It appears as though he removed the solder mask and added a paste mask
to the bottom layer. From what I understand, there is a thick layer of
substrate that typically is made of FR4 which results in low heat
conductivity between the top and bottom layers. Assuming his substrate
is FR4, Is it possible to heat the paste mask of the top layer through
the FR4? Or did he do something else to the layers?
I would not rate it as very likely that one could directly heat the top side solder paste from the bottom going through only FR4. It looks like he used two techniques in the video.
In one case he designed his board to have a large hole in it under the thermal pad. In that case he just soldered the thermal pad through the hole. He points to the holes at time code 1:13.
In the other case he made a pad on the bottom side that connected to the top using a bunch of thermal vias. You can see the vias when you look at the thermal pads on his board. They are filled but you can still see little circles.
At time code 3:27 you can see the pad for the IC (just above the big capacitor, on the bottom, just right of center). In the pad you can see four dimples for the thermal vias.
If you are designing your own board and can put in the thermal vias, then go for it.
If you can't do that then your best bet is going to be to use a small heat gun. You can get one for under $20. For example...
One issue with hot air guns (as mentioned in your linked video) is that they can put a lot of heat onto areas of the board that you don't want, which can de-solder other parts. The way I usually solve this is to take a piece of scrap sheet metal (say 6" x 6" or larger) and drill a hole in it that is a little bigger than the IC (say 1/4" or 1/2"). Place the metal over the PCB with the hole over your component. Aim the heat gun at the hole from about an inch away. The sheet metal will mostly shield the board from any stray heat, except where the hole is of course.
The thermal pad of the LED is supposed to be connected to a relatively
large pad on the top layer. Is that so that the copper can conduct the
heat away from the LED?
Yes it is.
From my understanding, the copper sits atop the substrate - how does
the heat escape if the substrate has low thermal conductivity?
Your LED part appears to make around 1W of waste heat + 0.8W of light with 1.8W of input power. If you have a copper thermal pad on your board, there are two heat paths.
- Heat can move through the board to a heat-sink on the back side.
The thermal conductivity of FR4 is about 1000X worse than copper. But if you have a wide thin slice it can still conduct useful amounts of heat. FR4 has a thermal conductivity of about 0.3W/m-k. For example a 1 inch square that is 60 mils thick has a thermal conductivity of 0.127 W / K. So 1W / (0.127 W/K) = 7.9 Kelvin temperature rise from front to back of board with the LED running at full power.
Using a grid of thermal vias located in the thermal pad of the LED you can substantially increase the flow of heat. By filling the thermal vias with thermally conductive epoxy it can be increased even more.
- Heat can move directly from the thermal pad into the surrounding air (without going through the board). Typical numbers would be around 40 K/W for a 1 inch square in "still" air that only has natural convection.
If I added a via in the middle of the mask, and there is nothing in
the bottom layer which connects to the via, is there a chance of
shorting the circuit?
There is always some chance of shorting the circuit. But as long as there is no metal connecting between the via and the plane then there will be no short. You can of course make a bottom side thermal pad as long as you keep it isolated from everything else.
I imagine that since the copper plate in the bottom layer is
surrounded by insulators, the resistance will be very high and no
current will flow, is that right?
FR4 has a very high (but not infinite) resistance. Some current will flow but its going to be below the pico-amp range, so it probably won't affect your application.