Category Archives: Screaming Circuits Blog

EMS provider Screaming Circuits offers tips and thoughts on design and manufacturing

Parts Substitution Gone Big

Posted on by

I’ve mentioned some cautions with parts substitutions before; wrong V values on barrier or flyback diodes, counterfeit parts and things like that. Here’s another example of something to watch out for if your supply is tight and choices are limited.

One of the things that I’ve run across a couple times, especially when hunting down capacitors, is the package size issue. Say, I need a 16uf, 10V cap on one of my boards. It’s not a critical app. I don’t particularly care about ESR, temperature or even much about tolerance. I just need a little head room in case of minor spikes or power line ripple. I’m not expecting a lot. I just want that safety margin.

But when I run over to my parts supplier, the specific cap I picked two weeks ago, when I started the design, is out of stock or jumped in price. I want to get building, so I just look through my parts drawers for something close. There it is, a 22uf, 50V. It’ll still work just fine. The problem is, of course, that I neglected to realize that the part  jumped up a notch in size. Bummer days.

I’ve run across the same problem, not due to a sloppy sub, but also due to picking the wrong footprint in my CAD package. I find that particularly easy to do with SMT electrolytic caps.

The other thing in these examples to watch out for is the open vias next to the pads. Granted, they aren’t in the pads, but they are close and without any kind of a break in the metal before the via. In the left pad of the yellow tantalum cap, I added in an example of a little solder mask dam between the pad and the via. That’s the way you should do it. Even though the vias are off pad, solder can still wick away and down the via – especially with leaded solder. Bad news if that happens.

Duane Benson
Have no fear, Underdog is here…

http://blog.screamingcircuits.com/

Newark Electronics and Eagle CAD — Interesting

Posted on by

I just read that Newark purchased CadSoft, including Eagle CAD. I guess it’s probably old news to everyone but me. The press release about it that Newark posted on its website, Element 14, has a date of Aug. 13. I find this purchase an interesting development and I don’t quite know what it means, or if it means anything.

I guess partnering is becoming a trend. Certainly, we’re involved in some good partnerships (Sunstone, Digi-Key, NXP, National Instruments) and Sunstone’s PCB123 connects up with Digi-Key parts. It does make sense. The engineer’s job has just gotten more difficult with this recession and the ensuing reduction in support staff. That’s pretty much what our ECOsystem partnership is all about — taking the disparate tasks involved in getting a prototype built up and reducing the steps and complexity involved in the process.

The Eagle / Newark deal does have me very curious. For one, I hope the CadSoft folks got a good deal. Their product has done a lot toward lowering the barriers to electronics design and they deserve a lot for that. The big questions are for the future. Will Eagle remain as accessible as it is? Will Newark throw a lot of resources into it and keep it moving forward? Will it get good attention or will it be treated as an impulse buy and not be given focus or direction? Hmmm…

Duane Benson
What about Element 32?

http://blog.screamingcircuits.com/

Ye Hardware Engineer on Quest for Firmware

Posted on by

I’ve been spending some time with the mbed here and I’m convinced that there are a lot of good uses for this little thing. One in particular popped in to my head with a lot of vigor.

Back in December of 2008, I listed 10 (Octal) things to do to help get through a lousy economy (read #3). If you’re pretty much a pure hardware engineer, now might just be a good time to develop some firmware skills.

In my mind, one of the biggest problems going from hardware to firmware isn’t the programming itself. That’s not really as tough to pick up as you might think. But it’s the environment. The tool chains. The configurations. Make files, environment variables, linked libraries, boot loaders, ICSP, flag bits… There’s a load of ancillary junk that gets in the way. Some micros require purchased proprietary compilers. Some use open source. Should you use C or C++ or ASM? Too many choices.

Well, here’s something that gets rid of all of that extra junk. Plug in an mbed and in minutes you can be experimenting with C programming on an embedded micro controller. Use the onboard LEDs and sample programs to get instant gratification. Plug in some external LEDs or a sensor of some type (maybe from sparkfun) as you get a little more versed in the language. Save the data to the FLASH and graph it in Excel or something.

My personal feeling is that a hardware engineer is much more employable these days with the ability to write firmware. I haven’t found a better way to get started then with an mbed. You can worry about all of the other details later, but use this little guy to teach yourself to code.

Duane Benson
It’s been a soft day’s night, and I’ve been coding like a frog

http://blog.screamingcircuits.com/

A Few Hints of the Centroid File

Posted on by

Every now and then, we get questions on the centroid file (aka pick-and-place file or SMT locations file). Most CAD applications will create one for you. If you use Eagle, download our ULP and run it to create a centroid from your board.

If you want to poke around and need some hints on what’s what you can download our guide on the centriod file. Here are a couple of illustrations from the guide. First, the point of origin needs to be centered in the part.

It should be centered in a box that contains the outline of the pins as well as the body of the part. The chips (above, left) are easy. The connector (above, right), is a little more ambiguous, but as you can see, it’s centered around the imaginary box containing the area. Top-side rotation goes counter-clockwise as shown (below, left) and rotation on the bottom side is simply a mirror image, left to right with clockwise rotation.

Diodes and other passives should have their zero rotation vertical, with pin one (if there is one) facing up. That would place the cathode up for diodes and the positive side up for electrolytics and other polarized two lead parts.

Duane Benson
If you get dizzy spinning counter clockwise, go to Australia

http://blog.screamingcircuits.com/

Via-in-Pad with Passive Components

Posted on by

Most of the via-in-pad writing I do concerns BGAs and QFNs. I do cover other parts from time to time, but the subject seems to come up most often with those packages. It is an important subject with passives too though. If you need to make your board smaller, putting vias in the pads of all of your passives may seem like a viable option to gain a lot of space. If you fill and plate over the vias, then, yes. It’s a good plan. If you leave the vias open, then no. It’s not.

Here are some via-in-pad guidelines:

Part type open via Solder mask
capped via		 6mil or smaller
open via		 filled and
plated over
BGA and LGA land pads Bad Bad Bad Good
QFN, TO-(power part) thermal pads Bad Acceptable Maybe Good
QFN signal pads Bad Bad Bad Good
Passive pads Bad Bad Bad Good

You can probably see a somewhat common theme in the table above.

All of these pictures show bad stuff. These are from the “don’t ever do this” camp. Open vias on passive parts can lead to tombstoning, poor mechanical connections, solder blobs on the back side of the board and crooked parts. Open vias on BGAs can also lead to the solder ball being sucked off of the the BGA. Bummer dude.

If you do use solder mask capped vias in a thermal pad, most manufacturers recommend the via cap be about 100 microns bigger then the via. This prior post here shows a decent example of using solder mask caps in the center thermal pad of a QFN (the rules from QFPs and DFNs are the same as for QFNs). And, I’m calling it a thermal pad in the center of the QFN, but the rules still apply of the pad is just for grounding and not for cooling.

Duane Benson
Where are we going? Planet ten
When are we leaving? Real soon

http://blog.screamingcircuits.com/

My MBED Is Up

Posted on by

I wrote about the new mbed development board a while back and mine just arrived over the holiday weekend. I have to say, true to it’s promise, It was the easiest piece of development hardware that I’ve ever brought up:

1. Take it out of the shipping box
2. Plug in the USB cable to the board and my computer
3. Wait a minute for it to be recognized and open up like a USB thumb drive
4. Double click on the web shortcut in the drive
5. Register
6. Click the Compiler link
7. Pull up a code sample and modify it a bit (I didn’t need to modify it, but I did anyway)
8. Click the compile button
9. Save it to the mbed as though it were a USB thumb drive
10. Press the reset button on the mbed board.

That’s 10 steps, but it’s only 10 steps. There was nothing else to do. Nothing. The longest step was No. 7, which took me about two minutes. I programmed a “Knight Rider” sweeper with the four on-board LEDs. I made one of those for my Jack-o-lantern back at Halloween, so it was the first test program that popped into my head.


I built the Jack-o-lantern sweeper with eight LEDs and an 8-bit PIC16F819. The PIC I used came in an 18 thought-hole DIP package, costing $3.22 at Digi-Key, and I hand soldered it all on an old perf board. It runs at 20 MHz, has 16 GPIO, 3.5K program code space, 256 bytes of flash and 256 bytes of RAM.

The 32-bit NXP LPC1764 runs at 100 MHz in a 100-pin LQFP and costs $8.70 in a quantity of one at Digi-Key. (The dev board, of course, costs more then that) It has 512K of program flash and 64K of RAM. The dev board can have up to 25 GPIO (the chip can have up to 70 GPIO with your hardware) along with the standard assortment of peripherals that can be configured, including six hardware PWM channels. The mbed dev board is like a breakout board configured as a 40 pin 0.1″ DIP so it will be easy to prototype with.

The processor, being a fine-pitch package really isn’t hand-solderable like the PIC except for by the most adventurous of folks, but that’s where Screaming Circuits comes in. Why wait for your custom hardware before starting on the software. Get one of these mbed dev boards to work on your software while the EE folks are designing the custom hardware. Then, when they’re done, we’ll assemble up the prototypes and you can integrate it all together. Take some time out of your development schedule that way.

I’ve wanted to try out an ARM processor for quite a while, but prior to this, haven’t found the right way to do so while keeping within the limits of my time availability and skill set, but this looks like it could very well do the job.

Duane Benson
Robots rule!

http://www.screamingcircuits.com/

10 Electronic Things to be Thankful For in 2009

Posted on by

It’s that time of year when we take stock of what’s good in our little worlds. Since I’m writing this on my work blog, I’ll keep my top 10 items focused on work-related thingys.

Number 10: Allocation!? Well, maybe. Nobody likes parts shortages and allocation, but maybe, just maybe, it means that we’re seeing the light at the end of the recession tunnel.

Number 9: The mighty QFN. Yes, I know the package can be a pain to layout properly, but the size reductions we can get with it are pretty cool. It used to take something like a TO-220 or D2Pak to drive an amp of current drain, but some of these new devices can do it in a little QFN (properly laid out, of course) form-factor.

Number 8: 99.47% on-time delivery in the last year. That’s less then one job late per month – and remember, if we’re one day late, the assembly is half off and if we’re two day’s late, the assembly is free.

Number 7: The Beagleboard being open source. It’s really opened up the world of high-end non-i86 embedded processors to a very large segment of the industry that just couldn’t quite get there before. Well done Beagleboard folks!

Number 6: The Internets. Back in the olden days when I was burning my fingers soldering up discrete transistors and plain TTL and such, I had a shelf of data books. I think I may still have an old purple National Semiconductor TTL data book buried in a box somewhere. It was always cool to page through those data books, and, of course, I didn’t need to be online in order to find what I needed, but heck, I can find it all now and even more without getting up and walking across the floor to my book shelf. In fact, I pretty much don’t have to move at all anymore thanks to the Intertubes.

Number 5: Google translator. Earlier today, I got an email written in German. Before online translators, I wouldn’t have been able to do anything with it and I would have missed a very big opportunity. The email was from a barrister in the tiny country of Togo. Apparently, he’s been looking for an heir to pass an inheritance to and can’t find one. He said that he went to the American embassy and they suggested me. If not for the Google translator, I would have missed out on this wonderful opportunity to get seven million dollars transferred right into my bank account.

Number 4: Level translators. It’s still a pain to deal with interfacing signals at different voltage levels; like a 5V I2C device to a 3V I2C bus to a 1.8V GPIO, but it was way more of a pain before easy to use level translator chips became widely available. Especially the bi-directional chips. Much more convenient.

Number 3: Better static protection built into chips. Yes, we still religiously use static ground straps. We have a conductive floor and wear foot straps and anti-static jackets and have anti-static stuff all over the place, but chips are so much more robust then they used to be. I can remember the old 4000 series CMOS chips. It almost seemed like if you breathed wrong, they’d get zapped.

Number 2: The LGA form-factor package. Just kidding. LGAs are annoying. Sure, there are some redeeming qualities: low profile, a RoHS part can go both leaded and unleaded, decent heat transfer. But, they also don’t flex as well as a BGA and the pads have the disdvantages of both BGA and QFN packages. Basically, they’re just annoying.

Number 1: And the number one electronic thing that I’m thankful for are these little Flash 8-bit microcontrollers like the PICs (that I use) and Atmels (like the Arduino uses). Holy mackerel, they make life a lot easier. All that GPIO, no support chips. And, self programmable flash. Ahhhh … Anybody out there still have a UV EPROM eraser?

Duane Benson
Embedded in my head

http://blog.screamingcircuits.com/

Tented QFN/QFP Via-in-Pad

Posted on by

Below is a pretty decent example of mask-tented vias in the thermal pad of a QFP. Most manufacturers recommend no more then 100 – 125 microns wider than the via to minimize voiding and thermal insulation in cases like this. This is a reasonably inexpensive way to handle vias in the thermal pad. Sometimes though, the tents will pop open, allowing solder to wick down through the via.

The mask over the center via on the right looks a little thin, so you’d want to give it an extra look over after reflow to make sure it’s okay. (We’d do that here, of course.)

We’d rather not see this technique on really small parts because it gets difficult for the fabricator to put down the mask with enough precision. With small parts, filling and plating over the vias is the preferred technique. Well, that’s always the preferred method. It’s just more important with smaller parts and BGAs. This method is acceptable for most QFPs and larger QFNs, though.

Duane Benson
All your via are belong to us

http://blog.screamingcircuits.com/

Pin BGA Interconnects

Posted on by

My “Speaking of Art in the Process” post used a photo of a point of load power module as an example. The specifics aren’t really relevant to this post though, but a commenter by the name of “Me” asked about the type of pins connecting the module to the main PCB.

“Do you know where to get those pins to attach two boards like that? I mean, do they sell just the pin for example on Digi-Key and give it a name, or is it just wire. Can’t see if they are pins with a lip to lift board to a set height.”

The part (above) came with the pins already on, so I don’t have a specific part number for the interconnect pins. I have some underside photos (below) that give a better view. They are basically solder-type terminal pins with a solder washer and BGA ball on one side (to attach to the main PCB) and either a press-fit or solder type side to affix to the module PCB.

I wasn’t able to find this exact part from Digi-Key or Mouser. Vector sells the solder washers and lots of interconnect pins of this sort, so they may be able to steer you to them with a phone call. This board uses the BGA style, but we’ve seen other POL modules of a similar type with thru-hole solder pins too. Digi-Key has lots like that. Here’s one example of some through-hole terminal pins from Mil-Max. You could use the solder-washers (like a T124 from Vector) to put some space between the module and the PCB.

Duane Benson

QFN Land Pattern Mix-Up

Posted on by

Another QFN oopsie here. The QFN looks good and the stencil cut-outs look good. The mask, though, as shown in the middle, is not good. In general, you shouldn’t put mask on the QFN land pad except to cover up open vias.

The QFN center thermal pad for this part should be free of solder mask. The stencil does give a good amount of paste — roughly 50% coverage; maybe a little more.

That’s the way we like it. But, with the mask on the board, there will be too much solder and too much with nothing to stick to. It will either flow to the small metal areas and raise the QFN up so the sides don’t solder or the excess solder may turn into solder balls and short something out.

There are some cases where it’s okay to only expose the metal in small areas of the QFN thermal pad (see here), but if you do that, the stencil has to match the pattern. Otherwise, like in yesterday’s post, you’ll probably end up with a gloppy mess.

Duane Benson
That’s the way aha aha your pcbs like it aha aha

http://blog.screamingcircuits.com/