application note AN257/0189 of the multiwatt package by r. tiziani thermal characteristics introduction this application note provides a complete thermal characterization of the multiwatt ? package (multi- lead double to-220 - fig. 1). characterization is performed according with reco- mandations included in the g32-86 semi guideline, by means of a dedicated test pattern. it refers to : 1. junction to case thermal resistance r th(j-c) 2. junction to ambient thermal resistance r th(j-a) 3. junction to ambient thermal impedance for sin- gle pulses and repated pulses, with different pulse width and duty cycle ; 4. thermal resistance in dc and pulsed conditions, with a typical external heat sink. most of the experimental work is related to the ther- mal impedance, as required by the increasing use of switching techniques. experimental conditions the thermal evaluation was performed by means of the test pattern p432, which is a 20k mils 2 die with a dissipating element formed by two transistors wor- king in parallel and one sensing diode. in order to characterize the worst case of a high power density ic, the total size of the element is 2k mils 2 with a po- wer capability of 20w. measurement method is de- scribed in appendix a. samples with the indicated characteristicswere pre- pared : package multiwatt 15 leads measurement of junction to case thermal resistance r th(j-c) is performed by holding the package against a water cooled heat sink, according with fig. 2. a thermocouple placed in contact with the slug mea- sures the reference temperature of the case. for junction to ambient thermalresistancer th(j-a) the samples are suspended horizontally in a one cubic foot box, to prevent drafts. both dc and pulsed conditions are used ; in the se- cond case the contribution of package thermal ca- pacitance is effective and transient thermal resistances much lower than the steady stata r th(j-a) can be found, according to pulse length and duty cycle. the effect of the external heat sink is quantified, using as test vehiclethe commercially available heat sink thm7023 especially developed by thermalloy for the multiwatt package, whose thermal resistance in still air is about 9 c/w. the measurement circuit shown in fig. a3 was used for all the thermal evaluations. junction to case thermal resis- tance the dependance of r th(j-c) on the dissipated power is reported in fig. 3. it is well known that the main contribution to r th(j-c) of power packages in given by the silicon die. figure 1 : multiwatt assembly. frame material copper slug thickness 1.5mm slug thermal conductivity 3.9w/cm c die attach soft (pbsn) 1/9
figure 2 : measurement of r th (j-c) . for devices other than the test pat- tern p432 the calibration curve of fig. 4 is needed. it shows the relationship between r th(j-c) and the dis- sipating area existing on the silicon die (power dio- des, power transistors, high current resistors), for different die sizes. junction to ambient thermal resistance in medium power applications (1.5-2w), the multi- watt package can be used without external heat sink, thanks to the significant size (about 3.5cm 2 )of its integrated thermal mass. its r th(j-a) has two contributions : the r th(j-c) , mainly due to the silicon die (as shown in fig. 4) and the ther- mal resistance of the copper slug r th slug . figure 3 : r th (j-c ) of multiwatt package vs. power level. figure 4 : r th(j-c) thermal resistance vs. die size and on die dissipating area. figure 5 : r th(j-a) of multiwatt package vs. dissipa- ted power. application note 2/9
fig. 5 gives the relationship between r th(j-a) and the power dissipation level for the p432 test pattern is still air, on pc board and on a commercial heat sink. in order to have an accurate value for other devices, with different die size and dissipating area, values of fig. 5 should be corrected through the cali- bration curve of fig. 4 correction term is always in the range of 0-2 c/w ; therefore, it affects the r th(j-a) of no more than 5% in still air or with the package mounted on pc board. transient thermal resistance in pulsed condition (without external heat sink) the effect of single pulses of different length and height for the multiwatt package without any external heat sink is shown in fig. 6. this behaviour is discussed in appendix b. due to a significant thermal capacitance (c = 2j/ c) and correspondingly long risetime ( t = 80s), single pul- ses up to 30w can be delivered to the multiwatt pac- kage for 1s with acceptable junction temperature increase. in order to have accurate r th (t o for other devices, with different die size and dissipating area, values of fig. 7 must be corrected as described in example 2 of the last section. repetition of pulses with definedp d , period and duty cycle dc (ratio betwen pulse length and signal pe- riod), gives rise to oscillations in junction tempera- ture as described in appendix b. the transient thermal resistance corresponding to the upper limit of the curve of fig. b4 (peak transient thermal resistance) is reported in fig. 7 and depends on pulse length and duty cycle. it can be noticed that dc becomes less effective for longer pulses. transient thermal resistance in pulsed condition (with external heat sink) characterization has been repeatedwith a commer- cial heat sink (thermalloy thm7023) in order to have an example of the effect of an external thermal mass on the impedance of the thermal module. relationship between transient r th and pulse length is reported in fig. 8. the effect of the increased thermal capacitance is evident in fig. 9, where thermal data of fig. 6 and 8 are compared : it can be noticed that the curves are definitely different for pulses longer than 1s, corres- ponding about to the rise time of the slug. the effect of the thermal mass is to keep low the heating rate of the silicon die thus allowing a better power ma- nagement of long power pulses. this conclusion has generalvalidity and can be applied to other heat sinks than the one considered in this note. figure 6 : transient thermal resistance for single pulse. figure 7 : peak transient r th vs. pulse width and duty cycle. application note 3/9
figure 8 : transient r th for single pulses, with heatsink. figure 9 : comparison of transient r th for single pulses, with and without heat sink. appendix a the thermal resistance evaluation is performed with the especially designed test chip p432 which has two bipolar power transistor and one sensing diode (fig. a1). the active area is about 2000 mils 2 on a 35000mils 2 chip. its lay-out was optimized in order to have a uniform temperature area, once the two transistor are powered ; the sensing diode is placed at the center of this area. the relationship between the forward voltage v f of the diode at the constant current of 100 m a and the temperatureis shownin fig. a2. the curve calibrates the junction temperature throughthe voltage drop of the diode. the measurement circuit is shown in fig. a3. a sto- rage oscilloscope or a fast digital voltmeter can be used for recording the v f value. figure a1 : test pattern p432 lay-out. application note 4/9
figure a2 : calibration curse (sensing diode). figure a3 : measurement circuits. appendix b - thermal management in pulsed condition thermal resistance and capacitance the electrical equivalent of heat dissipation, for a thermal module formed by the active device with its package and the external heat sink is shown in fig. b1. to each cell of the thermal chain are associated a value of thermal resistance r th ( c/w) and a value of thermal capacitance c th (j/ c). the former in- forms about temperature increase due to the ele- ment represented by the cell ; as, in the example under consideration, heat transfer is mainly based on conduction for the silicon, the copper integrated heat sink and to metallic body of the external heat sink r th can be calculated from the relationship : g kxs r th = where k is the thermal conductivity of the material, the length of the conductive path and s its section. thermal capacitancec th is the capability of heat ac- cumulation ; it depends on the specific heat of the material and on the volume effectively interested by heat exchange (this means that the parts which are not heated during heat dissipation do not contri- bute to thermal capacitance). thermal capacitance is given by : c th =dxc t xv where d is the density of the material, c t its specific heat and v the volume interested to heat accumu- lation. the last element of the network, assumed as purely resistive, is due to convectionand radiation from the external heat sink towards the ambient. application note 5/9
each cell has its own risetime t , given by the pro- duct of the thermal resistance and capacitance : t =r th xc th the value of the time constant determines whether a cell approachesequilibrium rapidly of slowly : if r th or c th increases, equilibrium is reached at a slower rate. the following relationship is valid for each cell : d t=r th xp d [1 - e -t/r ](1) typical values of r th ,c th and t for multiwatt appli- cation are shown in fig. b1. when power is switched on, temperature increase is ruled by subsequent charging of thermal capaci- tances while the value reached in the steady state depends on thermal resistances only. qualitative behaviour of the network of fig. b1 is shown in fig. b2. figure b2 : qualitative t j increase (network of fig. b1) for repeated power pulse. figure b1 : electrical equivalent of multiwatt package mounted on the external heatsink. application note 6/9
single power pulse when the pulse length has an assignedvalue, effec- tive t j can be significantly lower than the steady sta- te t j (fig. b3.). figure b3 : effect of a single power pulse. for any pulse length t o , a transient thermal resis- tance r th (t o ) is defined, from the ratio between the junction temperature at the end of the pulse and the dissipated power. obviously, for shorted pulses, r th (t o ) is lower and a higher power can be dissipated, without exceeding the maximum junction tempera- ture t j-max allowed to the ic from reliability consid- erations. fig. 7 and 9 of this application note give experimental values of r th (t o ) for the two cases of the multiwatt package without and with external heat sink. repeated pulses when pulses of the same height p d are repeated with a defined duty cycle dc and pulse length is short in comparison with the total risetime of the sy- stem (many tens of seconds) the train of pulses is seen by the system as a continuous source, at a mean power level of p davg =p d xdc the average temperature increase is : d t avg =r th xp davg =r th xp d xdc on the other hand, the silicon die ( t s = 1 3ms) is able to follow frequencies of some khz and junction temperature oscillates about the average as quali- tatively shown in fig. b4. the thermal resistance corresponding to the peak of the oscillation at equilibrium (peakthermalresis-tan- ce r th peak ) is now given by fig. 5, and can be obtai- ned if pulse length and duty cycle are known ; p dmax is derived from the same figure. figure b4 : junction temperature increase for repeated pulses. application examples example 1 - maximum p d for single pulse of assigned length problem : define the maximum p d for a single pul- se with a length of 20ms in the case of multiwatt pac- kage used without heat sink. ambient temperature is 50 c ; maximum temperature is 130 c. die size is 20k mils 2 , with dissipating area of 2k mils 2 (as in p432 test pattern). solution : allowed temperature increase d tis 80 c. having a r th(j-a) of 39 c/w, multiwatt package can dissipate about 2w in steady state. from fig. 7 the transient thermal resistance corresponding to one single pulse of 20ms is r th (20ms) p432 = 2.2 c/w. a peak of 80/2.2 = 36.3w can be applied to the circuit. example 2 - correction for die size and dissipating area problem : correct the results obtained in exam- ple 1, for assigned die size and dissipating area. practical case : ic having a die size of 35k mils 2 with a dissipating area of 20k mils 2 . solution : from fig. 5, thermal resistances of p432 and of the ic under consideration are r th p432 = 2.3 c/w and r th(j-c)ic = 1.2 c/w. application note 7/9
as the length of the pulse is 10-15 times longer than the rise time of the silicon, the die (first cell of fig. b1) can be assumed to have reached its equilibrium condition. r th (20ms) found in previous example has to be cor- rected in order to take into account the new value of r th(j-c) : r th (20ms) ic =r th (20ms) p432 - -r th(j-c)p432 +r th(j-c)ic = = 2.2 - 2.3 + 1.2 c/w = 1.1 c/w a single pulse of 80/1.1 @ 72w can be delivered to such device. when the pulse has the same order of silicon rise time t p432 is about 1ms) another type of correction is needed.in first approximation, t increase with dis- sipating area with the relationship : t ic = 20k ic/ 2k p432 x t p432 3.1ms expansionof the exponentialterm ofrelationship (1) limited to the first term term, is : r th ic (t o ) @ r th p432 (t o )/3.1 for t o =1ms: r th ic (1ms) = 1.05/3.1 c/w @ 0.34 c/w a single pulse of 80/0.34 @ 235w can be delivered to such device. example 3 r th with repeated pulses problem : find the peak power which can be dis- sipated by multiwatt package without heatsink, when power is continuously switched on 10ms and switched off 90ms. ambient temperature is 50 c, maximum temperature is allowed to be 125 c. solution : a maximum d t=75 c has to be con- sidered. fig. 5 indicatedthatfor a pulse width of 10ms and a duty cycle of 0.1, r th peak is 6.7 c/w. maximum p d is 75/6.7 = 11.2w, with an average temperature increase d t peak of 39 x 0.1 x 11.2 @ 43 c. application note 8/9
information furnished is believed to be accurate and reliable. however, sgs-thomson microelectronics assumes no responsibility for the consequences of use of such information nor for any infringement of patents or other rights of third parties which may result from its use. no license is granted by implication or otherwise under any patent or patent rights of sgs-thomson microelectronics. specification mentioned in this publication are subject to change without notice. this publication supersedes and replaces all information previously supplied. sgs-thomson microelectronics products are not authorized for use as critical components in life support devices or systems without express written approval of sgs-thomson microelectronics. ? 1995 sgs-thomson microelectronics printed in italy all rights reserved sgs-thomson microelectronics group of companies australia - brazil - canada - china - france - germany - hong kong - italy - japan - korea - malaysia - malta - morocco - the nether- lands - singapore - spain - sweden - switzerland - taiwan - thailand - united kingdom - u.s.a. application note 9/9
|
