Bought 3 of these fans. One of them behaved erratically since day one. After 20 seconds (exactly...) of normal operation it throttled back to very low speed and remained there. Popped the case open and found that there is a small "main" PCB with a PIC and some additional electronics
, forming a very sophisticated speed control system. In this particular case the surface-mount resistor R19 was dead (non-conducting) - it had a manufacturing fault so it might be a bad batch out there. If you have the exact same symptom replacing this resistor (5.1k SMD 0603) might fix it.
(To sum it up for the technically inclined: motor
speed signal is fed back via measuring voltage drop over a 0.5 ohm resistor in series with the motor
. This signal is formed via a network based on the quad op-amp LM224, and fed back to the PIC. If this signal is "out of bounds" the controller assumes that the fan has stalled/stuck and throttles back the 20kHz PWM signal used for speed control to 30%. If the feedback normalizes, speed will also return to normal. This is essentially an overload protection that kicks in after 20s. Regular speed settings are at 60%-80%-100% PWM ratio respectively. R19 is part of the signal forming network, it being faulty drives signal outside of normal limits, engaging the safety
Hope this might help someone