Delta CS64A - Manual (Page 19)

19
color sensor will require periodic calibration checks, this can typically be done at widely
spaced intervals with the CS64A. The drift of demo units is not representative of the
performance of production units.
Q. What are the limitations to the detection of colors?
A. The measurement consistency varies from application to application. Here are the
various error sources associated with any color sensor, including the CS64A.
Sample color variations: Often colors vary from sample to sample. Dark objects tend to
display the largest amount of variations. If possible, check multiple samples over time to
verify consistency before setting up the thresholds.
Presentation variations: The largest variation in multiple readings of the same object is
typically differences in the orientation of the object to the sensor. Not only does the
distance from the object to the sensor effect the readings, but so does the angle of the
object. The least sensitive objects are those with a matte or other non-glossy finish and
with a flat to slightly convex surface. The most difficult objects have a gloss finish and a
concave surface. The CS64A-00 handles these types of variation better than most color
sensors, in fact spectrophotometers often require contact measurements which eliminate
distance variations but also make them unsuitable for most on-line applications. Small
sensor-to-object variations (~1/4") typically only cause a few tenths of a percent variation
in ratios of the CS64A's readings--see spec sheet for worst case specs. Angle variations
are not specified and are dependent on glossiness of the object.
Environment: Dust collecting on the CS64A's window reduces signal strength. Use of
ratios eliminates errors from uniformly distributed dust, but eventually the loss of signal
strength will impact performance.
Noise: Electrical and optical noise is typically the next largest variation from reading-to-
reading of the same object. Noise is greater for green and blue since the green and blue
LEDs are less intense, and less for NIR and red. Optical noise is primarily caused by
changing ambient light conditions. Extraneous light in the sensing area is an error signal
and should be eliminated as much as possible. Low levels of constant light may be
tolerable, while sunlight is almost never tolerable. For lowest electrical noise, we
recommend connecting the CS64A analog outputs to differential inputs using shielded
twisted pair wires and a dedicated power supply, which allows the circuit common, and
case ground to be connected inside the sensor. This noise can also be effectively
eliminated in many applications simply by averaging multiple readings.
Temperature: Output voltage variation with temperature is specified and 100% tested.
The LED color also drifts slightly with temperature, but this is insignificant in most
applications.
Time: There is a certain amount of reading variation over time, typically a few tenths of
a percent per week or less. This can be dealt with by recalibrating more often in critical
applications.