Making Use of ae Charge Informaion in MOSFET and IBT Daa Shees Ralph McArhur Senior Applicaions Engineer Advanced Power Technology 405 S.W. Columbia Sree Bend, Oregon 97702 Power MOSFETs and IBTs have esablished hemselves as premier power semiconducors in a wide range of applicaions involving swiching or amplificaion. In order o use hese devices effecively i is necessary o undersand and make use of he gae charge informaion on he daa shee. This informaion appears in wo locaions: in he Dynamic Characerisics secion of lised parameers and in he figure showing gae-source volage versus oal gae charge. Hereafer he figure showing gae-source volage versus oal gae charge will be referred o simply as he gae charge curve. This applicaion noe is wrien wih MOSFETs in view, bu everyhing said here applies o IBTs as well. Jus exchange collecor for drain and emier for source. The discussion is based on a common MOSFET model, shown in Figure 1. The figure shows he MOSFET model, he oal gae resisance, and block elemens for he load impedance and he gae drive circui. Figure 2 shows a gae charge curve aken from a daa shee. I displays he gae-source volage as a funcion of charge injeced ino he gae. Charge is buil up in he gae as long as gae drive curren flows ino he gae. Noe ha he main power supply, V DD in he gae charge measuremen, serves as a parameer in he figure. Figure 1 shows where V DD is applied. The charge required o reach a cerain gae volage is differen for differen values of V DD. For a given V DD, if a cerain number of nanocoulombs are injeced ino he gae, he volage will rise o a cerain level. Use of Figure 2 in gae drive design has already been described. 1 ae Drive Circui R C D C S V DD D S Z L C DS Figure 1. Lumped elemen model for a power MOSFET The gae charge parameers called ou in he Dynamic Characerisics secion of he daa shee are as follows: Q g, he oal gae charge, Q gs, he gaesource charge, and Q gd, he gae-drain ( Miller ) charge. Figure 3 shows he pars of he gae charge curve from which hese charge values are aken. APT0103 Rev - 2001 Advanced Power Technology 1
Q gs is he charge from he origin o he firs inflecion in he curve, Q gd is he charge from he firs inflecion in he curve o he second inflecion in he curve, and Q g is he charge from he origin o he poin on he curve a which v S equals he peak drive volage. In Figure 3, v S equals he peak drive volage a he righ end of he curve. In Figure 2, Qgs equals abou 12.5 nc, Qgd equals abou 100 nc, using he V DD = 500 V line, and, assuming a peak gae drive volage of 10 V, Qg equals abou 205 nc, again using he V DD = 500 V line. The plaeau beween he wo inflecions in Figure 2 is fla, whereas he plaeau in Figure 3 is no. This reflecs he way i is in general. The plaeau is fla in some curves and no in ohers. More will be said abou his laer. be noed ha here are circuis in which separae drives are used for charging and discharging when he paricular applicaion imposes discharging requiremens ha differ from he charging requiremens. Even so, alhough he drives in ha case are differen, he physical principles involved are he same. ae-source Volage [V] Q gs Q gd Q g ae-source volage = peak drive volage Charge [nc] Figure 3. v S as a funcion of gae charge Figure 4 shows he MOSFET capaciances repored in he daa shees as a funcion of v DS : C iss, he inpu capaciance, C oss, he oupu capaciance, and C rss, he reverse ransfer capaciance. In erms of he inererminal capaciances in he model of Figure 1, C iss = C S + C D, C oss = C DS + C D, and C rss = C D. Figure 2. v S as a funcion of gae charge This noe will consider again he process of charging he gae in order o clarify imporan issues and provide a basis for discussion of various opics ha depend on hem: adjusing he swiching speed of he MOSFET, designing gae drive circuis, selecing commercial gae drive circuis, and esing he MOSFET o deermine is gae charge properies. These opics will no be covered in he noe; raher he noe will serve as a basis for discussing hem in he fuure. The process of discharging he gae a urn-off will no be reaed, since in principle i is he reverse of charging. Neverheless i should Figure 4. Typical capaciance as a funcion of v DS C iss, C oss, and C rss vary as a funcion of v DS, because C D and C DS vary as a funcion of v DS. C S on he APT0103 Rev - 2001 Advanced Power Technology 2
oher hand is consan. This is seen in Figure 5. The curren hrough C D and C DS depends on he ime derivaive of he produc of he capaciance and is volage. This will be discussed laer. Capaciance [pf] 1.00E+05 1.00E+04 1.00E+03 C DS 1.00E+02 0.1 1 10 100 VDS [vols] C S C D Figure 5. The inererminal capaciances as a funcion of v DS Figure 2 plos v S agains gae charge. When he measuremens are acually done, v S is ploed agains ime, as in Figure 6, and he way hey are done makes i possible o rea ime as he equivalen of charge. A consan-curren gae drive is employed in he gae charge measuremen. In order o deermine how much charge has been injeced ino he gae, he curren ino he gae mus be inegraed. For any given ime period, inegraing he gae curren from he beginning of he period o is end deermines he amoun of charge injeced ino he gae during ha ime. If we use i o represen he gae curren, Q o represen he charge going ino he gae, and b (beginning ime) and e (ending ime) o represen he ime period, he equaion is as follows: e Q = i d. b When a consan gae curren is employed, his equaion reduces simply o: where is he ime period over which he charge is o be inegraed, ha is, e - b. Therefore, in Figure 6, Q gs = i ( 2 0 ), Q gd = i ( 3 2 ), and Q g = i ( 4 0 ). In his way, consan curren in he gae makes i possible o rea he volage versus ime curve as a volage versus charge curve. Addiionally, consan curren in he drain eliminaes he effec of lead inducance from he measuremen. The four ime periods in Figure 6 are delineaed by cerain levels of v S and cerain changes in v D. The endpoins of hese ime periods are he origin a = 0, V T, he hreshold volage, V PL, he lef end of he plaeau in v S, V PR, he righ end of he same plaeau, and V DR, he peak value of he drive volage. When he drive circui is a volage source, his peak value is he nominal poenial a he oupu of he driver. However, when, as in Figure 6, consan gae curren is employed, he driver is a curren source, and he peak value of he gae volage is chosen o be 10 V, as saed in he daa shees. I is a he peak value of he gae volage ha he oal gae charge is deermined. ae-source Volage [V] 1 V PL V T 2 V PR 3 4 0 3 4 1 2 Time [ns] V DR Figure 6. Four ime periods in he v S versus ime characerisic curve Four Time Periods in he Rise of v S Q = i, The four ime periods in Figure 6 are described nex. Alhough i is helpful o break he charging ime ino four periods o explain he evens aking place, neverheless daa shees characerize his APT0103 Rev - 2001 Advanced Power Technology 3
process wih he use of he hree charge quaniies already menioned. Figure 7 shows he swiching waveforms of he MOSFET a urn-on for reference, as well. As he descripion unfolds, he evens can be observed in he figure as hey are called ou. The imes in Figure 7 are in µs, insead of ns, because hey were generaed using a low magniude (1 ma) consan-curren drive, a sandard way of measuring gae charge. The load for he gae charge measuremen in Figure 7 is a consan curren source (no an inducor). Volage [V] and Curren [A] Swiching Waveforms for a MOSFET v DS v S i D an end when v S = V T a = 1. To his poin, he drain curren has no ye begun o flow. I should be noed ha C S is no a funcion of volage, whereas C D is. Therefore, in his ime period C S is consan, bu C D is increasing slighly because he magniude of he poenial difference across is erminals is decreasing slighly. Recall ha as he volage decreases he capaciance increases. This can be undersood from Figure 5. As v DS decreases, C D increases. Also, apar from Figure 5, we know ha as v DS decreases, v D increases. Undersand ha, since V DD > v S, for mos of he range of values for v DS, v DS > v S. This means ha v D is negaive, since v D = v S v DS. Therefore, for v D o increase is for v D o become less negaive. This means is magniude is decreasing, and as v D increases (ha is, as v DS is decreasing and v D s magniude is decreasing), C D increases. 0 20 40 60 80 100 Time [us] 0 1 2 3 4 Figure 7. Swiching waveforms showing evens a imes 0 hrough 4. From Zero o V T v S rises from zero o V T ; i D does no flow; v DS remains unchanged. When v S is in his range of values, 0 < v S < V T, he MOSFET is off and v DS is consan a he level of he supply volage, V DD. A ime = 0, when he consan gae curren is applied, v S begins o rise. This rise in v S is brough abou by he charging of C S and C D. I is a misconcepion o hink ha only C S is being charged in his regime, for if he gae node is rising in poenial, hen he volage across C D is changing, and ha requires he flow of charge ino C D. A his unchanging value of v DS, C S is much larger han C D and, herefore, much more drive curren is flowing ino C S han ino C D. The rise of v S in his ime period is brough o From V T o V PL v S rises from V T o v PL ; i D begins o flow; v DS begins o decrease; Q gs is he charge injeced ino he gae from ime 0 o ime 2. Threshold volage is defined as he gae-source volage a which i D begins o flow. To say i his way means ha low-level, pre-hreshold curren is being negleced. Therefore a he beginning of his ime period, a = 1, i D begins o flow and v DS begins o decrease a some changing rae. The ime period ends a = 2 wih he appearance of he firs knee. This appearance of he knee depends on he rae of change of he produc C D v D wih respec o ime. The more general expression for curren in a capacior applies here, since he capaciance, C D, is varying. Recall ha q = Cv and ha, in general, he equaion for curren in a capacior is dq d( Cv) i = =. d d APT0103 Rev - 2001 Advanced Power Technology 4
This is he expression ha mus be used when C and v are boh varying in ime. This is exacly he case wih C D and v D in he MOSFET. v D = v S v DS, and is a funcion of i D and Z L, no i D only. In oher words, i is no correc o say ha he lef knee occurs when he drain curren reaches i maximum, as is ofen claimed. Any combinaion of i D and Z L ha makes v DS decrease fas enough, in urn causing v D o increase, will bring abou he knee. As saed previously, for v D o increase means ha i is becoming less negaive. In heory i D can acually reach is maximum afer he lef knee occurs, because he curren in C D, d(c D v D )/d, deermines he occurrence of he knee. A small value of i D and a large value of Z L can bring his abou. From V PL o V PR v S progresses from v PL o v PR ; i D reaches is maximum a some poin; v DS reaches is minimum a 3 ; Q gd is he charge injeced ino he gae from ime 2 o ime 3. As v S coninues o move, rising pas V T and V PL, because of charge delivered by he gae curren, he slope of v S decreases subsanially, in some cases decreasing all he way o zero. v S has hereby moved ino he region someimes referred o as he plaeau. This ook place a = 2 in Figure 6. The slope of his plaeau region, sricly speaking, depends on he division of drive curren beween C D and C S, and i is he es or circui condiions ha deermine his division. If C D v D increases quickly enough, v S in he plaeau region will be consan: v S = V PL = V PR. In his laer case, he slope is zero, which means ha none of he drive curren is flowing ino C S. Raher, all of he drive curren is being used o accommodae he change in volage across C D as C D v D coninues o increase. If he slope is nonzero, i means ha some of he drive curren is making is way ino C S wih he resul ha v S is rising, albei a a slow rae. Figure 8 shows wo gae charge curves for wo differen MOSFETs esed under he same condiions. One device exhibis low gae charge (he MOS 7 MOSFET from APT s laes generaion of low-loss MOSFETS), and he oher device exhibis high gae charge (he MOS V MOSFET from an older generaion of APT s MOSFETs). Here are wo real devices, one wih a non-zero slope in he plaeau region and one wih a nearly zero slope. The ime is in µs because a low magniude (1 ma) consan-curren drive was used o drive he gaes. Whereas in he firs and second ime periods here was no choice (some of he drive curren flowed ino C D ), here in he hird ime period, by providing he load wih sufficien impedance and providing he inpu wih large enough resisance, i should be possible o diver all (almos all) of he drive curren hrough C D. Tha is, he slope of he plaeau depends on C S, C D, and he drive and load circui elemens. However, regardless of he slope of he plaeau, he righ end of he plaeau occurs when he MOSFET reaches is fully on sae. ae-source Volage [V] 12 10 8 6 4 2 ae Charge Curves for Two Power MOSFETs LOW-LOSS MOS 7 APT50M75LL 0 0 50 100 150 200 250 300 Time [us] MOS V APT5010LVR Figure 8. ae charge curves for wo power MOSFETs showing differen slopes in he plaeau region For he enire duraion of his hird ime period, v DS is decreasing (which means ha v D is increasing) and C D is increasing. When v DS reaches he level of I D X R DS(on), i sops decreasing and C D sops increasing. Capial I in I D indicaes ha he peak value of i D is in view. This phenomemon ends he ime period and v S begins o increase a a new, higher rae. The righ end of he plaeau has been reached a v S = V PR and = 3. APT0103 Rev - 2001 Advanced Power Technology 5
From V PR o V DR v S rises from v PR o v DR ; i D says a is maximum; v DS says a is minimum; Q g is he charge injeced ino he gae from ime 0 o ime 4. Once v DS reaches is minimum, I D X R DS(on), v S rises ou of he plaeau region and increases a a new rae. Now, since v DS is no changing any more, he bulk of he drive curren once again flows ino C S. The slope is no as high as i was in he firs ime period, because C D is much larger han i was hen. I is much closer in value o C S. Now again boh capaciors are being charged, and boh capaciors are of consan value or nearly so, C S because i always is and C D because v D > 0. 2 Implicaions The foregoing has implicaions ha may also lead o beer designs and may improve he way hese issues are discussed. Consider he following. ae-source Charge, Q gs As already saed, Q gs is he charge i akes o bring v S up from zero o v PL a 2 in Figure 6. I is bes o undersand i as such raher han o hink, for example, ha i is he charge going ino C S, as he name suggess. By he ime v S reaches v DR a 4 in Figure 6, a grea deal more charge has gone ino C S han is represened by Q gs, which again is only he charge needed o raise v S from zero o v PL. I is also undersood ha some of he charge represened by Q gs is acually going ino C D. ae-drain Charge, Q gd Again, as already saed, Q gd is he charge i akes o bring v S from v PL a ime 2 in Figure 6 o v PR a ime 3 in Figure 6. Therefore, Q gd is aken as he produc of he gae curren and he duraion of he hird ime period, 3-2, in he gae charge measuremen. I is also bes o undersand Q gd in his way and no, as he name suggess, as he charge going ino C D. By he ime v S reaches v DR a ime 4 in Figure 6, more charge has gone ino C D han is indicaed by Q gd. Addiionally, in mos cases Q gd also represens some charge ha has gone ino C S. Toal ae Charge, Q g Toal gae charge depends on V DR, he peak value of he gae drive volage, on V DD, he volage supplied o he drain, and on I D, he peak value of he drain curren. These values are saed when he oal gae charge number is enered ino he daa shee. I is he charge i akes o bring v S up from zero o V DR a ime 4 in Figure 6, given he chosen values of V DD and V DR. The wo charge quaniies discussed previously, when added ogeher, represen an imporan piece of informaion. The sum of Q gs and Q gd is he amoun of charge ha mus be injeced ino he gae o bring he ypical MOSFET ino full conducion. Tha is, a he momen Q gs + Q gd has been injeced ino he gae, v DS = I D X R DS(on). A his poin he MOSFET is in full conducion and any more charge injeced ino he gae represens overcharge, charge ha does no change he oupu condiions. Manufacurers rouinely recommend ha he gae be overcharged, because his ensures ha enough charge is injeced ino each gae and i accouns for differences among ransisors in heir gae charge requiremens. For MOSFETs a common gae volage for overcharge condiions is 10 V. The charge injeced ino he gae o bring i o he overcharge level is he oal gae charge, Q g. Swiching Times These ideas have an impac on he swiching imes for a MOSFET as well. In paricular, he ime i akes o urn a MOSFET on is he ime i akes o APT0103 Rev - 2001 Advanced Power Technology 6
injec he wo daa shee charge quaniies, Q gs and Q gd, ino he gae. For any MOSFET he wo charge quaniies in quesion are is own Q gs and Q gd. These numbers may indeed be differen for wo MOSFETs wih he same par number. Once again, overcharge can render he differences unimporan. Insead of designing o he sum of Q gs and Q gd, if one designs o Q g, he differences should no affec he performance of he circui. In oher words, if one designs he gae drive o have he oal gae charge injeced ino he gae by he ime he sysem needs he MOSFET o be on, he swiching ime will be correc. This calculaion is easier when a consan curren gae drive is employed. The curren o be delivered o he gae is found as follows: i R C eff = I e, where I is he peak value of he gae curren; R is he oal gae resisance; and C eff is he effecive gae inpu capaciance. How o deermine C eff will be reaed in a subsequen applicaion noe and has been reaed by ohers, 3, 4 bu in general i will no be equal o he daa shee value of C iss. The preceding equaion may also be wrien, VDR R C eff = e. R i Now he gae charge equaion becomes Q g I =. s Q g s V = R 0 DR e R C eff d. Capial I has been used in I o show ha he magniude of he drive curren is in view, and s represens he swiching ime required by he sysem. When a consan volage drive is used, he peak value of he gae curren mus be deermined from he equaion, V DR I =, R where R is he sum of he driver s oupu impedance, he exernal gae resisance, and he series resisance of he gae iself. In Figure 1, his oal gae resisance has been lumped ino R. The proper value of R mus be deermined ieraively from he following gae charge equaion: Q g = s i d. 0 R is inheren in he variable i. To solve his equaion analyically one mus cas i in he form, The inegraion can be performed numerically, and i is easier o supply R and solve for s. If many ime poins are used, recangular inegraion should be adequae o he ask. The soluion of his equaion is ieraive, since one mus supply a value for R and hen solve for s. These wo seps mus be repeaed unil he desired s is found. Tha deermines R and, in urn, I. I remains only o provide a driver ha can source he peak curren equal o V DR /R. Since mos designers use consan volage drive, his is he process hey should follow in heir designs. ae Charge Measuremens How o measure he daa shee Q gd is a quesion as well. Figure 6 shows he imes 2 and 3 coinciden wih heir respecive ends of he plaeau. Sraigh line inersecions define he corners ha, projeced down ono he ime axis, idenify 2 and 3. This would seem o be he bes way o measure Q gd, alhough here are ohers. 5 Q gd is simply, hen, he consan gae curren muliplied by he difference, 3 2. APT0103 Rev - 2001 Advanced Power Technology 7
In like fashion, as suggesed earlier, Q gs is he consan gae curren imes he difference, 2 0, and Q g is he consan gae curren imes he difference, 4 0. Finally, since many designs now anicipae a V DD = 0.7V DSS, gae measuremens ough o be done a ha volage. V DSS is he raed volage for he MOSFET. The designer needs o know wha he Q g is going o be under he paricular circui condiions o be used, no he arificial and very common measuremen condiion of 0.5V DSS. 4 Erickson, Rober, Bill Behen, R. D. Middlebrook, and Slobodan Ćuk. Characerizaion and Implemenaion of Power MOSFETs in Swiching Converers. In Proceedings of Powercon 7, pp. 188-189. The Sevenh Naional Solid-Sae Power Conversion Conference, March 24-27, 1980, San Diego, California. Power Conceps, Inc., 1980. 5 Deparmen of Defense Tes Mehod Sandard, Semiconducor Devices, Mil-Sd 750D, February 28, 1995, Mehod 3471. 6 Dierberger, Kenneh. ae Drive Design for Large Die MOSFETs. Applicaion Noe APT9302. Advanced Power Technology. Conclusions Q gs and Q gd do no mean wha heir names seem o imply. Neverheless hey perform he useful funcion of idenifying ogeher he amoun of charge necessary o bring he daa-shee-ypical MOSFET ino full conducion. Since, however, he daa shee values are for ypical devices and do no accoun for he spread of values from MOSFET o MOSFET, proper gae drive design makes use of Q g o bring abou an overcharge condiion in he gae. Proper gae drive design also makes use of he correc curren inegraion equaion o deermine he curren requiremens for he gae drive circui. The principles in his applicaion noe, when combined wih proper circui layou 6 and power circui design and when accouning for he required swiching ime, will lead o opimal swiching performance. References 1 Use ae Charge o Design he ae Drive Circui for Power MOSFETs and IBTs. AN-944. Inernaional Recifier. 2 ran, Duncan A. and John owar. POWER MOSFETS Theory and Applicaions, p. 84. New York: John Wiley & Sons, 1989. 3 Pressman, Abraham I. Swiching Power Supply Design, Second Ediion, pp. 359-361. New York: Mcraw-Hill, 1998. APT0103 Rev - 2001 Advanced Power Technology 8