Tearing my hair out with a Chinese PCB company

[peter-h]

I've been buying a panel, 32 circuits on it, for many years.

The last company, China, ran for ~ 5 years and then vanished.

I found another. I sent them the gerber data, and for avoidance of any doubt I sent them by DHL a panel of the existing (production/assembly tested) board.

The idea is that they test the existing sample panel against the gerber data, to make sure there is no mis-interpretation of the data.

But they didn't get this at all.

After many emails, dealing with poor English, I agreed to pay USD 200 for them to make 1 panel from my gerber. They could then test that panel, and immediately test my sample panel, and if they both pass then obviously the two panels are electrically identical

No joy...

3 months and many emails later, no progress. It sort of appears that their panel probably does test ok. But they just don't get the principle of testing both panels to the same data.

Is there some other interpretation on this saga?

[mc172]

Can you get them to send you back your original sample and a production sample and test them yourself?

[thinkfat]

That's a tedious process. It'd be much easier if they ran both panels through their flying probes machine.

But this ordeal has only one conclusion: you cannot possibly work with this manufacturer.

[ataradov]

Why would they bother testing your other board? I don't think this is a typical request, and that's why they are probably confused.

You send gerbers, they send you the boards that match those gerbers. Why there needs to be another board involved?

And also, let's say they don't match, what do you want them to do? They can't be responsible for some third-party board.

[mc172]

That's a tedious process.

Not really, you just build one!

[peter-h]

There are various misinterpretations of gerber data possible, for example if there are tracks within planes. Yes, I know those should be done as polygon pours...

What I don't get is why they don't seem to understand the instruction. It is absolutely trivial. Just test their panel and immediately test my panel.

If they don't match then I have to abandon this order.

Populating a board is not a full test. There could be open circuits, etc.

[james_s]

I've never heard of someone sending a finished board to a PCB fab. You just have a minimal order made, then when you receive the sample order you evaluate the board yourself. It is not reasonable to expect them to set up and test a board they didn't make.

[langwadt]

if they both pass then obviously the two panels are electrically identical

if for some strange reason you don't think a pcb manufacturer can make pcbs from gerbers why do you think that they are capable of
making a test from the same gerbers that guarantee that they are "identical"?

[mc172]

Populating a board is not a full test. There could be open circuits, etc.

I obviously didn't mean just put the components on and be done with it.

You must have some sort of test procedure, fixture, etc., the outcome of which at some point must have convinced you that the boards from the original manufacturer were up to standard. So repeat that test with the new boards after populating one.

[ataradov]

There are various misinterpretations of gerber data possible, for example if there are tracks within planes.

On a gerber level it does not matter. All planes would be broken up at the time gerber is generated. Your gerbers need to be very bad to be misinterpreted with modern tools.

What I don't get is why they don't seem to understand the instruction. It is absolutely trivial. Just test their panel and immediately test my panel.

That sound like you want them to do extra work with no obvious benefit to them. Unless it is a very big order, I would not bother either.

[NorthGuy]

What you ask them to do is totally unreasonable. You want them to build boards using your gerbers, then if the result doesn't match the test board (which they have no control over) you want them to throw their work away and return you the money. They probably didn't understand what you're asking, because nobody in the right mind would ever agree to such conditions.

You probably can hire someone to test your board and see if the garbers match the board, but I would guess such job might cost more than manufacturing few PCBs. Even if your test board passes the fly test designed for the gerber, there's no guarantee that boards made with the gerber are identical to the test board. For example, some of the features of your test board might not be in the gerbers, not to mention possible SI differences.

[peter-h]

"What you ask them to do is totally unreasonable. You want them to build boards using your gerbers, then if the result doesn't match the test board (which they have no control over) you want them to throw their work away and return you the money. They probably didn't understand what you're asking, because nobody in the right mind would ever agree to such conditions."

Whole load of totally wrong assumptions there, but hey this is social media

I paid for the "tooling" plus the cost of making the one board. I merely asked them, when they BBT the panel they made, to also BBT the panel I sent them. Extra labour = seconds. If the two don't match, then there is a problem, I lose the NRE charges plus the panel cost, but detecting that is better than getting a few k boards back, populating them, and finding they don't work.

Many years ago I got a few k boards back from a UK company. BBT passed of course. Populated them all. First one tested blew up the PSU. Turned out that all the vias were plated to all the inner layers Yes this is rare nowadays, but the risk is very high. Especially when one has struggled buying the various chips at extortionate prices of today, and even then managed to get only just enough.

I have used this process for 100% board validation before, and never had a problem.

The reason I posted this was not to get sympathy it was to find out whether there is some issue with BBT which prevents two panels being tested from the same data. Of course I can't see how that could be possible. I am 99% sure these people just struggle to google translate between Chinese and English.

[ataradov]

The deal is that you send gerbers, they send you board that match those gerbers. Why would they do anything extra? It is not seconds, it is never seconds. No vendor in their right mind will do what you want. This opens up so much liability that it is not worth it.

[thm_w]

If all the vias were plated to inner layers, then the flying probe test would not pass. Either they lied or they tested based on their bad import, or there is some other information you are leaving out.
They can just lie and say "yes it matches your board" and you will be in the exact same situation here.

You should build up 1-10 boards and see if they work, before building 1000. And if you can't trust them to be making decent quality PCBs, you shouldn't be using them in the first place.

[james_s]

Bottom line is modern PCB fab houses work a certain way. Either follow the standard procedure or you're going to have problems. It doesn't really matter how difficult or easy it is to do things your way, that isn't the way they work. I'm confused by what you're trying to get them to do and why so I'm not at all surprised that they're confused as well or don't want to waste time trying to work with you. They've got piles of other customers who just send them gerbers and don't try to get them to jump through extra hoops.

[thm_w]

Bottom line is modern PCB fab houses work a certain way. Either follow the standard procedure or you're going to have problems. It doesn't really matter how difficult or easy it is to do things your way, that isn't the way they work. I'm confused by what you're trying to get them to do and why so I'm not at all surprised that they're confused as well or don't want to waste time trying to work with you. They've got piles of other customers who just send them gerbers and don't try to get them to jump through extra hoops.

He's done this in the past too. Expecting JLC to manufacture to his complex notes instead of just using industry standards:
https://www.eevblog.com/forum/eda/jlpcb-having-an-interesting-attitude-to-reading-text-on-layers/

[Mangozac]

He's done this in the past too. Expecting JLC to manufacture to his complex notes instead of just using industry standards:
https://www.eevblog.com/forum/eda/jlpcb-having-an-interesting-attitude-to-reading-text-on-layers/

Wow, that thread gives this one so much more context (and not to OP's benefit)!

[mikeselectricstuff]

The reason I posted this was not to get sympathy it was to find out whether there is some issue with BBT which prevents two panels being tested from the same data.

For example, they may add tooling holes or other panelisation details that their process needs to load into their BBT system. It is quite possible that they make your panels as multiple images on a larger panel, so testing a single panel may be a fiddle, both mechanically and in dealing with the data.

I really don't understand why you didn't just order a panel and test it. It's not like that is expensive these days.

[thinkfat]

That's a tedious process.

Not really, you just build one!

It's not about just one. You'd have to build a whole panel of them and verify them all. I guess the OP tried to avoid the time and material to spend on this and thought of a quicker, more automated way to verify that the new panel is electrically identical to his reference. But as others have pointed out, it might not be as simple for the manufacturer as he thought.

But what I pointed out is a different issue: How can you possibly work with a manufacturer that you cannot communicate with. It seems impossible for the OP to get across his request and just as impossible for the shop to explain why it won't work for them. I'm not doing finger pointing, I'm not calling anyones demands unreasonable, but clearly, this is not going to work for either of them.

[peter-h]

When the character assasination squad is finished...

"quicker, more automated way to verify that the new panel is electrically identical to his reference. "

Exactly. Every other PCB company has done this previously. It is an easy way to ensure that you can go to a new PCB mfg and then go straight to production. The cost of a screwup on say 1000 PCBs is likely to be well into 5 digits.

"But as others have pointed out, it might not be as simple for the manufacturer as he thought."

I struggle to see what is hard about
- testing the panel they made
- testing my reference panel right after that

Maybe they don't have in-house BBT? That would be weird, but not impossible. A lot of stuff is subcontracted down there.

"How can you possibly work with a manufacturer that you cannot communicate with."

That is normal in China. Comms are often borderline. The typical Chinese scenario is a firm of say 100 people, and just 1 is educated in the West and he/she is their "voice" for all foreign customers.

[free_electron]

Ever thought that their panel size may be different from yours ? They may simply have no way of mounting your panel in their tester.
Your 'panel' may be a sub-panel for them. So what you supplied is smaller than what they run.

Your data typically has stuff added : tooling , impedance and process control coupons and what not.

Once the board has passed final testing then it is removed from its manufacturing panel. They typically do not reconfigure the tester for any arbitrary shape. the tester is set up for their fab panel size.

[themadhippy]

That is normal in China. Comms are often borderline. The typical Chinese scenario is a firm of say 100 people, and just 1 is educated in the West and he/she is their "voice" for all foreign customers.

How many uk firms with a staff of 100 ,dealing with foreign customers have a member of staff fluent in mandarin or Cantonese ?

[langwadt]

When the character assasination squad is finished...

"quicker, more automated way to verify that the new panel is electrically identical to his reference. "

Exactly. Every other PCB company has done this previously. It is an easy way to ensure that you can go to a new PCB mfg and then go straight to production. The cost of a screwup on say 1000 PCBs is likely to be well into 5 digits.

"But as others have pointed out, it might not be as simple for the manufacturer as he thought."

I struggle to see what is hard about
- testing the panel they made
- testing my reference panel right after that

sure if it someone doing manufacturing in a shed and testing is just some guy with a stack of pcbs he runs through a test manually. It is much more likely a highly automated (posssbly ISO certified) assembly line so what you want required stopping production to stick in your board, and that it assuming it will even fit 

[Kean]

As mikeselectricstuff and free_electron have pointed out, and from what I've seen myself of BBT machines in some factory tour videos, the BBT machines will often work with a larger production panel rather than a customer "sub-panel" for higher efficiency.  Less manual handling.

I actually had a video sent to me of testing a board produced for me.  It had passed all BBT but had a shorts on an internal layer.  It turned out they had a bug in their gerber processing tool for combining boards into production panels.  The short was basically a line that ran right through the programming header, but only on some of the "sub-panels".  Because the short was in the production panel gerber data, it passed testing.  It was visible with back lighting, but annoyingly I'd already populated a few boards which failed before discovering it.

[Mangozac]

I really don't understand why you didn't just order a panel and test it. It's not like that is expensive these days.

Exactly this. You order a prototype from JLC for $2. You build a sample and qualify it. When you're happy and want to order the production quality you go back to JLC and click "reorder" and enter in 100/1k/whatever quantity. That way you know that the same data has been used to fab both boards.