Best way to solder prototype pcb with thermal-pad ICs?

Am 27.02.20 um 10:29 schrieb Winfield Hill:

I usually solder the power pad first using hot air and a surprisingly small amount of tin. Using too much will make the tin protrude from underneath the package. Then I measure if there are unwanted connections. I everything is OK, I solder the other 8 connections.

My experience is with ADA4898-2, and it's good. My Aoyoue (?) hot air station also removes the chip easily if needed. And I love that Metcal 5xxx. I think it would also do the job.

When there are a lot of GND vias under the chip that may behave differently. Then a hot plate could be required from

board above the glass temperature.

I have a lab-scale vapor phase machine but it was never ever needed up to now.

regards, Gerhard

I put the solder paste on with a stencil now, I've tried using an air powered sysringe type thing and you just can't control the amount of paste properly. Stencils are cheap compared with boards and time. Use leaded solder if you can. Use decent solder paste and don't keep it out of date.

The best way to reflow is to use a proper temperature controlled IR thing. Ideally you want a belt type job with several zones - there are cheap Chinese ones but I didn't dare buy one. I bought a CIF FT03 from Farnel or Rs (forget which) - it cost loads compared with Chinese (about

take ? The downside of the CIF is that it doesn't have forced cooling.

If you want to spend less then use a cheap hotplate, I used to get this to temperature and place the board on it and watch the paste reflow. Dont' attempt this with lead free - the margins aren't big enough to get away with it. On prototpyes I much believe in putting vias through thermal pads, it lets excess solder out, it lets you check if the pad is soldered (visually and with a needle probe (does work - I've done it)) and it makes rework easier.

If you need to rework then use temperature controlled air heating from

heat.

MK

reflow. On our board with extended pad you might be able to reflow with an iron.

Dont't know if that's the best way, but failures are quite rare and usually due to rushing and burning components with air.

With a bigger 8 SOIC pac. I had two plated holes where the pad was and hand soldered it from the back side.

George H.

Yes, I've done that before, seemed to work OK. Just forgot about it this time - getting used to doing it right for automated assembly. I appreciated and liked the advice from the three previous answers. I need to step up my game, look into more and better soldering equipment, but the hole-under trick is much easier.

Looking at my layout, it was poorly done, perhaps even causing the malfunction. Here's a redo from scratch.

The design is crammed into an 11mm slot. The inductor's current paths are better above the IC rather than below. The two FB resistors are supposed to be close to the IC, and with the up-to-1MHz speed and up-to-100V input of the converter, it's a good idea to have a ground plane under.

I've been known to tin the pad on the board, then use a flux pen and hot air. It's an easy way to control the solder volume.

If I'm stuffing more than a couple of parts, I use a ~120C hot plate to help out.

Cheers

Phil Hobbs

I'm using these to go from 24 or 48 volts down to 12.

It can do up to 72 volts in.

My tiny pulse generator can be powered from 24 or 48 volt warts, so this gets me down to the internal 12 volt rail, for secondary regulators and fans.

I think it will also do +48 to -12. Gotta try that.

That sure has a lot of personality.

On the

Maybe use a baby-board, so if it fails, it doesn't have to be unsoldered. Unsoldering power pads by hand is about impossible.

Do you need to solder the power pad? Maybe use a bit of silicone grease or gap-pad if power is low.

A hot plate plus hot air usually works OK. Fancier hot air tools have rectangular nozzles for different package sizes.

Cheers

Phil Hobbs

There are assembly houses that will work these prototypes for way less than what it costs for you to do it, if you can wait a few days. Lots of really cheap vendors in China too.

Or thermally conductive double-stick tape. That pad is electrically isolated.

On 27.02.20 9:29 am, Winfield Hill wrote: > After soldering the chip's 8 pins, [?] thermal pad. > [?]

I've made good experiences doing it the other way round: Apply a little bit of solder directly onto the exposed pad with the iron. Then, add copious amounts of flux and use hot air to solder the pad while lightly pressing the chip down with tweezers to make sure the contact is good (and any excess solder is ejected to the side). Then, solder the pins as usual with the iron (either drag-solder, or just one by one for something fairly big like an SOIC-8).

If there are any doubts about connectivity (or shorts), just reflowing the entire chip with hot air in the end tends to magically correct any misalignment, even with a thermal pad. Of course, this is predicated on having gotten the footprint right and using a sensible amount of solder.

Like this, I never really had any issues soldering things like the 2 mm x 2 mm 6-pin LFCSPs that some of the ADPxxxx regulators come in, a number of TI's SOIC/TSSOP op amps, etc. I haven't needed to manually solder anything excessively sensitive to thermal stress in a while, though.

If the thermal conductivity of the PCB is on the high side (heavy copper, aluminium substrate, etc.), doing the above with the board on a hot plate to pre-heat it evenly makes all the difference, and also helps to avoid thermal stress.

? David

That calculation was based on a 15V input, where the upper high-Ron MOSFET, 725mR, is on most of the time. W/o the thermal pad connected, I measured 46C package temp for the chip and 49C for the inductor, 1.3 ohms, at 300mA. But when I raised Vin to 60V the chip temp rose to 100C and the inductor to 65C. So apparently the LM5163's switching losses mean it does indeed need its thermal pad connected.

I've also done MLF/LFCSP by putting vias in the thermal pad under the chip and a large square pad on the back side of the board with a solder mask opening all over it. There were 5 plated through holes in a domino pattern in the thermal pad, ideally large enough to fit fine solder wire (0.25mm?) down the holes (0.4mm?). I would flux everything, then tack the part in place by soldering the pins, then flip it over and heat the big pad on the back with a Metcal iron and feed solder down one of the plated holes until it comes up through the other ones. I figure that if solder goes down one via and comes up the others, then there is enough solder on the thermal pad.

If you have to, and provided the board is designed as described above, you can remove small LFCSP/MLF packages by reheating the thermal pad from the back with an iron, and eventually the solder on the pins should melt and you can pull it off, but hot air would be better as it would be quicker.

The only parts I can recall damaging by heating were some LTM DC-DC converter modules that I failed to bake before reflowing the board. They were already soldered to the board when I wanted to add another BGA by vapour phase reflow. I baked the new parts that I was adding but it had not occurred to me to bake the populated board as well. Nearly all of the LTM parts died. Next time we baked them.

Exactly right, at least two holes so you can see the solder flow via the other hole. And hey, then the big vias are filled with solder.

George H.

you might be waiting more than a few days for assembly in China this month.

Win,

You could try something like this: Chipquik SMDLTLFP

This is a no-clean solder paste made from a tin, bismuth, silver alloy with a melting point of 138degC, available from Mouser, Digikey, Farnell etc.

John