Professional Documents
Culture Documents
AN 5361
Rev. 5
MilkoScan™ Mars
Milk Calibrations
The Milk application is suitable for cow,
buffalo, sheep and goat raw milk and processed
homogenised milk.
It is expected that a slope and intercept
adjustment is needed for each of the different
milk types you want to analyse, for example one
adjustment for raw milk and another for
pasteurised milk.
Fat, Protein, Lactose, TS, Solids non Fat and
Freezing Point are important quality parameters
in milk. The analysis may be used for payment,
calculation of the mass balance, for
standardisations purposes as well as to verify
the end products.
Calibration samples have been collected from various parts of the world to build a robust
calibration with measurement values shown in the above table.
Lactose
Raw milk contains lactose as the only carbohydrate and the calibration model is more correctly a
"Total Carbohydrate" and not a "true lactose" calibration. This means when a matrix of
carbohydrates is present in the milk such as in low lactose products with a combination of lactose,
glucose and galactose the calibration model is not able to distinguish only the lactose sand will
include all carbohydrates.
The conclusion is that the calibration in raw milk for lactose is based on the principle that lactose is
the only carbohydrate present. The calibration model will not predict correctly when other
carbohydrates are found in the milk. This means it is not possible to measure lactose in low lactose
products.
Table 2 Validated accuracy, repeatability, and correlation
The absolute accuracy, Acc.(abs), is expressed as Standard Error of Prediction (SEP). The absolute
repeatability, Rep (abs) is repeatability expressed as pooled standard deviation of replicates. The
relative accuracy, Acc. (rel) is expressed as Acc.(abs) relative to the mean of the reference values.
The relative repeatability Rep.(rel) is expressed as Rep.(abs) relative to the mean of the reference
values.
The accuracy of individual validation sample sets will depend on sample handling, the reference
method used and the range of the components found in the set. For example, the accuracy for Fat
validation will be much better if the reference method is based on Röse Gottlieb compared to a
Gerber method.
Note: The performance examples outlined in this note should be regarded
as a guideline only for the expected performance of each installation and
not considered an instrument specification.
7
10
6
Protein (Reference)
Fat (Reference)
8
5
6 4
3
4
2
2
1
0 0
0 2 4 6 8 10 12 0 1 2 3 4 5 6 7 8
Fat (MilkoScan Mars) Protein (MilkoScan Mars)
11 20
SNF (Reference)
TS (Reference)
10 18
16
9
14
8
12
7
10
6 8
6 7 8 9 10 11 12 13 8 10 12 14 16 18 20 22 24
SNF (MilkoScan Mars) TS (MilkoScan Mars)
5.5
-0.35
Lactose (Reference)
FP (Reference)
5
-0.4
4.5
-0.45
4
-0.5
3.5
3 -0.55
3 3.5 4 4.5 5 5.5 6 -0.55 -0.5 -0.45 -0.4 -0.35 -0.3
Lactose (MilkoScan Mars) FP (MilkoScan Mars)
Fig. 1