Resistor power ratings

[OM222O]

I was shopping for some high power smd resistors with decent temp co and I came across these 2:
RHFH4Q001R0FE (data sheet at: https://www.mouser.co.uk/datasheet/2/611/rhf-series-1074351.pdf)
RW5S0FA1R00FE (data sheet at: https://www.mouser.co.uk/datasheet/2/303/res_rw-1265480.pdf)

to me this makes no sense  how can the much much smaller package (DPAK , not even D2PAK!) be rated for 35 watts, while the much larger package, wirewound resistor be only rated for 5? similar packages are rated in the same region: 1 to 5 watts, so I'm not sure how they claim a 35 watt figure?
is that under the assumption of infinite heat sink or do they mean 35 watts without any heat sink (like any other manufacturer states the power ratings)?

even assuming a heat sink (a realistic one, not an infinite case under refrigerant or something), a DPAK would pretty much de solder itself under a 35 watt load, given how small the package is! Am I missing something here?

edit: after checking the notes on the datasheet, they claim 2 watts without heat sink, but I'm still not convinced it can pull 35 watts under any reasonable heat sink!

[wraper]

is that under the assumption of infinite heat sink or do they mean 35 watts without any heat sink (like any other manufacturer states the power ratings)?

Do they? First thing I see in the datasheet:

POWER RATING ¹
(with heatsink)

¹ 2W on simple solder pad

And then again:

Power Rating (with heatsink) 35 W ( 2W on Simple Solder Pad )

Did you even spend 30 seconds reading datasheet?

[OM222O]

yes, my bad. I had already edited the first post though.
I'm still not convinced you can cool 35 watts on a DPAK with any heat sink!
I did a quick search on mouser (only stocked options) and the best ones they have are rated for about 3 to 4 c/w

that is about 140 degrees, assuming 0 thermal resistance from the resistor to the pads! add the package thermal resistance and you're already at double that, about 300c! as I suspected before, solder melts at those temperatures!

[wraper]

edit: after checking the notes on the datasheet, they claim 2 watts without heat sink, but I'm still not convinced it can pull 35 watts under any reasonable heat sink!

If you attach in directly to heatsink (only leads soldered) it could do that just fine, though I would not recommend. Package itself is basically TO-220 with a screw tab cut off (still can be used with spring clip). Otherwise the best you could do is using aluminium PCB mounted on heatsink.

that is about 140 degrees, assuming 0 thermal resistance from the resistor to the pads!

Thermal resistance is 3.3 K/W, That means 115.5^oC rise at 35W. Max internal resistor temperature is 175°C. That's not unreal if you keep the tab cool. Though note that max spec is not something you should use constantly.

[wraper]

Then also look on power derating curve, you can run it at full power if keep the tab below 50^oC.

[OM222O]

Thermal resistance is 3.3 K/W, That means 115.5^oC rise at 35W. Max temperature is 155^oC. That's not unreal if you keep the tab cool.

That is the "junction to case" resistance. you must add the heat sink resistance too! also the temperature range specified is only upto 155c! that means with a heat sink that has zero thermal resistance (impossible but lets just assume we have that), at an ambient of 25c we're already at junction temperature of 140c! ouch! 

that basically means the maximum allowed thermal resistance for PCB + heat sink + whatever means of cooling the device should be less than 0.4 c/w! very unrealistic!
to me this really sounds like misinformation since you MUST use sub zero cooling (chilled water or some kind of refrigerant) to get to that 35 watt rating, which is incredibly impractical for a single resistor! I even looked at TO-220 resistors and most of them don't go above 8 watts (there a few claiming 100 watt but I assume they are using the same technique) which confirms 35 watts for such a small package is unreasonable!

[wraper]

hat is the "junction to case" resistance. you must add the heat sink resistance too! also the temperature range specified is only upto 155c! that means with a heat sink that has zero thermal resistance (impossible but lets just assume we have that), at an ambient of 25c we're already at junction temperature of 140c! ouch! 

Read the freaking datasheet  . Operating temperature and Max internal temperature is not the same thing. What ouch, it's a resistor, not semiconductor, although some of those also can operate at such temperature. For example you could put it directly on a heat pipe, then running it at low tab temperature is not a problem.
Then just open some power MOSFET datasheet and look on those power ratings. Max power rating at certain conditions does not mean it's intended or is practical to be used so.

[OM222O]

"The maximum internal resistor temperature is 175°"
that is a direct quote from the datasheet.

What ouch, it's a resistor, not semiconductor, although some of those also can operate at such temperature. For example you could put it directly on a heat pipe, then running it at low tab temperature is not a problem.
Then just open some power MOSFET datasheet and look on those power ratings. Max power rating at certain conditions does not mean it's intended or is practical to be used so.

it's not about the resistor itself, the plastic packaging is also likely to melt at similar temperatures. it can't be mounted on a heat pipe since it's a SMD package, you must use heat sinks with ridiculously low thermal resistances (you don't seem to differentiate between junction to case thermal resistance and case to ambient thermal resistance, rather combine them into a general "thermal resistance" term). mosfets are meant for switching loads, not continuous power dissipation! that's a completely different story (the data sheet for the resistor claims 400 watts for peak power!).

I had a DPAK darlington pair fail with a constant 3w load, which had a 1c/w junction to case thermal resistance. that is 3 times less than the resistor package since they advertise "Isolated Back Plate" which adds significantly to the thermal resistance. as you mentioned, under 35 watt load, with an ambient of 0c, with infinite heat sink, the resistor is already sitting at 115c, at which point it must be de rated to just about 50%! I had to use a D2PAK package (much much larger than DPAK) with a lot of thermal vias to cool just 3 watts without the device getting too hot and failing. this 35 rating doesn't make any sense with the numbers pulled out of the same datasheet! please do the calculations yourself and you'll realize this is not realistic at all.

[wraper]

That plastic will never melt, only disintegrate at extremely high temperature and 175^oC is nothing to it. It's not a thermoplastic. If it would melt, you would be left with a pile of plastic over PCB after a reflow at 260^oC.

you must use heat sinks with ridiculously low thermal resistances (you don't seem to differentiate between junction to case thermal resistance and case to ambient thermal resistance, rather combine them into a general "thermal resistance" term). mosfets are meant for switching loads, not continuous power dissipation! that's a completely different story (the data sheet for the resistor claims 400 watts for peak power!).

If you have a had above your shoulders, you should find it's not practical to be used at max rating. Max specs in the datasheet are given according to what can be done, not what is practical to do.

(the data sheet for the resistor claims 400 watts for peak power!)

So what, look at duration.

[wraper]

under 35 watt load, with an ambient of 0c, with infinite heat sink, the resistor is already sitting at 115c, at which point it must be de rated to just about 50%! I had to use a D2PAK package (much much larger than DPAK) with a lot of thermal vias to cool just 3 watts without the device getting too hot and failing. this 35 rating doesn't make any sense with the numbers pulled out of the same datasheet! please do the calculations yourself and you'll realize this is not realistic at all.

Nonsense. It does not need to be derated unless TAB temperature is above 50^oC. Yet again you replace one spec with another and make nonsense conclusions  .

[wraper]

BTW it's TO-263 = D2PAK, not DPAK. Seems they made a typo.

[OM222O]

under 35 watt load, with an ambient of 0c, with infinite heat sink, the resistor is already sitting at 115c, at which point it must be de rated to just about 50%! I had to use a D2PAK package (much much larger than DPAK) with a lot of thermal vias to cool just 3 watts without the device getting too hot and failing. this 35 rating doesn't make any sense with the numbers pulled out of the same datasheet! please do the calculations yourself and you'll realize this is not realistic at all.

Nonsense. It does not need to be derated unless TAB temperature is above 50^oC. Yet again you replace one spec with another and make nonsense conclusions  .

before answering with such certainty and aggression, I suggest you read the difference between different thermal resistance meanings  I used the tab tempreture which can be calculated by "T_ambient + (R_j-c + R_c-a) * Power" and assumed an ideal Rc-a of 0 as well as 0c ambinet. the tab will reach 115C under the limitations of the package itself, EXCLUDING any external factors. Have a nice day.

A quick side note: R_j-c + R_c-a is the same as R_j-a. in case of this device, the R_j-c alone is 3.3c/w 
Here is an example from a random N channel fet data sheet:

As you can see , the package itself has a limit of 0.52c/w. that would be your thermal resistance with an ideal heat sink!
the more realistic figure would be the 62.5c/w which is calculated without any heat sink!

This is really huge compared to any heat sink (again, about 3 or 4 c/w) so you can just add the 0.52 to the thermal resistance of your heat sink to get your overall thermal resistance. multiply that by the power and you will have your temperature rise. add that to your ambient temperature and there is your "tab temperature".
Again, the tab is NOT isolated in most cases, which gives you a figure between 0.5 and 1c/w for junction to case resistance. in this case they are isolating the tab which leads to a 3 times higher figure of 3.3 

[T3sl4co1l]

yes, my bad. I had already edited the first post though.
I'm still not convinced you can cool 35 watts on a DPAK with any heat sink!

I can't say I would recommend it.  But consider what they're really rating.

Case Temperature = 25°C

Doesn't matter how you get there, doesn't matter how you have to do it, if it's a bath of boiling freon (which most times, is exactly what they do!).

Shitty, but that's how it is.  Anyway, they can't give you an arbitrary number for just any heatsink, that wouldn't be very helpful either.  What is useful is the RthJC, and with the Tj(max) and some maximum ambient, and some heatsink Rth, you can calculate the real power dissipation capacity, for any heatsink, given you have these numbers.

Tim

[OM222O]

yes, my bad. I had already edited the first post though.
I'm still not convinced you can cool 35 watts on a DPAK with any heat sink!

I can't say I would recommend it.  But consider what they're really rating.

Case Temperature = 25°C

Doesn't matter how you get there, doesn't matter how you have to do it, if it's a bath of boiling freon (which most times, is exactly what they do!).

Shitty, but that's how it is.  Anyway, they can't give you an arbitrary number for just any heatsink, that wouldn't be very helpful either.  What is useful is the RthJC, and with the Tj(max) and some maximum ambient, and some heatsink Rth, you can calculate the real power dissipation capacity, for any heatsink, given you have these numbers.

Tim

I knew there was something fishy, even mentioned sub zero cooling 
It is very misleading and I was quite surprised when I first saw it. this is honestly the first time I see this done for a resistor since manufacturers use practical figures since they know if you have spent 2$ on a resistor, you don't want to be spending 50$ on a cooling setup! I'm by no means a professional, just a hobbyist. maybe this practice of over rating components is more common in the industry than I expected. I had only seen mostly on transistors before.

[wraper]

before answering with such certainty and aggression, I suggest you read the difference between different thermal resistance meanings  I used the tab tempreture which can be calculated by "T_ambient + (R_j-c + R_c-a) * Power" and assumed an ideal Rc-a of 0 as well as 0c ambinet. the tab will reach 115C under the limitations of the package itself, EXCLUDING any external factors. Have a nice day.

Before answering, recheck if you are understanding what you are answering. It's funny how you used internal thermal resistance between tab and resistor to calculate tab temperature outside the device which has nothing to do with it.

A quick side note: R_j-c + R_c-a is the same as R_j-a. in case of this device, the R_j-c alone is 3.3c/w 
Here is an example from a random N channel fet data sheet:

As you can see , the package itself has a limit of 0.52c/w. that would be your thermal resistance with an ideal heat sink!
the more realistic figure would be the 62.5c/w which is calculated without any heat sink!

This is really huge compared to any heat sink (again, about 3 or 4 c/w) so you can just add the 0.52 to the thermal resistance of your heat sink to get your overall thermal resistance. multiply that by the power and you will have your temperature rise. add that to your ambient temperature and there is your "tab temperature".
Again, the tab is NOT isolated in most cases, which gives you a figure between 0.5 and 1c/w for junction to case resistance. in this case they are isolating the tab which leads to a 3 times higher figure of 3.3 

You put different specs into one pile. To not confuse yourself, using max junction temperature and junction to case thermal resistance, calculate what max temperature tab is allowed to reach. Then select heatsink and thermal interface accordingly to keep it at lower temperature. Instead of calling part flapping in the breeze "realistic".

[wraper]

Again, temperature of backplate, not internal resistor.

[OM222O]

The temperature of back plate and "internal resistor" are not two separate things in a different universe. depending on the power dissipated by the "internal resistor" the temperature rise to the tab will be that power times 3.3! adding that to ambient temperature will be the final tab temp. it's honestly simple math, not rocket science. unless since last time I checked the definition of thermal resistance and power have changed ... Again I'm not sure where the hostility comes from. some input from other users won't be bad either. please let them respond rather than shouting the same thing over and over again 

[wraper]

The temperature of back plate and "internal resistor" are not two separate things in a different universe. depending on the power dissipated by the "internal resistor" the temperature rise to the tab will be that power times 3.3! adding that to ambient temperature will be the final tab temp. it's honestly simple math, not rocket science. unless since last time I checked the definition of thermal resistance and power have changed ... Again I'm not sure where the hostility comes from. some input from other users won't be bad either. please let them respond rather than shouting the same thing over and over again 

Oh, my.
Tab temperature = (heatsink temperature) + (thermal resistance of thermal interface, say thermal paste) * (power dissipated).
3.3 * power is resistor temperature above tab temperature. NOT TAB TEMPERATURE!

P.S.
Heatsink temperature = (ambient temperature) + (heatsink thermal resistance) * (power dissipated). If heatsink has zero thermal resistance, it stays at ambient temperature regardless of power dissipated. Thus in ideal case tab temperature is only determined by thermal resistance of thermal interface (thermal paste) and dissipated power. Specs in this datasheet other than tab size have nothing to do with either of those. But often you can find Case-to-heatsink thermal resistance spec which is basically a ballpark of what you could expect. But it really depends on thermal interface you will use.

[OM222O]

ah yes, you are right. I got that backwards  the temperature of resistor will be higher not the tab. 
disregarding the de rating, the max temperature of the resistor is uncomfortably close to the operating point, even with an ideal heat sink at 35 watts. as tim mentioned, those figures are probably measured under a liquid freon bath.

[wraper]

ah yes, you are right. I got that backwards  the temperature of resistor will be higher not the tab. 
disregarding the de rating, the max temperature of the resistor is uncomfortably close to the operating point, even with an ideal heat sink at 35 watts. as tim mentioned, those figures are probably measured under a liquid freon bath.

If you mount it on a modern air cooled CPU heatsink with heat pipes, it will stay below 50^oC easily. Again, max spec is max spec, not something necessarily practical to do. It shows limits of the component (which BTW is not something extreme in this case). Not that you should push it to this limit. Usually it's more practical to put several parts in parallel and put them on usual heatsink rather than put one part and cool it with liquid nitrogen.

[Siwastaja]

The datasheets do not do the design for you. They give you all the data relevant to the part to do the thermal calculations, but you need to have external data to calculate/simulate the whole system, including the PCB.

For through-hole devices, this was easy as there were no custom parts in the thermal circuit: just add together Rth of junction-to-case, Rth of thermal interface material, and Rth of the heatsink.

But for SMD devices, PCB is always highly relevant; either it is your heatsink, or even if you add an external heatsink, the PCB acts in series and often has significant thermal resistance. This tends to be the missing piece, because you are the designer of the PCB, and hence no one else can produce that number for you.

Device datasheets tend to give you two extreme Rths, one wildly optimistic (device attached to an infinite heatsink with an extremely good thermal interface material), one wildly pessimistic (device mounted on a single-side FR4 PCB with minimal footprint, no internal layers, no thermal vias)... For your actual PCB, you'd either simulate it with some expensive thermal simulation software, or do your own simple guesstimates. I do simple back-of-the-envelope style thermal models in Excel.

Note that aluminum core PCBs are a real possibility. Coupled with a massive heatsink and highly thermally conductive and thin thermal pad as the interface medium, dissipation near 35W from a large DPAK/D2PAK size SMD device is certainly not impossible.

For standard FR4 without paying extra for copper-filled vias, it will be very difficult to get below around 10 K/W for the case-to-heatsink, as you need to go through thinly plated vias, and you can't extend the via array to infinity, as you'd face increasing thermal resistance on your top layer before getting to these vias.

This calculator: http://circuitcalculator.com/wordpress/2006/03/12/pcb-via-calculator/  gives you the Rth of your vias.

[David Hess]

Film resistors can have a much lower element-to-case thermal resistance than a wirewound resistor for a given size.  That high power rating only applies up to a 50C case temperature so suitable power derating must be added to that which depends on the mounting and heat sink.