How much Delta E* is visible?

Quality Control is all about numbers.
As technology becomes more complex, tolerances become tighter and harder to manage.

Old School methods would take the Mass of an object, say a 500g block of butter, and set the tolerance to plus-minus 5 g. As long as a batch sample mass was between 495 to 505 g it was compliant. Colour on the other hand does not work in the same way as a property like Mass or Size.

Colour is not a property of the object, but the perception of colour under set conditions. As the conditions change, say the light source, the colour will be perceived to have altered.
Has the object actually changed colour? Well, yes to the observer it has, but only because the conditions of observation were changed. If we altered another condition, say the distance from the sample, we will again notice that the colour is perceived to have changed.

Judging colour quality by a standard “Upper or Lower Limit” is not suitable. After a certain variance in perception values, i.e. L*a*b, a standard C.I.E “Colour Space”, the colour is now simply another colour.

Customers often ask us to help with tolerances settings.

Typically, we are not in a position to decide what their customer, business, or consumer would accept. It very much depends on the customer’s priorities.

For instance, the Automotive Manufacturing sector has the best tools and the tightest colour quality specifications, yet none of our Auto Manufacturer customers has ever lost a customer due to poor colour matching. Customers expect the match to be perfect so don’t even look, and if they see some anomalies, are likely to assume they will be fixed later.

The Auto Refinisher, or panel beater as we may say, is in the opposite position. Using the weakest technology was at the same time the customer is at their most critical. Auto Insurers constantly report that customers reject a repair job on colour mismatch because now they are really looking closely at getting back to 100% perfection.

Take food as another example. Incoming ingredients, process control and final quality inspection are not really important to the customer UNLESS one batch has a number of variables, suggesting that somewhere in the process, the process failed. Even artisan bread need to comply with a “Regular irregularity” to be acceptable.

Each stage in each company has a seller and a buyer, either internal or external and these stages need to meet logical standards. If an ingredient has a 5% error and is going to be diluted 10 times, the error will be reduced to 0.5% or less in the final product.

Consistency, on the other hand, is critical, as demonstrated in the image below:

The object above is a Tolerance setting, all of the samples, Green, Blue, and Red are all within the red tolerance target.

However, the Green samples are all close together, and perhaps the blue sample is on the boundary, but the Red samples are nearly double the required tolerance from the Green samples. In this case, these outliers will appear to look different, even if within the overall target acceptance value. Batching, therefore, is another valuable aspect of colour quality control.

To conclude, if we take a popular Food CIE Colour Space, L*a*b, we can say as far as visual perception is, the following.

The CIE L*a*b colour space includes a Δ (Delta or difference value) which is the average difference of the *ab values.

  • L = the degree of whiteness to darkness, ignoring colour per se. 0% = Black and 100% = White.
    Our eyes are less sensitive to the colour lightness or darkness and up to a CIE L value difference of 1 may be hard to discern, especially in very bright Yellow and Red shades.
  • a + (Red) and *a – (Green) can perceive to change from 0.3 difference. If it’s redder in a yellow, or greener in a blue, this is more noticeable.
  • b + (Yellow) and B – (Blue) are also more noticeable in contrast colours and 0.3 may be visible to the consumer’s eye.

If Consumers expect apples to vary from blue-red to scarlet red then wider tolerances will be acceptable, bearing in mind that after a difference of Δ*ab of 3 will actually be quite different. Acceptable in Jam, but not your new suit.