Silencing PWS-741P-1R power supply
Disclaimer: the following article explains an unauthorized modification of an electrical appliance which runs at high voltage and contains a lot of chemical stuff which doesn't play nicely at high temperatures. The manipulation explained in the article voids the warranty and modifies the product in a way which wasn't tested by the manufacturer. Unless you have enough experience, DO NOT ATTEMPT TO DO THE SAME MANIPULATION WITH A SAME OR A SIMILAR DEVICE. Not only you'll void the warranty, but you would be at risk of permanently damaging the product or compromising your own safety.
I have a bunch of SuperMicro 2U chassis 825TQ that I use for the tasks which don't require a lot of CPU power.
The original chassis makes a lot of noise. When I say a lot, I mean it. You can't possibly stay in the room when this thing is on. I originally replaced the three 80×80 fans, which made the chassis way more quiet that it used to be. Once I did it, the next loudest source of noise became the small 80×80×28 fans of the redundant power supply. So it was time to replace those ones as well.
- The machine is heavily underused. The usual power consumption averages 40 W, while the power supply is rated 740 W.
- Its fan stays at its lower speed all the time.
- The power supply has an over-temperature protection.
- The machine is used in a ventilated environment, which only occasionally reaches 35 °C.
It is for those reasons that I took the risk to replace the fan: if some of the factors were different, I believe that taking the risk of tampering with a power supply wouldn't be a good idea. It should be noted that the original fan is rated at 40 m³/h, while the one I will be using is rated 9,4 m³/h, which is a substantial difference.
Choosing the fan
Unlike with the replacement fan for an UPS, I didn't have much choice here in terms of the fans: all fans with high static pressure are noisy, so for quiet options, Noctua NF-A4x20 PWM seems to be the only option. I took the 4-wire (PWM) variant, because the original AVC DB04028B12U-128 fan is a PWM fan.
Figure 1 The video showing the noise made by the two fans.
Wiring the fan
Figure 2 The original AVC DB04028B12U-128 fan.
Figure 3 The fan connector.
The original fan uses a custom connector, so some soldering was necessary. By the way, the colors of the wires were all wrong: usually, PWM fans use black-yellow-green-blue wiring (A.), or black-red-yellow-blue variant (B.), therefore the blue wire is always the control one; this fan, however, was using black-red-blue-yellow coloring (C.), that is, the blue wire was a tachometer signal one.
Figure 4 The standard color codes when wiring the fans.
I didn't know that, so I soldered the wrong wires together, and was welcomed with an amber LED light of the power supply, indicating that a fault prevents it from starting. Found the issue; soldered the correct ones, but this time, the fan was starting for a fraction of a second, then stopping, and the power supply LED would remain amber.
Finally, I figured out that the problem was the difference in the speeds of the two fans. The original one is rated 13.000 RPM, while Noctua's is 5.000 RPM. Given that the power supply is underused, it ensures the fan turns at its minimum speed: the original fan did exactly that, while the new one just stopped, which alerted the power supply through the tachometer that something went wrong.
The solution was to disconnect the PWM cable. This means that the Noctua fan is turning now at its maximum speed, but this shouldn't be a problem: at maximum speed, it is still much quieter than the original fan at its minimum.
Figure 5 The matching of custom colors with the standard ones.
Figure 6 Noctua NA-EC2 extension cable cut to be soldered with the original cable.
Figure 7 NA-EC2 soldered to the original cable.
After I did the change for one of the two PSUs, I let it run for a few days, just to make sure it works correctly. The PSU didn't complain, and the exhaust air temperature (measured with a temperature probe positioned at the back of the server) didn't show anything suspicious. I then replaced the fan in the second PSU.
Figure 8 Connecting the fan to NA-EC2.
Figure 9 Noctua NF-A4x20 PWM.
Figure 10 The fan, with custom screws.
What's nice is that this 2U chassis is now one of the quietest devices out there. I can't hear it when I turn it on, given the ambient noise produced by the primary server (with its seven fans), the UPS devices, and the switch.