2
\$\begingroup\$

I had 5 two-layer boards fabricated by a popular assembly house in China and I'm curious what folks thoughts are based on what I received. Here is a photo of the 5 boards which has a red square around all 5. enter image description here

Essentially each board has a MCP23017 GPIO extender and the base control board at the bottom controls these MCP's via the I2C bus. The firmware I'm running just turns on all the LEDs as part of my test.

Interestingly 2 of the 5 boards has one or more issues with an LED not lighting up when it should. After investigation of these two faulty boards I have confirmed that the LED's are not burned out and light up properly when ground is applied to them. The LED's are controlled via a MOSFET and the GPIO extender actually controls the gate of the MOSFET. I'm using a SOT-23 2N7000 MOSFET for each LED.

I've determined that in the cases where the LED does not light up, there is 3.3V at the proper pin on the MCP23017 but there is no voltage on the gate of the corresponding MOSFET.

The boards have been very carefully handled (although I hope to use these in a high school classroom electronics course so I'm hoping they are not super fragile).

Here is a video of proper operation for 3 of the 5 boards.

https://www.dropbox.com/scl/fi/6ld5bgfg8yf9t606u19fd/A001_12050806_C003.mov? rlkey=hbtr2kooy8hqdm4af0vmsefei&dl=0

Here is a video showing the top boards being replace by the other 2 faulty ones.

https://www.dropbox.com/scl/fi/7iig1s7rpj40yh2zgroi8/A001_12050805_C002.mov?rlkey=3ucqg8cdrtrksmiyr4vhw3ybq&dl=0

I'm curious what folks think about this? Is this a production issue that I should hit up the manufacturer about or might something else be going on?

I would think this failure rate of SOT-23 to LED (80 pairs across all 5 boards) with 5 failures out of 80 that rate seems way too high to me.

\$\endgroup\$
15
  • 1
    \$\begingroup\$ Yes, in each failed case there is no continuity between the output of the MCP and the SOT-23 \$\endgroup\$ Dec 5, 2023 at 14:45
  • 2
    \$\begingroup\$ Also, i guess the LEDs does not draw that much current? The MOSFET is vaguely defined for Vgs below 4.5V. \$\endgroup\$
    – Linkyyy
    Dec 5, 2023 at 14:58
  • 1
    \$\begingroup\$ @Linkyyy They are a little bright, I was thinking on my next run I will up the resistor value. I also wasn't aware that the 2n7000 may not be the best choice for a 3.3V system. Is there another MOSFET better suited for 3.3V? \$\endgroup\$ Dec 5, 2023 at 15:00
  • 1
    \$\begingroup\$ Why are there no caps on the boards? At a bare minimum there should be a 100nF next to that IC. \$\endgroup\$
    – Lundin
    Dec 5, 2023 at 16:03
  • 1
    \$\begingroup\$ @KevinMcQuown Aah, hehe that explains a lot :) Dang hard to see those little buggers IRL, let alone on a pic :) \$\endgroup\$
    – Lundin
    Dec 5, 2023 at 16:20

1 Answer 1

4
\$\begingroup\$

My reaction would be contingent on the type of failure encountered.

As you mentioned, these failure rates might be acceptable in hobby projects, but they are certainly not suitable for professional applications. It's essential to remember: you get what you pay for.

(1) PCB Issue:

If the trace between MCP and GATE (either the trace or a via) is broken, I would never order a single PCB from them again. A 100% flying-probe test is de facto standard, and for good reason! A two-layer PCB can perhaps be manually fixed, but anything with four layers runs a high risk of ending up in the bin — money not well spent, not to mention the effort involved in tracing down the problem.

(2) Soldering Issue:

If the soldering on either the MCP or the FET is the cause of the problem (due to issues like bad board finish, old solder paste, or an incorrect profile for the paste), I would also reconsider ordering from them again. While a faulty solder joint can occur, especially with BGA or fine-pitch SMD components, it requires special equipment for rework. The primary motivation for ordering, in the first place, is to avoid the investment in these tools.

(3) Layout:

If you used uncommon/self-made footprints for your parts, the soldering problem might be caused by this. Double-check your footprints against the recommended dimensions in the datasheets. If there is a significant difference, please post images of both so we can verify. However, this seems unlikely.

(4) Component Issue:

If they used the cheapest offer on the Shenzhen market today as components—which can happen—you may end up with faulty components. Personally, I would either supply the parts myself, especially for more complex or expensive ones, after purchasing them from a reputable distributor. However, this approach significantly increases lead time.

(5) Schematic Issue:

It could very well be that the FET you are using is not suitable for a 3.3V system. However, if there is no continuity, as you stated, this does not appear to be the main problem here. For Rev2, I would consider using a different FET with Vgs < VOut,min MCP x 0.8.

\$\endgroup\$
5
  • \$\begingroup\$ Soldering looks ok visually far as I can tell, but it would be nice to see the component side too. \$\endgroup\$
    – Lundin
    Dec 5, 2023 at 16:14
  • 1
    \$\begingroup\$ I suppose there is an outside chance that a via opened up after flying probe, in transport, storage, or especially during soldering. Which could potentially be blamed on the layout (solid planes trap moisture leading to bubbles and delamination; perforated planes can be used when this is a concern), but it's rare that this is even a problem, and it's possible such a failure could be traced back to a faulty plating process, or lack of bake-out. \$\endgroup\$ Dec 5, 2023 at 16:17
  • \$\begingroup\$ @TimWilliams, I agree with you when it comes to transport and/or storage. However, I expect a key-turn solution to have processes refined to a degree where they don't damage their own boards during soldering. Regarding the solid planes trapping moisture: Do you happen to have a source on that? I could not find an appnote/abstract/... via Google, but I'd like to learn more about it. \$\endgroup\$ Dec 5, 2023 at 16:32
  • 1
    \$\begingroup\$ Ah, blistering or delamination are better keywords. Here's one result, worldsway.com/… though I'm not too keen on its origin (yet another manufacturer promoting industry terms), had to scroll down a ways to find it. Most of the results seem concerned with fab defects (presumably, moisture content in the prepreg and core themselves). I recall seeing this in one of the IPC-600s as well, but that might be old data by now. I'm not in manufacturing, so I don't have any idea how common this is. \$\endgroup\$ Dec 5, 2023 at 16:55
  • \$\begingroup\$ @TimWilliams Thank you fine Sir! \$\endgroup\$ Dec 5, 2023 at 18:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.