Discoloration? "max" surface temperature for a PCB/SMT resistor?

[incf]

When sizing banks of SMT resistors there are tradeoffs to be made.

In my case I want size it "just right" so that the resistors hit a certain surface temperature. If the temperature is too low, I'm wasting valuable space. If the temperature is too high, I'm compromising reliability.

• But, how do you chose an "ideal" surface temperature?

I feel like 30C above an ambient temperature of 70C seems "plausible" (ie. surface temperature of 50C at room temperature and 100C at the max operating temperature)

The above is just a "gut feeling" and is not based on anything. I'd like to somehow justify that a 30C temperature rise (or any number really) is a "good" target.

I don't have any long term reliability data to work from.

I also wonder if it matters how many hours per day the resistor operates at its max temperature. For example it spends 1% of its time at max temperature vs 100% of its time.

In my case, all of the components, including capacitors are rated to last a very long time at high temperatures. I think my main concern is thermal expansion causing mechanical stress.

*surface temperature = the highest temperature on the surface of a resistor when measured with an IR camera

--------

Addendum: what is the maximum temperature that a PCB can take without discoloring?

[TimFox]

The linear expansion temperature co-efficient of SMT parts is in the ballpark of 10 ppm/K.

[Smokey]

That's a big temp delta.  Better get good tempco resistors or their values will be all over the place depending on operating condition (or measure temp and compensate in software).

I typically try not to push the edge of temp rise on smd resistors.  If it's going to be dissipating appreciable power then I'll just size it up until the dissipation is well under it's rating.  This is usually a pretty easy decision since resistor values for this type of thing tend to be odd ball values and not common with any other BOM items so I don't feel bad about not tying to consolidate. 

You can always add parallel footprints and share the power between multiple parts too.  If it turns out to be unnecessary, then just don't populate those parallel parts.

[incf]

It's a bank of paralleled 2512 power resistors in a power supply application whose value are not very critical.

edit: with copper pour heatsinks around each resistor

Which type of temperature coefficient are you referring to? If it's resistance temperature coefficient that's not a problem.

[Kurets]

2512 even at 50% of their rated power will run so hot that they discolour the board. With time I expect that board and solder joint reliability is degraded.

In general I think avoiding 85C chip temp in worst case ambient temp is a good idea. But this cannot be just based on resistor rating/datasheet, your board will heat up based on all dissipators on the board. So in the end probably max delta T for a single device is something like max 10C.

[incf]

2512 even at 50% of their rated power will run so hot that they discolour the board. With time I expect that board and solder joint reliability is degraded.

In general I think avoiding 85C chip temp in worst case ambient temp is a good idea. But this cannot be just based on resistor rating/datasheet, your board will heat up based on all dissipators on the board. So in the end probably max delta T for a single device is something like max 10C.

Any idea what exact PCB temperature results in board discoloration? (my resistors have a copper pour heatsink and run cooler than one might expect)

Cheap FR4 has a Tg of 135°C. And allegedly can "operate" (maintain electrical characteristics) at >110C? [1]

[1] https://electronics.stackexchange.com/questions/110972/what-is-a-suitable-temperature-rise-for-pcb-traces
[2] "When selecting the right FR4 PCB, you should always allocate at least a 20°C margin. For example, a lower Tg FR4 of 130°C should have an operating temperature limit of 110°C." ("assypcb: What's the Highest Temperature Can PCB FR4 Withstand")

[Smokey]

Anything that goes through lead free reflow (which is essentially everything that gets machine assembled) I use minimum of 170C Tg boards just for that. 

If changing the design for those resistors an option?  Moving to a heatsink maybe?

[incf]

Anything that goes through lead free reflow (which is essentially everything that gets machine assembled) I use minimum of 170C Tg boards just for that. 

If changing the design for those resistors an option?  Moving to a heatsink maybe?

The architecture is set in stone - and in short term testing it works well in our system.

I'm in the process of revising the design to have a larger temperature rise spread over fewer resistors so that I can reuse valuable real-estate on the PCB for other purposes.

My fear is that I might run into issues long term by being too aggressive with the temperature rise. I lack any concrete figures about temperatures and longevity - which makes it difficult speculate about whether a certain temperature is too high or not.