More Beagle CAD Paws

Continuing on from my last post...

As I said, I do everything I can to avoid re-using the package footprint when adding the the parts library in Eagle CAD. The schematic symbol can be a different story though. It still takes a lot of caution, but it's less risky (in my opinion) than reusing the package footprint.

Eagle version 6 made some improvements in the way copy and paste works. It's still a little different from your typical word processor, but it's not that difficult.

Eagle footprint menu bar 3 buttonsBut before I get to that, I want to mention one item that caused me a fair amount of confusion early on. And that's the way all of this fits together. There are three buttons you will need to worry about. From left to right in the green oval are; the device, the package footprint, and the schematic symbol. In my last post, I pointed out the package footprint and today I'm talking about the schematic symbol.

Really, you only build the footprint and the schematic symbol. Then you connect the two up to create the devices. And, you can build the footprint or schematic symbol in either order, but you have to have them both before the last step (the icon in the green oval with four little AND gates).

If you're using a chip that comes in a couple of different packages (e.g. DIP28, SOIC28, TSSOP28) you most likely only need to make one schematic symbol. You can make the multiple footprints and connect them up in the device section as different variants of the same part.

There are a few exceptions though. Sometimes QFN, QFP or BGA parts will have a few extra pins. In those cases, it may be better to create a different schematic symbol.

Duane Benson
This solder paste stencil glows blue when goblins are around

Speaking of Reference Designators...

In my prior post about BOMs, I gave a few examples of reference designator formats in the BOM. BOMs are a common item that have standards but no standards as are reference designators. There are actually a number of standards covering reference designators, but I still find people referring to documents published in the 1970's!

Some aspects are pretty obvious. They are a code letter followed by a sequential number. Each and every placement on the PCB has to have a unique reference designator. The code letters are somewhat standardized, in practice. Some vary based on the particular user. Pretty much everyone uses "R" for resistor and "C" for capacitor. The mostly standard designator for an integrated circuit chip is "U", although I've seen "IC" used enough times. Crystals and oscillators are supposed to be "Y", but I've also seen "X", "Q" and "U" used. Check this page over at Mentor Graphics for their recommendations.

Things start to get sticky when people have more than ten of a given type of component or when putting together a family panel (several different designs on the same PCB panel). Let's say you have 15 resistors. You could designate them as R1, R2, R3 - R15. But maybe you're a little OCD and you want them to all have two digits. In that case, you might have R01, R02, R03 - R15. To a human, "R1" and "R01" might very well be exactly the same thing. But to a surface mount robot, they are two different things. The robot would be happy with R01, R02, R3, R03, R4... but that could cause problems for a human reworking or maintaining the circuit later. It's best to be consistent. Basically, the assembly systems see reference designators as text items, not numerics.

Let's take the example of a family panel. One board has C1 and C2 are a 10uf, 24V tantalum cap. The other board has C1 as a .01uf, 50V ceramic and C2 as a 220uf 24V metal can electrolytic. If you were having them built separately, there wouldn't be any problems, but the two of those on a surface mount machine in a family panel and you will have bad news.

First you could avoid running your boards as family panels. That's not always practical though. Second, you could just start numbering the second design where the first one stops: design one: R1, R2, R3, R4. Design two: R5, R6, R7, R8. That makes a lot of sense for a family panel. Just treat it all like one big design. That can get confusing though if you later run them individually or need to do some rework. Some poor tech could go crazy looking for R1 on design two. Even worse would be: design on: R1, R3, R7. Design two: R2, R4, R5, R8. Again, fine as a family but darn confusing when separated.

Personally, I would probably go with something like: design one: R101, R102, R103. Design two: R201, R202, R203.

Duane Benson
You know the nearer your designator, the more you're silk screening away

Let's Get Small, as in 0.3mm

Not long ago, I wrote about a 0.3mm pitch wafer scale BGA we received and were asked to place. The gist of that article was that those parts are very small and we d0n't yet have a process that we feel will give the quality, reliability and consistency that we want to deliver. That means officially, we don't, at the moment, support that form-factor.

However, as it turned out, we went ahead and built it and the x-rays all said it looked good. Whew! We still don't officially support it, but we're working on it. If you have one of these things, you can always give us a call and see if it's something our manufacturing engineers are comfortable with. If they say "sure, send it in", It will be a non-standard, essentially, experimental, operation so our normal guarantees won't apply. It will be "we'll do our best."

But that's not the point. The point is that there are still a number of unanswered questions with 0.4mm pitch, and now we have a smaller one??!!

I've only seen 0.3mm pitch in two places: some data from Amkor, and the data sheet for this part.The part in questions is a Maxim MAX98304 Mono 3.2 Watt Class D amplifier. The entire package is just 1mm x 1mm.

There is still a lot of difference of opinion on solder mask defined (SMD) vs. non solder mask defined (NSMD) at super small pitch like this. For BGAs 0.5mm and lager, the general consensus and IPC recommendation is NSMD. At 0.4mm, the Beabgleboard folks at Ti recommend SMD to reduce bridging. But I've had other folks say they get good results with NSMD. For 0.4mm, we've had best results with SMD. It's more than just that though, you also need to religiously follow the manufacturer's recommended pad sizes and such.

Shrinking BGA pitchFor this part, the datasheet shows the pad size (0.18mm), but doesn't cover the SMD vs. NSMD question. Instead, it refers to a Maxim app note (#1891) for that bit of information.

Of course, this is where it gets sticky. That app note, as of this writing, shows 0.5mm and 0.4mm, but no 0.3mm. It does reference IPC-7351, which is a very good thing, but I don't think IPC-7351 has 0.3mm pitch covered yet. Ugh. The 0.3mm part we placed used SMD pads.

Duane Benson
It's not just Facebook where you can designate something: "It's complicated."


It (.3 mm) Finally Happened

Back in January of 2012, I wrote about the possibility of 0.3 mm pitch BGAs being used here and there. I predicted that in a year, we'd see some 0.3 mm pitch BGAs showing up. I was about three month's off. Almost to the day.

I delivered a session at PCBWest last month and asked if anyone had used a part with that pitch yet. One hand went up. That actually surprised me. What surprised me even more was when one of them (a .3mm pitch BGA, not a hand) arrived on our shipping dock in a parts kit earlier this week.

0.3mm pitch trimFor comparison, the land pattern for an 0402 passive component is about one millimeter long. This specific part is just shy of a millimeter square. Even as small as it is, this part can supply 750 mA continuous. The olden days are so very long gone.

We do many, many complex parts and PCBs. We've put 5,000 parts on a single PC board. We've built boards to be shot up in rockets and dunked way down in the ocean. Some very crazy stuff has come though our shop, but we don't do everything. We don't do 01005 passive components at the moment. Our machines have the technical capability, but we don't rework them, which has to go along with the assembly capability, so we don't support that form factor for now. 0.3mm pitch components pretty much fall into that camp. Our machines can physically pick up and place the component, but until we've developed to process to assemble those parts with the quality people expect from us, we won't be supporting them.

I expect we'll be getting more and more requests for the form factor, so we'll be looking at it. Keep checking back. One of these days, we'll have the process down and reliable.

Duane Benson
It's (Huey mm, Dewey mm, and Louie mm)/10

Missing Mars Probes

Back in ancient times when multi-legged beasts ruled the earth, there were a lot more standards. Or maybe there were just fewer total things resulting in fewer total variations, which looks like more standards.

In any case, if you got a 7408 IC from one manufacturer, it was pretty much equal to a 7408 from any other manufacturer. Even connectors were more or less standard. If you plugged in one PCB mount DB25, you could plug in just about any PCB mount DB25. There were variations, just not as many as now. Today, though, there are a very large number of variations to a standard footprint. For example, while the pin footprint on most Ethernet jacks matches, I've probably seen a dozen different arrangements of mounting and alignment pins.

Another area that can throw monkey wrenches all over is the dreaded metric v. SAE units.

Metric vs imperial

This seems to pop up most often with connectors, as in this image, but it occasionally shows up on other types of parts as well. The footprint here is for a .1" (2.54mm) pitch connector. The connector has 2.5mm pitch. It would be fine for three pins, maybe four or five. But beyond that, it's just not going to fit.

I don't really understand the logic in 2.5mm pitch. If .1", which equals 2.54mm weren't such a ubiquitous standard, 2.5mm would make sense, but as it is, it's just too close. It's close, but they aren't the same. 2.5 != 2.54.

Duane Benson
It doesn't seem like much difference in mm, but in beard-seconds, it's 4,000* units off

*By some definitions, including the Google converter, it would be 8,000 units off


More Thermal Examples

Speaking of thermal relief... Here's an interesting example I ran across the other day.

Thermal relief can be a pain. If you've got a high current device, you may want more than just the thin little connections, one per side, that you get with thermals. You might feel the need for greater current capacity or you may need all the copper to distribute heat. You might have one pad, like in this image, that lands on a plane but not the other one.

In this particular board, the designer just made a few parallel traces coming out of the pad rather than one thick one.

Multi-via passive

The other side of this passive part sits right on the ground plane and has the standard thermals so the other reason this might have been done is to keep the amount of copper trace coming into the pad to be equal to that on the other side. It doesn't have exactly the same amount of copper going into both pads, but it's much close than if just one thin trace had been connected on the left pad.

Duane Benson
One thin trace rides away

Passively annoying

Passive components can be kind of offensive sometimes. I can understand them in analog circuits or charge pumps. But the fact that we need to put them all over our digital logic is just rude. Technically, I Common ground 0402s schunderstand why they have to be there, but philosophically, they violate my basic principles of life.

Back in the early days of personal computers, there allegedly was a company that had it's engineers remove bypass caps one by one until the motherboard stopped working. Then they'd add the last one back in and smile about the short-term cost savings. Well, that was a bad idea. The reality is that we need them.

I've written about some of the problems that can show up because of passives (or other small two lead parts like LEDs and other diodes). Like here, here and here. That last example has popped up recently and I have some more thoughts on it. Essentially, I'm talking about multiple two-lead components that have one lead tied together. That's a pretty common scenario with bypass caps or LEDs (or the LED current limit resistor).

There are a couple of ways to do this. Some error prone and some not. First, the general rule of thumb for two lead passives is, if at all possible, to have the same amount of copper going into both sides. That means that if you have one 8 mil trace going to one pad, have one 8 mil trace going to the other. Also make sure that you have solder mask stopping the solder from going off pad.

Passively annoying bad way AMethod A here is bad. It might just barely meet IPC standards, but it still is really not manufacturable. First, there are no thermals. That makes the solder melt much slower on the right side which can lead to unreliable solder joints or tombstoning.

Second, even though the theoretical solder mask openings don't touch and the keep-out (it's not shown but is just a hair narrower than the mask area) areas don't touch, they are close enough that you might not have any mask between the parts on the thermal pad. That can lead to components shorting.

Passively annoying bad way B Passively annoying bad way B1Method B is also bad. You have your thermals in there so that's good, but the parts are still so close together that you might not get any mask between them, leaving a path of bare copper between the parts that can cause them to drift around and mess things up.

Method B1, on the right here has the same issue. Likely no solder mask between the parts and a bare copper path between the parts.


Passively annoying bad way CMethod C here is fine. The parts are still at risk of not having mask between them, but there isn't bare copper running straight between them. There will be mask between the parts and the pad so there isn't any way for solder to bridge or the parts to drift.




Passively annoying bad way DMethod D here is also okay. You do need more room to spread the parts apart. That's a bummer, but sometimes "bummer" is the cost of reliability. Here, there will be solder mask between the parts and there are thermals. Everything is happy.

Use method C if you have a little side to side room to play with or method D if you have a little top to bottom spare room.



Duane Benson
Prevent flanking maneuvers.
Don't be like the Solders at Thermopylae

Electrolytic Ambiguity

I've written about ambiguity a few times before, like this post about fiducials. But I'm not talking about the PC board today. I'm talking parts. More specifically, I'm talking about silk screen markings for your parts on the PCB.

CapacitorsDiodes have a lot of opportunity for ambiguity, as you can read here. There are many ways to mark parts, but fewer ways to clearly mark them. Take a typical electrolytic capacitor. It can be thru-hole, smt metal can, tantalum, or a few other form factors. The capacitor manufacturers aren't doing any of us any favors where markation is concerned.

Check out this image. Yikes! In all cases shown here, I've oriented positive on the left, which, according to IPC is pin 1. This is also the zero degree rotation for the centroid value. But, isn't it nice of those component manufacturers to put the identification bar on the positive side for tantalum capacitors and on the negative side for metal can electrolytics? Not!

So, how should you mark this in the silk screen on your PCB? For an electrolytic capacitor, the best approach is to mark the positive sided with a (+), plus sign. If you mark pin 1, with the number 1, it can easily be mistaken for the minus sign. If you mark the negative side with a minus sign, it can easily be mistaken for pin 1.

For a metal can capacitor, it is also acceptable to put the notched outline in silk screen. We still recommend that you place the (+) plus sign on there too.

Duane Benson
I'm just positive I put the negative right on the left

Counting once, counting twice...

Panel single scLet's say you have two options: First, you could send in your PC boards for assembly as individuals. Second, you could send them in a panel. That's all fine and dandy. For a few, send individuals. For a bunch, panels might make more sense. But, when you do go to quote and order, how do you count the parts?

Let's take this example. As a single, this board has 32 line items on it's bill of materials. That's 32 unique parts. Counting all of the individual part placements, there are 56 total parts: 42 smt and 14 thru-hole. So, naturally, if you quoted the assembly of 20 of this board at Screaming Circuits, you would enter your desired board quantity as 20, Panel 4-up sc32 total unique parts, 42 smt and 14 thru-hole.

But what do you do if you send it in panel form? How do you count? It's actually not as difficult as it seems. In this example, it's in a panel of four. There are still only 32 BOM line items, but there are four times as many placements. That means that if you quoted this, as a panel, you would enter 32 total unique parts, 168 smt and 56 thru-hole parts. If you still need 20 of the final boards assembled, you would enter 5 as your desired board quantity.

In the end, you will have 20 assembled boards. In case you are wondering about the cost, there won't be a difference. As long as the final number of boards (after the panel is broken apart) are the same, your cost will be exactly the same for panel vs. one up. You don't save any money by sending in singles. However, if your board is panelized and all of your parts on on reels, full or partial, you can save money by ordering Short-Run production.

Duane Benson
50 Years ago today
Robert Rushworth flew the X-15 to Mach 5.03 at 100,400 feet altitude

Fiddling with Fiducials again

I recently posted a note about fiducials but I didn't have any images. Here's a couple of examples:

IPC acceptable fiducialsThis first example shows what IPC would like to see. If this is an individual board, this would be it. If it were part of a panel, you would follow the same pattern on the panel rails and also put it on each individual board in the panel.

As I wrote in the earlier post, we don't require these, but it's always a good idea. You'll need them once you go into volume manufacturing anyway.

The next example won't make IPC happy, but it will make Screaming Circuits happy:

Also acceptable fiducialsIt only uses two fiducial dots, but it isn't reversible. Reversibility is okay for jackets, but not for circuit boards. Since one of the dots is offset, it can't be placed on the machine and recognized as correct in any way except in the proper orientation.

The important aspect of both of these examples is that they remove ambiguity. There can't be any uncertainty, which is good because uncertainty is your enemy. It's a subtle enemy. It might not bother you 99.9 times out of a hundred, but then, when you're not looking, it can strike. So, give a hoot and stomp out ambiguity.

Duane Benson
False data can act only as a distraction. Therefore, I shall refuse to perceive.