Deengineering PCB 2

[dino40]

Good day all,
I am presently trying to find a design flaw inside a PCB. It looks to be a trace problem but I need to get inside to see. Does anyone know what chemical will allow me to delaminate the board a layer at a time in order to find this problem without destroying the traces? Any suggestions will be very helpful. Thanks all Dino40

 

[Comcokid]

There are several methods of tracing a fault internal to a PCB. Tracing down a problem requires knowledge of the insturments and it's limitations. Common insturments are: Milliohm meters (primarily to find shorts),tone-ohm meters (audio equivalent to a milliohm meter), current tracers (for shorts), and TDR (time domain reflectometers for finding shorts or opens). More exotic methods consist of: X-ray of the board, or finding shorts by applying a current-limited source to the suspect circuits and using a thermal imaging camera.

The typical method of repairing internal traces to a board other than trace cutting-jumpering around the problem is to carefully grind away over the known fault to correct the problem but this is done only on very expensive boards where trained personnel and copies of the artwork are available.

 

[VEBill]

If it is a design flaw, then you'd want to start with the design documentation (layout).

 

[dino40]

All the tests listed have been done. The problem is either trace spaceing or a portion that was not etched the whole way. I need to get into see it so we can get the board house to address the problem. I really need to get inside.
Thanks again

 

[VEBill]

If two traces are shorted internally, then you could demonstrate the fault with an ohmmeter. Then you could put the problem back to the PCB manufacturer and let them sort it out.

 

[sreid]

If the artwork says you are OK then it's a fabrication problem. At least a few percent of PCBs will have opens or shorts. The right solution for this is to have bare board bed of nails testing done.

 

[MikeHalloran]

The smart solution is to re-do the board with no traces inside.

As for your quest, a good machinist with a surface grinder can expose the inner layers. Ask him to stop a few thou short of where the copper is supposed to be. At that point, you may be able to image the copper through the remaining resin.



Mike Halloran
Pembroke Pines, FL, USA

 

[zeitghost]

I've had boards that were allegedly tested that still had faults.

Trivial things like bus tracks shorted with a whisker... and the real classic was an offcenter drilled via that shorted all 4 planes together.

Class.

So what's the theory behind bare board testing?

Make little impressions on the pads so the customer thinks that you've done something?

Or is that too cynical?

 

[kf6bbl]

zeitghost, sounds like you have a dishonest vendor, and need to find a new one.

In general, my decision as to paying for bare baord test is based on the complexity of the board and what the value is fully loaded with parts. If both are low, then the first test any component sees, both board and electronics, is in-circuit or functional test.

-Bill
CE Designer Forum

http://www.cedesigner.com

 

[IRstuff]

You say that you have a "design flaw," which means that it should be apparent on either the CAD layouts or the proofs for the board masks. Did your board layout guy run a design rule check? Is your board supplier actually able to build to your design rules?

TTFN

 

[felixc]

We do not know a lot about the type of product that you have but trying to stripe the layers to see what's going on sounds like quite a very extreme step. Unless you're trying to reverse engineer a design... Is there only one board with the problem? or all of them?

If it is just one, this is a manufacturing problem. If all the boards show the problem, if still can be a manufacturing problem. Has this design worked in the past? If not it can be a pcb design issue. Is it a short, an open, can you identify which electrical connections are faulty?

You could tell with the CAD tools, from the PCB designer side, or with the CAM tools (those used by the PC fab to generate the manufacturing masks) if there is any potential problem with a given trace or electrical node. You can evaluate all the clearances and impedances. No reason to stripe off the layers to find a bug. IF teh PC fab does not cooperate, go to another place next time.

 

[itsmoked]

I always felt that the PCB board tests were a joke and a form of sucker insurance. No test but if you come back screaming about the bad stuffed boards they promptly cough op "calming money". Rather like insurance, they get money from lots of customers for "doing nothing" and give some of it back to the hopefully few customers who get stung.

But then this is just my experience, hopefully most board houses really do tests.

Ah who are we kidding

Keith Cress
Flamin Systems, Inc.- <

http://www.flaminsystems.com>

 

[zeitghost]

It's 10 years ago now, but I could never figure out how shorts between lines on the processor bus could have been missed.

And missing 4 planes shorted together is beyond belief in my opinion.

Unless there's something about bare board testing that I'm unaware of.

 

[felixc]

There are different levels of electrical testing of PC boards. Some of them only do single-contact tests, after a visual inspection has deemed a first board to be "good". So it can guarantee that all your boards will have the same flaws... and I just don't know how it can test for shorts. You can do automated line impedence tests on test traces etched at various layers. There are higher-end tests where bed-of-nails and the pcb netlist are used. You can set ohm thresholds for on and off conditions. After a thousand 10-layer CPU boards with controlled impedence, I never had a shorted or open trace on my boards. So there's a way to make good boards and to trust the electrical tests. When you have several hundreds of dollars on a board with no way to probe once it is assembled (BGAs), you can not not have the board electrically tested. On high-density boards the internal reject rate is quite high, a serious PC fab just can not ship untested boards. There are crooks anywhere.

To tell the truth I did find two boards with a shorted trace. They happened to be shorts that have been detected by the test machine, but that have not been properly reworked... I saw the rework and under the varnish there were the shorts.

 

[MikeHalloran]

Design rule checks don't catch stuff like a perfect image of a hair, in copper, crossing several tracks.

It was a short hair, of course. ;-)



Mike Halloran
Pembroke Pines, FL, USA

 

[IRstuff]

The OP called it a "design flaw."

TTFN

 

[MacGyverS2000]

Mike, was it curly? I imagine someone getting snatched by the short and curlies after letting that one through []


Dan
Owner

http://www.Hi-TecDesigns.com

 

[MikeHalloran]

Sorry, I guess the intended pun was too obtuse. SHORT hair?

To answer your question, the hair was straight, probably from the fur of rattus rattus, a stable population of which lived in our pcb production facility.

The hair's image was actually present in the photomask, so it appeared in several hundred circuit boards. Because they were only two-sided boards, it was possible to rework them while replacements were being made. They were memory boards fully covered with big SRAMs, and of course the flaw was under a socket.

It was embarrassing that the boards had passed a board- level test, and a functional test after assembly, and were being used in production. We just couldn't get them to save and restore data in a prototype in the R&D lab, and we had a stubborn ex-Navy tech who knew how to use an ohmmeter.




Mike Halloran
Pembroke Pines, FL, USA

 

[Comcokid]

As the saying goes S*^& Happens - especially with PC boards.

At one time I worked for a company that had one of the early low cost personal computers. The computers were produced at several locations in Europe. I setup test equipment for the North America customer service repair. One computer that came back never had the board etched at all. Just drilled, plated, masked, and silkscreened - and yet it got completely through assembly and test at a production facility, shipped to a customer, and returned!

 

[itsmoked]

Oh man you could have some real fun bustin chops with that bugger!! Think of how many hands it went thru each and every one of which should have recognized a problem.

I once got a linear power supply. Triple output with 20A on the 5V. Installed it into our product. Didn't work! I pulled it out opened it up, the controller IC a Usomething or ruther (Unitrode) wasn't even present. I called the supplier up and said, "I have one of your supplies here. It was DOA." The guy interrupts me and sez,"that cannot happen since we test each one before shipping". He had a bad day following that.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com