Online proceedings - EDA Publishing Association

24-26 September 2008, Rome, Italychamber box, air temperature can not be changed as fast asin the center of the box where the air rotates. The greater isthe difference between room temperature and test chambertemperature, the greater is the error.Another difficult problem is the dissipation of the cellsunder great current loads. Air counts as isolator on theviewpoint of temperature, and if the cool air is not rotatedproperly in the box, working battery cells can heat up the airlocally around them, and make the measurement false. Thecooler is the air in the chamber, and the greater is the loadcurrent for the cells, the greater the problem is. At greatercurrents the self-heating of the cells can reach such anextent, that the rotating air can not carry the energy awayfast enough, and the inside temperature of the cell increasessignificantly. Also the pulsing of the air motion gives afailure in the measurement, as the cell’s output voltageswings, showing the cell's great dependence on temperature.Figure 2. A 1500mAh Li-Po battery under different temperature conditionsTo decrease these effects, we applied the followingrecommendations. Cells had been placed inside the chamberas far as it could be reached, to minimize the coherencebetween them during the measurements. Coolers had beeninstalled at both sides of the flat cells to increase the heattransmission, and we improved heat pass between the coolerand the cell with heat pasta. This way the batteries couldtransmit their heat faster, and inside temperature increasedue to the big discharge currents could be minimized.Additional ventilators have also been set up inside thechamber to raise the speed of the rotated air.V. EVALUATION OF THE RESULTSConstant current discharge measurements have beencompleted during different regulated temperature conditions.Measurement results show significant dependence onambient temperature. Mostly the offset of the curvesdecreases with the cooling of the ambient. Curves shift lowerwhen a lower temperature is set. As we mentioned earlier,the middle and the final transient periods have got a sharplimit point. The beginning of the transient period is afunction of the loaded output voltage. If it reaches a value(mostly near 3.6V), the output voltage starts to decreasefaster until the cell gets fully discharged. At lowertemperatures, curves shift lower but this limit does not varyas much. This implies that the limit is reached earlier and thebattery cells get discharged earlier in time, as it happens athigher temperatures (e.g. at room temperature). Fig. 2 showsthe limit points of the middle and final transient phases fortwo different temperatures under the same current load.VI.CREATING MODEL FROM THE DATATo evaluate and handle the data fast, we have chosen anappropriate function and fit all the measured curves to it.This function has to follow the curve's shape at the rangefrom the first measured point until the last one, but whathappens outside the range is of no importance (we don'tcalculate with that part). As we described earlier, there arethree different parts in each discharge curves, but at leasttwo, which could be exactly separated, and the chosenapproach function has to contain these discrete parts.We chose the following function:A 2f ( x)= + + Cx + D ⋅ x + E(1)B + xwhich consists of a sum of a hyperbola and a quadraticfunction. By the fit, hyperbola represents the final transientof the cell at the end of the curve, and the positive coefficientquadratic function's plays role by the sweepness of the curveat the initial transient and the slope of the middle part.With this form, flat initial transient and more peaked,steeper final transient curves can be fitted, which is passingto the rest of the curves given by the results of ourmeasurements.This form is beneficial because it consists of only 5parameters, which means that fitting a whole constantcurrent discharge curve results in 5 real numbers thatrepresent the fitted measurement. Figure 3. shows ameasured curve and the fitted function plotted in onediagram to demonstrate, that the error is negligible. Table 1.shows the parameters yielded by the fitting algorithm.©EDA Publishing/THERMINIC 2008 130ISBN: 978-2-35500-008-9