Panel Rails — What Are They?

I referred to “panel rails” in my blog about V-score panels, but I didn’t explain the “whats” and whys” of panel rails. You might find yourself asking “what are panel rails and why would I want to use them?”

Well, first of all, for our Full-Proto service, we don’t require panels or panel rails. We’ll take just about any old board that’s bigger than 0.75″ x 0.75″ and smaller than 14.5″ x 19.5″ and run it through our machines. For our short-run production service, we only require panelization for boards less than 16″ sq.

That being said, panel rails do have a purpose. They give the machines a spot to grab onto without coming close to components. They’re also a convenient place to put fiducials (more fiducial info here).

As you can see in the image below, the panels give a clear area for handling the panel.

Tab routed panel

There are two important things to note about this panel. First, look closely at the four outside corners. You can see the scoring for easy separation of the rails. This designer made sure that there isn’t any copper where the scores are. That’s the right way to do it. The V-Score blog shows a panel rail done the wrong way – with copper across the cut.

Next, this board has fiducials. Good. But, the fiducials are in a symmetrical pattern. Not so good. IPC-7351b-3-10 specifies a non-symmetrical pattern so that the board can only be processed in one orientation.

Duane Benson
Once I build a panel rail, now it’s done
Brother can you spare a diode

http://blog.screamingcircuits.com/

V-Score Panelization

V-score top view

My last post talked a bit about panelization, in general. Today, I’m taking a look at V-score panelization. V-score is created by running a V-shaped blade across the top and bottom of the panel without cutting all the way through. The board in the mini-image of my prior post is V-scored. Shown above is a closeup of the V-scoring.

(Note that the cross-hatched area is not in the active circuit portion of the panel. It’s in the rails. You’d never want to cut through copper like that in part of the board that will be used. Even here, it would be best not to have copper in the path of the v-scoring blade.)

You’ll note that it’s all straight lines. V-score can only separate rectangular panelized boards. For curves, you’ll need to use a different technique.

V-score edge onThe next image down, on the left, shows an edge-on view of the V-score. You can clearly see what I mean by “without cutting all the way through.” The cut leaves enough material to hold the boards solidly together during processing, but easy to separate.

V-score de paneled edgeBy the way, we generally don’t just snap them apart. We’ve got a special tool – a bit like a pizza cutter in a fixture – specifically designed to separate them without stressing or bending the board. If we feel there’s any risk of over-stressing, we’ll use the tool.

The next image, here on the right, shows a board edge after depanelization. Note that it’s not a smooth, flat edge.

In contrast, the next image down, on the right, shows a flat milled edge. Generally, though, you can’t visually tell the difference without close examination. You can, however, feel it if you run your finger lightly along the edge. Just be careful to not get slivers.

Next time, I’ll examine tab-routing, which will permit non-rectangular shapes.

Milled edgeDuane Benson
I saw two Buffalos, two Buffalos,
Buffaloes on my lawn,
Romping all around and stomping on the ground
And all of my grass was gone.

PCB Panel Routing Technique

Most PCBs we receive are individually routed; i.e., not panelized. That doesn’t mean that, sometimes, sending them in a panel isn’t a good idea, or required. Generally, we don’t require panels (sometimes called a pallet), but there are some cases when we do.

V-score panelIf the individual PCB destined for Full Proto service is smaller than 0.75″ x 0.75″, it needs to be panelized. If a PCB needing Short Run production service is less than 16 sq. in., it needs to be in a panel of at least 16 square inches to qualify for Short Run.

So, you ask, why else might I want to panelize my PCBs? Keep reading and I’ll tell you why.

  • First, if you’ve got a lot of small boards, it’s easier to handle and protect then when they’re in a panel. A few panels can be more safely packed coming and going from our shop here.
  • You may be able to get the through our factory faster. If you have a really large number, and need them super fast, panelizing them may enable that fast turn. With a lot of boards, sometimes, it simply isn’t physically possible to put them all on the machine, run them and take them off, in a short turn time. Panelize them and the machine will be running longer for each board change, which reduces the total run time.
  • It may also cost you less. If you use leadless parts like BGAs, QFNs or LGAs, you can usually reduce your cost a bit by panelizing the boards. Leadless parts cost a little extra because of the X-Ray test needed, but the extra handling is mostly per board, rather than per part. One panel of ten boards with ten BGA, in total, will cost a little less than ten individual boards with one BGA each.

Stay tuned for my next few posts where I’ll cover the pluses and minuses of different panelization techniques.

Duane Benson
I looked outside my window and what do you think I saw?
The strangest sight I’ve ever seen you’ll never guess just what I mean,
I can’t believe it myself

Component Packages — Let’s Get Small

I’ve been on a bit of a package binge lately. First talking about metric vs. US passive sizes, and then a very tiny ARM Cortex M0 from Freescale.

The Freescale BGA part checks in at 1.6 x 2mm. That’s cool and I’m almost always in favor of making things as small as possible, but, as I wrote in my prior blog on the subject, it’s not always possible. The 0.4mm pitch BGA is problematic unless you can spend a lot of money on the raw PCBs, or will have super high volume.

All is not lost, though. You still can use a tiny ARM Cortex M0 part. Just not quite as tiny. That same part also comes in a 3 x 3mm QFN package. You lose four pins (16 vs. 20) going from the BGA to the QFN, but if you can handle that, it’s a very viable option that doesn’t require any exotic circuit board technologies.

A few years ago QFNs were scary, but not so much any more. I’ve designed a few of them in using Eagle CAD. Just be sure to pay attention to the footprint. A 6 mil trace is more than small enough for a 0.5mm pitch QFN.

Duane Benson
Strive at all times to bend, fold, spindle and mutilate

http://blog.screamingcircuits.com/

0.4mm Pitch BGA is Awesome

I recently had a conversation with a friend about 0.4mm pitch BGAs. The specific part is the FreescaleFreescaleKL03KL03 ARM Coretex-M0+ microcontroller in a 1.6mm x 2mm, 04.mm pitch package. That’s a 20-ball wafer scale BGA form factor.

I don’t have an actual part to photograph next to a grain of sand, but trust me (or don’t), it’s really small.

Ti 0.44 pitch dimensionsThe challenge, and the reason I suggested a QFN form factor instead, is the costs involved. If you have the extra budget money for more expensive PCBs, then go ahead and use this form factor. You probably won’t be able to use this package in cost-constrained situations.

The simple reason is that you can’t escape route the inner six pins without using super small vias between pads, or in pads and filled and plated over. The page on the left is from a Ti doc, but any variations in geometry will be minor.

You can see that you can’t put a trace between the pads. Maybe a 2 mil trace, but maybe not. There just isn’t much room. The recommended method is to put microvias in the pads and have them filled and plated over at the board fab house. Never put a via in a micro BGA pad unless it’s filled, plated over, and flat.

Duane Benson
There are more things in heaven and earth, Horatio,
Than are dreamt of in your philosophy.
But open vias in pads aren’t one of them

http://blog.screamingcircuits.com

When is an 0201 Not an 0201?

I’m working on a special project here that involves some 0402 LEDs and 0201 resistors. When doing such a thing, you should always check the footprint you’re using against the datasheet. When using extra small parts, like this, I recommend making a custom footprint unless the one you picked is exact, and I mean exact. There just isn’t an margin for error at these geometries.

Take a look at the table on the right. The dimensions are in mm. Spot anything a bit off? Counter to most datasheets, the sizes listed in the “Type” column are metric sizes. At DigiKey, the package was listed as “0201 (0603 Metric).” I see that all the time, but for some reason, most datasheets show the package name in US size while listing the dimensions in metric.

The first table was at the front of this data sheet (page 5). The second table was on page 35 – the opposite end of the datasheet.

We do occasionally get boards with metric size pads for a US-size part, or vice versa. Sometimes we can make it fit, but not always. Bottom line, is to check and double-check. I caught this one because the dimension 0.54 mm is about 21 mils, too small for an 0402. That, and the fact that the table doesn’t list an 0201 size.

Duane Benson
Is it Bigfoot or Sasquatch?

http://blog.screamingcircuits.com/

Open The Pod Bay Doors, HASL

Does anyone use HASL (hot air surface leveling) anymore? It’s also known as HAL.

Prior to the RoHS days, HASL was probably the most common surface finish. You can get it lead-free, but most boards seem to use immersion silver or ENIG (electroless nickel immersion gold). HASL has traditionally come at a lower cost than those other two finishes, but immersion silver can generally be found at the same price now.

Our friends at Sunstone.com, for example, charge the same for silver and tin/lead HASL. ENIG is still more expensive no matter where you go, though.

One of the chief disadvantages of HASL these days, is the lack of planarity on the surface. (Note the bumps on the BGA land pattern in the image on the right.) With through-hole or large components, an uneven surface doesn’t matter so much. With the increasingly smaller BGAs and QFNs, however, surface irregularities can cause big problems.

Both immersion silver and ENIG have nice flat surfaces. OSP (organic surface preservative) has a pretty flat surface too, but it’s not used much except in high volume consumer goods or specialized applications.

Duane Benson
Oh, the pain! Save me, William.

http://blog.screamingcircuits.com/

ODB++ Plus, Plus, Plus

I wrote a bit about ODB++ back in October. Nothing has really changed much since then. I’m just clarifying a few things.

First, I want to put more emphasis on the use of ODB++. In addition to being beneficial to the manufacturing process, it can make your job a little easier. If you send ODB++, you do not need to send either the centroid or Gerber files. The ODB++ replaces both.

Eagle CAD does not have an ODB++ export. However, the Eagle .brd file will work too. You can send the .brd instead of the centroid and Gerber files.

If you can’t send either of those formats, we as an EMS still need the centroid and Gerbers (top copper, bottom copper, solder paste stencil, silkscreen and solder mask layers).

Duane Benson

Number Six
I am not a number, I am a free man!

http://blog.screamingcircuits.com/

More Fun File Facts: ODB++

In my last post, I wrote about the up and coming IPC-2581 PCB manufacturing file format. While IPC-2581 may be looked at by PCB fabricators and assemblers as a holy grail of sorts, it’s not yet widely adopted by CAD software. But, that doesn’t mean that Gerbers are the only option.

ODB++ was developed by Valor in the waning years of the last century as an improved method for getting manufacturing data into their CAM systems. Valor and, hence, ODB++ was purchased by Mentor Graphics in 2010. ODB++ is still widely available, however there’s concern in some circles that it’s not truly open. That concern is where IPC-2581 came from. In fact, IPC-2581 is somewhat derivative of ODB++.

I can see how a CAD software developer might fear the use of something owned by a rival. However, my understanding is that Mentor does it’s best to treat it like an open standard and has made it available more or less as though it is open.

The history isn’t really important. What is important is that ODB++ is a more complete format than the Gerber and is widely supported. Pretty much everything good that I said about IPC-2581 in my prior post also applies to ODB++.

The bottom line is that, regardless of whether Screaming Circuits is your fab (through our partner Sunstone) and assembly (through our factory right here) provider, ODB++ is a good thing. It makes the job easier and more accurate than does use of Gerber files. Both “easier” and “more accurate” help keep costs down and keep ambiguities to a minimum. As you know, ambiguity is the bitter enemy of both accuracy and quality.

Unfortunately, for all of you Eagle users, Eagle does not yet support ODB++. If anyone out there is really good with Eagle ULP scripting, you might want to create a on ODB++ and/or IPC-2581 creation ULP.

Duane Benson
I was ionized, but I’m better now. 

http://blog.screamingcircuits.com/

Fun Facts about Manufacturing Files

Circuit boards live and die by their manufacturing files. Without complete and accurate information, the board fab house can’t fab the boards, the assembly house can’t assemble your boards and nobody can buy the parts.

Our old standard, the Gerber file, has been around since about the time King Arthur pulled the inductor out of the solder pot. It’s old. We all use it because it’s familiar, but it’s day is done. It’s time to pass the torch.

IPC-2581 is the new standard in manufacturing files. It hasn’t been fully adopted, but it’s showing up in more and more CAD packages. The IPC-2581 format is much more advanced and has the complete data set in one file. While we still work with Gerbers every day, we can also accept IPC-2581 manufacturing files.

I’ve been called the champion of bad analogies, but I’ll try one out anyway.

Imagine, if you will, a map of the city. All of the streets are there. All of the houses are there. What’s missing are all of the street names. No street names, no numbers and no landmarks of any sort are labeled.

Given that information, find John Smith, at 1620 SW 14th Avenue. There is a house at 1620 SW 14th Avenue. There are a dozen or so houses at 1620 something. You just don’t know where 14th is, or which direction 14th runs, or where the street numbering starts.

You can physically walk each and every street until you find John’s name on his mailbox, but it’s not an easy nor error-safe process. And, hopefully, the town only has one John Smith. That’s a Gerber file.

IPC-2581, on the other hand, is an electronic map, with everything clearly labeled, and a GPS guiding you. Which would give you more confidence?

Duane Benson
IPC-2581 is like shatter-proof glasses for Henry Bemis

http://blog.screamingcircuits.com/