Tag Archives: Duane Benson

What Route Do You Take?

Posted on by

There are a lot of polar opposites in the “what is my philosophy” world: Mac vs. PC, on shore vs. off shore manufacturing, Ford vs. Chevy, Atmel vs. Microchip (well, maybe not that one so much any more), auto router vs. hand route…. Yes, I’m specifically avoiding political opposites.

Routing is what I’m really interested in today. The conventional debate is hand vs. auto route. CAD companies spend a lot of time and money on autorouters, but there’s definitely a line of thought that says it’s not ready for prime time yet. This shirt designed by Chris Gammel, on Teespring pretty much says it all.

But, it’s more complex than that. Most auto-routes end up requiring some hand work, either to finish routes that can’t be found automatically, or to clean up a few less than efficient choices. There are differing techniques for complete hand-routing as well.

I often find myself looking at a layout project a bit like a chess game. I don’t just start at one end of the board and work my way to the other side. I tend to focus on specific parts or critical requirements first, like signal paths that need to be short, or sections with more critical grounding requirements. (The image above isn’t mine. It’s from the Beagleboard.)

When it gets to the mass, I tend to try and think ahead, projecting moves out, as though it were a chess game. When I’m looking for the best route for signal path A, I try and think ahead to how it will impact B, D, D… as far ahead as I can go.

I’m not sure if doing it this way is easier, of if it would be better to just start routing and then re-route as I run into roadblacks. What about you? How do you approach a complex layout?

Duane Benson
Holy cow. I Googled “Trust no one” to get some ideas for my signature
Never do that. It’s going to take a week to shake off all the negativity

blog.screamingcircuits.com

Building Boards for the Intel Edison

Posted on by

I’ve recently spent some time getting familiar with the Intel Edison. The Edison has a dual-core 500MHZ Intel Atom processor, with built-in Wi-Fi and Bluetooth. It comes with 1GB of RAM, 4GB of eMMC internal storage, and a USB 2.0 OTG controller. It doesn’t bring any of the connectors (power or signal) out in a usable form. Rather, it’s designed to be plugged onto another board through a 70-pin high density connector from Hirose.

I designed a small board with I2C (both 5V and 3V connectors) and a micro-SD card slot. My board still doesn’t have the power or console connectors. For that, I’m using a base board from Sparkfun.

Figure 1

Step one of the assembly process, is, of course, to design and layout the board. Using the Sparkfun open source designs as a jumping off point, I ended up with the nice, compact layout (1.2″ x 1.75″) shown below in Figure 2.

Figure 2

After getting the files ready and placing a turnkey order on our website, I followed the board through with my camera. Here it is after offline setup, with the parts ready for robot pick-and-place:

Figure 3

In one of our Mydata My500 solder paste printers:

Figure 4

On the pick-and-place machine, with solder paste, but before any components are placed:

Figure 5

The parts plate in the machine:

Figure 6

 

With most of the components placed:

Figure 7

Through the reflow oven, prior to final inspection:

Figure 8

The final product, top view:

Figure 9

I abbreviated the process a bit, but those are the major process steps along the way.

Duane Benson
Happy birthday (month) Nikola Tesla

Where to Put Panel Tabs

Posted on by

Many small quantity PCBs are ordered individually cut. They come to us as a set of unconnected boards. For small quantities of reasonable size boards, it makes the most sense to order them this way. However, for really small boards, and larger quantities (50 or more), purchasing boards in a panel (also called an array) is more appropriate. It reduces errors and assembly time.

There are a few additional factors to consider with panelized boards.

  • First, don’t try to create a panel in your CAD software. Just lay it out as a single board and have the fabricator put it in a panel. You’ll get the most efficient use of PCB space that way, and the fabricator will create the files in the format that the assembly shop (Screaming Circuits) needs.
  • Avoid family panels. A family panel is when several different boards are put onto the same panel. The boards in family panels often repeat reference designators, which causes problems at assembly. See this blog article on how to properly assign reference designators on a family panel.
  • If you have overhanging parts, like the increasingly common micro USB connector, make sure that the panel tabs aren’t placed near the overhanging them.

This blog article gives some background on the connectors.

Some components, such as the connector in the link above, have protrusions that will keep them from laying flat on a panel tab. In all cases, even without the protrusions, the operation of separating the panels with a component on the tab can weaken the component solder joints, or even pop it off the board completely.

How not to do it:

Figure 1

Instead, make sure that the tabs don’t end up under your overhanging component. Have the tab moved like this:

Figure 2

You can put this instruction in the document layer of your CAD file, or in a separate document covering fab instructions. In the CAD image below, the overhanging component has a keepout area. The document layer has instructions to keep panel tabs out of the area.

Figure 3

If in doubt, don’t hesitate to contact us or Sunstone Circuits directly to ensure that your instructions are clearly communicated.

Duane Benson
Wood paneling as a wall covering is really out of style

http://blog.screamingcircuits.com/

Top 5 Things to Know When Moving from Hand Assembly to Robotic Assembly

Posted on by

A lot of factors go into the decision to hand build or outsource circuit boards. I hand build my own sometimes, simply because I enjoy the challenge. Of course most of the projects I design are for my own use, so timeliness isn’t that important. When I do design something that will go out to a customer, like my electronic business card holder, I will send the board through our shop. In those cases, quality is important, as is delivery, and the quantity is often too high to hand build. Machine building also allows me to use smaller and more complex parts. 

That same decision — hand build or outsource — takes place in the heads of designers all over the country. When the decision is to outsource, there are a few important things to consider. Some things that work fine when hand soldering may stand in the way of quality, repeatability, and reliability when machine assembling.

Here are five of the most important considerations when changing from hand-built to outsourced at a place like Screaming Circuits

1. Use solder mask and silk screen. A good solder joint needs the right amount of solder in the right place. Solder will tend to flow down bare copper, bleeding outside of the area it belongs, and down exposed copper traces and vias.

The main purpose of solder mask is to keep the solder where it belongs. It also protects the traces, but that’s a longevity issue. Solder bleeding is a manufacturing and reliability issue. This isn’t a problem when hand soldering. In fact, it can even work to your advantage when hand-soldering really small parts. It gives you more room for your soldering iron to hit metal.

Not so with solder paste and machine assembly. Use solder mask.

2. Avoid the pseudo panel. Keeping small boards in a panel is the recommended best practice in the manufacturing industry. We appreciate it and, while not always necessary, it can reduce your costs. We sometimes see what we call a “pseudo panel.” This is a board where multiples of the board are put in the same PCB, like a panel, but unlike a panel, the boards don’t have routing or V-score between them. Sometimes the designer will put a bunch of vias to outline the board, or just ask that we use a band saw to separate them.

That’s a time-consuming, expensive and potentially damaging process. The vibration of the saw can crack solder joints, and, you’re unlikely to get boards that are all the same size. Have small boards panelized by your board house.

3. Family panel (pseudo or not). Similar to the pseudo panel is the family panel. A family panel is a case where a project is made up of several different PCBs, and they are all laid out together, as though they are one design. If the board isn’t routed between the designs, you’ll have the pseudo panel problem described above.

The bigger problem, though, comes with reference designators. We typically see family panels with duplicate reference designators. Each design, for example, will have its own C1, R1, Q1, etc. We use the reference designators as position identifiers: If you have three different parts labeled R5, our machine programmers will have a problem with it. It’s even worse if the values differ; on one design, C1 is a 0.1uf capacitor, while on another design, it’s a 22pf cap.

If you’re making a family panel, give each and every placement a different reference designator. One way would be to us extra digits. For example on one design on the family panel could have C100, C101, C102… The next would be C200, C201, C203, and so on.

And don’t forget the routing or V-score between the designs.

4. QFN — hole  in the middle. A common technique in the hand soldering world, for QFNs and other parts with thermal pads underneath, is to put a big via in the middle of the center pad. By doing so, you can stick a soldering iron and some solder down through the hole and get a good solder connection on the bottom pad.

This doesn’t work with machine assembly. the solder paste will flow down and out the hole in the reflow oven. You’ll end up with a poor connection (or no connection) to the thermal pad, and solder slop on the back side of the board.

5. Parts and the bill of materials (BoM). When I build my hobby projects, I often get a bit carefree with the bill of materials. It’s not good practice, but I do. I’ll put a part in the BoM that I used before, and not check to see it’s still in stock. I’ll put parts in the BoM with just the values and not any part numbers. Things of that sort require tribal knowledge, which only the designer has.

When building, sometimes I’ll just grab a part that’s close. If I need an 0805 1uf, 10V capacitor, I can grab a 16V, 25CV, etc. I can even make an 0603 part work. You as the designer may know that something close will work, but an outside house can’t know. You need to tell them exactly what the part is.

Before sending anything through our shop, I do clean up the BoM. In order for us, or any manufacturer, to build the boards, the BoM needs:

  • A unique reference designator for each part placement.
  • The quantity of each part used on the board.
  • The manufacturer.
  • The manufacturer’s part number.
  • DigiKey part numbers can be used as well.

Here’s our website page explaining the BoM format in more detail.

The transition from hand building to outsourced machine building can be an intimidating one. But, with a few considerations, it can be an easy and rewarding transition.

Duane Benson
Put the right part in
Put the wrong part out
Put the right part in
But please don’t shake it all about

blog.screamingcircuits.com

USB Type-C Connectors

Posted on by

It wasn’t terribly long ago that pretty much every cellphone came with its own custom charging cable. It was a major step forward when they all (except Apple) standardized on the USB micro-B connector.

However, there are a number of limitations with the. First, it takes a minimum of three attempts to get the orientation right when trying to plug in a cable. Second, it’s limited in maximum current carrying capacity.

Now, along comes the USB 3.1 Type-C cable and connector. It’s similar in size, universally polarized (the connector and the cable can be plugged in any end to any end and in any orientation), it has much higher data throughput, and it’s spec’d to carry up to 3A. Further, it has alternate modes for other standards, such as DisplayPort and Thunderbolt.

The connectors are larger than the micro-B, as you can see in Figure 1, micro-B, Type-C with only surface mount connections, and Type-C with both surface mount and through-hole wiring, and a US dime. The size difference won’t be an issue in most cases, but it could be in really small devices. My guess is that we’ll be talking about a smaller, Type-D connector, not long from now.

Figure 1. Micro-B, Type-C with only surface mount connections, and Type-C with both surface mount and through-hole wiring, and a US dime. Figure 2. Micro-B connector with tabs formed from the same sheet metal as the shell.

All three of the surface mount connectors shown above have through-hole mounting tabs. That adds strength, but it does bring one caution with it. Looking at the micro-B connector in the image on the right, you can see that the tabs are formed out of the same sheet metal as the shell.

You can also see that the tabs don’t stick all the way through the PCB. This can lead to some deception when soldering. Without the tabs protruding, it’s easy to believe there’s not enough solder in the connection. If more solder is fed in, it will likely wick along the tab, and end up inside the receptacle, preventing the cable from being plugged in. If hand soldering or reworking these type of connectors, keep a close watch on the amount of solder used.

Duane Benson
Fester Bester Tester is alive and well and living where?

http://blog.screamingcircuits.com

Fiducials and Odd Boards

Posted on by

One of the handy aspects of getting boards assembled at Screaming Circuits is that we don’t require fiducial marks for standard process boards. I would say that we build far more boards without fiducials than with. That’s cool, but there are sometimes when fiducials really are a good idea. In fact, if you’ve got room on the PCB, they’re always a good idea (just because something isn’t required doesn’t mean that it’s not a good idea).

Some boards are more in need of the marks than others. For example, not long ago, we got a rigid flex board in. It had three separate rigid boards connected by flex, designed to be folded into a stack. It looked pretty similar to the mockup in this image:

The boards didn’t have any fiducial marks. Normally, what we do, is find a via hole, through-hole pin hole, or some similar feature to use as a fiducial. That usually works, but not always. In this case, the length of the flex varied slightly from board to board. The PCB color was also very low contrast, which made it difficult for the machine to consistently recognize any mark we picked.

That meant our machines had a hard time finding the “home” spot, and we had to reset for each of the connected boards. Finding a spot on one board did not guarantee that we’d know where to place parts on the other two boards in the set.

In this case, it would have been far better if the boards were a consistent distance apart, and if each of the three boards had a set of fiducial marks.

What makes a good fidicual?

Most CAD packages have fiducial marks in their components library. Basically, it needs to be a metal dot surrounded by an area without any copper or solder mask. More than one is best. It should be an asymmetrical pattern that can only be oriented one way.

I’ve got some more details in this article here.

Duane Benson
Routed up like a fiducial
Another rigid flex in the night

http://blog.screamingcircuits.com

Happy St. Patrick’s Day!

Posted on by

And some trivia. You may have noticed that the soldermask used on most PC boards is green, as is the paint used on most steel truss bridges. Why is that? And what do the two things have in common? Why green PCBs and why green bridges?

To answer, I brought in color expert expert Patty O’Patrick O’Dell, who stated: “Many bridges and PCBs are green because they absorb red and blue light, only reflecting the green.”

That wasn’t quite what I was getting at, but close enough. The important thing, is that, generally, in commercial products, the PC boards are hidden, so the color doesn’t matter that much. With prototypes and a lot of the hobby or development boards, that is not the case, so many companies have chosen to use a different color as a part of their identity.

Arduino products are blue, as are most boards from Adafruit. SparkFun makes theirs red. Ti Launchpads are red as well. The Beaglebone uses color, essentially, as a model number; Beaglebone black, Beaglebone green. This is possible because major PC board fab houses have made different colors more economical than they used to be.

I’ve been asked if the color makes any difference electrically. In general, no. If you’re dealing with super high speeds, RF, or other exotic conditions, it’s always best to ask your board house. In those fringe areas, a lot of things have the potential to make a difference. Other than that, if you can afford it, and want to make a statement, go for it. You can often get different color silk screen legend too. Just make sure there’s contrast between the two. White silkscreen on white soldermask would not be the best choice.

Duane Benson
Beware the monsters from Id

blog.screamingcircuits.com

An Electronic Business Card Holder

Posted on by

I design and build electronics at home, late at night when the spiders are out, and by day, I put my hours into Screaming Circuits. My job here doesn’t involve building things. I’m the marketing department, but I like to keep as much manufacturing smeared all over me as is possible. Here’s one way I do that.

Business cards are a bit of an anachronism today. I don’t give out many, this being the 21st century and all, but I still need some on my desk – I guess to look businessy or something. No one’s ever given me a cheap card holder with their logo on it, and I don’t want to just scatter cards around. So, why not combine my need to display business cards on my desk with my compulsion to create electronic things? With that thought in mind, I decided to build an electronic business card holder. Of course, I first had to decide just what an electronic business card holder would be.

Here’s what I came up with:

  • It should be small, about the size of a business card
  • It should have a lot of blinky lights
  • It should do something when a card is removed
  • It should have a long battery life
  • It should use tiny parts to show off our manufacturing capability a bit
  • It should be 100% buildable within our electronics manufacturing process (meaning it should be just electronics; no bolts or case)

That’s not a long list, but does involve a few decisions. I’m pretty familiar with Microchip PIC processors, so that would be a logical choice to drive the thing. Arduino compatibility would be cool, but I’d have more trouble with battery life, and the PIC microcontrollers come in some pretty inexpensive forms.

I’d recently been using a variant of the PIC18F46k22 on another project. I comes in a 5 x 5mm QFN package and can be purchased for less than $3 in small quantities. it has plenty of I/O and can be set to a very low power sleep mode. I settled on that MCU and a CR2032 coin cell battery for power.

Rather than add any extra hardware to hold the cards, I came up with an arrangement of pin headers and small push-button switches. (as in the photo below right). One header is the six-pin Microchip in-circuit programming (ICSP) header, and the other is a six-pin I2C/SPI header. Not that I need I2C or SPI, but with that, you could turn this into a robot business card holder or something.

I considered a light sensor to detect when a card is being picked up, but that would require leaving the A to D powered up, and it would be less reliable due to changes in lighting. I looked around my junk box at home, and found a Freescale MMA8452 3 axis accelerometer in a 3 x 3mm QFN package. It also has a decent low power mode, and can be talked to over I2C.

Some 19 GPIO pins remained open, so naturally, I had to put in 19 LEDs.

Stay tuned for my next installment, where I’ll go through some of the design decisions. At the end of this series, I’ll be giving out 10 of these, so stay tuned to see how you might be able to get one.

Duane Benson
If you dreamed you saw the silver spaceships flying
That’s a okay. They’re RoHS compliant

http://blog.screamingcircuits.com/

Raspberry 6.283185307 Zero

Posted on by

AKA a second post on the Raspberry Pi Zero.

It’s been two months since the release of the $5 Raspberry Pi Zero, and I still haven’t been able to buy any. As I discussed in my prior blog about it, there is plenty of discussion around the fact that, out of the box, it’s not real useful without adding enough accessories to make it as expensive as any other Pi model. I certainly understand that point, but here’s another way of looking at it.

If you want to learn software, buy one of the other Pi models. If you want to learn about hardware design, buy the Pi Zero and download some CAD software. Then go online and get the Pi Zero dimensions and start designing accessories for it. You can start with one of the many open source Pi Zero accessory designs, or come up with your own. Don’t look at it as a system that’s missing too many parts. Look at it as a base for a different type of learning.

One of the scariest things about designing a plug-in/on board for a bigger computer is the possibility of a mistake that will fry the expensive board. With the Pi Zero, you’re risking $5.

Like I said, I still don’t have one, but I’ve drawn up my for Pi Zero accessory:

It will plug right on to a Zero as a rechargeable Li-Poly power supply. Not at all a complex circuit, but it’s only the first in a series. As a small board, it doesn’t cost much to get fabbed, so for about the price of one PCB sized to fit the bigger Pi boards, I can get two of these.

Next, I’ll design a motor driver, and then possibly an IMU, or sensor board.

Duane Benson
If you have your Pi calculate Pi, would that Pi be Pi enough for Pi?

http://blog.screamingcircuits.com/

Predictions for the End of the Decade

Posted on by

Half a decade ago, back in January 2010, I wrote up a list of predictions for the end of the decade. You can read that list here. It’s still 2015, so I can plausibly say that we’re halfway there, which is a good time for a status update.

0000: In 2010, I said: By the end of the decade, 50% of all passives will be embedded passives and 20% of all PCBs will have 90% or more of their passives embedded.

In 2015, I say: This doesn’t look to be coming true, but it still might. As mobile devices and wearables get smaller, or more powerful, more electronics will need to be stuffed in progressively smaller areas. Those passives need to go somewhere. That somewhere could be into the PCB, or into the chips. I think the PCB is more likely.

0001: In 2010, I said: By the end of the decade, quad stack PoP (package-on -package) will be commonplace.

In 2015, I say: Quite likely. Double layer POP is showing up on low cost devices, like the $5 Raspberry Pi Zero. If it can go there, it can go pretty much anywhere. It won’t be long before double stacking won’t be enough. Although, the layers may end up being inside the chip package, rather than individual chips as layers.

0010: In 2010, I said: By the end of the decade, Each individual human will have their own IP address. Several of us will have more than one. That way, we can jury rig accelerometers into our hands and feet so we can wirelessly know where each of our extremities are at all times. Cats will have them too.

In 2015, I say: Yep, and then some. I already carry one in my pocket. In five years, we’ll likely see personally assigned IP addresses that won’t be device-dependent. We’ll be able to buy IP-enabled clothes, like gloves, which will do a lot more than just know where each finger is. The pet ID chips that today use NFC will be available in wireless Internet connected versions.

0011: In 2010, I said: By the end of the decade, solder paste will be used less often than not when assembling components on to PCBs.

In 2015, I say: We will be seeing welded copper, additive embedded 3D printing processes, conductive glue, and other non-solder methods of assembly, but nowhere near to the degree I was thinking back in 2010.

0100: In 2010, I said: By the end of the decade, nearly all hydraulics and pneumatics in new motor vehicles will have been replaced by electrics.

In 2015, I say: This is already well on the way. With electric and hybrid electric cars growing in numbers, and with weight and fuel mileage being such a concern, this has to happen.

0101: In 2010, I said: By the end of the decade,the first semi-autonomous passenger vehicle will be on display on the auto-show circuit. Hobbyist built semi-autonomous cars will already be on the road.

In 2015, I say: I may have missed the boat on this prediction, in the pessimistic direction. Part of it has already happened. I haven’t yet seen hobby kits, but most of the major car manufacturers have shown models. Tesla has a really good driver assist “auto pilot,” and is promising fully autonomous vehicles for sale within two years of this writing.

0110: I said: By the end of the decade, “airline pilot” will generally be a really, really, really boring job. That’s a bit of a problem.

In 2015, I say: The necessary level of automation required for this prediction to come true is already installed in most airliners. The only real question remaining, is how long before it changes from “Pilot primary, systems secondary” to “Systems primary, pilot secondary.”

0111: In 2010, I said: By the end of the decade, most military “foot action” will consist of two soldiers in command of a squad of robots and those two soldiers will as likely be in Fort Lewis, Washington as in the combat zone.

In 2015, I say: Sadly, I still think this will happen. Not sad that fewer humans will be shooting and getting shot, but sad that we as a species will still consider war important enough to be throwing large quantities of money and resources at.

1000: In 2010, I said: By the end of the decade, the president of the US will be promising health care reform as the highest priority.

In 2015, I say: Yep. The president, presidential hopefuls, senators and representatives will still see this as a hot issue. One side will be trying to make quality healthcare more accessible, the other side less. One side, more publicly funded, the other side, less. I’m not really sure which side will be doing which, but I’m certain that each side will say they want to fix it and the other side wants to destroy it. Ugh.

1001: In 2010, I said: By the end of the decade, routine bioengineering will be, well, routine. Very scary.

In 2015, I say: I’m not so sure about this one. When I wrote it, I was thinking that home bioengineering would be happening and a class of bio-hackers would be emerging. That may still happen, but it won’t be common. Governments, agriculture, and medicine will be doing a lot more of this, but I’m not sure the term “routine” will be accurate.

1010: In 2010, I said: By the end of the decade, the 2019 recession will be looming large and all the people who have forgotten about the 2009 recession and the 2001 recession and the 1985 recession and the 1975 recession … will be freaking out again.

In 2015, I say: Is there any doubt? Does this ever not happen?

1011: In 2010, I said: By the end of the decade, lead will be gone from 98% of new electronics. Bummer.

In 2015, I say: Exemptions are going away. This will happen.

1100: In 2010, I said: By the end of the decade, four of the substances that replaced the substances removed from electronics due to RoHS and similar regulations will have been found to be significantly more harmful to the environment and the people recycling the materials than are the substances that they replaced.

In 2015, I say: I was being tongue-in-cheek, but it still might happen. The only caveat is that if it does happen, the data will be so obscured by politics that it likely won’t be possible for anyone to come to an informed opinion.

1101: In 2010, I said: By the end of the decade, the world of intellectual property will be in even more of a mess than it is today. Virtually everything will be accessibly for easy theft and cheap replication. (This is pretty much a big “duh.”)

In 2015, I say: This is still well on the way. Any industry that designs things will need to adapt to keep competitive. The patent world will still be a mess. Copyrights will be more of a litigation attack weapon than a protection tool. The best defense against pirates will be faster innovation. On the positive side, a lot of IP sharing will be intentional (by the inventor) and many businesses will be built based on collaborative innovation.

1110: In 2010, I said: By the end of the decade,building your own multipurpose robot will be as easy as building your own PC was in 1988. Hardware components and operating systems will be off the shelf, but standards will be pretty loosely defined, interoperability will be more theory than reality and applications will be sketchy and buggy.

In 2015, I say: This will happen, but it may be a little later than the end of the decade. The technology will very much exist for this to happen, but the capability of the hardware will probably be advancing so fast that even the limited amount of standardization needed for this won’t be possible.

1111: In 2010, I said: By the end of the decade, still no flying cars and personal jet packs, dadgummit!

In 2015, I say: And, still no real hover boards.

Duane Benson
http://blog.screamingcircuits.com/