How much current can thermal reliefs handle?

[Clear as mud]

I guess this is a beginner question, and one I didn't see answered elsewhere on the forum.

I've been designing a PCB with some relatively high currents in parts of it.  Some of the through-hole pads will have 10 amps, a couple of them will even have 20.

I've used solid connections for all these high-current pads, although I know it will make it harder to solder the parts on.  The question is, how much current can a thermal relief handle?  And at what point, in terms of expected current through the pad, do you start designing the board with solid connections instead of thermal reliefs?  I've been specifying 10 mils (0.254mm) thermal clearance and 20 mils (0.508mm) thermal spoke width.  So basically the pad would be connected to the surrounding copper by four traces 10 mils long and 20 mils wide.  I think the default is 20 and 20 instead of 10 and 20.

Using 2-ounce copper, plugging some numbers into an online PCB trace width calculator, I get about 27 degrees temperature rise at 10 amps.  Does this seem accurate in real-world conditions?

[thm_w]

In reality expect less than 27deg as the heat is transmitted to the nearby copper planes.

https://web.pa.msu.edu/hep/atlas/l1calo/hub/hardware/components/power/synqor_thermal_relief_study.pdf
https://www.eevblog.com/forum/eda/high-current-traces-and-thermal-relief/
https://electronics.stackexchange.com/questions/19479/how-to-design-high-current-through-hole-solder-connection-to-large-planes
youtube.com/watch?v=zbWwdEyCOUE

4x20mil spoke should be fine as that is 80mil total. But there may be other considerations here, eg thermal transfer from the lead, etc.

[T3sl4co1l]

It will be far less than the temp rise of a trace, because a trace assumes lateral heat flow only.  The spokes are heatsinked by the pin and pours.

You can calculate it yourself!  The resistance of copper is known, the dimensions are known, the current flow is known, and the thermal conductivity is known.  You can make the assumption that current and heat flows are uniform (lengthwise) within the spoke, which won't be a bad guess.

Small spokes are fine at 10A, that's nothing special.  I'd worry about direct connect at around, hrm, 60A let's say?  (Scaling the spokes along the way, e.g. say, 10x10 mil x 4 spokes for 70 mil o.d. pads under 5A, 10x15 to 10x20 for 70-120 mil pads 5-15A, etc.  A 60A pad is going to be pretty big, like 250 mils or more o.d., so will have quite a lot of copper coming off of it.  Though I don't like to use direct connections up there, anyway; terminals are better.)

Tim

[floobydust]

I don't use thermal reliefs for higher current or mains.

I've seen them act as fuses, if you have any inrush or overloads that could happen, the spokes melt.
If there is mains involved, an arc can then start which then burns everything up. You say 10A but it better be fused or limited somewhere. 80mil is kinda thin.

Components such as relays, terminals, fuses do generate their own heat due to their resistance.
For best cooling, solid copper-pours give some needed heatsinking to the components.  I have seen one appliance product recall where the relay's skinny pcb pad made extra heat which started a failure that caused fires. Who'd have thought that.

If a tech has to whine and complain about desoldering parts now without thermal reliefs, just turn up the iron's heat.
https://www.ultracad.com/articles/reprints/codreanu.pdf {Onderdonk’s formula has typo}

[tszaboo]

I wouldnt place thermal reliefs. They will use selective wave soldering or wave soldering for the board, the only parameter they need to adjust is the time. I assume that you dont really care about the yield, if it takes them a few seconds per board extra, just pay it.

On the other hand, funny stuff it, the heat conductivity and electrical conductivity of materials are closely correlated. If you decrease the heat conductivity, you decrease the electrical conductivity. If it overheats, fails, it will ned up on your desk.

No thermal relief -> someone else's problem
thermal relief -> your problem

You can calculate it yourself!  The resistance of copper is known, ...

Most of us dont have a finite element simulator, and gerber to model converter lying around.

[T3sl4co1l]

Most of us dont have a finite element simulator, and gerber to model converter lying around.

You really need a whole-ass simulator for a single finite element? Come on, dude!

Tim

[Clear as mud]

Some of the pads are after a fuse, some are un-fused, so if thermal reliefs can act as a fuse, I wouldn't want them everywhere.

It's actually the same board I posted about before, a low-quantity project that someone is going to hand-assemble.  However, my customer already said he thought it would be OK without thermal reliefs, so I don't think I'll go back and add them now.

I wrote a whole wall of text for my most recent previous post about it, so I didn't get any replies, but there are pictures there of the project status a week ago.  It's pretty much done now.