MOSFET transient dissipation
in [Design]
|
Prev: Voltage controlled high current source problem with low valuesense resistor
Next: SMD identification: please help!
From: Jim Thompson on 4 Dec 2009 14:31 On Fri, 04 Dec 2009 12:25:21 -0700, Jim Thompson <To-Email-Use-The-Envelope-Icon(a)My-Web-Site.com/Snicker> wrote: >On Fri, 4 Dec 2009 23:00:39 +0530, "pimpom" <pimpom(a)invalid.com> >wrote: > >>Wimpie wrote: >>> On 4 dic, 12:25, "pimpom" <pim...(a)invalid.com> wrote: >>>> For the drag race Christmas tree project I opened for >>>> discussion >>>> several days ago, I intend to use IRFZ44N MOSFETs to switch >>>> car >>>> headlight bulbs. The MOSFET has an ON resistance of 17.5m? >>>> (say >>>> 25m? warm). The lamps are 12V/45W used in small cars and >>>> should >>>> be just enough to get the transistors slightly warm to the >>>> touch >>>> in the on state without a heatsink. >>>> >>>> However, since incandescent lamps draw a high turn-on surge >>>> current, I wondered if it's advisable to bolt the MOSFETs on >>>> to a >>>> larger thermal mass such as a strip of thick aluminium plate. >>>> To >>>> those of you with experience in that kind of application, >>>> would >>>> you say that such a precaution is - >>>> 1. essential >>>> 2. not likely to be needed but good insurance >>>> 3. a waste, not needed at all. >>>> >>>> Switching frequency is expected to be no more than once in >>>> several seconds at most - essentially a single pulse. I've >>>> looked >>>> at the transient thermal Z of the transistors, but I have no >>>> clear idea how long the turn on surge period is except that >>>> it's >>>> a small fraction of a second and is probably an initial peak >>>> with >>>> an inverse exponential decay. I haven't had time to rig up >>>> something to measure it with a scope. Anything other than a >>>> wild >>>> guess will be welcome. Thanks. >>> >>> Hello, >>> >>> Steady state dissipation is about 0.4W, as Rthjc = 62C/W, this >>> results >>> in 25K temperature rise. So the steady state doesn't require a >>> heat >>> sink. >>> >>> Assuming a cold resistance of 12 times below hot resistance, >>> the >>> mosfet has to supply about 50A. This would result in about 63W >>> (0.025 >>> Ohm), but for a very short time as lamp temperature rises >>> rapidly and >>> dissipation is proportional with I^2. Let us assume that 63W >>> is >>> dissipated during 0.1s, that is 6.3J. Look to figure 8 of the >>> datasheet and extrapolates the graph to 100ms. Then you will >>> see that >>> you are close or above the SOA limit (for Dutch speaking >>> readers, Safe >>> Operating Area). >>> >>> When the inrush current reduces to 60% of peak value well >>> within 0.1s, >>> you are within the safe operating area (at 60%, the dissipation >>> drops >>> to 36% of max. dissipation). >>> >>> Do you need a heat sink for the inrush current? >>> Assuming 0.8 gram of copper in the tap results in an average >>> temperature rise of 20K in case of 6.3J added heat. For the >>> transient >>> energy, a heat sink is not required. >>> >>> So my first conclusion is: heatsink is not required as long as >>> inrush >>> current to reach 60% is well below 0.1s (based on 50A peak >>> current). >>> When inrush current > 0.1s, use another mosfet. I think you >>> should >>> measure the inrush current versus time, or try to get reliable >>> data. >>> AFAIK, halogen head lights have more inrush current. >>> >>Thanks for the input. I also considered biasing the lamps with a >>pre-heating current to just below incandescence (a dim red glow >>wouldn't interfere with the application), but that would add >>complexity and also consume considerable standby power from the >>battery. >> > >RAMP the MOSFET's ON ?? > > ...Jim Thompson Or a current limit at maybe 2X operating current... avoids the typical 10X+ surge. ...Jim Thompson -- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | You can never be too prepared for the REPRESSION!
From: pimpom on 4 Dec 2009 14:55 Jim Thompson wrote: > On Fri, 4 Dec 2009 23:00:39 +0530, "pimpom" > <pimpom(a)invalid.com> > wrote: > >> Thanks for the input. I also considered biasing the lamps with >> a >> pre-heating current to just below incandescence (a dim red >> glow >> wouldn't interfere with the application), but that would add >> complexity and also consume considerable standby power from >> the >> battery. >> > > RAMP the MOSFET's ON ?? > Well, a linear ramp would raise the MOSFET's turn-on dissipation - just the thing I want to limit. Pulsed drive with a rising duty cycle will stretch out the lamp's turn-on time which is also unacceptable. Since the lights are used for timing the racers, they have to be lit up as nearly instantaneously as possible. What I meant by pre-heating was either by pulsing the MOSFETs at perhaps 25% duty in standby or with a resistor in parallel with the MOSFET. But either scheme would waste battery power.
From: Dave Platt on 4 Dec 2009 15:01 >Already covered. > >Regarding the turn-on surge, I've drawn a load line on the MOSFET >characteristics curve with an assumed cold filament resistance of >0.25 ohms. It intersects the 10V Vgs curve at about 45A Id and >0.7V Vds. That's more than 30W dissipation for a brief moment >(ignoring gate drive rise time). The fall to the steady-state >dissipation of less than 0.4W will be non-linear. What I'm >concerned about is the thermal inertia during that brief period. How about slowing down the turn-on pulse, and "warming up" the filament (and raising its resistance) more gradually? Run your gate drive through an RC with a reasonable time constant, and then feed it to the gate through a gate-stopper resistor. Depending on the time constant of the RC, and the thermal time constant of the filament, I imagine that you can probably bring the lamp up to full brightness in (e.g.) 100 milliseconds or so, without the current rising to more than double its steady-state value at any point. This might result in a better-looking power dissipation curve. You may increase the service life of the bulbs, too. If you do decide to bang the MOSFET all the way on as fast as possible, it wouldn't surprise me if the MOSFET's ability to get rid of the transient heat pulse is limited by the rate of heat conduction through the package and tab. The current surge may be over, and the dissipation settled down to its steady state, before much of the heat has been conducted out to the far side of the tab. If that turns out to be the case, adding a heatsink (for additional thermal inertia at this point) might not buy you much. -- Dave Platt <dplatt(a)radagast.org> AE6EO Friends of Jade Warrior home page: http://www.radagast.org/jade-warrior I do _not_ wish to receive unsolicited commercial email, and I will boycott any company which has the gall to send me such ads!
From: pimpom on 4 Dec 2009 15:09 Jim Thompson wrote: > On Fri, 04 Dec 2009 12:25:21 -0700, Jim Thompson > <To-Email-Use-The-Envelope-Icon(a)My-Web-Site.com/Snicker> wrote: > >> On Fri, 4 Dec 2009 23:00:39 +0530, "pimpom" >> <pimpom(a)invalid.com> >> wrote: >> >>> Thanks for the input. I also considered biasing the lamps >>> with a >>> pre-heating current to just below incandescence (a dim red >>> glow >>> wouldn't interfere with the application), but that would add >>> complexity and also consume considerable standby power from >>> the >>> battery. >>> >> >> RAMP the MOSFET's ON ?? >> >> ...Jim Thompson > > Or a current limit at maybe 2X operating current... avoids the > typical > 10X+ surge. > That will slow down the lamp's build up of light which is unacceptable. Even the normal turn-on time of an incandescent bulb is not completely satisfactory - at least in theory, but the error it creates in the racers' recorded elapsed time should be minor. We'll just have to accept that for the coming event. For later races, it should be possible to adapt the control unit to LEDs with little or no modification.
From: pimpom on 4 Dec 2009 15:17
Dave Platt wrote: >> Already covered. >> >> Regarding the turn-on surge, I've drawn a load line on the >> MOSFET >> characteristics curve with an assumed cold filament resistance >> of >> 0.25 ohms. It intersects the 10V Vgs curve at about 45A Id and >> 0.7V Vds. That's more than 30W dissipation for a brief moment >> (ignoring gate drive rise time). The fall to the steady-state >> dissipation of less than 0.4W will be non-linear. What I'm >> concerned about is the thermal inertia during that brief >> period. > > How about slowing down the turn-on pulse, and "warming up" the > filament (and raising its resistance) more gradually? > > Run your gate drive through an RC with a reasonable time > constant, and > then feed it to the gate through a gate-stopper resistor. > Depending > on the time constant of the RC, and the thermal time constant > of the > filament, I imagine that you can probably bring the lamp up to > full > brightness in (e.g.) 100 milliseconds or so, without the > current > rising to more than double its steady-state value at any point. > This > might result in a better-looking power dissipation curve. You > may > increase the service life of the bulbs, too. > A slow build-up of light is unacceptable for the reasons I gave in the reply I just posted to Jim Thompson's post. Please see that. > If you do decide to bang the MOSFET all the way on as fast as > possible, it wouldn't surprise me if the MOSFET's ability to > get rid > of the transient heat pulse is limited by the rate of heat > conduction > through the package and tab. The current surge may be over, > and the > dissipation settled down to its steady state, before much of > the heat > has been conducted out to the far side of the tab. If that > turns out > to be the case, adding a heatsink (for additional thermal > inertia at > this point) might not buy you much. That's what I'm concerned about and is really what this thread is about. Hmmm |