Ubiquiti UniFi Switch US-8-150W missing 3.3V on mainboard

Question

Unifi Dream Machine

Ubiquiti's most popular all in one UniFi OS console including 4-port gigabit switch, AP and security gateway.

1 个问题查看全部

Flemming Steffensen @fsteff

声誉积分: 1

3

发帖于: 2022年12月27日

更多选项

Ubiquiti UniFi Switch US-8-150W missing 3.3V on mainboard

My Ubiquiti UniFi Switch US-8-150W (USW8P) suddenly stopped working and became 100% unresponsive. It's just a month out of warranty, but nothing to do from the official site.

Dreading the common PSU problem, I have verified that the PSU is outputting 53V DC and that I can easily draw current from it.

On the mainboard, I have verified that the 24V POE voltage is available on the far end of the board, for all 8 ethernet ports. But on J9, I notice the 3V3 isn't available, which I guess is the main problem. There is no apparent damage to be seen anywhere.

I've been searching the net for schematics and clues as to what might be wrong but found nothing.

Can anyone here help?

-------------------

Update:

Thank you Alisha C @flannelist for your nice answer, where you suggest I update the question with some pictures.

The USW8P board looks like this:

On the lower left 53V DC is coming from a tiny but capable 300W PSU.

Lower mid is a connector marked "3V3 R T G" which I presume is for debugging, but people also use it to run a small fan. 3V3 is supply output and G is ground. The function of R and T is unknown.

The 3V3 on this connector is missing.

On the right-hand side of the board are the 24V POE power circuit, where both 53V (input) and 24V (output) are present as expected.

I started tracing backward from the missing 3V3, identifying components on the way, and ended up with this:

The chip closest to the connector with missing 3V3 is an AZO1212, EZBuck 3A Simple Buck Regulator. Its input appears to be cascaded from another chip, the LM5071 Power Over Ethernet PD Controller with Auxiliary Power Interface, where the circuit follows the single Isolated Output application note to a reasonable degree, except for no actual POE supply, only an auxiliary supply):

The big EP13 transformer looks very much like the PA1367NL, and appears to be functioning.

Studying the PCB very closely I discovered a tiny amount of black residue on one side of C2571 and was able to determine that it was short-circuited (zero ohms, also after desoldering) - in the application note this corresponds to one of C7, C8, C9, or C10. With the shorted capacitor removed, the resistance across the remaining capacitors became 1 ohm, which also appears to be way too low for anything healthy.

With power connected to the board again, I measured absolutely no change anywhere.

I'm pretty sure the error exists somewhere in the circuit around the LM5071, but I have no idea how to locate it further. Switchers are pure magic for me.

Any ideas for how I can continue will be highly appreciated.

Also: Happy New Year everyone. :-D

回答此问题我也有这个问题

这是一个好问题吗？

是否

得分 0

添加一条评论

1个答案

Alisha C @flannelist 声誉积分: 32.6千 24 35 32 · Answer 1 · 2022-12-27T20:35:28-07:00

Schematics would certainly help. But are unlikely to be available. A missing voltage might be missing for a few reason.

If there is voltage, it's just low, there may be a short circuit (to ground) on this line. You can check for low resistance to ground on this line to see if a short is possible.

If there is no voltage it's likely because its source is faulty. Or its source is not being told to generate it. In which case knowing where that voltage comes from would be helpful. Even if you don't have schematics you can look at the components you may be able to identify to try and locate it.

Overall your best tool at first, is going to be your eyes. Look for any board components that seem bad. Upload pictures and attach them to your original question if you'd like extra sets of eyes on your issue.

UPDATE 12/31/2022 -

The thing I have learned in my time troubleshooting circuits is that although it's cool and rewarding to understand how they work, it's more important to understand how they fail. These are related, but not the same. Also your legwork on finding the datasheets for these is 👌 chef's kiss. All the applause. Sorry for the wall of text that's about to ensue.

Regarding the missing 3.3 pin. LX outputs are part of a buck circuit and would be the primary output of this chip. It probably connects to one of the nearby inductors, which it would use to convert to a lower voltage.

If this isn't outputting, There could be three things at fault: it's not getting power in, it's not getting power on the enable line, or it's not getting proper feedback. I'll hand these one by one.

Power input (VIN) which is pin 7, as you also noted you traced back to the LM5051. This is like plugging a computer into the wall outlet. Is voltage present on this line? If no, then there's your problem. If voltage is low, also a problem. I suspect this is your issue, but I'll get to that later.
Pin 6 is for enable. If VIN is plugging your computer into power, this is pressing the power switch. The datasheet mentions this can be linked to VIN if a separate enable is not required. So. Are pins 6 and 7 continuous? Check for voltage here as well. Enable is set for "voltage high" meaning it's off when low or no voltage is detected.
Feedback is essentially a computer POSTing. The output loops through a voltage dividing resistor back into the chip so it can check that it's operating correctly. This doesn't apply to you since you're not getting any output, but including it for anyone else who might stumble into this for a similar issue.

Back to your issue. If you've traced the input for the AOZ1212 back to the diode near the failed cap, that line still seems like it's likely to be the problem. Short to ground means low or no voltage where it's needed. I would check the line's resistance to ground.

If you're checking resistance across the caps while they're on the board, and the line is short to ground, then it's going to be low resistance since both sides of the cap are effectively ground. The tricky bit is going to be finding the issue causing the short.

Hopefully this helps a little.

Thank you, Alisha. I have updated my question as you suggest. — Flemming Steffensen
@fsteff I updated my answer so I wouldn't be limited by character count. Keep us posted! — Alisha C
Thank you for your new-years update on this! It's highly appreciated!

I'm very confident that the problem lies either with the feedback circuit of the LM5051 or on the line between the LM5051 and the AOZ1212 (and probably other circuits powered by this voltage).

The cap I discovered that was shorted is zero ohms (also in-circuit), but once it's been removed the resistance to ground is exactly 1 ohm, no matter where I measure, where I would expect to measure in the 10-40k ohm range (judging by the feedback voltage divider).

I guess the only way to really locate the problem is to start pulling capacitors off the board. :-(

I'll keep you posted. — Flemming Steffensen
@fsteff Not a problem. I tend to get super invested in stuff like this too :)

I would also expect a much higher resistance on that line. So I'm guessing you've still got a short circuit somewhere. It's just a matter of finding where.

Typical short detection methods would dictate looking for heat. The easiest way to do this is to use a thermal camera, but obviously not everyone has one of those kicking around. Otherwise you can try and use isoproyl alcohol to see if a particular spot on the board is generating a lot of heat (it will boil, or evaporate off super quickly). Or freeze spray, if you want to go down that route, just again, not something everyone has at the ready.

Shorts tend to generate a lot of heat as all the available current passes through the shorted component to ground.

Normally I would say you could feel around, but I wouldn't suggest that on a board with bigger caps. — Alisha C
@flannelist I don't think it's possible to detect this error looking for heat, as I suspect the switcher isn't operating. Using a multimeter to measure the voltage across the capacitors (where one of them was shorted), I measured dead zero Volt.

As you suspected, I don't have a thermal camera lying around, but I do have a laser temperature meter with a decent resolution, that I could probe around with, but it turns out, nothing differs temperature-wise in that area. As expected there is a slight warmup in the POE supply area, so something is working.

I also haven't tried probing around with an oscilloscope on the primary side of the LM5051, but I suspect I should see some switching attempts on the OUT pin and on the transformer. (I better read the datasheet first.. :-) ) The 53V input voltage is rock steady, so it can't be anything that draws a lot of current.

Hope to get some time away from the kids tomorrow... then I'll dive in deeper. :-D — Flemming Steffensen

Unifi Dream Machine

Ubiquiti UniFi Switch US-8-150W missing 3.3V on mainboard

1个答案

添加你的答案