Accurately locating SOT-23 parts

Den mandag den 16. december 2013 23.23.55 UTC+1 skrev Tim Wescott:

if placement was 0.25mm off a 0402 would be marginal, a 0201 wouldn't hit the pads, and a 0.5mm pitch IC would be pure luck

-Lasse

IME that is about right.

0.25mm is going to be a stretch for SOT23. If you optimized the heck out of the land pattern and dialed in the reflow process meticulously, maybe. But I would not bet the farm.

I can only speak for tin-lead here.

If your sensors would come in SC75 that would make life easier. Another option might be glue so it won't float off during reflow. Pick-and-place machines can get to within less than 0.1mm accuracy. You'll have to talk with your assembly house and see if their machines can do this.

--
Regards, Joerg 

http://www.analogconsultants.com/

Placement is always better than that. I just went to my sample board drawer and looked. SOT23 easily floats 0.5mm out of position during reflow. Orientation is always good, meaning they don't rotate. But they sashayed or, as line dancers would call it, they did a shuffle.

--
Regards, Joerg 

http://www.analogconsultants.com/

A BGA will self-centre to about +- 2 um. Interestingly, you can make it better if you put in a slight intentional mag error, i.e. making the pad array 0.25% bigger in all dimensions than the ball array.

The reason is that if the two match exactly, the restoring force that tends to null out the position error is cubic in the position error, so it has a huge broad null near zero. If there's a mismatch, you have one side pulling against the other at all times, so the restoring force is linear in the position error.

(It can't go as an even power, because then the direction information is lost.)

With a SOT23, if you make the pads the same size as the leads, and stretch the pattern a little in the same sort of way, you should get within a thou or two.

Cheers

Phil Hobbs

If you make the pads smaller than normal, it may help keep the part more accurately centered. If the pads are big, the part can be a bit off to one side and not be dragged back by solder surface tension. Probably .25 mm accuracy will be pretty easy for a commercial fab house to maintain.

Jon

Only with smaller pads. But then you normally get a DFA flag from the assembly house which results in a CAM hold. Better to discuss that with themn upfront, before wrapping up the layout.

--
Regards, Joerg 

http://www.analogconsultants.com/

Since the leads are only .5x.2mm, I would sure as hell *hope* you can get them down much closer than +/- 1mm. ;-) You'd better be able to do *much* better than .25mm, or you're going to be doing a lot of rework.

Once the process gets tweaked in, I don't see any difference in the quality of the process. The big thing that I saw, going to RoHS was that a four-stage oven wasn't really good enough. It could be done but it was really picky. An eight or eleven stage oven cured all problems. I still hate RoHS, though. Rework on some components is almost impossible. It makes everything more difficult than it needs to be but some it just makes impossible.

My experience with hand soldering indicates little difference in solder-ability per se; ease of use very similar, wetting the same assuming good flux, superior looks and connectivity with tin/silver solder (NO DAMN COPPER!!!).

Here is my take on this:

Now, if you want more uniform placement, make the solder pads virtually the same size as the part pads (flat part of the leads). Capillary action of solder will pull each lead to the center of its pad; larger pad in X or Y will allow slipping of lead. Yes, it seems that slight oversize of pad will allow a meniscus of solder beyond the lead, but that tends to pull the lead (if small oversize). Then you have the problem of meniscus pulling at "heel" of lead (where lead comes down from part, curves and becomes flat to contact PCB) "fighting" the meniscus at the far end of the lead. Guaranteed to be stronger at the "heel" if your pad allows that.

Variations of lead curvature, etc between parts might "kill" the precision you would like to have; use parts from the same lot and for "low" quantity, from the same reel may help.

Nice. That is going to get me fiddling with the footprints of a couple of the parts I use, that should be well centred.

--

John Devereux

If you can stand some manual assembly, make your pads big squared off U shapes with no copper under the SOT pins. The pins rest on the bare board material in a copper-sided ravine. This gives a positive location for the part and soldering over the top of the pins is easy.

Even better if you can mill out the ravines a little.

Cheers

--
Syd

Not really. I routinely see boards where the SOT23 parts are more than

0.5mm off center and everything works nicely. Mostly all are scooted in the same direction, but not always. [...]

--
Regards, Joerg 

http://www.analogconsultants.com/

You may be able to help the assy house out by placing local fiducials around the part (and telling them about them) rather than depending on panel-level reference features. As well as the pad design, you might want to pay close attention to the solder mask (and solder mask expansion dimension).

--sp

I think that could be problematic, if the pins are level with the package you can't have tracks under the package and what about solder mask? you can't have solder mask in the ravines or they will above the copper so you can't have solder mask under the package either

-Lasse

The leads are off the pads and it still works?

The pads need to be longer than the leads, but not wider. The center "void" keeps the parts centered length-wise but the longer pad is necessary on most parts, for the fillet. There are also pad tricks on smaller parts (0402s, etc.) to keep them from tomb-stoning, too. The IPC standard footprints are a good place to start.

I only use wireless transistors and diodes :-)

But seriously, why should 0.5mm veer cause it to be off pad? See page 26 top:

--
Regards, Joerg 

http://www.analogconsultants.com/

Counter-productive.

Those pads are pretty wide but even using that footprint, .5mm puts the leg off the pad. If your process is that bad, you're in a world of hurt.

If the pins are level with the package, it won't work. I haven't seen a package where that is the case.

Cheers

--
Syd