Re: OT: Soldering SMDs (was Re: [avr-chat] isp-usb)

[Don Ingram]

> One day I must try the upturned (clothes) iron method, where the PCB is
> slid onto the hot iron, and off again when the solder has reflowed,
> after a few seconds. It's only going to work for the one side, I guess,
> but so far I've only put a few componenets on the back. They could be
> hand soldered afterwards. (And now, more easily. :-)
>
> I like the idea of most of the heat reaching the solder before the
> chips.

Be very careful with this!!!!!!

As the board heats it is constrained by glass mat in the X & Y axes but not in
the Z axis. Epoxy expands considerably when heated above glass transition
temperature and the Z axis is the only way it can go. If overheated or heated
for too long the FR4 will swell and tear the barrels out of the via's &
component holes. When the board cools they return to size and provide
intermittent connectivity with the pads leaving a marvellous little box of
intermittent tracks and pads.

Skillets are in a similar boat, the heating pattern can be very uneven. Just
toss in some water & see where it boils. Usually in a narrow band above the
actual curved heating element. Compensating for one area can fry another area, 
if you are lucky there will be a reasonable consistent zone in the centre, 
particularly if the surface is quite thick. Similarly a test of the hotplate can 
reveal some areas broad enough to handle a board with a consistent level of heating.

Done correctly the IR toaster can be very effective, but check the heat 
distribution first, a bit of alfoil around the inside can shape the temp 
distribution a bit better and remove the cold sinks from the walls. Component 
choice / mass / location is always an issue but this is part of the designers 
job to ensure that it will be manufacturable. Sometimes it is just plain easier 
to manually fit the last few bits at the end of the process for very low quantities.

Either of the above "kind of" works but just don't be surprised if the results
are occasionally rather disappointing. The problem described above is not an 
infrequent occurance on an overheated PCB. A test run with a low mass 
thermocouple affixed to a scrap PCB with high temp solder can give you some 
useful feedback. Similarly the temp sensitive plastic provided with the proto 
stencil kits can provide useful feedback. Toss the time/temp data into excel and 
have a look at the reflow profile that you see on your rig & compare it to a 
published profile from one of the IC manufacturers, this will allow you to tune 
your process to get a more consistent and reliable result.

SMD lines in factories achieve incredible levels of reliability due to good 
attention to consistency in their processes. A little of this attention to 
detail at home can provide some pleasing results too.

--
Cheers

Don