Author Archives: Duane

About Duane

Duane is the Web Marketing Manager for Screaming Circuits, an EMS company based in Canby, Oregon. He blogs regularly on matters ranging from circuit board design and assembly to general industry observations.

How to Fill a Via

Posted on by

We here at Screaming Circuits keep telling people to fill and plate over vias in BGA pads. The same goes for small passives too, like 0402s and 0201s. As I closed in this post, best option is to fill and plate over. Bigger parts can use some of the solder mask techniques, but BGAs — especially 0.5 mm and small pitch — and tiny passives really do require filled and plated vias.

The question of the day, however, is: “Just how do I get those vias filled?” Here again is another place where our old Gerber file format falls a little short.

There are quite a number of different methods for filling vias: Copper plugs, copper plate, surface material (i.e.  HASL), conductive epoxy, non-conductive epoxy. Maybe others. But, the Gerbers don’t have a method for specifying this.

So here’s what you do.

1) You need to contact your fab house to determine what techniques they have available and what they recommend for your fab project.

2) You have to create a set of fab drawings or special instructions that point out which vias need to be filled and with what material.

Duane Benson
Maybe raspberry jelly filling?

http://blog.screamingcircuits.com/

Via in Pad – Tenting the Bottom Side

Posted on by

Speaking of vias in pads, here’s a reason we’re not terribly fond of the technique of using solder mask to tent the bottom side of a via in a BGA pad.

As I mentioned in my last post, capping the bottom of the via with soldermask is sometimes an acceptable method for dealing with a via in pad. It’s not a desirable method, but it does sometimes work. However, with BGAs, the top-open via can still wick solder down to fill the space. At the very least, you will end up with a hefty void in the BGA solder ball. At worst, you will end up with an open, as illustrated here.

Best option: Fill and plate over the via.

Duane Benson
Does cheap mass production = this perfect day?

http://blog.screamingcircuits.com/

Large Via In Pad

Posted on by

I haven’t written about via in pad in a while, but the issue hasn’t quite gone away yet. This particular QFN, below, has the vias tented, which is good. However, it could be better. If you look close, you’ll see that they’re tented on the bottom of the board with solder mask.

Tenting on the bottom will usually prevent solder spillage out on the back side of the PCB, but with vias this large, the solder will probably flow down into the space, leaving quite a bit of voiding under the part. Sometimes outgassing will pop open the little tents too causing the spillage. And with immersion silver boards, outgassing can cause corrosion in the vias if you have the bottom tented and the top also sealed — like by the part.

If it’s a low speed, low temperature QFN that just needs a little ground connection to the center pad, that voiding might not matter. But, in most cases with QFNs, you need minimal voiding for thermal or noise reasons.

The best option for manufacturability is always to have the vias filled and plated over at the board house, but that can be expensive. If you are going to tent with solder mask, this next image illustrates the three ways to do it.

A is the best: a cap on the component side about 100 to 125 microns bigger than the via. B, a larger cap on the component side, or C, a cap on the bottom, will also work but both come with a greater risk of excessive voiding.

Duane Benson
Do solder mask tents need a rain fly?
In Oregon – probably yes.

http://blog.screamingcircuits.com/

Underness as a Common Theme Again

Posted on by

I didn’t mean to, but here I seem to be with a common theme again. Like yesterday, and again not too long before that — parts under parts seems to be the thing of the hour.


Maybe the caps are shy and are just trying to hide. Or maybe we’ve been writing too much about package-on-package lately. Dunno.

I do wonder though, how many of these issues are caused by parts substitution (the cap originally chosen was narrower), the wrong CAD library footprint, or just a rushed layout. Who wants to vote on that one?

Duane Benson
Just push on it harder. We’ll make it fit.
Maybe…

Underlapping

Posted on by

Sometimes, it’s the simple things in life that we must stop and appreciate. Like the fact that just because there aren’t any pads in a spot, doesn’t mean that you can use the space.

Take this PCB, for example. U3 is placed so that one and a half of its legs will be underneath a big RF module. The module isn’t shown in the picture, but the pads and outline are shown. U3 is right into the lower left corner of the module as you can see by the rows of pads and the silk outline. The module that U3 steps under is one of those module on PCB using castellated mounting holes.

If U3 were your hot little sports car and the big module were a dump truck, you most certainly wouldn’t do this. You’d look first. That’s what mirrors are for.

C6 up on top is pretty darn too-close to the module too. Move it a few mils to the right please.

Duane Benson
Mirror mirror on the wall
Who is the overlappingest of them all

‘PC Load Letter’

Posted on by

I remember back in the olden days when we didn’t have any space and everything needed to be abbreviated or put into obscure symbols. We had eight characters for file names, plus a reserved extension. So, for example, on disk, this photo might have to be called “APTC6010.BMP.”

And, years had to be saved in databases as “82” instead of “1982” because space was expensive. I think what happened, is that we all got in the habit of being obscure and obtuse because of that. We’ve since passed that habit to newer generations too, so that today, even when there is room available, we still do the “PC LOAD LETTER” thing instead of just being nice and clear.

This cap footprint, pictured to the left, does have a polarity indication in the silk screen. Either that, or the silkscreen had some printability issues and was cut off on one side. It’s probably positive up, but just to be sure, why not put a little + sign on the positive side to remove that bit of ambiguity? There’s plenty of room on this board. If space was tight, I could see leaving it off, but what’s wrong with a bit of extra safety when you can get it?

Duane Benson
Baseeball been very very good to me


http://blog.screamingcircuits.com/

Don’t Do It

Posted on by

Friends don’t let friends wire-wrap.

Do people still do this? Do you still do this? When’s the last time you did some wire-wrap?

Digi-Key still sells an assortment of wire-wrap wire and wrapping tools, including the simple little hand tool that I own priced at — holy mackerel! — $34.35! I should sell mine on eBay. Somebody must still be doing this if they’ve got such a nice selection. Amazing.

Duane Benson
Who wants spaghetti?

http://blog.screamingcircuits.com/

Panelization

Posted on by

A lot of small quantity PCBs come individually routed these days, but when they get too small or come in larger quantities, panels can be very nice to have. When you have your PCBs panelized, what’s your preferred method? And why?

We get questions about this reasonably often: “What’s the best way to panelize my boards?” For our shop, we have some guidelines on how to go about it (make sure to follow the specific guidelines from whatever manufacturer is assembling your boards), but the guidelines don’t specify whether you should use V-score or tab routed. That’s a decision left to you.

What if you don’t know? Well, it depends then, but you can easily eliminate a few options. For example, if you have curves in your board outline, you can’t V-score. V-scoring only works in straight lines. With curves or odd shapes, you have to use tab-routed. If your outline is a pure rectangle, V-scoring tends to require less board-edge, so you can get a bit more out of your PCB real estate. But it’s more difficult to deal with on very thin boards and V-scoring leaves a rougher edge after snapping the boards apart.

Two of the key disadvantages of tab-routing are the greater waste area and the nubs that stick out after separating the boards. You can leave the nubs, sand them down or use a clean-up router.

Here’s my take on it: A) If it truly doesn’t matter, use whichever method is less expensive or that you think looks prettier. B) If you have curves or other odd shapes, you’ll probably need to go with tab-routed. C) If your boards are rectangles and you can deal with the less-smooth board edge, go for the V-score.

Duane Benson
Tab. Not Diet Coke.

http://blog.screamingcircuits.com/

1.0, 0.8, 0.5, 0.4, 0.3 …

Posted on by

As pitch gets smaller, things get more difficult. Did that blinding flash of the obvious hurt your eyes? Mine too. It’s not always just more difficult though. Sometimes it’s different as well.

Like BGAs, for example. With 0.5 mm and larger pitch, we and pretty much everyone else recommend NSMD (non solder mask defined) pads. However, once the pitch drops to 0.4 mm, some manufacturers are recommending solder mask defined pads to prevent bridging between solder balls. Make sure your fab house can nail the mask registration.

With LGAs and QFNs, IPC recommends NSMD pads for 0.5 mm pitch and larger. Once the LGA or QFN (DFNs too) pitch drops below 0.5, does everyone suggest solder mask defined pads like with BGAs? No. Actually, IPC tells you to go ahead and remove the solder mask web between your land pads with 0.4 mm pitch parts. With spacing that close, the solder mask can’t reliably stick to the PCB. That would be messy. As to why a solder mask defined pad would not be a good thing, I can only speculate at this point. My assumption is that without the solder ball like the BGA has, the whole land area is needed to ensure a good solder joint.

Duane Benson
Fight Uni!

http://blog.screamingcircuits.com/

Weekend Wondering: Is Anything Really New?

Posted on by

Last year, Screaming Circuits started closely following the Ti OMAP processor and it’s package on package (POP) form factor. The OMAP 35XX processors are very nice on their own, being a very hi-performance ARM jobby, but the package on package made it something for us to take real notice of. POP had been done before, but this seemed to be the first broad-audience application of the technique.

But is it really the first common package on package application? I’ve seen some unintentional package on package, like the capacitor-under-connector pictured on the right in this post. That’s not so fun, but it can keep your board size down — as long as you don’t want it to actually work. And then, many, many network cards used to have a chip placed right under their ROM socket. Would you call that package on package?

How about the old Ti SN754410 motor driver? It’s pretty common for robot buildersto stack a pair of them to get two amps of drive from a pair of one amp chips. That’s probably more a case of actual package on package than is the network card example. Maybe the network card should be called “package under package”. I know, the OMAP is a BGA, but that might actually make it easier to manufacture. With the 754410, all of those leads have to be hand soldered. The OMAP, we just put it on our machines and they do all the work for us.

The other “new, but really old” subject I’m thinking about is cloud computing. Yes, it’s the newest rage in the software and application world right now, but is it really that new? Or has just the name been changed so some pundit can claim to have invented a new concept? I learned software development in a cloud computing environment – back in the early 1980’s.

Duane Benson
The story you are about to hear is old; only the names have been changed to protect the egos.

http://blog.screamingcircuits.com/