EEVblog #1259 - PCB Manufacturing Options EXPLAINED

[EEVblog]

What are all those PCB manufacturing options in the checkout when you order your board?
Dave explains each option one by one, from Aluminium to Z-Axis Milling.
Including obscure stuff like Organic Solderability Preservative.

[KaneTW]

What are the reasons to not get a glossy black PCB? I've never worked with one of those so no idea what the pitfalls are.

[maginnovision]

They're difficult to make anything out. It may as well be glass.

[wilmer]

 

I will love a video comparing different suppliers. Time of shipment and service support. I learned that PCBway is expensive if you live in any place that DHL mark as rural Australia.

[EEVblog]

I will love a video comparing different suppliers. Time of shipment and service support. I learned that PCBway is expensive if you live in any place that DHL mark as rural Australia.

And the video would be obsolete 6 months later, and a completely different result depending upon the country you are in.

[Syntax Error]

Dave, thanks for the video > an hour of informative joy 

Regarding the UL/RU symbol. Can a PCB designer 'self-certify' the board with the logo before it goes to the FAB house? On most online wizzard pages, the FABs state they will add the logo with their certificate number to the board. Are they only allowed to issue it or are they just cashing in on regulatory red tape?

On our own prototype creations, should we also silk the UL94 flammability code, the dreaded (but otherwise pointless) CE Mark and, the WEEE-ly bin logo?

[Rutger]

Thanks Dave for the great explanation of all the PCB fabrication options, there are so many...
I always went with the default options and they seem fine for me and I guess for most hobbyist.

 

[Psi]

Small goof at 27:35.
The board does have a few blind via's.
You can see them by the two tone colors inside the via.
Between Red L1 and  Purple L9.

[KaneTW]

Interesting. I've gone with glossy blue in the past but matte actually sounds like a pretty nice option.

[Syntax Error]

As per the guy holding the panel to the light in Dave's video, the OSH Park purple solder mask is pretty classy. Especially with the gold ENIG finish. Their dual sided silks are very sharp too.

There was a company in the UK called Ragworm that used a 'spectacular' bright red-orange solder mask.

[golden_labels]

Considering that chances of failed PCBs are low, wouldn’t it be cheaper to produce more boards and skip the flying probe test? In some configurations I checked that decreased the price by up to 40%. Not feasible for complex boards (your time is not free!) or with larger orders, but for simpler amateur projects where you may actually need one working board this could be a better option.

Comparing manufacturers is a poor idea. To make it fair and avoid any bias, reasonably objective requirements would have to be set — good luck doing that for a task with so many variables. And I guess Dave has more entertaining things to do than collecting numbers from tens of websites. Finally, at least some of those manufacturers are spammers and it would look silly if EEVblog would feature companies that abuse the forum.

What are the reasons to not get a glossy black PCB? I've never worked with one of those so no idea what the pitfalls are.

They are great for “protecting your intellectual property”, along with rubbing off part numbers.

[EEVblog]

Considering that chances of failed PCBs are low, wouldn’t it be cheaper to produce more boards and skip the flying probe test?

Most certainly.
But of course depends upon the BOM cost of the assembled components and the risk calculation.

[AndyC_772]

Considering that chances of failed PCBs are low, wouldn’t it be cheaper to produce more boards and skip the flying probe test?

Absolutely no!!!

The first you'll know of a failed PCB is when it's built up, tested, and doesn't work. By that point you've wasted time, effort, and a complete kit of parts, and must waste even more time and effort trying to locate the fault and fix it (if, indeed, it can even be fixed, which may not be possible depending on where it is).

Failed PCBs aren't nearly as uncommon as you might think, and 100% bare board testing is the way to *usually* weed them out before they get assembled into complete products.

[ali_asadzadeh]

Dave please add the different manufactures links,that you told

[EEVblog]

Regarding the UL/RU symbol. Can a PCB designer 'self-certify' the board with the logo before it goes to the FAB house? On most online wizzard pages, the FABs state they will add the logo with their certificate number to the board. Are they only allowed to issue it or are they just cashing in on regulatory red tape?

They are not in the certification business, no requirement on them to verify you have the right to use the logo, so they will print anything you want. It's the company selling the product that takes the legal responsibility of using the logo.

[AndyC_772]

I usually put a box on the drill drawing layer with an instruction "PCB manufacturer to add UL Recognised Component 'UR' mark on top side silk screen in this area".

[EEVblog]

Considering that chances of failed PCBs are low, wouldn’t it be cheaper to produce more boards and skip the flying probe test?

Absolutely no!!!
The first you'll know of a failed PCB is when it's built up, tested, and doesn't work. By that point you've wasted time, effort, and a complete kit of parts, and must waste even more time and effort trying to locate the fault and fix it (if, indeed, it can even be fixed, which may not be possible depending on where it is).

Sure you can, it's just a risk calculation.
Take even half the 40% figure golden_labels said, 20%.

If a board costs $5 you save $1 per board.
If your component BOM cost is $100 you only need better than 1 in 100 failures to break even on that. And that's assuming it's board you wouldn't rework.
Small extra for the little bit more processing involved, but that's neither here nor there.
If it's not a dense board, like say 8/8 you wouldn't even bother testing. But if you are running say 3/3, the odds of failure are significant higher.

[EEVblog]

I usually put a box on the drill drawing layer with an instruction "PCB manufacturer to add UL Recognised Component 'UR' mark on top side silk screen in this area".

Why not just put the UL logo yourself?

[AndyC_772]

Sure you can, it's just a risk calculation.

This is the point at which I re-tell this story, in which BBT *should* have picked up on a process fault which resulted in a 25% failure rate for bare PCBs.

I say "should" because the operator actually doing the testing was incompetent, and failed to pick up on the faulty boards. IMHO that's not a reason to avoid BBT, but instead, a reason never to hire idiots, nor tolerate them anywhere in your supply chain.

It was an expensive lesson for all concerned. Given the way in which PCBs are mass produced, I can easily see how a process issue could cause the same fault to exist on multiple PCBs, and without BBT it might not be the odd random board here and there which must be scrapped, but a whole batch could have a defect in the same location.

Why not just put the UL logo yourself?

Historical reason, really... it's what I've always done. Plus the PCB manufacturer is already required to put their own identifying mark and date code (which I can't include on the artwork, obviously), so it's an opportunity for them to apply the mark if they're licensed to do so, or flag up to me that they can't comply if they're not.

[golden_labels]

AndyC_772:
There are a few assumptions I have made, and I have clearly stated them. Two of them are, let me quote:

• “Considering that chances of failed PCBs are low”
• “Not feasible (…) with larger orders”

Of course, if you are expecting the supplier is likely to deliver bad boards, you can’t apply that idea. If you need a larger quantity, and therefore even 1% of failures will cost you a lot of time and annoyance, that makes no sense too. If you order for production, you want to minimize failure rate and you will also not save money on testing. But I made it clear in which scenario it may work — and many hobbyists projects and prototypes fall into that category.

I kindly remind you that merely a decade ago it was common to do such things using unreliable DIY termotransfer or UV etching, with no inspection other than your eyes and a few measurements at critical paths. And it didn’t ended in a catastrophe. Do not panic — trust the data (just do not neglect variance).

[SPCBD]

Thank you once again for the wonderful comments regarding the Saturn PCB Toolkit. 

____________________________________
Kenneth J. Wood
President
Saturn PCB Design, Inc.           
60 Spring Vista Dr, Ste B
DeBary, FL 32713-3269
www.saturnpcb.com
sales@saturnpcb.com

     

[Ice-Tea]

Are you sure you can get alu boards with 2/4 layers? I thought the vias would short out with the aluminium?

[golden_labels]

Are you sure you can get alu boards with 2/4 layers? I thought the vias would short out with the aluminium?

As for shorts: they may be unimportant for many designs for which aluminium substrate is used.

You may have 2-layer boards without vias. A LED PCB may have LEDs on one side, the driver on the other and those are joined by thick wires.

You may have 2-layer boards with vias that short to the aluminium substrate, if they are all the same node. A practical example: a LED PCB may have LEDs on one side, the driver on the other, positive driving voltage going through vias and negative being supplied directly from the flashlight’s metal body to both sides.⁽¹⁾

This photo is showing a similar configuration. It employs a single layer aluminium PCB, because third-party suppliers were used for both LED PCB and driver PCB, but this is a limitation of what I have at hand, not because technology limits it that way:

This is a stack of three elements:

• A Cree LED PCB, on aluminium;
• A copper disk used to supply negative voltage to the whole stack;
• An AK-47A LED driver, which is coupling to the positive terminal of the battery.

Note that negative wire has just a blob of solder on it, without any regard of shorts to the aluminium substrate or copper disk below it — simply because they can are the same electrical node anyway. The LED PCB and the copper disk act as a heatsink for AK-47A. The only isolated part is the red wire on the other side.

Going to a more general case: as far a I know multilayer aluminium PCBs are actually stacks of thin FE4 boards on bot sides of a thick aluminium board. Not multiple alu boards. And vias through the aluminium are still possible, but that requires covering them with dielectric. I have no slightest idea how that process looks like and how much it costs, but I imagine they first drill holes and then add dielectric, instead of using boards with dielectric on them.

____
⁽¹⁾ Typically the battery positive terminal is easy to couple with the middle of the PCB, while the negative one goes to the body’s tail and the return path is through the body.

[VK3DRB]

A few points:

1. Regarding panelisation, if you are outsourcing a complete PCBA assembly, if is often false economy to do panelisation yourself, because you don't know what the PCB manufacturer's panel handling requirements are. Get them to do it, but you might want to put on the mouse bites on your PCB yourself.

2. Please DON'T say "mil" when talking "thou". "Mil" is short for millimetre in the metric world which covers most of the known world. Avoid this ambiguity!

3. Just a suggestion. I tend to use blue mask for prototypes, and green for production boards.

4. Castellations are VERY useful. Anyone who has had is use Silicon Labs (formerly Bluegiga) WT12A or similar Bluetooth series modules will know why avoiding cast2llations to save a few dollars is a very bad idea.

[VK3DRB]

I don't think embedding general information text in a layer is a great idea, other than locking the info in with the Gerber. I think a README checklist as a Word document or PDF zipped up with the Gerbers and drill files is a better idea overall when you weight up the pros and cons. I have used both formats over the years, and PCB fabricators have never had problems with a README PDF added into the package.... I have always got what I needed with few, if any, questions asked.

Reasons:

1. All too often, text in the layer is not amended when there is a change, so it conflicts with PCB info. Eg: Mask says red in the embedded text, but the user selected green when buying the boards. A common mistake too is versions and dates. Miss changing a review date in the embedded info and then you may have re-spin the whole Gerber package if you have already released it, which will make you look silly.

2. Lack of ease of amending the embedded text. If you want to change the mask from red to green, you have to release a new Gerber package, which is time consuming and often a pain in the arse. Editing a README Word doc is much easier. It can be version controlled with the version controlled zipped file name.

3. Anyone can read a README Word or PDF document. Not everyone has the tools to read a embedded layer on a Gerber. All computer literate people can read a Word and PDF. But not everyone in you company (eg: Purchasing Department) can read embedded information other than what is printed out or exported to a hard-to-read PDF.

4. Sometimes, reading embedded text is painful on the eyes and always slower to edit compared to using a Word document.

5. The auto generated layer stack-up diagram can be easily exported to the Word document.

6. A Word document is portable and proforma. Not so with embedded text from one design to another, without a few hoops and hurdles.

Of course it makes sense to point to specific regions of the PCB using embedded text in a Gerber layer. So, what justification other than locking the embedded text to the design, is there to use general embedded text in a Gerber layer, over using a README file?

[KaneTW]

I've seen thou used in mechanical engineering circles, but almost exclusively mil in electrical engineering. I've not found it too ambiguous in actual use, but I get what you're saying.


How would I go about making a PCB where one of the prepregs has a cutout?

[Ice-Tea]

golden_labels: appreciate the time you've taken to explain. Learned something today

VK3DRB: I think you'll find that anyone who does PCB design will know exactly what 'mills' are. I understand your dislike, but that's just the way it is.

[nctnico]

Of course it makes sense to point to specific regions of the PCB using embedded text in a Gerber layer. So, what justification other than locking the embedded text to the design, is there to use general embedded text in a Gerber layer, over using a README file?

The problem is that the people processing the gerbers will need to be able to understand what is in the readme file. And the preparation process needs to include reading the readme file during processing.

What works best is to adhere to what the board manufacturer prescribes. ODB++ is a nice container format for production data. In there you can also specify the layer built-up for example.