-e Menn et aı. 
SVP-BRST Fiducial Reference Network 
1.0 
9,0 
7,0 
- +— 
DD 632% =6" 
» 50 +— 
3,0 
—*— Measured (T - TO) 
—u— Theoritical (T - TO) 
N 
1.0 6% 
——— 17 
16,09— —m——PR0— — —{— —A40— —— —m—— — — — 16,0 
t(s) 
FIGURE 5 | Determination of the response time of MoSens module n° 4656 
The uncertainty of the bath stability and homogeneity. 
According to the standard deviations of the measurement 
;eries and to homogeneity measurements made on the 
bath, the standard uncertainty on the bath stability and 
homogeneity pay can be assessed to be 0.3 mK. 
The reproducibility S and repeatability Sep of MoSens sensors 
measurements. The reproducibility is evaluated according to 
the standard ISO 5725-2 (1994), by calculating the variance of 
the deviations obtained during the verifications of calibrations. 
or the prototypes of MoSens n° 4656 and n° 4658, this gives, 
respectively, S = 1.7 mK and S = 0.9 mK. The repeatability 
Srep Can be assessed by calculating the average of the standard 
deviations of the temperatures measured by MoSens sensors. 
For the two sensors Syep = 0.3 mK. This repeatability is 
strongly correlated to the bath stability. 
According to this budget, the model used to calculate the 
combined uncertainty on the deviations D is: 
D=T+ Öryep + Öyeprod - Tyef + Öpath (12) 
In this equation, T is the average of the series of temperature 
values given by the sensor under calibration, T,ef is the average 
reference temperature, ö,ep is the short term variation of the 
sensor temperature, ö,epod Is the long term variation of the sensor 
temperature and Öpg:4 Is the difference in temperature due to the 
stability and the homogeneity of the bath which introduce small 
errors between T,.f and T at the time of measurements. Applying 
the GUM method (BIPM, 2008) to relation (12) and assuming a 
correlation coefficient of 1 between ö,ep and Spa yields: 
un — Uryef + Sep +t S +t Zn + 2Upath Srep (13) 
The expanded uncertainty (Uc) on the deviations obtained 
during the calibration can be calculated by the relation: 
Uc=2 | Uhef + (Syep tHpay) + S? (14) 
Table 2 shows the uncertainty budget and the results of relation 
(14) for the two buoys. For the n° 4656, Uc = 4.0 mK, and for the 
n° 4658. Uc = 2.8 mK. 
trontiers in Marine Science | www.frontiersin.or 
Measurement of the Response Time of the 
MoSens Module 
As the MoSens sampling rate is only 1 s, it has been necessary to 
fix the initial deviation Tsw - To of Equation (9), close to 10°C, 
to allow the assessment of t. The rt value has been calculated (see 
Figure 5) and it gives 0.200 s°C 7! for the n° 4656 instrument and 
0.206 s°C7! for the n° 4658 instrument in nearly static exchange 
conditions. The time to obtain 99.99% of the final response can 
be calculated with the relation fog. 99 = t In(1 - 0.9999)/1000, and 
it gives about 1.85s. The graphical determination of the time to 
obtain 63.2% of the final response is made with an uncertainty 
close to 75 ms. It gives a maximum uncertainty value on t of 
17 ms°C71, 
The measured values of t are about 2.5 times the theoretical 
value given in Table1 in low flow speed conditions and for 
the HRSST sensor alone. This can be explained by taking 
into account the heat exchange of MoSens module, in its 
PVC housing, with the water by convection and radiation, and 
with the sensor by conduction. This hypothesis is reinforced 
by drawing the theoretical response curve of HRSST sensor 
{rom the relation (9), in which tT7 would be equal to 0.2s 
(Figure 5). It appears that the slope of the temperature increase 
is more important that the measured slope. The response 
of MoSens module cannot be represented by the simple 
relation (9) and it doesn’t represent exactly the response of 
RSST sensor. 
The PVC housing and the self-heating of the electronic board 
of MoSens modules have another effect. It takes longer to reach 
the final temperature to within +2 mK. The time to obtain 
T- To = 2 mK is close to 35 min when Tsw - To is close 
to 10°C. This time is not representative of the response time 
at sea because the MoSens devices are integrated in the buoys 
without their PVC cylindrical housing and it is always at a 
temperature that is relatively stable or slowly changing. In other 
words, the operation carried out here to estimate the sensor 
response time is sub-optimal, and would need repeating with 
the sensor integrated in the buoy, but this would pose other 
practical issues. 
Qantembear 2019 I Valııme A 1 Article R7£