Teensy 3.1 Changes To Green PCB

Status
Not open for further replies.

Paul

Administrator
Staff member
PJRC is changing Teensy 3.1 to green solder mask.

teensy31.jpg


Since launching Teensy 3.0 on Kickstarter in late 2012, we started using black solder mask. It looked different, which seemed like a great idea at the time.

Over the last year and a half of making Teensy 3.0 and 3.1, we've been working on improving the quality and production yield. With any PCB, a small percentage fail. We test every Teensy throughly, so any shorts or opens are detected. The bad ones are saved for diagnosis and then we just toss them. (no, do not ask if we'll sell the bad boards at surplus prices, because we won't ever do that)

It turned out some of the failures could be traced back to the black solder mask. What seemed like it ought to be the same, just a different color, in fact isn't the same at all. The black material just doesn't have quite a high a resolution as the green. I was very surprised to learn this, but it's true.

So we're changing Teensy 3.1 back to green. The other PJRC products with black solder mask will also switch back to green as their stock runs out and we make more boards.
 
You can get white or yellow. However, I recently read an article describing problems with white and the coating flaking off. A problem that was attributed to the pigments necessary to achieve the white color.
I guess there's a reason that I yet have to find another color than green in any consumer or industrial device I've ever taken apart.
 
I will vote for crystal.

Btw I like the black one's but as my mother say's about the color of cars, it has to work apart this the color stays on the 2nd place and no I won't buy a car in pink. And at least inside the car you won't see the color.
 
Black and white also have the problem that it is difficult to see the traces under it. Green and read are better.
 
Clear would be awesome.

But Teensy 3.1 is a 4 layer board. Layer #2 (the one just underneath all the components) is a ground plane. That layer is opaque, except near the edges and between the pins where a via blocks the plane from reaching. In the photo, you can see board is much lighter near the edges, since this photo was taken on a light table with some light shining through the PCB.
 
Wouldn't that make a clear board end up looking almost like raw copper? That might be kinda cool, too...
 
I find this hard to believe. First, I haven't been able to find any reference to this being an issue. Where did you hear this? Second, motherboards and 3D cards and sound cards and all that fun stuff are way more densely populated than your boards and they're all kinds of crazy colors. If there were an issue with a particular color causing a significant increase in the failure rate, I kinda doubt we'd see any 3D accelerators or motherboards with black PCBs.

Are you sure the issue isn't just your PCB house? I mean, wouldn't any PCB house worth their salt be electrically testing the traces to make sure they're okay before they populate the boards, thus averting this issue? Could it be they're just doing a quick visual inspection and it's harder to see the traces on the black PCBs?

I use Gold Phoenix to manufacture my boards and their website says they do electrical testing on all boards with more than 2 layers.

[edit]

Actually I did find this one page just now which lists inks used for photo resist:
http://www.taiyoink.com.tw/en/products/index_01.html

It does list two different black inks, one with "excellent" and one with "good" performance. Perhaps your PCB house is using a poor quality ink? Still, I'd think they'd catch any bad boards with electrical testing.
 
Also please don't go with clear. That looks so cheap. It's worse than green. :(

I guess there's a reason that I yet have to find another color than green in any consumer or industrial device I've ever taken apart.

Really? You've never seen those tan boards with no solder resist that they stick inside stoves and washing machines where they only have a few through hole components? And you've never looked inside a PC where the motherboard is blue, and the 3D card is black, and the IO expansion card is red?

I assume they use green inside most consumer devices because in the past it's been the least expensive. But anytime they expect the consumer to see the PCB, as when upgrading their PC, they use other colors because green looks cheap.
 
I've started taking electronics devices apart 35 years ago. Not sure what I was thinking when I wrote that I had not seen other colors. But that is actually somewhat beside the point really.

Based on previous posts it's clear that Paul's manufacturer of choice is local and within driving distance. For a small company such as Paul's that does not make 10000 thousands of boards a month, with limited resources in terms of personnel and finances and operating on a very small margin, that is a must. When problems occur you can react very fast without having a whole lot of product in your logistics chain that could potentially be bad. So there is not that much of a choice there.

When you have enough data and have identified the color of the solder mask as a problem that reduces yield it is just common sense to address that problem. Unless you can convince your manufacturer to use a better product for the solder mask, there are really only two more options. You can change the solder mask to something that has more reliable over the last 40 years (because back then everything was either green or reddish), or you select another manufacturer with the very high risk of having startup problems because of their different problem. But then, again, if you're operating on a small margin that does not allow for much room to play.
 
Quite a few behind-the-scenes improvements have been made over the last several months.

I actually sent quite a bit of time looking into solder mask alternatives. There are a LOT of options if you're willing to pay more. But the last thing I want to do is raise the price of Teensy.
 
For me, the green pcb's will be useful, as I haven't yet ordered Teensy 3.1's, and the green vs. black pcbs should give me more of a clue of which target I'm going to. However, I suspect sooner or later, I will forget to change the processor type.
 
Teensy Loader parses your .elf file, if it exists in the same directory as the .hex file. The architecture (avr vs arm), stack starting address and a couple other symbols are used to figure out which processor you selected when building the code.

If the wrong board is connected, you'll get an error message explaining the mismatch. It won't load obviously incompatible code. Well, unless you delete the .elf file, in which case the .hex is blindly trusted and written onto your Teensy.
 
Based on previous posts it's clear that Paul's manufacturer of choice is local and within driving distance. For a small company such as Paul's that does not make 10000 thousands of boards a month, with limited resources in terms of personnel and finances and operating on a very small margin, that is a must. When problems occur you can react very fast without having a whole lot of product in your logistics chain that could potentially be bad. So there is not that much of a choice there.

I designed a microcontroller of my own with a budget of around $22K. I spent 8 months working on it full time, and spent the last $9K I had on a run of 150 of them. I know what operating on a small margin is like.

I also know that it costs less than half as much to get boards made in China as it does in the US.

To give you an example, the assembly cost for those 150 boards was $500. In the US it cost me $750 to have another 50 assembled. And it was only that low because they gave me a deal after I told them I'd been using Gold Phoenix. When I later emailed another one of their representatives asking for a quote for another 50 boards the price for assembly had jumped to $1,600.

And lest there be any confusion... That's $1,600 for 50 boards, not 150. So you're looking at $3-$5K to have 150 boards assembled in the US versus $500 for the same in China.

If you're operating on a tight budget, and you have to choose between rapid turn times or saving 85% on assembly, it seems foolish to choose the faster turn time. A couple weeks delay is probably not going to make or break you, but spending thousands of dollars more for your product might.
 
Last edited:
Teensy Loader parses your .elf file, if it exists in the same directory as the .hex file. The architecture (avr vs arm), stack starting address and a couple other symbols are used to figure out which processor you selected when building the code.

If the wrong board is connected, you'll get an error message explaining the mismatch. It won't load obviously incompatible code. Well, unless you delete the .elf file, in which case the .hex is blindly trusted and written onto your Teensy.
I figured there were checks. It is good to know, I won't cause problems when I get to multiple teensy versions.
 
Some more explanation on resolution

Hi Paul,

Can you please elaborate a more by what you mean as resolution that is available in case of green PCB and not black PCB ?

I basically want to know the difference that will be observed between the Green and Black PCB , Is it the visiblity of tracks or the layout or something related to electronics. Request you to please explain this with an example.

Thank you.
 
Parth:
The only difference between the two that you'd notice is the color. He's making the switch because he's getting too many PCBs back from the manufacturer that have broken traces or shorts, and he believes the color of the solder mask has something to do with this.

I'm not sure I understand how that could be the case though. Isn't the solder mask applied after they etch the board, like a stencil? If so, unless the problem were solder mask overlapping pads, which seems extremely unlikely, I don't see how the resolution of the solder mask could affect the traces and result in PCBs that don't work.
 
Can you please elaborate a more by what you mean as resolution that is available in case of green PCB and not black PCB ?

I wish I knew. PCB fab vendors can be remarkably cagey when it comes to process details, especially when having to admit something isn't up to spec.

I basically want to know the difference that will be observed between the Green and Black PCB , Is it the visiblity of tracks or the layout or something related to electronics.

That too is a good question. I know a bit more on this, but again, this involves some interpretation from what a PCB assembly vendor is willing to admit. (Two different vendors here, the "PCB fab" vendor actually makes the circuit board, the "PCB assembly" vendor solders all the parts onto the board)

The solder mask is a non-conductive film applied over the entire PCB, except the pads where you're supposed to solder parts. Its main purpose is to prevent the solder from flowing down the traces, away from the pad. A secondary purpose is providing a very low surface adhesion to the molten solder, so if any solder does accidentally get incorrectly applied off the pads, it will tend to be pulled towards the pad. The solder adheres to the plated metal surface on the pad, and has very little "stick" to the surface with solder mask.

With the black solder mask, we were seeing defects in the mask material. Again, I don't have a good feel for the numbers (going back to the difficulty of getting vendors to be forthcoming with quality problems that expose problems in their process), but it's probably something like 1% to 5% of the boards had these issues, in varying degree. I also don't have really good info about what the defects really were, but some I saw appears to be missing solder mask, or perhaps really thin spots in random places. It's remarkably hard to know after the board is built.

The board can still be soldered and work perfectly, even without the solder mask.

By far the worst problem we saw was difficulty inspecting the boards. Our assembly vendor uses a camera system. I haven't personally used this system or even seen it in operation (they did offer to give me a demo, but the timing hasn't worked out... I'm extremely busy and I tend to work odd hours). So this is only based on what I've been told, so you're hearing it 3rd hand. Apparently the camera views the PCB from multiple angles and image recognition software analyzes the solder joints. The problem is only an ideally shaped solder joint passes. A perfectly fine solder joint that has the solder distributed differently, because some of it "escaped" onto the trace underneath (especially to the large center pad on the QFN chip) or down the trace away from the part, will "fail" on the camera inspection system. Then a real human needs to inspect the board.

The vast majority of the cost on Teensy comes from 2 places: the silicon and the PCB assembly (software development isn't counted in the cost per board). If anyone ever wonders how Sparkfun is so successful, consider they do all their own assembly work and they do very little software themselves! Their main cost is the silicon, and they can afford to pour a lot of money into website development, which clearly shows in the awesomeness of their site. PJRC operates on a very different, not nearly as profitable model, where most of my focus is on developing software. Teensy3 is a 4-layer board, which adds cost too, but not on the scale of the chip and assembly. Much of the reason we can make Teensy affordable is because I work pretty closely with a couple local PCB assembly vendors. These troubles with the solder mask were mostly about automated inspection.

In extreme cases, solder mask problems can lead to unconnected pins or even shorts. If the solder that was supposed to connect the pin all flows away because the connecting trace is exposed, it's possible to end up with an unconnected pin. But that's pretty rare. The pad and the pin are plated with metals that optimize solder adhesion, and the solder is already touching them. For an open to happen, this needs to coincide with some other problem, like the part wasn't placed accurately or the PCB wasn't perfectly flat, which just don't tend to happen. It's also possible to end up with shorts between the pins. But again, for that to happen there needs to be solder deposited between the pins, which ideally shouldn't happen if the stencil is correctly aligned when the solder paste is applied.

If any part of this sounds well researched, let me assure you that is only in hindsight. The reality of dealing with these manufacturability issue is far less certain. We would get a batch of boards in, or a partial batch if we were close to running out (even though they would run all the panels on the pick-and-place and reflow machine at once, they were struggling with inspection which limited the rate we'd get them), but often we'd test and package them as needed to ship orders. Soldering the pins to boards takes a lot of time, which is why we charge $3 more for the pins soldered, and it puts a limit on how quickly we can test and package a batch of incoming boards. That limits how fast we can get feedback to the PCB assembly vendor. The pain they suffered having to visually inspect so many boards is largely forgotten and turns into a sign of relief when we report that virtually all of them actually passed electrical testing here. Remember, they're not not always very forthcoming about limitations on their end, with much of the communication issues internally between their production people and their management, so inspection issues didn't become known to me until recently. On this side, Robin, Erin and I are highly focused on delivering good products, shipping orders fast and a huge number of day-to-day things that come with running a tiny 3-person company. Optimizing a production process, where only a tiny percentage of boards actually fail, well, just isn't as high a priority as a mountain of other stuff. Until mid-2013, to be honest, a lot of Arduino libraries still needed porting for Teensy 3.0, so I just wasn't able to focus on much else that wasn't extremely urgent. The easy answer is to just write off and discard the small number of bad boards. It's taken about 1 year and especially the last several months of working closely with our assembly vendor to really refine this process. On both sides, we've really increased our attention to optimizing the manufacturing process. The same happened years ago with Teensy 2.0 and Teensy++ 2.0. Most of this stuff is very mundane details you'll never see. Another upcoming change is a very small adjustment of the size of the center pad under the Mini54 chip, and perhaps to the 4 smaller cutouts in the stencil that put 70% solder paste coverage on that big pad. Usually these refinements are invisible details. Much to my surprise, one of the problems turned out to be black solder mask. The change to green is pretty easy to spot. ;)
 
Last edited:
I should also mention high quality solder mask is possible in different colors, from premium PCB vendors. They tend to have premium pricing.

Laen from OSH Park helped me approach a couple fabs he's used. They have excellent quality and could probably do black solder mask very well. They certainly can with purple. But their prices are much higher. Using them would have meant raising the price of a Teensy 3.1 to about $20.50 to $21.00.

Likewise, the silk screen printing resolution can also vary quite a bit. If you look at a Teensy 3.1 under a microscope, you can see the resolution of the white printing. The Arduino people have absolutely beautiful printing and solder mask on their boards from the last couple years.... I believe starting around the era of Uno R2. I have an original Uno that has silk screen similar to Teensy and most circuit boards. Obviously they've invested a lot of optimization in the aesthetics of their boards.

My goal has been to keep Teensy as affordable as possible and sustain as much software development effort as possible, which is obviously only possible due to the sale of Teensy boards. Making circuit boards with distinctive colors seems to be the norm in this market. A few people suggested this and it seemed like a great idea for PJRC to try, so we did it starting with the Kickstarter for Teensy 3.0. In hindsight, doing it well with ordinary low cost boards isn't easy. I certainly didn't know this in mid-2012. PJRC is going to stay with the mainstream green PCBs (once the remaining black ones are sold... I believe only the audio board and wiz820 adaptor are left) and hopefully we'll distinguish PJRC from the many other companies in this market with superior software and support rather than the color and other aesthetics of the circuit boards.
 
Last edited:
hopefully we'll distinguish PJRC from the many other companies in this market with superior software and support rather than the color and other aesthetics of the circuit boards.

Spot on Paul!

Functionality and software are far more important, at the end of the day most of these boards will probably end up inside some form of enclosure and won't be seen anyway.

I probably should know something about solder resist as my old man set up and ran a very successful PCB ink company for over 25 years. All I do know is that some colors were harder than others to get right, I think sometimes it was just the amount of pigment, rather than what pigment was used. It's kind of logical if you think about it, you might have to add twice as much white pigment as green to get a good color, twice as much powder added to the resin mix might make it more brittle or not stick as well.

Anyway, I couldn't care if the Teensy was sky blue pink with yellow polka dots, it's what it does that counts!.

Regards,

Les
 
I recently received a green Teensy 3.1 and was confused for all of about 3 seconds. Once that was cleared up, I put the Teensy 3.1 to work and never thought about the color again (aside from now).
 
Status
Not open for further replies.
Back
Top