SMD reflow with a blowtorch, try #3

See try #1 and try #2.

First, I have tried to fix the board from try #2 by pointing the torch vertically down, with the board ~20 cm from the tip of the flame, for ~90 s. There was no difference. It will become clear why after considering my second attempt.

Second, I have tried to solder another SLG46620V onto an identical breakout board, using the same procedure as described above. It was immediately clear that this won’t work—the package hasn’t even centered itself over the footprint, which meant that most of the solder definitely didn’t liquefy. Despite the success in try #1, it seems that ultimately pointing the torch down at the board is at best very unreliable.

I suspect that the primary reason is that the rather hot air is drawn upwards by its buoyance much faster than I expected, and most of it does not actually reach the board.

Third, I have heated the board again, but from underside, keeping the tip of the flame ~15 cm under the board, and soldering for ~90 s as well. This makes it much easier to observe the package, and it centered itself over the footprint along with the solder becoming reflective at ~60 s into the cycle.

This second SLG46620V was programmed to output a 25 kHz square wave on every pin instead of 1780 Hz to make any soldering failures even more visible:

1
2
3
4
5

module top(output [16:0] o);
  wire clk;
  GP_RCOSC #(.OSC_FREQ("25k")) rcosc(.CLKOUT_FABRIC(clk));
  assign o = {17{clk}};
endmodule

In a way this has worked even worse; all board pins except 16 and 17 have a perfect square wave but board pins 16 and 17 are not connected to the package pads whatsoever. (The signal appears as it supposed to on the exposed sides of the QFN pads, which can be carefully probed.)

In conclusion, I will shelve this method as unreliable unless I discover that something else in my process is at fault.