How to set your tolerance levels for master colour targets or standards?

Customers new to colour measurement often get confused with the “Target” and “Sample” procedures.
This blog is a quick guide as to why and how you set these Target Tolerances.

Firstly, QA labs are used to “Upper and Lower” limits. These apply to properties of the object like Mass, Dimensions and other fixed properties.
Colour however is a property of the Observation (And its properties), not a fixed property of the object, so the normal procedures have to be modified a bit.

Colours don’t have Upper and Lower Limits. Either the colour is the same, or different.
From a human visual point of view, the point at which a colour becomes different, is when the observer can visually discern a difference. As people differ both in visual perception and the accuracy of visual perception, there is no fixed rule to set a tolerance value.

How do instruments describe a colour?

To better describe a colour, the C.I.E have over the years developed standard Colour Space Algorithms which describe colour in more detail, using mathematics. A typical Colour Space we use is the L*a*b colour space.

L denotes lightness or darkness
+a* denotes REDNESS and -a* denotes GREENESS
+b* denotes YELLOWNESS and -b* denotes BLUENESS

These three values can be used to accurately describe true colour.
The three dimensional data is like a GPS co-ordinate with longitude, latitude and altitude and is often depicted as a sphere.

NOTE: Click on IMAGES to enlarge them

L values are between 0 (Black) and 100 (White).

The + or – * a or *b values denote the intensity of the colour.
The image depicted left may not apply to all COLOURANTS as they may not be able to cover all known colours, so this depiction is accurate subject to the actual colourant’s being used at the time.

The L*a*b algorithm can also be used to compare two colours very precisely, using the mathematical difference (Δ) values. For instance a standard Target to Sample comparison renders not only the L*a*b values, but also the ΔL*a*b values. In this manner not only can you tell that a colour is different, but exactly why.

Setting Tolerances for Master Colour Targets?

As a practical example, we will now compare two yellow ceramic tiles which are similar but not identical in hue.
The Larger Tile on the left will be used as the “TARGET” colour.

1. Measure the desired object as a TARGET:

 

The software records the Target L*a*b values, but as yet, no Tolerance has been set.
This will be decided after further measurements, but initially a DEFAULT Tolerance will be used which is discernible by the trained eye easily, and also by the casual observer if they look more closely.

This default tolerance is 0.8 to – 0.8 (Total 1.6) for the L value, *a value and the *b value.
There is also an option to add an AVERAGE dE*ab target.

2. Find and change the Tolerances in the Tolerance Setting menu:

By Right-Clicking on the selected Target, a menu will appear the allows you to open the Tolerance Settings. Once you open that menu the screen on the left will appear.

This allows you to change your tolerance settings for that specific product.

Note: Typically all four settings are selected
You can have different values for each setting should you wish depending on which parameters are most important to you. Some customers might place a higher weighting on the lightness and darkness (L*) of the colour while other might place more weighting on the actual colour differences.

After clicking on APPLY in the bottom right corner, the target tolerance settings are enabled.

3. Note the tolerance settings applied to the Target.

After setting the default tolerance, tolerance boxes appear on the target box for *ab values and on the L value box. This box defines the area that a sample needs to fall into inorder to PASS or meet the specifications and tolerances of this Target.

No Judgement (PASS/FAIL) has taken place as no sample has been measured as of yet. We therefore have not compared the sample to the Target.

4. Measure the SAMPLES against the defined TARGET.

Sample #1 is a measurement of the Large Yellow Target tile.

As this was used as the TARGET,  you would expect that this sample passes with very close readings to the original. (There are always some slight variations in the decimal part of the reading due to fluctuations in temperature etc.) These are not considered to be a major difference in colour.

Sample #2 is a reading of the Small Yellow Sample tile.
Here it is clear that this tile is out of the default tolerance setting and it is judged as FAIL.

At this point you can make one of two judgements. These will be based on YOUR requirements and your customer expectations, and at this stage we can not tell you what you and your customers prefer.
You can either say the Target is not negotiable and sample #2 has FAILED or alternatively say that the appearance of Sample #2 is still acceptable for you and your customer, and you can increase the tolerance to allow the smaller tile to pass.

5. Changing the Tolerance levels after accepting that the two tiles are close enough that they should PASS.

By returning to the Target and Right-Clicking on it, you can define the new Tolerance values so that sample #2 would be considered a a PASS.

After Clicking APPLY, these new settings are enabled. In this example we have increased the Tolerance by 100% or have doubled the tolerance levels.

6. Repeat the process until you find the point where visually noticeable colour differences are rejected but samples without any visual difference will PASS. Repeat until you and your customers can agree on a tolerance level.

In this case the Default Tolerance was doubled, but sample # 3 (The small yellow tile) is now closer to passing but the overall judgement is a FAIL. The Tolerance Boxes also indicate WHY the colour still fails by comparing what the actual differences are between colours.

The tolerance can now be further widened if the shade is acceptable and the judgement is not, or it can be judged FAIL.

By repeating the procedure, and widening the tolerances further, it is now possible to have a tolerance acceptable to the customer, and repeatable with further measurements.

7. Confirm Tolerance Settings and Define the new default tolerances.

In this case with Sample #4 (The small yellow tile) the measurement passes and is acceptable to both supplier and customer as “Close Enough” for the purpose of this product / colour standard.

These last settings will be used as your normal default settings.

By this process, you come to the tolerance standards that work for you.

It is possible to have different tolerances for the L, *a and *b values, as well as the Average *ab values should one value be critical, while the others less so. In all cases we use the Δ values to compare, not the absolute values.

Caution: If tolerances are increased, the batch to batch variance will also increase.