JLCPCB alters soldermask without any notice

[KL27x]

I know what you're expecting. That I used a stock soldermask layer with clearances too fine for the manufacturing process, and JLCPCB simply changed it to make it manufacturable within their listed specifications.

Nope. The soldermask layer I sent has a nice and easy solid polygon covering entire area of a quad flat IC, extending about half a mm beyond the edge of the chip.

I've ordered from them before, done the same thing, and they made it no problem.

This time they changed it without any engineering query. Instead of one large polygon, they made the sodlermask opening as 4 small slits for each row of ICs and a fifth hole for the ground pad.  The problems with this are multiple. First off, it doesn't work. When trying to reflow a chip to this board, the soldermask opening ends right at the edge of the chip (if not slightly before!), which doesn't even meet the specification for this part. Datasheet specifies pads to extend 0.3mm beyond the edge of the chip. No matter how I try, a filet does not form to the sidepads. So short of having X-ray inspection, you can't even verify a connection is made. Worse, connections are NOT reliably made. The bottom part of the pads is too small for that to occur, reliably. Especially when there's that extra thickness of coverlay that was added around the ground pad!

This is besides the point that I purposely increased the exposed copper slightly larger than the standard footprint so I could solder these chips with an iron. Even if the altered soldermask worked properly when reflowed, I would not use them. I'd have them remade. But at least I could have verified the board works properly in other regards.

Worst of all, this board was a relatively minor change to a previous board, so I did not order prototype quantity. I ordered a fairly large quantity. And not they're all trash.


Now am I being unreasonable? All I asked was for them to be remade with my soldermask layer, and I would bite these boards and pay again. So what happened?

I spent the next week receiving a free lecture from their engineer as to how he fixed my boards. After that waste of time, I finally received instructions to include instructions next time to use my soldermask layer and include the magic words "I accept all risk of solderbridges and losses caused by soldermask." So I hit reorder, added this message, checked out, and what happens?

I get message that I have to start order process all over from the beginning. Adding more salt, I paid them extra to panelize my board for me. Now I have to put the specs and quantities in all over again, reupload gerber, and pay to panelize my board, again. Good grief. I might as well order from a different manufacturer at this point.


This isn't a "don't use JLCPCB" rant. I was happy with them before. Just a caution that you must apparently add a note to request they actually bother to use the gerber files you sent them, especially the soldermask layer. They have a new engineer there who is very bright and eager to improve your boards without notice, and he will also give you free lectures.                 




 

[thm_w]

Photo would help.

I'm surprised they bother to look at any gerbers that don't obviously violate the rules. It seems like a waste of time to me.
Just have the disclaimer "we make your design if your design is bad no refunds."

[wraper]

This time they changed it without any engineering query. Instead of one large polygon, they made the sodlermask opening as 4 small slits for each row of ICs and a fifth hole for the ground pad.       

I've experienced Chinese fabs messing up with stencils in a similar way by reducing aperture in different ways, even when explicitly requested not to. Probably because many hobbyists just leave 100% aperture everywhere.  Why in hell you'd mess with solder mask though. Picture would certainly help to see the issue as it's hard to understand what it was and what they did. Row of ICs or did you mean row of pads? 

[Smokey]

There is a check box on the order page labeled "Confirm Production File".  You should probably check that and do your review of their production gerbers in the future.

I wonder what would happen if under "PCB Remarks" notes section you just put something like "Gerber files MUST NOT BE MODIFIED!!!  Boards MUST be made EXACTLY AS SPECIFIED IN GERBERS!!!" 

[thm_w]

There is a check box on the order page labeled "Confirm Production File".  You should probably check that and do your review of their production gerbers in the future.

Looks like you are right that would help, https://jlcpcb.com/help/article/582-how-to-confirm-the-production-file
Then download the files and do a gerber comparison using some free tool: https://www.altium.com/gerber-compare/

Shouldn't have to but.. good to know

[KL27x]

Thanks, Smokey. I'll have to look for that. I don't recall it. They definitely have the correct gerber. The engineer sent me a picture of it, the way it is supposed to be, to show me where it "would short-circuit" and why he changed it.

Wraper, that's funny. But I typed that wrong. They made 4 small slits for each row of pins on the quadpack. So in addition to now covering parts of the copper that are necessary to form the side fillet and see a connection, they also don't connect properly. If you set the chip on the board, you can't see any exposed copper, at all. That's how small they changed the openings.

Either some pins don't connect, or if you try to use more solder you get short circuits. I tried populating some with hot air, and I have yet to succeed.

The worst part is the delay. When you ask "how do I get these made with my soldermask layer and shovel more of my money at you?" You wait and wait to be told by the engineer why his board is so much better and how your board will end the universe. And still no answers to any of my questions.

[HwAoRrDk]

There is a check box on the order page labeled "Confirm Production File".  You should probably check that and do your review of their production gerbers in the future.

Beware that JLC only give you a finite period to review of 24 hours before auto-confirming. Orders don't sit at the confirmation stage indefinitely until you act.

I got caught out by this recently - submitted an order late in the evening (my time), and the "please confirm" email came through about an hour later. Didn't see the email until next day, but something urgent came up and couldn't do it straight away, and by the time I could it was already auto-confirmed.

They don't mention this anywhere I've seen. I suggested to their customer service that this is far too short a period, and at least 48h would be more suitable.

[JLCPCB Official]

Thank you for your feedback! I’ll forward this to our support team. If necessary, feel free to provide your order number, and I’ll look into it with our engineers.

[Smokey]

There is a check box on the order page labeled "Confirm Production File".  You should probably check that and do your review of their production gerbers in the future.

Beware that JLC only give you a finite period to review of 24 hours before auto-confirming. Orders don't sit at the confirmation stage indefinitely until you act.
...

When it's important, I'll open a chat with a real person (as far as I can tell anyway) and make it very clear that an order is not to be started without gerber review and confirmation.

[KL27x]

Last I have heard they are remaking my boards, now. Agent also acknowledges this shouldn't have occurred and claims it won't happen again.

It took me some time to even remember all of the reasons I altered the soldermask openings on some of the other component footprint. All of which were altered.

Just one example: ICSP pogo pin header. I removed the soldermask between pads, turning the opening into a single big rectangle. So I'm silently staring at my tstop layer, wondering the purpose. But I know I don't want it altered.

Couple weeks before recall kicked in. I did this because it makes it easier to clean the pads with a Q tip, if necessary. It's just I haven't use this header, myself, in many years. When I need to do ICSP, it requires a more stable, hands-free connection. Because I'm writing new code, and that never works on the first try. So I hardwire a better connector. But pogo pin connection IS used further down the road. If I didn't have a problem populating the board, I probably wouldn't have noticed this change. This would have just made someone else's days slightly more annoying!

That's my free lecture. Manufacturing is complicated, and reasons for weird decisions are not always obvious even to the person who made them.

[thm_w]

Sharp pointed or crown tip pogo pins shouldn't require cleaning of the board. But interesting info regardless.

[KL27x]

It was nearly ten years ago since this last occurred, but I have been known to leave fatal bugs in my code and spend a weekend on-site helping to reflash boards.

Most of the time, you don't need to clean pads. Sometimes you do, because of stray flux splatter or oxidation. Or just because the flash failed the first 1-2 times and you have to rule it out.

Having been on this end of things, in that scenario where you might be flashing boards for the next several hours of your life, nonstop? This small tweak can make your life easier. If you're an EE who works only behind a computer, it's easy to believe that everything is black and white, either 100% works all the time or it doesn't, and that's rarely the reality. And with sufficient quantities, even fringe cases will add up and become significant.

That said, +1 agree on spear or crown-points.


I feel like I dodged a bullet, finally asking for a refund. Even including the suggestion I might as well start over with a new manufacturer. My previous favorite seems to have gone out of business since COVID. And JLC prices are hard to even compete with, for these boards, specifically. That was basically "now I'm finally so irritated, I'm talking nonsense." The reason I was originally happy to pay again to get my boards done correctly, is because JLC is good.  Going elsewhere was a bluff, and I would have eventually resubmitted a new order after pouting for a bit!

[Infraviolet]

I had an issue once where JLC cut out a milled slot in a weird way such that it was at an angle to what my gerber specified. I was using the confirm-before-production option, but the gerber I confirmed for was correct, so this error got introduced after my review and confirmation and it wasn't known by me until the boards showed up. JLC had a quality complaint facility in their website, I sent them a photo of the PCB slot that arrived vs what it looked like in my original gerber and they gave a refund (money off next order) on those boards from among a larger order. With luck you might be able to get that from them for this soldermask ****-up.

[KL27x]

I'm still waiting on confirmation that my replacement order is even started. When this first began, I wasn't in a hurry. But weeks have passed for nothing.

While popping back onto my account to check, I noticed something I missed before. They didn't even alter my soldermask. They just decided it would better to use my paste layer as the top soldermask layer.  When I initially read that they used "data from your paste layer" and showed how my "soldermask layer would cause short circuits," I assumed they knew what those words meant and that they altered my soldermask based on their expertise on how the paste layer looked.   He was maybe just repeating the names I gave these layers and assumed they were a cryptic inhouse naming system one wouldn't pay anymore attention to than the name of a racehorse?

Now I see how this could happen with no query. They don't have strict rules for layer extension ".xxx" at JCL. I suppose engineer that doesn't speak english might just pick the one he likes, even though my layers are clearly titled in english. And it should not have made sense, because the boards uses both top and bottom soldermask layers (matching/complementary titles and extensions) and only top paste layer. But I can see if you make that blunder, now you don't need an EQ.

In the future, I'll be sure to remove any unnecessary layers. 

[xvr]

> They don't have strict rules for layer extension ".xxx" at JCL.

They have preferable naming (and it described on its site). If your files match it there is should be no problems. Otherwise they will do their best, but not always right

Very strange that they won't come back to you for approve 

[KL27x]

Thanks!

That's was one of my first questions that was never answered. How to get boards made using my original soldermask layer? Why did this happen and how do I avoid this from happening again? Did I do something wrong?

Even though they're remaking them, they never answered this. They just told me to not worry, it will never occur, again. So I'll be sure to look these preferred extensions up.

I don't expect any given tech to know all the words that might be used for a paste/cream/stencil layer in 100 different languages. You'd think "soldermask" is more univeral, if they actually read the layer names. Maybe they don't always or ever do that.

[KL27x]

After uselessly poking around the website, I finally found it using Google. Thanks.
https://jlcpcb.com/help/article/592-gerber-files-preparation

They don't even have an extension for what I idiosyncratically call my paste layer. I don't think I've ever ordered a stencil from JLC, come to think of it. It's just in my default gerber CAM job for the occasion I need it.

[Kean]

After uselessly poking around the website, I finally found it using Google. Thanks.
https://jlcpcb.com/help/article/592-gerber-files-preparation

They don't even have an extension for what I idiosyncratically call my paste layer. I don't think I've ever ordered a stencil from JLC, come to think of it. It's just in my default gerber CAM job for the occasion I need it.

Their file naming conventions are pretty common, e.g. as used by Altium (with my OutJob settings at least) and Eagle/Fusion Electronics.

The paste filenames would be boardname.GTP and boardname.GBP

I've had approaching a hundred stencils from JLC using that filename convention and the layer files have always been automatically detected correctly.

[Smokey]

Layer Markers:
One thing I do is below the actual board layout, I add text with the layer name on the actual layer.
So "TOP COPPER" is actually text on the top copper layer.  Do that for all layers in a neat grid.  "TOP PASTE", "TOP MASK", etc.

This marks each layer with what that layer is right on the gerbers.  As long as it's outside the board outline, they will just cut all that off when they make the boards (like FAB notes and title block and all the other stuff they just cut out).

No more guessing what layer does what.

About the paste layer:
One thing to be careful of is that JLCPCB will make a paste stencil using whatever you give them on your paste layer.  So if you don't have your apertures right, they won't fix it for you.  This is a bit of a shock, especially if you only have outsourced stencils to assembly CMs previously.

[KL27x]

Very helpful, guys. Thanks so much.

I also got a shipping notification from DHL, so I take it my boards WERE made VERY FAST.

I take it the "P" is for paste. And I see you guys also call it so, at least when talking with me. I'm glad I'm not a complete idiot. I'm largely self taught!

[HwAoRrDk]

About the paste layer:
One thing to be careful of is that JLCPCB will make a paste stencil using whatever you give them on your paste layer.  So if you don't have your apertures right, they won't fix it for you.  This is a bit of a shock, especially if you only have outsourced stencils to assembly CMs previously.

They can and do change paste layers. I know, because they did it to one of my recent stencil orders. Changed the shape of pads on my QFN and FFC for no apparent reason (from rounded corners to completely circular ends).

As I understand it, they universally apply "solder beading treatment" (adding notches to pad apertures of 0805 and larger parts), and will tweak aperture sizes according to the stencil thickness (which they choose unless you specify). And if you don't include a paste layer at all, will make it from solder mask layer.

So no, JLCPCB don't make stencils 'as-is'.

[Smokey]

About the paste layer:
One thing to be careful of is that JLCPCB will make a paste stencil using whatever you give them on your paste layer.  So if you don't have your apertures right, they won't fix it for you.  This is a bit of a shock, especially if you only have outsourced stencils to assembly CMs previously.

They can and do change paste layers. I know, because they did it to one of my recent stencil orders. Changed the shape of pads on my QFN and FFC for no apparent reason (from rounded corners to completely circular ends).

As I understand it, they universally apply "solder beading treatment" (adding notches to pad apertures of 0805 and larger parts), and will tweak aperture sizes according to the stencil thickness (which they choose unless you specify). And if you don't include a paste layer at all, will make it from solder mask layer.

So no, JLCPCB don't make stencils 'as-is'.

True.  I should have been more specific.   
They might change your paste layer to make it easier for THEM to make the stencil, but they won't change it to fix obvious mistakes you made in aperture size or shape.

[KL27x]

Got my boards...

And now I realize I didn't hit "fill" after I loaded up my PCB files and sifted through the proper layers. 

There are definitely some spots where my mask looks very bad, and their engineer was surely trying to help me out.

One is the ICSP header. The ground plane creeps into the opening and it looks like an obvious short waiting to happen. Of course there's no solder there (even though my paste layer shows it). If my crystal ball is working correctly, this is a complete non-issue.

There's an additional footprint where a sliver of ground plane is exposed and it looks like a mistake. I might optimize that a bit better if there's a future order, but solder will not go anywhere near that. This larger opening is still an improvement, making it easier for me to populate that part.

There's another spot I enlarged an opening for a passive. I believe I did this in order to make it easier to take a variety of different size packages, including ones that are technically too large for the pads. This mistake is as dumb as it looks, because it reveals some of the ground plane around a power pad. This is the one thing that is not an actual improvement. But I don't even think it will cause extra time/work during population, using the correct sized parts.

So my board looks like a nightmare to a PCB engineer. But it works, now. No major issue. Customer happy. And I'll be using the correct
Gerber extensions here on out.

[xvr]

> And now I realize I didn't hit "fill" after I loaded up my PCB files and sifted through the proper layers.

My condolences 

Just note from me - in KiCAD I ALWAYS run DRC after any final change (no matter how insignificant they may be). KiCAD automatically refill all polygons and performs check. Couple of times I've got clearance violation on this stage, despite the fact that I only moved some traces with turned on run time DRC check.