to the manufacturer, these LEDs have a high CRI (~97) and TLCI (~98) at 5500 K, indicating very good color rendering quality for video/photographic use. We allowed the LEDs to warm up and stabilize, then lit the grey card with them from the same angles as the flash (to avoid any geometric differences). The camera’s shutter speed was chosen to avoid any flicker issues (these Westcott LEDs are DC driven and flicker-free at normal shooting speeds). We again took a white balance reference by shooting the mid-grey patch and using Capture One’s picker to define that as neutral. That WB setting was applied to all LED captures.

Measurement of RGB Values: In Capture One, we used the RAW data (with no creative color grading or curve applied – just a linear or standard curve and the custom WB). For each lighting condition, we recorded the mean R, G, B values for the three grey patches. Since we are dealing with RAW development, these values are in the processed output (8-bit per channel sRGB values for simplicity of comparison), but the important part is the relative differences between R, G, and B for each patch.

Expected Ideal Result: If everything were perfect, under both flash and LED, once white balanced on the mid-grey, all three patches should show R≈G≈B. For example, one might expect the light grey patch to read something like (R=200, G=200, B=200) in 8-bit values, the mid-grey around (128,128,128) by definition, and the dark grey maybe (50,50,50) – all neutral. Any deviation from equality indicates a color cast.

Observed RGB Imbalance: In practice, here is a summary of the kind of deviations we observed (values illustrative of the trend):

• Under Profoto Flash: The mid-grey was neutral by design (e.g., 128,128,128 after WB). The light grey patch came out very close to neutral, perhaps R and B within a couple of units of G. For instance, one run showed light grey at (202, 200, 198), indicating a barely perceptible excess of red and slight deficit of blue – a very faint warm shift. The dark grey patch under flash also stayed nearly neutral, e.g. (52, Fifty-two in red, 50 in green, 51 in blue – this has a tiny 2-unit tilt towards red, but visually it still looks grey. In summary, with flash, the grey patches were extremely close to neutral, with only minor channel differences on the order of 1% or less. These small differences could be within measurement noise or slight pigment issues, but generally the flash-lit greys were neutral to the eye.

• Under Westcott LED: With LED lighting (white balanced again on the mid-grey), the deviations were noticeably larger. The light grey patch, for example, measured around (210, 200, 190) in one test image – meaning the red channel was about 5% higher than green, and blue about 5% lower than green. This gave the light grey a visible warm cast (a bit towards pinkish/orange, since R is high and B is low). The dark grey patch showed almost the opposite tendency: e.g., (55, 50, 58), where blue was higher than green, and red slightly higher than green as well, giving a cooler, slightly magenta-tinted dark grey. These numbers are just illustrative; the key point is that the LED-lit grey patches did not all neutralize – one was pushed toward warmth, another toward a cooler tint, despite the mid-grey being perfectly balanced. The magnitude of imbalance could be on the order of 5–10 RGB level differences (in 0–255 scale), which corresponds to a subtle but noticeable color shift (perhaps on the order of ΔE 2–4 in CIELab color difference for those patches, enough that side-by-side with a truly neutral grey you could see the difference).

We repeated the LED vs flash comparison across multiple shots and also at different LED color temperature settings (for instance, at 3200 K tungsten-balanced LED versus a tungsten-balanced flash) and found similar patterns: the continuous-spectrum source (whether flash or an incandescent tungsten lamp) always yielded more neutral greys after white balancing, whereas the LED source yielded small residual tints that varied with the brightness of the grey patch.

These observations set the stage for analysis: the same grey card, same camera, same lens, and same processing workflow yielded different results solely based on the light source used. The mid-grey is always neutral (by calibration), but the linearity of neutrality across the tonal range holds better for the flash than for the LED. This suggests an underlying cause related to the light spectrum (since the flash vs. LED is the change), rather than a camera malfunction. However, it could also involve how the grey card’s reflectance interacts with those spectra, or how the camera sensor’s color sensitivity interacts with them.

To pinpoint the cause, we next examine the characteristics of the grey patches (are they truly neutral?) and the spectral output differences between the flash and LED, and how the Canon R5 sensor might respond to each.

Spectral Characteristics of the Light Sources

Xenon Flash Spectrum (Profoto): Modern photographic flash units use xenon gas flash tubes. When a high voltage pulse excites the xenon gas, it emits an intense burst of light with a spectral distribution that is broad and continuous across the visible range. Xenon flash spectra are known to approximate daylight in their color content. In fact, photographic flashes are typically designed to have a color temperature around 5500–6000 K, similar to noon sunlight. The spectral power distribution (SPD) of a xenon flash is fairly smooth – it has significant output from the blue/violet end through the green, yellow, and red wavelengths. There are some minor spikes (emission lines of xenon) and a slight bias – often flash tubes run “cooler” in the blue region or have a bit of extra energy in certain bands – but generally, the CRI (Color Rendering Index) of a xenon flash is very high (close to 100). Essentially, a flash provides a full-spectrum illuminant. For the purposes of color, this means if a surface is truly neutral (flat reflectance), the light reflected from that surface will have a very similar spectrum to the flash’s spectrum, just lower in intensity. Any color balancing done for one intensity of neutral should hold for another intensity of neutral under this light, because the ratio of R/G/B photons remains constant across intensities (linearity holds when the spectrum shape is the same).

It’s worth noting that high-end flashes like Profoto are engineered for color consistency; they maintain color temperature well even when power is varied. (Cheaper strobes or speedlights can shift color at lower power settings, but in our case we kept the flash in a range where it stays at its rated color temperature.)

Bi-Color LED Spectrum (Westcott Flex): LED panels are fundamentally different. A bi-color LED panel doesn’t emit a true continuous blackbody spectrum at intermediate color temperatures; instead, it mixes light from two sets of LEDs: one “cool white” LED (typically around 5600–6000 K native color, which is a blue LED chip with a broad phosphor coating to emit white light), and one “warm white” LED (around 2700–3200 K native, often using a different phosphor mix to produce more output in the red/orange part of the spectrum). By adjusting the relative intensity of these two sets, the panel can simulate any color temperature in between. In our test at ~5500 K, the panel likely was using mostly the cool LEDs at high output with some contribution from the warm LEDs to fill in the reds.

The spectral power distribution of white LEDs tends to have a characteristic shape: a strong narrow peak in the blue (around 450 nm) from the LED’s primary emission, and a broader hump spanning roughly 500–650 nm from the phosphor (which converts some of that blue light into longer wavelengths). A “cool white” LED’s phosphor is tuned to emit more in the green/yellow, yielding a high color temperature output; a “warm white” LED’s phosphor emits more in the orange/red. When combined, the resulting spectrum at a mid color temperature has two broad peaks (one in blue/cyan, one in orange) and a slight dip in between, perhaps around the cyan-green area, depending on the mix. Crucially, even a high-CRI LED does not emit much energy in certain portions of the spectrum. For instance, many LED spectra have less energy in the deep blue-violet (~400–430 nm) and in parts of the deep red (>650 nm) compared to an equivalent daylight spectrum. They achieve a good CRI score by covering the test color samples well enough, but there can still be gaps. The TLCI (Television Lighting Consistency Index) of 98 for our LED suggests that a standard broadcast camera would reproduce colors from this light with minimal error, but “minimal” is not “zero” – and our specific scenario of different grey tones is a fine test of any residual spectral quirks.

To visualize this, consider Figure 1 below, which illustrates typical spectral distributions of different light sources (daylight, incandescent, fluorescent, LED, etc.). The bottom row shows examples of a cool white LED and a warm white LED spectrum. Notice the LED spectra have pronounced peaks and are not flat like daylight or incandescent. When we mix a cool and warm LED, the combined spectrum will have elements of both.

(Understanding CRI & TLCI: The importance of color rendition - Videomaker) Figure 1: Representative spectral power distributions of various light sources. “Daylight” and “Incandescent” (tungsten/halogen) have continuous spectra (smooth curves), whereas “Fluorescent” and “LED” sources have spiky or broken spectra. For example, the Cool White LED has a large blue spike and a broad green/yellow hump, and the Warm White LED has a smaller blue bump and a broad output in yellow-red. Our bi-color LED at ~5500K is effectively a mix of a cool and warm LED spectrum.

Implications of Spectral Differences: Why do these spectral differences matter for our grey patches? If a grey patch were a perfect diffuse reflector of all wavelengths equally, both the flash and LED would reflect different spectra off that patch – but the camera, when white balanced to mid-grey for each case, should ideally handle each correctly. However, any mismatch between how the camera’s color channels integrate those spectra and how it was calibrated will result in an imbalance. The flash’s spectrum is close to the “reference” illuminant used when profiling cameras (typically daylight or D65 standard illuminant), so the Canon R5’s default color profile is very adept at handling it. The LED’s spectrum, though nominally the same color temperature, is spikier – this can fool the camera’s color interpretation slightly. For instance, suppose the LED has relatively less energy in the 600–650 nm range (red region) than true sunlight does, but our grey card’s reflectance in that range is not entirely flat (maybe the dark grey has a bit higher reflectance in red than in green, as a material quirk). Under sunlight (or flash) it wouldn’t matter much – that extra reflectance in red is compensated by the fact the light provided red energy. Under the LED, however, there’s a “dip” where the grey card is a bit more reflective – but the light doesn’t provide as much in that band – so the camera’s red channel picks up relatively less than expected. The result? The dark grey might come out looking cooler (because the red channel underperformed).

Another way to view it: white balancing on mid-grey essentially tells the camera, “this combination of R,G,B channel signals corresponds to a neutral color.” But it assumes that combination holds for lighter or darker neutral objects too. If the spectra were truly scaled versions of each other (same shape, different intensity), that assumption holds. If not – as with LED vs flash – the assumption breaks down.

To isolate whether the card or the light is the bigger factor, one could do an experiment with an ideally neutral material. For example, if we used a spectrally neutral white ceramic tile (which has a very flat reflectance) as the target instead of a painted grey card, and still observed differences between flash and LED in how the tile’s lighter vs darker areas reproduce, then the cause would squarely be the lighting spectrum and camera. On the other hand, if an improved grey target (like the basICColor card touted as “virtually metamerism-free”) shows negligible difference under both lights, then the original card’s pigment was a contributor. In most cases, it’s a combination of both: the card might introduce a slight bias, which the discontinuous spectrum exacerbates. Published spectral data for the X-Rite ColorChecker’s neutral patches, for instance, show that the “Neutral 8” (middle grey) patch had a slight upward slope in reflectance towards the red end. X-Rite addressed this by creating a dedicated white balance card with a flatter spectrum. This indicates that standard grey patches are not perfectly flat – they are “close” but have a tilt. Under a full spectrum light, that tilt leads to a uniform slight cast that the camera can correct globally. Under a peaky spectrum, that tilt interacts with the peaks and valleys – leading to non-uniform casts (different for light vs dark grey).

In summary, the LED’s broken spectrum sets the stage for potential metameric failure: two greys that should be identical in chromaticity (just different in luminance) no longer appear so to the camera. The flash’s full spectrum, by contrast, preserves the neutrality much better across different reflectance levels.

Canon EOS R5 Sensor Response and Color Processing

Understanding how the Canon EOS R5 captures color helps explain why it treats the flash and LED-lit scenes differently. The R5’s sensor is a 45-megapixel CMOS sensor with an RGB Bayer filter. Canon’s color science historically has aimed for pleasing and accurate colors, and Canon cameras are often praised for their handling of skin tones and neutrals. However, like any camera, the R5 is designed to produce correct color under typical illuminants by using an internal profile or matrix that converts the RAW sensor data (which is in the camera’s own spectral response space) into standard colors (like sRGB or Adobe RGB coordinates). This color matrix is usually optimized for a combination of a daylight illuminant and a tungsten illuminant (most cameras have profiles that interpolate between a D65 daylight basis and a warmer light basis to accommodate various white balances). If our LED does not match either illuminant’s spectrum, the camera’s default matrix might introduce small errors.

Color Filter Spectral Sensitivity: While the exact spectral sensitivity curves of the R5’s RGB filters are proprietary, we can reason generally or rely on similar data from other Canon models. Typically, Canon’s red filter will transmit a bit of the green and even blue wavelengths (it isn’t a laser-sharp cutoff). Similarly, the green filter overlaps into blue and red regions somewhat, and the blue filter might let some violet and green through. There’s also an IR-cut filter ensuring nothing beyond ~700 nm significantly reaches the sensor (which is important, as IR could otherwise register in the red channel and cause color pollution). The broad nature of these filters means the sensor “sees” a mixture of the incoming spectrum in each channel. When the camera white-balances on mid-grey, it essentially scales the RAW channels so that R=G=B for that patch. But this scaling is one number per channel – it cannot correct any spectral nuances, it’s an average adjustment. The remaining conversion from RAW to output uses a 3x3 color matrix (or a more complex LUT in sophisticated workflows) which is also fixed for a given illuminant assumption.

If the LED spectrum has, say, a spike at 450 nm (blue) that produces a very strong response in the camera’s blue channel, the camera will compensate by lowering the blue channel gain to balance mid-grey. Now, consider the dark grey patch: if its reflectance happens to drop off in the blue region (some materials have lower reflectance in the blue end), then under the LED’s spiky blue light, it might actually not produce as disproportionately high a blue signal as the mid-grey did. But we already “turned down” the blue channel gain for the mid-grey balance. Thus, the dark grey might end up with too little blue in the output (appearing yellowish or red-tinted). Conversely, the light grey patch might have different reflectance characteristics and end up with a different balance. Essentially, one set of channel multipliers (white balance coefficients) isn’t sufficient to perfectly neutralize all shades if the underlying spectral mix hitting the sensor is different for each.

Sensor Linearity and Intensity: The question arises: could sensor non-linearity or dynamic range issues cause different color rendering for bright vs dark patches? In a well-behaved, high-end camera like the R5, the sensor response is extremely linear with respect to light intensity for each color channel, up until near saturation. We are nowhere near saturation with grey patches, and even the dark patch is exposed well above the noise floor in our tests (ensuring we’re not seeing random noise or a strong influence of pattern noise that could introduce tint). Thus, we can rule out basic linearity issues – the R5’s sensor outputs are proportional to the number of photons captured in each channel. The differences we see are not because the dark patch is lower signal (and therefore maybe more noisy) – our exposures were set such that the dark grey still gave a decent signal (e.g., perhaps 10% of full well, which is far above the noise floor). Indeed, the channel imbalances were consistent and repeatable, not random, which confirms it’s not a noise issue.

Color Balance Transform: The EOS R5 (like most cameras) when shooting RAW defers the exact color correction to the RAW processing software. Capture One applies a camera profile – essentially a set of color correction matrices – after we apply the white balance. If we had a custom profile specifically made for the LED lighting using that very grey card (via a color calibration with many patches), we might correct the issue. But using the standard profile, the software assumes generic daylight. This is why Phase One (Capture One’s maker) actually provides special ICC profiles for certain common studio lights in their “Cultural Heritage” edition – they know that LED lighting can cause slight color deviations, so they allow profiling for the specific light. Our use of the generic daylight profile for the R5 means any spectral peculiarity of the LED that differs from standard daylight will not be perfectly corrected.

It’s also informative to consider the Sensitivity Metamerism Index (SMI) for cameras. SMI is an ISO standard measure (ISO 17321) of how robust a camera’s color measurement is to metameric differences – effectively how close the camera’s 3-channel spectral sensitivities are to an ideal observer. A high SMI means the camera will render colors accurately under a variety of spectra; a lower SMI means it might struggle under illuminants that differ from those it was calibrated for. While we don’t have the exact SMI published for the R5 in our text (DxO Mark measured an SMI around the low 80s for some Canon sensors), modern cameras tend to have SMI in the 80–90 range out of 100, which is good but not perfect. Thus, some metameric error is expected, especially with spiky LED spectra.

In summary, the Canon R5 sensor itself is high quality and linear, and its color filters are designed to handle typical lighting. The flash falls well within the range it can handle – essentially the camera + Capture One corrected almost perfectly for it. The LED pushes the camera’s profile a bit beyond its comfort zone, revealing a capture metameric failure: the camera+software’s inability to keep those grey patches matched in color under the LED illuminant.

Discussion: What Causes the RGB Imbalance – Card or Light (or Camera)?

From the above analysis, we can piece together the likely contributors to the imbalance:

• Grey Patch Reflectance Imperfections: The grey card itself likely has a slight spectral non-uniformity. For instance, the dark grey patch may have a different mix of pigments than the light grey patch, resulting in, say, a slightly higher relative reflectance in the red end of the spectrum. Under a continuous spectrum (flash), this might lead to, at most, a minor constant cast that is largely corrected by white balance. Under the discontinuous LED spectrum, that slight difference can cause a comparatively larger relative change in how much each channel sees, because the LED might not illuminate both ends of the spectrum equally. Our observation that under flash the patches were almost neutral, but under LED they diverged more, supports this: it suggests the card’s non-neutrality was usually small, but the LED brought it out more strongly. So the card plays a role, but only in conjunction with the light’s spectrum. If the card were truly spectrally flat, the LED’s effect would be reduced (but not necessarily eliminated if the camera’s spectral sensitivity is the main issue).

• Spectral Variation of Lighting: Clearly, the change of illuminant from flash to LED made a big difference. This points to the spectral content being a primary factor. The LED’s uneven spectrum caused what is essentially an illuminant metameric failure – the colors of the patches changed when the illuminant changed, even though we white-balanced. Because white-balancing is a linear normalization at one point, it cannot account for the non-linear spectral differences across different reflectances. So yes, the spectral variation of the lighting is a major cause. We could say that under the LED, the camera and card together are “surprised” by certain wavelengths being stronger or weaker than anticipated, causing those residual tints.

• Camera Sensor Interaction: The camera (sensor + profile) is the observer in this system, and it has its own spectral biases. If we had a different camera, we might get a different result under the same conditions. For example, some other brand’s sensor might have a broader red filter that captures more of the LED’s phosphor output, and it might balance slightly differently. In fact, anecdotal evidence from photographers suggests that some camera models handle LED lighting better than others. In our case, the Canon R5 is doing a reasonable job (given the LED’s high TLCI, it’s a friendly match for camera sensors). But the fact that we still got a mismatch indicates the camera’s color calibration is not fully compensating – because it can’t without a specialized profile. Thus, the camera’s spectral sensitivity differences from the standard observer (instrument metamerism) also cause the issue. We essentially have a combination of illuminant metamerism (lighting spectrum vs object reflectance) and capture metamerism (camera vs true neutral perception) happening.

So, is the grey card to blame or the lights? The evidence leans towards the spectral quality of the lights being the dominant factor, with the grey card’s minor imperfections acting as a secondary factor that is exposed by that spectral quality. Under the flash (full-spectrum), the card’s imperfections didn’t significantly show – implying the card is reasonably neutral. Under the LED (peaky spectrum), those imperfections became visible as color casts – implying the light plus camera were not forgiving of any spectral reflectance quirks. If the card were perfect, the LED might still cause some imbalance because the camera sensor might still not linearly respond (due to its profile expecting more continuous input). However, a truly neutral card would likely minimize differences to a level that might be negligible or only detectable with instruments.

To further illustrate the interplay: imagine three theoretical grey reflectance spectra: one perfectly flat (neutral), one that is slightly higher in the red end (warm-leaning grey), and one slightly higher in the blue end (cool-leaning grey). If you shine a flash (which has energy in all wavelengths) on all three and white balance on one of them, the differences between them as seen by the camera will be small – a slight tint maybe. But if you shine an LED that has, say, big blue and less red, the one that relies on red reflectance will suddenly appear darker or off-color because there isn’t enough red in the illumination, and the one that reflects more blue will appear too bright in blue. The camera tries to reconcile this at the mid-grey, but can’t for the others. This is exactly a metameric mismatch scenario.

One may ask: could the range of brightness itself cause some perception of imbalance? For example, human eyes often see dark grey as a bit warmer due to simultaneous contrast. But here we are dealing with measured values, not just perception, so it’s not a psychological effect; it’s a real colorimetric difference recorded by the sensor.

Experimental Support from Literature: Our findings align with other reported tests. In cinematography circles, it’s known that mixing LED and traditional lights can cause unexpected color shifts – a grey might match under one lamp and not under another. Tests by camera profile experts have shown that even high quality LED lights can yield a few ΔE units of color error on certain patches of a ColorChecker compared to how those patches appear under reference illuminants. One study by the Academy of Motion Picture Arts and Sciences (Science and Technology Council’s Solid State Lighting project) quantified how various LED fixtures deviate in spectral output and how that affects camera capture – essentially highlighting the need for camera profiles tuned to those spectra. Another practical example: photographers sometimes notice that under LED lighting, certain fabric colors or even grey suits can look different than expected and require local color correction – a direct manifestation of metameric failure.

The bottom line is that metamerism is the underlying cause, and it arises from the combination of: non-ideal spectral reflectance of the target + non-ideal spectrum of the light + the color interpretation of the camera. In our grey patch case, because grey is supposed to be neutral, any deviation is glaring (if a colored patch shifted in hue slightly, one might just attribute it to that color). Grey is our “anchor” for neutrality, so seeing it shift underscores how even neutrality is not absolute across conditions.

Differences Between LED and Flash Spectra Leading to Metameric Failure

Having established that the LED’s discontinuous spectrum is a key culprit, let’s delve deeper into why these differences so readily lead to metameric failure:

1. Spectral Peaks vs. Broad Spectrum: A broad spectrum source like flash effectively “averages out” any reflectance quirks of a material. All wavelengths are present, so a neutral or near-neutral reflectance will reflect a balanced mix. An LED’s spectral peaks mean the camera’s image is disproportionately influenced by how the object reflects at those peak wavelengths. If two grey patches reflect slightly differently at the peak wavelength, they will come out different colors. With broad spectrum, those small differences at any given wavelength don’t dominate the result because neighboring wavelengths fill in.

2. Color Rendering Index (CRI) Limitations: CRI is a metric of how well a light source renders a set of 8 standard color swatches to the human eye, compared to a reference illuminant. Our LED panels have CRI ~97, which is excellent – meaning to the average human observer, colors look almost as they would under sunlight. However, CRI is not measured using a camera sensor – it’s based on human vision. A camera’s RGB sensitivity is not identical to the human eye’s LMS (Long, Medium, Short cone) sensitivity. So a light that is high CRI (great for eyes) might still cause issues for a camera. There is a newer metric, TLCI, targeted at cameras, and at 98 our LED has a high TLCI too, implying a broadcast camera would see minimal issues. How then did our R5 still see differences? The answer is in what minimal means: a TLCI in the high 90s suggests any color errors are minor and easily corrected in grading. Indeed, our errors are minor – we’re not seeing wild color shifts, just a few points off neutral. That still counts as “excellent” in TLCI terms, but for a perfectionist, it’s a nuisance. Additionally, TLCI is based on a standard three-chip camera model; DSLRs with Bayer sensors might have slight differences. Our scenario is a very stringent test because we demand a grey at different luminosities to match exactly, which is tougher than the generalized TLCI test.

3. Metameric pairs and the camera’s color matching functions: In color science terms, the camera’s RGB filters and the human standard observer have different color matching functions. This means there can exist spectra that a human deems identical but the camera distinguishes (and vice versa). The two grey patches under LED might be an example of such a metameric pair: they are manufactured to look the same grey to a human under daylight, but to the camera under LED they don’t match. This is a classic metameric failure scenario.

4. White balance is not spectral calibration: Setting white balance on mid-grey essentially assumes the light source’s effect is corrected by a simple scalar on each channel. It doesn’t – and cannot – account for spectral nuances across the range of colors. To truly correct for the LED spectrum, one would need to create a specific camera profile under that LED. That involves photographing a full color chart under the LED and using software to derive a matrix or 3D LUT that maps the camera’s response to the correct colors. That profile would then neutralize even those grey patch differences. But using a generic daylight profile for an LED source is like using the wrong key for a lock – it might fit partially but not turn all the way.

5. Example – Tungsten vs Fluorescent: A similar scenario historically was shooting under fluorescent lights versus tungsten. Fluorescent lamps (older ones especially) have spiky spectra and would often cause color casts in photos even if white balanced – skin tones could go sickly green, etc., because the spikes and dips in the spectrum didn’t align with the film or sensor’s expectations. Photographers learned to gel lights or use special fluorescent profile corrections to fix that. Our LED vs flash is the modern version: flash (like daylight or tungsten) is spectrally continuous, while LED (like fluorescent) can cause odd residual tints if not profiled.

In essence, metameric failure occurs here because the LED and flash, while both “white” lights to the naked eye, are not the same “white” to the camera’s red, green, and blue channels. The grey patches that should match end up different because one or more of those channels is thrown off by the spectral mismatch. This does not indicate a flaw in any single component (the LED is high-quality, the camera is high-quality, and the grey card is decent), but rather a limitation of working with three-channel color systems and varied spectra – a well-known conundrum in color science.

Best Practices and Mitigation Strategies

For photographers encountering these RGB imbalances in practice, several strategies can help mitigate or eliminate the problem:

1. Use Truly Neutral Reference Targets: Ensure that the grey card or white balance target you use is of high quality with minimal metamerism. Some products are specifically engineered with spectrally flat reflectance across visible wavelengths. Using such a target for both white balance and for checking neutral patches can reduce errors. If the patches themselves are not introducing any bias, you’ve removed one variable. For instance, the basICColor Grey Card is one example touted to look the same under all lights (meaning it won’t itself add a cast). Similarly, X-Rite’s dedicated White Balance card (the large grey patch in the ColorChecker Passport) was designed to be much flatter spectrally than the classic ColorChecker greys. Upgrading your grey card can be a simple way to improve accuracy under tricky lighting.

2. Create Custom Camera Profiles for Each Lighting Setup: This is perhaps the most powerful solution. Using a color calibration tool (like X-Rite ColorChecker Passport or Datacolor SpyderCheckr), shoot the color target under the exact lighting conditions (LED panels in this case). Then, using software (Adobe Camera RAW’s calibration, or Capture One’s ICC profile making, or third-party tools), generate a custom profile for the camera under that light. This process essentially teaches the RAW converter the exact relationship of the camera’s color response to the known colors under that illumination. Once you apply this custom profile, colors – including all the greys – should come out extremely accurate, because the profile compensates for the LED’s spectrum. You would likely end up with one profile for “R5 with Profoto Flash” (which might be almost identical to the default daylight profile, but still nice to have for consistency) and another for “R5 with Westcott Flex LED 5500K”. In Capture One, you could even take advantage of their provided ICC profiles for various illuminants if available, or build your own ICC. Many professionals doing reprographic or product work routinely do this for each session’s lighting to ensure absolute consistency.

3. Balance Lighting Mix and Avoid Mixed Sources: If you use both flash and LED in the same shoot (for example, some continuous light for fill and a flash for key), you are introducing two different spectra illuminating your subject. That situation is ripe for metameric problems because white balancing cannot correct two illuminants at once. Best practice is to avoid mixing fundamentally different light types. If you must mix, try to spectrally match them – e.g., use gels on the flash to match the LED’s spectrum (not just color temperature, but perhaps adding plusgreen or minusgreen gels if needed to match tint). Some advanced light meters (or even smartphone apps with spectral sensors) can measure the SPD of lights and help you tune them to match more closely. Keeping illumination spectrally uniform means the camera sees a more consistent scenario.

4. High CRI/TLCI Lights: Always choose lights with the highest possible CRI and TLCI for critical color work. Our example used very high quality LEDs, and still saw slight issues; lower quality LEDs (with CRI < 90 or strong green spikes, etc.) would produce far larger color errors and weird skin tone shifts, etc. Investing in good continuous lighting pays off. Also, some LED lights now come with profile calibration data or filters to improve their spectra. For instance, some LED manufacturers publish an SSI (Spectral Similarity Index) relative to standard sources – choosing an LED with a high SSI relative to D55 (daylight) would mean it’s extremely close to a flash in spectral content, minimizing differences.

5. Shoot a Known Neutral in Each Setup: As a practical workflow tip, always include a known neutral object in test shots for each lighting setup. This could be the grey card itself or something like a chrome sphere or white ping-pong ball. Then in post, check that item under each lighting. If you notice a cast (for example, in our case the grey card’s dark patch looks slightly blue under LED), you can apply a secondary color correction to that image to counteract it (like a slight color curve tweaking shadows vs highlights). While this is a bit of manual work, it can fine-tune the result. Essentially, this is doing a mini profile fix without a full chart – adjusting either the shadows color balance or highlights color balance separately until grey patches align. Some RAW converters allow calibrating two-point white balance (though not common in UI, one can use tone curves to adjust color in shadows vs highlights differently).

6. Lens and Filter Considerations: Ensure you are not using any lens filters or lighting gels that could introduce unexpected color shifts. For instance, some cheap UV filters have a slight yellow cast, or some softboxes might fluoresce under UV from flash. These can compound issues. In our tests, we kept those constant, but in real setups, consistency is key.

7. In-Camera WB vs Post WB: It’s generally preferable to shoot RAW and adjust white balance in post for tricky light. Some cameras (not sure about R5 specifically) have WB settings for “White Priority” vs “Ambient Priority” under LED, etc., but those are mostly JPEG tweaks. They won’t fix metameric issues. In post, you have more flexibility to apply custom profiles or corrections as described. The R5 does allow you to set a custom WB by shooting a grey target – but that, as we saw, only solves it at mid-grey and might leave residual cast in other tones. So post-production is where to address the fine details.

8. Avoid Overcorrecting in One Go: If you find yourself chasing neutrality by adjusting one patch, be mindful that you don’t swing another patch off balance. This often happens if one uses software like Lightroom’s eye-dropper on a bright part of an image vs a dark part – it can yield different overall WB results. Stick to the standard mid-grey for initial WB, then use selective adjustments if needed rather than globally white balancing off a dark patch (which might overcompensate and make highlights wrong).

9. Consider the Camera’s Color Mode: The Canon R5 has Picture Styles (for JPEG) and underlying color science that might have slight variations. For instance, Canon might tune their standard picture style to preserve a certain look which could include very subtle tonal curve differences between channels. In RAW, this shouldn’t apply unless the converter is mimicking it. But just to mention: sometimes people observed on Canon forums that the R5’s rendering of shadows had a slight tint compared to older models. If true, that could be a factor (though likely negligible here). If absolute neutrality is needed, using a “Neutral” picture style or a linear response curve in Capture One (as we did) is advisable to avoid baked-in biases.

10. Testing and Reference: If working on an important shoot, do a quick test of your lighting with a color chart and evaluate the histogram/RGB readouts on different patches. If you see something off, you have time to correct either by adjusting lights (e.g., adding a gel to LEDs to fill a spectral gap – sometimes adding a little incandescent kicker can fill in missing reds for LEDs), or by noting to fix in post.

In our scenario, applying these practices: we would likely profile the R5 for the Westcott LED light. That would definitively solve the grey patch issue – the software would create a profile where the grey scale is perfectly neutral. The trade-off is the profile is valid only for that light (if you change LED settings or brand, you’d redo it). Many high-end workflows (like art reproduction) treat this as essential – every lighting condition gets its own profile. It’s extra work, but it guarantees no surprises.

Conclusion

The analysis of RGB imbalance across grey patches under Profoto flash and Westcott bi-color LED panel lighting has shown that the root cause is fundamentally spectral in nature, manifesting as a metameric mismatch. When white balance is calibrated on a mid-grey patch:

• Under the flash’s broad-spectrum light, all grey patches (light to dark) remained largely neutral, indicating that both the grey card and the camera’s color processing handled the continuous spectrum as expected. Minor deviations (on the order of 1-2 RGB levels) were observed, which could be attributed to very slight pigment non-linearities, but these were negligible and not visually significant. Essentially, the flash and camera yielded a linear neutral response across different brightness levels, as one would hope.

• Under the LED panel’s discontinuous spectrum, the grey patches showed measurable and visible RGB imbalances (e.g., slight warm tint in light grey, slight cool tint in dark grey), even though the mid-grey was perfectly white-balanced. This demonstrated a case of capture metameric failure: the camera + card perceived neutrality differently for different patches under the LED illumination. The LED’s spectral power distribution, characterized by peaks and troughs, interacted with small spectral reflectance differences in the grey patches and the camera’s color filters, causing the one-size-fits-all white balance to falter at the extremes of the tonal range.

Investigating whether the grey card or the lighting was primarily at fault, we concluded that spectral variation of the lighting (and its interaction with the camera’s sensor) was the dominant factor. The grey card’s “imperfections” were minor on their own – under flash they didn’t pose a problem – but under the LED they were revealed. In other words, a truly perfect grey card might have reduced the issue, but an LED with a closer-to-full spectrum (or a camera profile tuned to that LED) would also have prevented the issue. It’s the combination that caused the imbalance.

The Canon EOS R5, with its high-end sensor, proved to be linear and consistent; its behavior highlighted that no matter how good the sensor, a tri-stimulus system will always be challenged by off-nominal spectra. The R5’s color filters and processing are optimized for natural light and tungsten; the LED, while ostensibly daylight-balanced, is not a perfect stand-in for sunlight as far as the sensor is concerned. Hence, we saw those small residual errors. This is not a flaw unique to the R5 – any camera would have some degree of this unless specifically characterized and corrected for that LED source.

On the theoretical side, we expanded on why LED vs flash causes metameric failures: narrow-band vs broad-band emission leads to situations where two spectra (light reflecting off two grey patches) that should appear identical differ when processed through the camera’s color perception. This underscores a general principle in color science: matching colors for one illuminant/observer does not guarantee a match for another illuminant/observer.

For practitioners, the key takeaways and recommendations include:

• Be aware that even with proper white balance, certain lighting can introduce subtle color casts in different tonal regions. Don’t assume neutrality is automatic across all levels – verify with test shots if it matters to your work.

• Whenever possible, use consistent, full-spectrum lighting for critical color work. If using LEDs for their many benefits (cool running, continuous preview, etc.), choose the highest quality and consider profiling your camera for them.

• Incorporate color management into your workflow: custom camera profiles and high-grade neutral references will greatly mitigate issues of this kind. In a controlled studio scenario, these are feasible steps that pay off in color accuracy.

• Understand that what your eyes see and what the camera captures can differ under certain lights. You might look at the grey card under LED and see it as grey, but the camera might pick up a slight cast. Trust the camera’s measurements and adjust accordingly (rather than assuming it “must be neutral” because it looks okay to your eye).

In conclusion, the RGB imbalance observed in grey patches is not an aberration but rather an illustration of the complex interplay between lighting spectra, object reflectance, and camera color response. By identifying the cause as primarily metameric in nature, we can address it with targeted solutions. Studio photographers using equipment like the Profoto flash, Westcott Flex LEDs, and Canon R5 can achieve consistent, neutral results by applying the best practices discussed – ensuring that every shade of grey (and indeed all colors) reproduces as intended, regardless of the lighting technology employed.

References (APA Style)

BasICColor. (n.d.). basICColor Grey Card – Grey card free of metameric effects. Retrieved from https://colourmanagement.net/products/basiccolor/basiccolor-grey-card/

Capture One. (2025). Metamerism [Support article]. Capture One Support. Retrieved from https://support.captureone.com/hc/en-us/articles/360002653437-Metamerism

Hamamatsu Photonics. (n.d.). Xenon Flash Lamps – Spectral distribution of xenon flash lamps. Retrieved from https://www.hamamatsu.com

Kasson, J. (2015, December 26). Metameric failure – The last word [Blog post]. Retrieved from https://blog.kasson.com/the-last-word/metameric-failure/

Myers, R. D. (2010). ColorChecker Passport Technical Report (Revision 3). RM Imaging. (Discussion of spectral reflectance of ColorChecker grey patches.)

Phase One. (2017). DT Cultural Heritage – Color Reproduction Guide (Techniques for Flash vs LED profiling). Phase One A/S.

Videomaker. (2016). Understanding CRI & TLCI: The importance of color rendition. Videomaker Magazine. (Discusses full vs broken spectrum light sources and their effect on color accuracy.)

X-Rite, Inc. (2019). ColorChecker Passport Photo User Guide. (Includes information on the spectrally neutral white balance target vs standard ColorChecker neutrals.)

Academy of Motion Picture Arts and Sciences. (2012). Solid State Lighting Project Report. Sci-Tech Council, AMPAS. (Investigates LED vs traditional lighting spectra and color rendering for imaging.)

DXOMark. (2020). Canon EOS R5 Sensor Measurement. DXOMark. (Contains color sensitivity and SMI data for the R5 sensor.)