In my PCB design, I have 0402 components connected to polygons. To reduce the risk of tombstoning, I connected them using thermal relief.

My guiding principle is that if heat dissipation is not symmetrical on both sides of the component, and the component is 0603 or smaller, thermal relief should be applied. I differentiate between three cases:

Both pads are connected to polygons. (See Image 1: C14, C15, C16, C17)

Only one pad is connected to a polygon. (See Image 1: R15)

Neither pad is connected to a polygon. (See Image 2: C12, R11)

In the third case, I made an effort to route the traces symmetrically to the component to ensure even heat distribution. This follows recommendations I found online to prevent tombstoning—see Image 3 (top: preferred routing, bottom: non-preferred).

Another consideration is whether the components will be soldered or desoldered manually, such as jumpers.

Additionally, I assume that for TH components, thermal relief should always be used when they are connected to a plane/polygon.

My Questions:

Are my assumptions correct? Should thermal relief be applied in each of these cases?

If thermal relief is used for thermal reasons, is it important that the number of connections (or their total width) be symmetrical?

When considering symmetry, should factors like polygon size, nearby vias, and other thermal paths be taken into account? Is this generally based on estimation or engineering judgment?






Thanks in advanced for any answer,
Eyal

 

Picture 1 U3 has no thermals on pin 4 and 8.

I make high end boards where there is high current and extreme temperatures. The boards are mostly copper.
The proto boards have no thermals. None. It is hard to hand solder with no thermals. The machines have no problem.

Manufacturing called me from Singapore and said the parts will not solder. The solder will not flow on these parts. I knew they heat the entire board to the milting point, and they do solder. (I know the book said they will not.) I solder here with a small oven, and it works well. I challenged manufacturing to run a board and see what happens. I asked for 10% more time (heat) than normal. They ran the boards at normal time and speed and all the parts were good. (granted I have very little 0402, mostly big parts, and wild mixes of small and large)

Taking one of your pictures, this is how my boards look. I don't think the parts will twist. At 90 to 95% copper the board will cool down even. IMO


I want good connections to Ground. I often have more than one via to ground. No thermals. Traces as wide as I can to pull out the heat when operating. I certainly would not have a trace dance around the board making a long inductive and resistive trace. (Somes time I do if I want delay in that signal)

I know this is not "normal". My boards are tested at temperatures near the melting point of solder. I don't have tombstoning but I might just be lucky.

 

@ronsimpson, thank you for your reply!

This board is intended for machine soldering only. All components are SMT (that means that it would be reflow soldering only, correct?)

When you say "thermals" do you mean thermal reliefs?

Do you suggest that U3\4 and U3\8 don't have thermals? I'm not sure I understood

Regarding U10, don't you think it is better to apply thermal relief at the upper pad in order to make both pad a little more thermal symmetric, instead of the upper connected to a large polygon and the bottom is connected only to a trace?