# Rules for choosing power rating of a resistor

Let's say we're designing a system, and within it we have some resistor $R$ with a voltage difference $V$ across its terminals. Then it dissipates power $P = V^2 / R$.

Of course the resistor has a certain tolerance, so its value lies between $R - \Delta R$ and $R + \Delta R$.

The voltage has also a certain "tolerance" (maximum ripple over a regulated nominal voltage), so its value lies between $V - \Delta V$ and $V + \Delta V$.

When deciding what power rating to pick for the resistor, I believe it makes sense to put ourselves in the worst possible situation, so the highest possible dissipated power would be $P_{max} = (V + \Delta V)^2 / (R - \Delta R)$.

If we assume we precisely know the values for $\Delta R$ and $\Delta V$, how much bigger than $P_{max}$ should the resistor rating be? We already are in the worst possible scenario, should we give ourselves some more error-space anyway?

Also, is there a more robust method to compute power rating than the one I mentioned?

• 20%. For engineering. – Ignacio Vazquez-Abrams Apr 28 '16 at 8:09
• If you want to do this exactly right, you might want to consider derating for the specific environment, or again for the worst case environment. Or you take a gut feeling 25% and call it a day. – PlasmaHH Apr 28 '16 at 8:10
• Heat causes damage but, if the voltage is AC, the time lag is typically enough that you don't consider the peak, only the RMS. This is a bit of a spoiler - you have to consider thermal capacity and where do you draw the line for this resistor or that resistor i.e. at what low frequency do we consider peak instead of RMS. – Andy aka Apr 28 '16 at 8:15
• I calcuate the resistor to know what power is dissipated, then I choose higher rating. – Marko Buršič Apr 28 '16 at 8:16
• I like my resistors to run cool for reliability, so I tend to pick resistors whose power rating is roughly double the average dissipation that I've calculated. – Dave Tweed Apr 28 '16 at 11:27

In engineering for commercial test equipment, we had a default rule that resistors should never be run beyond 50% of their rated power. This rule meant you didn't have to consider their contribution to the overall equipment MTBF, when the resistor manufacturers rate the dissipations for ambient temperature and a X0000 hours lifetime. Every so often, we couldn't meet this generous over-design, and then we'd need to do detailed calculations, including ambient temperature.

For military or auto under-hood electronics, you would derate even further for reliability.

• if you had to put a number to that(military/auto), what would that be? – seetharaman Apr 28 '16 at 15:56
• Less than 50%, perhaps 25%. <insert tongue in cheek>Bear in mind that auto electronics has to be waaaay more reliable than military. If I build 100 military radios, 10 go on a mission and one fails, that defect is lost in the noise of battle. If I build 1m EMUs, 1 million go on the road, if 100 fail I get business-killing publicity and a big recall </remove tongue> – Neil_UK Apr 28 '16 at 18:29
• hmm.. But I thought military standards are more stringent than automotive!! – seetharaman Apr 29 '16 at 13:46
• @seetharaman that's exactly it, you'd like to think so wouldn't you! But commercial beats operational. – Neil_UK Apr 29 '16 at 16:37
• if I understand correctly, you mean to say that military standards are more stringent than automotive, but only on paper. Right?? – seetharaman Apr 29 '16 at 17:10

As Military was mentioned I will specifically address it.

I am in Mil/Aero currently and aerospace applications for commercial aircraft have stringent reliability requirements as does military equipment.

Take the venerable CRCW series of surface mount devices; these derate linearly from 70C.

In these applications, we are conduction cooled (convection cooling loses meaning in an unpressurised avionics bay as the molecular density doesn't support much thermal transfer), so we take the card edge temperature (typically we have to withstand about 85C there) and add the thermal impedance of the PCB, so we usually end up with an environment of about 90 to 95C at the component.

At that temperature, the rated power of the part is about 65% of the nominal rated power at 70C. For reliability we then derate that power by 50%, so the effective rating of the part in this environment is about 30% of the nominal rated power.

Note that even at that derating value, it still forms part of the reliability analysis (usually done to MIL-HDK-217).

Note: For this series of parts, the thermal rule is you cannot exceed 155C within the part itself (see datasheet).

If you are running at close to the rated power, consider the environmental temperature , and the effect that temperature may have on other parts.

But generally I stick to 75% of the rated power.

Running at rated temperature (or even 75%) requires flawless cooling in ideal operating conditions - the rating assumes that. Even then your resistor will have a surface temp of 600-700F which is enough to set a lot of things on fire.

I am not bashful about overrating by 400 percent. On a project where I did just that, I still had a 260F rise. It was worth the peace of mind of not having to worry about that component in decidedly non-ideal operating conditions.

If I want reliability and the resistor in question is dissipating sufficient power that it's worth considering (rather than "small signal") I would derate by at least 50%. Even quite low apparent heat levels have an insidious effect on reliability over the long term. I would also give as much "breathing room" in terms of space and airflow as possible.

I just dismantled for refurbishment a power amplifier I built 29 years ago (boy, I feel old) and was rather gratified to find no symptoms of overheating anywhere.