IR is insane
in [Design]
|
From: Glen Walpert on 20 Mar 2010 10:15 On Sat, 20 Mar 2010 07:24:54 +0000, Don Klipstein wrote: >>On Thu, 18 Mar 2010 15:58:08 -0700, John Larkin >>> >>>>Just got a short-form IR mosfet thing in the mail. >>>> >>>>They have a PQFN 5x6 mm package they rate at 104 amps. And a D2PAK >>>>rated for 340 amps. > > (And, IIRC, in that article John Larkin also mentioned: > http://www.irf.com/product-info/datasheets/data/irfs3006pbf.pdf) > >>>>John >>> >>>The only thing that actually matters is whether you can stay under the >>>power curve of I^2*Rdson(T). At 1milliohm max 340 Amps gives you 115 >>>W. You need to make sure to check the C/W rating of the package and >>>cool it enough. >> >>Rds-on is spec'd as 2.5 mohms max at 170 amps, and will go up at higher >>currents as it self-heats. 340 amps would fuse the source lead. > > I checked out that datasheet, and Rds(on) at 170 amps is 2.5 mohm max, > 2 mohm typ at 25 C. The datasheet also says that thermal resistance is > .4 degrees C per watt, and the absolute maximum junction temperature is > 175 C. > > If the case is held to 25 C, the temperature difference is 150 C, > meaning 375 watts is flowing through the thermal resistance inside the > package out the heatsinkable surface of the package. > > Given the thermal conductivity of copper and aluminum, I have trouble > envisioning a practical or semi-practical heatsink with thermal > resistance less than .4 degree per watt - > unless one does something extreme, along the lines of grinding and > polishing a flat region onto a copper pipe that the heatsink tab gets > soldered to, and then cold water flows through the pipe. > At that rate, the heatsink tab has an area around 66 mm^2, and > slightly > ground down copper pipe wall may be 1.5 mm thick... Thermal > conductivity of copper is about .4 W/(mm-k). Thermal resistance of the > pipe wall is .056 degree C per W on paper - plus a bit for the solder > joint, plus something for the water. > Suppose this extreme hetasink with its solder joint adds only .1 > degree > C per watt of thermal resistance (which I consider optimistic) to the .4 > of the package... maximum power dissipation is down to 300 watts, at > 175 C junction temperature and 25 C water temperature. > > Fig. 4 says that normalized on resistance at 175 C junction > temperature > is typically slightly over double that at 25 C. Looks to me about 2.04 > times as much. How about multiplying 2.04 by that maximum 25-C-figure > of 2.5 milliohms. I see 5.1 milliohms. Ohm's Law says that 300 watts > is dissipated in .00504 ohm when current is 244 amps. > > So I see ability to pass 244 amps *maybe* with a fairly extreme water > cooled heatsink, and assuming that maximum on resistance at 175 C does > not exceed the maximum 25 C value times what that curve in Figure 4 > indicates, and assuming that things are going to be fine and dandy and > life expectancy and reliability will be sufficiently good when the > junction is at 175 C. > > I would rather plan for 125 C junction temperature. Of course, the > maximum current will be only a little less for heating the junction to > 125 C than to 175 C... > > At .5 degree C/watt (device plus a rather extreme watwer cooled > heatsink), 125 C junction, 25 C water temperature - this means 200 > watts. > > Figure 4 indicates resistance at 125 C very nearly 1.7 times that at > 25 > C. This indicates 4.25 milliohms. At this rate, Ohm's Law indicates > 217 amps. Assuming resistance does not exceed 1.7 times the maximum 25C > value (which I do not consider a safe assumption), and assuming ability > to achieve a .1 degree C / watt heatsink for a D2PAK even with running > water. > > There is still the fact that the datasheet says the bonding wire > limits > safe continuous current to 195 amps, when case temperature is > sufficiently low to have 195 amps not causing the junction temperature > to exceed 175 C. (I calculate 96.75 C based on square relation between > current and temperature difference between junction and case, and 270 > amps being the "silicon limit" for 25 Ccase and 175 C junction. Figure > 9 makes this maximum case temperature for 195 amps continuous appear (to > me) slightly lower, to my eyeballs maybe 94-95 C. > > As large as these figures are and as small as a D2PAK is, it feels > adventurous to me to go much past the current that many of the > characterizations of this device are valid for (170 amps). > > 195 amps squared times .005 ohms (high-side but short of > guaranteeable- > maximum resistance at junction temp. of 125 C) is 190 watts. At heat > discharge point temperature of 25 C and junction temp. of 175 C, this > means thermal resistance of .79 degre C per watt. .4 of that is in the > device itself, leaving .39 degree C per watt of thermal resistance > maximum combined for the heatsink and the bonding of the device to the > heatsink. That sounds to me like something quite on the beefy side and > maybe requiring a little optimism even as far as fan-cooled heatsinks > go. > > If figuring on 150 amps absolute maximum, things get easier. At > junction temperature of 175 C with resistance .005 ohm high-side (but > not guaranteed maximum), I find 112.5 watts. If heatsink temperature is > 50 C, then thermal resistance is 1.11 degree C per watt - minus .4 for > in the device itself, leaves .71 for the heatsink and the bonding of the > device to the heatsink. It sounds to me like a more agressive modern > CPU heatsink and fan can do that or almost can, assuming good bonding > such as some sort of solder joint between the heatsink tab and the > typically aluminum heatsink (gallium alloy, while using 600 grit > sandpaper to sand the contact area of the aluminum heatsink while that > area is immersed in in gallium alloy?) > > This is still with the heatsink being fan-cooled or something of more > monstrous size as far as audio power amp heatsinks go. > >>There's no practical way to heatsink a D2PAK to dissipate 375 watts. > > I agree here. I seem to think semi-practically optimistically 300 > watts, and realistically practically 112.5 watts is sounding to me like > quite a notably big number. > > One thing my father tells me is that engineering is an economic > science. > I seem to think that there are alternatives more economical than > water-cooling a D2PAK, including likely using larger or paralleled > devices to avoid resorting to water-cooling. > > The gate charge of just one of these beasts is very considerable... > > 200 typical, 300 maximum nanocoulombs for gate delta V of 10 volts, and > drain delta V of 30 volts. (To whatever extent this matters, the > current being switched here is 170 amps.) > 37 typ of this 200 typ nanocoulombs is in gate-source capacitance > alone, > indicating 3.7 nanofarads gate-source capacitance. *Typical* 60 of > those nanocoulombs are "Miller", indicating 1.5 nanofarads of "Miller > capacitance" from 40 volt change in D-G voltage. I wonder how the > remaining 103 typ nanocoulombs is accounted for here. > 200 typical, 300 maximum nanocoulombs gate charge with gate voltage > change of 10 volts means that the gate looks like a 20 typ 30 max > nanofarad capacitor *as averaged through voltage swing* between 0 and 10 > volts. > If the gate has voltage changed through this range in 1/2 microsecond, > then the average current over that .5 microsecond is 400 mA typ, 600 mA > max, with peak being noticeably higher. > > It appears to me that this is a big MOSFET in a small package. This > MOSFET appears to me to have many of the requirements typical of MOSFETs > of similar voltage and current ratings and larger package sizes. > > Any comments at this point, anyone? ISTR a previous thread where someone turned up a maximum current test procedure where the device was submerged in a "phase change fluid" for the peak current test. Not a fluid cooled heatsink; direct immersion of the device. Probably some Freon-like refrigerant, with boiling point controlled by pressure regulation, possibly high velocity forced circulation and a hefty metal fixture, also submerged, making electrical connections. I suppose you could use these figures as a relative figure of merit when comparing devices tested in the same manner; other than that they are useless. You always need to do the thermal analysis for your particular situation, select your operating junction temperature based on your reliability requirements, and determine the current the device can handle in your situation, which will pretty much always be *way* lower than the rated peak except in low rep rate short pulse applications. IMHO no one should consider the top of the data sheet bold print to be useful data, you always need to read the fine print and then verify with testing. "The large print giveth and the fine print taketh away." I have not yet found an exception to this rule, except perhaps that the omitted print sometimes taketh away even more :-). There is research being done in phase change fluid cooling for high power modules for inverters, where the switching devices and diodes are located in a sealed container full of a refrigerant, where the container might be a big aluminum heat sink for instance. No other cooling methods can match the low junction to heatsink thermal resistance of phase change cooling - by a large margin. Availability of modules with phase change cooling within this decade has been predicted by one of the trade rag pundits, although I have forgotten where I read that.
From: Baron on 20 Mar 2010 11:35 Don Klipstein Inscribed thus: > In article <QKGdnSW78tT5rz7WnZ2dnUVZ_uIAAAAA(a)posted.localnet>, Robert > Baer wrote: > >>John Larkin wrote: >>> Just got a short-form IR mosfet thing in the mail. >>> >>> They have a PQFN 5x6 mm package they rate at 104 amps. And a D2PAK >>> rated for 340 amps. >>> >>> John >>> >>...in that case, i have some #30 wire rated at 1,000 amps and will >>guarantee it or double your money back! > > #30 wire sent back for refund under the warranty achieves refund at > sold cents/foot for portions of the purchase that remain discernably > defective wire after verifyingly endured 1,000 amps RMS for 1 > continuous > second! (Warranty does not apply for lower currents.) > > ==================== > > Warranty on the XZXZFUFYOU13 xenon flashtube, secondary safe > operating > area thereof, and the flashtube is a linear one having overall > diameter less than 3.5 mm and overall length less than 25 mm: > > Secondary warranty on this flashtube guarantees minimum of 10,000 > flashes provided anode-cathode voltage is at least 1,000 volts and > less than self-firing voltage, ratio of flash frequency to flash > energy does not exceed 1.6 watts, and flash energy is at least 100 > megajoules. > > Warranty provides for refund of purchase price of the flashtube if > the > flashtube fails to successfully flash in these parameters. Warranty > is only valid if the flashtube is actually flashed within these > parameters and fails to repeat successful attempt at a flash within > these parameters. Warranty furthermore requires that the failing > flashtube be shipped to the engineering department of the supplier > (ADDRESS BELOW) in order to qualify for refund of purchase price. > > Warranty covers nothing other than refund of purchase price of the > flashtube. Warranty does not cover damages to property other than the > flashtube in question, or injuries (whether physical or mental or > both) (whether to human beings or other life forms or both) or killing > any life form with lack of detectable injury (including but not > limited to vaporizing-from-existence any life form or property in > question). > Warranty does not cover anything else, such as fire/blast damage to > property other than the waranteed flashtube, or impairment of sexual > or other physical or psychological or even partially psychological > performance of a human being or other organism exposed to any damaging > effects of a flashtube failing in a way qualifying for this warranty, > or "the like". > > - Don Klipstein (don(a)misty.com) Does that mean that the warranty is worth what I paid for it. Nothing. Since you gave it to me in the first place ? -- Best Regards: Baron.
From: Nico Coesel on 20 Mar 2010 11:46 John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >On Thu, 18 Mar 2010 18:02:27 -0500, Damon Hill ><damon16ONE(a)comcast.not> wrote: > >>John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote in >>news:cpb5q5p2013r34ainnmcrdfjml4ifvp03g(a)4ax.com: >> >>> Just got a short-form IR mosfet thing in the mail. >>> >>> They have a PQFN 5x6 mm package they rate at 104 amps. And a D2PAK >>> rated for 340 amps. >> >>Hmm. Define 'continuous'. (a >very< short pulse rating I'd believe) >> > >The flyer doesn't mention pulsing. The D2PAK datasheet pulsed current >rating is 1080 amps. > >http://www.irf.com/product-info/datasheets/data/irfs3006pbf.pdf > >The flyer claims 340 amps. The datasheet says 270 amps but "package >limited" to 195. I don't believe any of them. > >The datasheet also claims 375 watts power dissipation... in a D2PAK! That reminds me of that 1MW laser producing 1ns pulses and a 1 minute recharge time. -- Failure does not prove something is impossible, failure simply indicates you are not using the right tools... nico(a)nctdevpuntnl (punt=.) --------------------------------------------------------------
From: John Larkin on 20 Mar 2010 13:06 On 20 Mar 2010 14:15:15 GMT, Glen Walpert <nospam(a)null.void> wrote: >On Sat, 20 Mar 2010 07:24:54 +0000, Don Klipstein wrote: > >>>On Thu, 18 Mar 2010 15:58:08 -0700, John Larkin >>>> >>>>>Just got a short-form IR mosfet thing in the mail. >>>>> >>>>>They have a PQFN 5x6 mm package they rate at 104 amps. And a D2PAK >>>>>rated for 340 amps. >> >> (And, IIRC, in that article John Larkin also mentioned: >> http://www.irf.com/product-info/datasheets/data/irfs3006pbf.pdf) >> >>>>>John >>>> >>>>The only thing that actually matters is whether you can stay under the >>>>power curve of I^2*Rdson(T). At 1milliohm max 340 Amps gives you 115 >>>>W. You need to make sure to check the C/W rating of the package and >>>>cool it enough. >>> >>>Rds-on is spec'd as 2.5 mohms max at 170 amps, and will go up at higher >>>currents as it self-heats. 340 amps would fuse the source lead. >> >> I checked out that datasheet, and Rds(on) at 170 amps is 2.5 mohm max, >> 2 mohm typ at 25 C. The datasheet also says that thermal resistance is >> .4 degrees C per watt, and the absolute maximum junction temperature is >> 175 C. >> >> If the case is held to 25 C, the temperature difference is 150 C, >> meaning 375 watts is flowing through the thermal resistance inside the >> package out the heatsinkable surface of the package. >> >> Given the thermal conductivity of copper and aluminum, I have trouble >> envisioning a practical or semi-practical heatsink with thermal >> resistance less than .4 degree per watt - >> unless one does something extreme, along the lines of grinding and >> polishing a flat region onto a copper pipe that the heatsink tab gets >> soldered to, and then cold water flows through the pipe. >> At that rate, the heatsink tab has an area around 66 mm^2, and >> slightly >> ground down copper pipe wall may be 1.5 mm thick... Thermal >> conductivity of copper is about .4 W/(mm-k). Thermal resistance of the >> pipe wall is .056 degree C per W on paper - plus a bit for the solder >> joint, plus something for the water. >> Suppose this extreme hetasink with its solder joint adds only .1 >> degree >> C per watt of thermal resistance (which I consider optimistic) to the .4 >> of the package... maximum power dissipation is down to 300 watts, at >> 175 C junction temperature and 25 C water temperature. >> >> Fig. 4 says that normalized on resistance at 175 C junction >> temperature >> is typically slightly over double that at 25 C. Looks to me about 2.04 >> times as much. How about multiplying 2.04 by that maximum 25-C-figure >> of 2.5 milliohms. I see 5.1 milliohms. Ohm's Law says that 300 watts >> is dissipated in .00504 ohm when current is 244 amps. >> >> So I see ability to pass 244 amps *maybe* with a fairly extreme water >> cooled heatsink, and assuming that maximum on resistance at 175 C does >> not exceed the maximum 25 C value times what that curve in Figure 4 >> indicates, and assuming that things are going to be fine and dandy and >> life expectancy and reliability will be sufficiently good when the >> junction is at 175 C. >> >> I would rather plan for 125 C junction temperature. Of course, the >> maximum current will be only a little less for heating the junction to >> 125 C than to 175 C... >> >> At .5 degree C/watt (device plus a rather extreme watwer cooled >> heatsink), 125 C junction, 25 C water temperature - this means 200 >> watts. >> >> Figure 4 indicates resistance at 125 C very nearly 1.7 times that at >> 25 >> C. This indicates 4.25 milliohms. At this rate, Ohm's Law indicates >> 217 amps. Assuming resistance does not exceed 1.7 times the maximum 25C >> value (which I do not consider a safe assumption), and assuming ability >> to achieve a .1 degree C / watt heatsink for a D2PAK even with running >> water. >> >> There is still the fact that the datasheet says the bonding wire >> limits >> safe continuous current to 195 amps, when case temperature is >> sufficiently low to have 195 amps not causing the junction temperature >> to exceed 175 C. (I calculate 96.75 C based on square relation between >> current and temperature difference between junction and case, and 270 >> amps being the "silicon limit" for 25 Ccase and 175 C junction. Figure >> 9 makes this maximum case temperature for 195 amps continuous appear (to >> me) slightly lower, to my eyeballs maybe 94-95 C. >> >> As large as these figures are and as small as a D2PAK is, it feels >> adventurous to me to go much past the current that many of the >> characterizations of this device are valid for (170 amps). >> >> 195 amps squared times .005 ohms (high-side but short of >> guaranteeable- >> maximum resistance at junction temp. of 125 C) is 190 watts. At heat >> discharge point temperature of 25 C and junction temp. of 175 C, this >> means thermal resistance of .79 degre C per watt. .4 of that is in the >> device itself, leaving .39 degree C per watt of thermal resistance >> maximum combined for the heatsink and the bonding of the device to the >> heatsink. That sounds to me like something quite on the beefy side and >> maybe requiring a little optimism even as far as fan-cooled heatsinks >> go. >> >> If figuring on 150 amps absolute maximum, things get easier. At >> junction temperature of 175 C with resistance .005 ohm high-side (but >> not guaranteed maximum), I find 112.5 watts. If heatsink temperature is >> 50 C, then thermal resistance is 1.11 degree C per watt - minus .4 for >> in the device itself, leaves .71 for the heatsink and the bonding of the >> device to the heatsink. It sounds to me like a more agressive modern >> CPU heatsink and fan can do that or almost can, assuming good bonding >> such as some sort of solder joint between the heatsink tab and the >> typically aluminum heatsink (gallium alloy, while using 600 grit >> sandpaper to sand the contact area of the aluminum heatsink while that >> area is immersed in in gallium alloy?) >> >> This is still with the heatsink being fan-cooled or something of more >> monstrous size as far as audio power amp heatsinks go. >> >>>There's no practical way to heatsink a D2PAK to dissipate 375 watts. >> >> I agree here. I seem to think semi-practically optimistically 300 >> watts, and realistically practically 112.5 watts is sounding to me like >> quite a notably big number. >> >> One thing my father tells me is that engineering is an economic >> science. >> I seem to think that there are alternatives more economical than >> water-cooling a D2PAK, including likely using larger or paralleled >> devices to avoid resorting to water-cooling. >> >> The gate charge of just one of these beasts is very considerable... >> >> 200 typical, 300 maximum nanocoulombs for gate delta V of 10 volts, and >> drain delta V of 30 volts. (To whatever extent this matters, the >> current being switched here is 170 amps.) >> 37 typ of this 200 typ nanocoulombs is in gate-source capacitance >> alone, >> indicating 3.7 nanofarads gate-source capacitance. *Typical* 60 of >> those nanocoulombs are "Miller", indicating 1.5 nanofarads of "Miller >> capacitance" from 40 volt change in D-G voltage. I wonder how the >> remaining 103 typ nanocoulombs is accounted for here. >> 200 typical, 300 maximum nanocoulombs gate charge with gate voltage >> change of 10 volts means that the gate looks like a 20 typ 30 max >> nanofarad capacitor *as averaged through voltage swing* between 0 and 10 >> volts. >> If the gate has voltage changed through this range in 1/2 microsecond, >> then the average current over that .5 microsecond is 400 mA typ, 600 mA >> max, with peak being noticeably higher. >> >> It appears to me that this is a big MOSFET in a small package. This >> MOSFET appears to me to have many of the requirements typical of MOSFETs >> of similar voltage and current ratings and larger package sizes. >> >> Any comments at this point, anyone? > >ISTR a previous thread where someone turned up a maximum current test >procedure where the device was submerged in a "phase change fluid" for >the peak current test. Not a fluid cooled heatsink; direct immersion of >the device. Probably some Freon-like refrigerant, with boiling point >controlled by pressure regulation, possibly high velocity forced >circulation and a hefty metal fixture, also submerged, making electrical >connections. I suppose you could use these figures as a relative figure >of merit when comparing devices tested in the same manner; other than >that they are useless. You always need to do the thermal analysis for >your particular situation, select your operating junction temperature >based on your reliability requirements, and determine the current the >device can handle in your situation, which will pretty much always be >*way* lower than the rated peak except in low rep rate short pulse >applications. IMHO no one should consider the top of the data sheet bold >print to be useful data, you always need to read the fine print and then >verify with testing. > >"The large print giveth and the fine print taketh away." I have not yet >found an exception to this rule, except perhaps that the omitted print >sometimes taketh away even more :-). > >There is research being done in phase change fluid cooling for high power >modules for inverters, where the switching devices and diodes are located >in a sealed container full of a refrigerant, where the container might be >a big aluminum heat sink for instance. No other cooling methods can >match the low junction to heatsink thermal resistance of phase change >cooling - by a large margin. Availability of modules with phase change >cooling within this decade has been predicted by one of the trade rag >pundits, although I have forgotten where I read that. I haven't done the math on this, but I suspect the source lead would vaporize at 195 amps, much less 340. If it were immersed in a boiling liquid, it would probably be OK. I think that's how they test them. John
From: Archimedes' Lever on 20 Mar 2010 13:29
On Sat, 20 Mar 2010 10:06:13 -0700, John Larkin <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >I haven't done the math on this, We can tell. > but I suspect the source lead would >vaporize at 195 amps, much less 340. You always seem to have suspicions. We have a few about you. > If it were immersed in a boiling >liquid, it would probably be OK. I think that's how they test them. Why would the liquid have to be boiling, Johnny? You *think* that is how they test them? It is clear that you ASSume, as opposed to thinking. |