If you have ever built a single-cell lithium battery protection circuit, you have used this pair: a protection IC watching the cell, and a dual N-channel MOSFET doing the actual disconnecting. The MOSFET is almost always an FS8205A. It's in phone banks, vape pens, e-bike packs, and a large fraction of the hobby projects on this site.
I put one in my pen. Then I nearly told a factory their correct part was defective.
The substitution request
I'm running 100 units of a smart pen, ESP32, ST7789 color display, custom four-layer board, 120 × 19 mm. Partway through, my contract manufacturer asked to substitute the battery protection MOSFET for an equivalent from a different brand, and sent the datasheet.
This is routine. Most people approve it without reading.
I read it, and it looked wrong:
| Pin | What my board was routed for | Their part |
|---|---|---|
| 1 | S1 | S1 |
| 2 | G1 | D |
| 3 | S2 | S2 |
| 4 | G2 | G2 |
| 5 | D | D |
| 6 | D | G1 |
Pins 2 and 6 swapped. Gate where drain should be. Mounting that on my board would connect a MOSFET gate straight to the drain node, and the protection circuit would do nothing at all.
I wrote the rejection message. Specs are fine, pinout doesn't match, please don't use this part.
The check I almost didn't do
Before sending it I went looking for the genuine article's datasheet, mostly to have a stronger citation. Fortune Semiconductor makes the original FS8205A.
Fortune only makes it in TSSOP-8. There is no Fortune SOT-23-6 FS8205A. There never was one.
So I checked a third: FuxinSemi's SOT-23-6. Pin 2 = drain, pin 6 = G1, identical to the part my manufacturer had proposed.
That's when it landed. Every SOT-23-6 device sold as an "FS8205A" is a second source, and they all agree with each other. The pin arrangement I had routed for is the one from the TSSOP-8 original. I had put an eight-pin part's pin assignment onto a six-pin outline.
The manufacturer was right. My board was wrong, and it had been wrong through every review I had done on it.
What would have happened
The board would have been fabricated and populated. It would have passed visual inspection, the part sits flat, the solder joints look correct, nothing is bridged. It would have passed bring-up, because the pen charges and runs off the cell regardless.
It would have failed the first time the protection circuit actually had to do its job: over-discharge, or a short across the pack. Which is the one moment a lithium cell needs something standing between it and the outside world.
A hundred units, all looking perfect.
The fix
Only the copper around U6 changed. Same part, same position, same rotation, the BOM and the placement file were untouched. DRC came back to the same baseline as the released package, five accepted violations and zero unconnected. New fabrication files went out the next day, and nothing had been assembled yet.
The manufacturer had paused the run waiting on my answer, which is the only reason there was still time.
What I'd actually change
The footprint and the part number are two independent assumptions, and I had only checked one. My footprint came from a library, my part number came from a reference design, and I never verified that the specific part I was going to buy came in the package I had drawn. Those are separate facts and they can disagree silently.
So: pull the package drawing from the datasheet of the exact part you will actually purchase, and check it against your footprint pin by pin. Not the part the library entry is named after. Not the part in the reference schematic. The one on the purchase order.
And if a vendor pushes back on something you're sure about, check anyway. The whole reason I found this is that someone asked me a question I thought I already knew the answer to. I was one send button away from being confidently, expensively wrong, and the only thing that saved me was going to look for better evidence to prove a point.
I'm seventeen and this is my first production run. I expect there are more of these.
Logan Holby is the founder of Nexus Pen LLC, building an AI-powered smart pen. Board files available on request.