Why a kelvin difference is not a filter shift
The reciprocal colour-temperature scale expresses mired = 1,000,000 ÷ CCT in kelvin. Equal changes on this scale do not correspond to equal kelvin differences. For example, a shift from 5000 K to 4000 K is 200 to 250 mired, giving +50 mired. The same +50 mired applied to 3200 K gives approximately 2758.62 K, rather than a further 1000 K reduction.
For a source and target, the signed shift is Δmired = 1,000,000/target K − 1,000,000/source K. A positive shift gives a lower nominal CCT; a negative shift gives a higher nominal CCT. A 3200 K to 5600 K change gives −133.928571… mired, displayed as −133.93 mired.
CCT means correlated colour temperature, a description of light colour relative to a reference locus. It is not the physical temperature of the LED, lamp housing or filter.
Applying a nominal manufacturer shift
With a known source CCT and supplied nominal shift, the inverse arithmetic is target K = 1,000,000 ÷ (source mired + shift). The denominator must remain positive. At 5000 K the source is 200 mired: a shift of −200 would give a zero denominator, not a finite temperature; a more negative value would not give a valid positive kelvin result.
LEE’s colour-temperature tool and filter records illustrate why nominal values and source conditions matter. The 204 Full CTO record describes 6500 K to 3200 K with a nominal +159 mired shift. The exact reciprocal difference between those two stated temperatures is approximately +158.653846 mired. Applying exactly +159 mired to exactly 6500 K gives approximately 3196.46 K. The small discrepancy follows from rounded product values; the calculation does not force them to agree.
Internal calculations use the entered values at full available precision. The displayed mired shift is not rounded and then fed back into the inverse calculation. Ordinary displayed values use two decimal places, so a product specification’s own measurement uncertainty and rounding still matter.
A reciprocal shift does not describe the whole spectrum
A source CCT and nominal filter shift provide reciprocal-temperature arithmetic. They do not supply the source’s spectral power distribution or the filter’s wavelength-dependent transmission. An arbitrary LED and filter combination can therefore differ from a result obtained under the manufacturer’s stated source conditions.
Green–magenta tint, CRI and other spectral properties are separate from this one-dimensional change. Averaging two lights’ kelvin values, or averaging their mired values, does not predict the colour of their mixture. A measured spectrum and an appropriate colourimetric method are needed for that different task. If a filter also has neutral-density loss, the Exposure and ND Calculator can apply a separately supplied attenuation to the camera settings; mired shift alone does not specify that attenuation.