CISCO SG250-26P Noctua Fan replacement problems - Fans spin up, then stop after a few seconds

[XpDesk]

Hey there dear STH-Community,
a few months ago I was on the look out for a 19" rack mountable manage switch with PoE+ and came across a dirt cheap CISCO SG250-26P which I scored for 100€. It was (and still is) in perfect condition, but upon powering it up, what I've not taken into consideration were the ultra loud jet-engine type fans which are not near acceptable for my environment (though I also know that this switch wasn't made for homelabs, either).

So I put the switch aside and researched on how to replace the fans and quickly came to the conclusion that it's not easy as swapping them out with Noctuas.
What I've came to learn is that the 12V and ground pins are swapped, as well as the tacho pin not being compatible with the signal of the PC's Noctua PWM wire.
I followed the few guides, YouTube videos and forum/ reddit posts and memorized the pinouts, and tried to replace the fans.
I bought two of each the PWM (4-pin) and the FLX (3-pin) 40mm Noctua fans, because I couldn't find a definite answer on which to use (I ended up using both, without luck, more on that shortly).
Using the included Noctua extension cables and Scotchlok adapters I spliced the wires to match the pinout on the Switch's mainboard. I had different configurations:
First I tried the FLX fans. I wired the 12V (red) pin on Switch to 12V Noctua wire and Ground pin (black) to ground Noctua wire, and left the blue wire (tacho?) out at first.
I know that there were complications with Fan errors in the Web panel or PoE disabled, but people seemed to find a way around that by grounding the blue tacho wire to ground (or something like that?).
I got the Noctua fans to spin up, but they only spun for like 5 seconds, then stopped forever. The Switch seemed to boot up however, since all lights on the ports lighted up.
I also tried to use one Noctua fan with my modified wire-switching cable as well as an original fan to see if maybe the firmware expects both fans to be present during startup, but it didn't change anything.
I also tried to wire the blue wire to ground shorting it on the mobo, but it also didn't make a difference.
Using the exact same configuration, but with the PWM version of the Noctua fans, also with and without shorting tacho to ground, but the fans only spun up again for 5 seconds (or similar), then stopped, so I tuned off the unit.
I should add that I measured the fan header pins for 12V and ground on the Switch's with the Noctua fans attached with a multimeter, and they only seem to deliver 6-7 volts.

It seems like some people have got it working, though I only seem to find guides for SG200 or SG300 or higher.

I don't care about the warranty since I already bought it used, and I don't expect a high security risk regarding overheating because I'm not planning to nearly use the 195W PoE budget of the switch (max. powering 3 to 5 Pis over PoE, if I even can get it working).

Do you have any experience, suggestions or recommendations?
Thank you in advance and stay safe.

 

[Tusai]

Dead post but I'll offer what I know. Long and short is I'm running into a very similar issue. Most of my conclusions come from my own experience, not thorough research. Take what I say with a grain of salt.

From what I've read, most of these switches are 3 pin fans with a weird pin-out (as you mentioned). Along with this, the internal fan controller must measure the current draw of these fans to some extent. I think it is expecting a current draw of around 0.24 A probably with a minimum around 0.1 A. the Noctuas under WORST case can only draw up to 0.05 A. Now I see two ways to go about this, the exact model number escapes me at the moment but some of the Delta 40 mm fans will work and are quieter. In my case, I have a couple of two pin fans that I was able to get running in test-bench conditions, and I may order more of them for real use. You can also probably find similar 3 pins from Delta or Sunon or any of the others that will be quieter. Along with this, you could probably add a parallel resistor to the positive for the Noctua fans.

Assuming 12V 0.05A draw & 0.1 required:
r=v/i
r=12/0.1
r=120 ohm

Two things:
Use a 100 ohm as its just plain easier.
The resistor may heat way up, so getting a big one with at least some mass to it may be necessary to keep it from burning.

Finally, the other option is to tap into the 12V power wire directly. I may just do this, as I already have the Noctuas and it is just easy to do. The switch may get mad over it but honestly, I don't care. There is probably some Cisco CLI shenanigans that need to go down to fix config issues etc.
Good luck!

 

[Tusai]

Dead post but I'll offer what I know. Long and short is I'm running into a very similar issue. Most of my conclusions come from my own experience, not thorough research. Take what I say with a grain of salt.

From what I've read, most of these switches are 3 pin fans with a weird pin-out (as you mentioned). Along with this, the internal fan controller must measure the current draw of these fans to some extent. I think it is expecting a current draw of around 0.24 A probably with a minimum around 0.1 A. the Noctuas under WORST case can only draw up to 0.05 A. Now I see two ways to go about this, the exact model number escapes me at the moment but some of the Delta 40 mm fans will work and are quieter. In my case, I have a couple of two pin fans that I was able to get running in test-bench conditions, and I may order more of them for real use. You can also probably find similar 3 pins from Delta or Sunon or any of the others that will be quieter. Along with this, you could probably add a parallel resistor to the positive for the Noctua fans.

Assuming 12V 0.05A draw & 0.1 required:
r=v/i
r=12/0.1
r=120 ohm

Two things:
Use a 100 ohm as its just plain easier.
The resistor may heat way up, so getting a big one with at least some mass to it may be necessary to keep it from burning.

Finally, the other option is to tap into the 12V power wire directly. I may just do this, as I already have the Noctuas and it is just easy to do. The switch may get mad over it but honestly, I don't care. There is probably some Cisco CLI shenanigans that need to go down to fix config issues etc.
Good luck!

Well! News! I spent over 3 hours with an oscilloscope, Noctua 40mm and my SG350 and here are my conclusions.

1. - The pin-out is very strange. (as expected)
left to right with tabs facing up:
black (-*), blue (FG PWM), and red (+*)

2. - PWM on the power wires?!?!
Power is fed between black and red, but I see a PWM signal when I probe these two wires. I'm pretty sure it is in reference to the positive wire, and the PWM signal is actually fed on the negative. I know for a fact it is PWM and not voltage controls (due to waveform) and I tried running the existing fan without the blue wire. Neither the fan nor the switch even noticed. Speed control continued to work on the fan, and the switch didn't show any errors. Why Cisco or Delta are doing this is beyond me. This is probably why the Noctuas won't even spin when plugged in, they probably need constant power on the power leads to run.

3. - Reference to positive?!?!
It is normal for tachometer signals to be reference to positive, both Noctua and some Deltas do it. However in this fan, the PWM is also referenced to positive? I was able to get a Noctua to run by pulling parasitic power directly off the 12V PSU. It would run, but when I added the PWM wire to either the negative or the FG PWM wire on the switch, the fan actually did the INVERSE of what it was supposed to. When the switch requested a ramp up, the fan would ramp down. I think this is due to the wierd reference behavior, as the fan is seeing the off section of PWM as on due to the polarity swap.

4. - It VERY MAD.
I was unable to get the switch to NOT be mad at me for removing a fan, I tried a 560 ohm resistor (I only had quarter watts and didn't want to make magic smoke) and it was mad no matter what. The switch also didn't want to provide PWM or power unless there was a resistor or device present. Strangely enough it seems like the Noctua when all three wires are plugged in are no voltages period.

I'm scratching my head over this and truthfully, I don't care that much. I'm in love with Unifi for my home network, I don't need Cisco's granularity. I just wanted to learn it, and I'll still work with that switch just for giggles, but it won't run 24/7. Way to loud. It is possible that more work with a two watt resistor and maybe an Arduino or Attiny85 to inerpret and re-create a real deal PWM signal and emulate the fan for the switch could make it work. Just ideas for the next person to stumble here.

TLDR: Get a fanless, or tap in to 12V and let the switch be mad with no speed control.
Also I don't think all Cisco switches are like this, just be ready if yours is.