Cured my GPS2303 PSU rattle.

[Gyro]

Hi,

I finally got fed up with listening to the rattling noise from my GW Instek GPS2303. The rattle showed up after it had been running for a while at low current output, and at low fan speeds.

Initially I thought that the rattle must be from the fan bearings and took it apart in preparation for finding a suitable quiet PC fan, then I realised that it is a 24V fan! The bearing actually felt fine, it's a double ball type with no slop, the spec. indicates that it is on a par with the quieter ones anyway.

Next I investigated the drive waveform - a 555 generated rough PWM with fast rise time and 65Hz repetition.  I could feel the vibration in the fan rotor. The only thing filtering the fan supply after the switching transistor is a little 10uF electrolytic (C56) which, as far as I could trace, is straight on the switching transistor output (certainly straight across the fan), so not having a happy life. At low fan speed the PWM beats with the fan comutation too, causing more repetitive rattle.  The 'rattle' is actually structural resonance in the fan, coupled into the case.

I didn't want to take the power board out, so I chopped into the fan lead to add an RC filter. After some experimentation I've settled on 10R 1W and 220uF (100uF works nearly as well), and fixed  them to the rear panel. The result is that the PSU is now silent apart from transformer hum, the fan cuts in smoothly and quietly. I was initially worried that the sharp PWM pulses were needed to start the fan but it actually seems to cut in a bit earlier, and is running slowly by the time the heatsinks hit 40'C. The voltage drop across the 10R resistor (1V at full speed) doesn't affect full speed operation, in fact it seems to run better with the smoother supply. I haven't managed to top 55'C on the heatsinks with both outputs shorted and set to 3A. I've also snipped out C56 to reduce peak current through the switching transistor (Q51) which gets hot enough anyway.

There is obviously fan noise and 'whoosh' at higher speeds and some resonance into the metal structure (I may investigate resilient mounts), but at normal loads it is now virtually silent. 

I hope this helps someone. Photo attached.

[kendalll]

Nice detective work thanks for posting!

[jitter]

Good work, but I would like to make a few remarks.

I was initially worried that the sharp PWM pulses were needed to start the fan but it actually seems to cut in a bit earlier, and is running slowly by the time the heatsinks hit 40'C.

Good because some fans get harder to get going as they age. That's why PC fans are often started at full power and then drop down to a lower speed.

I haven't managed to top 55'C on the heatsinks with both outputs shorted and set to 3A.

Shorting the output isn't going to create as much heat dissipation as one might think. Ohm's Laws explain why.
You can find a sweet spot for torturing a PSU by slowly increasing the output voltage to the point where it has just switched to the highest transformer tap. Then you will want to increase a load to the max output current of the PSU with it right at the point where it will go from CV into CC mode. This is what will heat up the PSU the most as the voltage the linear regulator has to drop is highest, and it is that drop multiplied by the current that creates the heat.

[Gyro]

Thanks.

Good because some fans get harder to get going as they age. That's why PC fans are often started at full power and then drop down to a lower speed.

Agreed. The 'rattling' (rather akin to 'Diesel knock') is caused by mechanical resonance in the fan / case from the very steep rising edge. The fan still gets its PWM at any meaningful mark-space, the time constant of the new RC is only about 5ms, It just takes off the sharp leading edges (not useful and probably more damaging to the fan's ball bearings and its electronics), something that the original bare 10uF cap was put there to try (and fail) to address.

Shorting the output isn't going to create as much heat dissipation as one might think. Ohm's Laws explain why.
You can find a sweet spot for torturing a PSU by slowly increasing the output voltage to the point where it has just switched to the highest transformer tap. Then you will want to increase a load to the max output current of the PSU with it right at the point where it will go from CV into CC mode. This is what will heat up the PSU the most as the voltage the linear regulator has to drop is highest, and it is that drop multiplied by the current that creates the heat.

True, it's not stressing the transformer, caps etc as much. It comes close enough for this test as the output transistors are still dropping the entire voltage of the lowest tap at full current (the taps are fairly evenly spaced) and the fan control circuit is only monitoring the output transistor heatsinks. Close enough in terms of monitoring the fan control PWM anyway. At full speed, shorting to 10R resistor doesn't result in any significant speed difference (not to be tried at low pwm, by the way, as the surges could possibly take out the switching transistor).

[Shock]

For noise reduction you can use heat resistant silicone tubing as washers or split it open and run around the fan edge.

[Gyro]

I received my ebay silicone rubber screw replacement fan mounts today. Easily fitted once the rear panel was unscrewed and there's just enough clearance between the fan and the heatsinks. A further improvement as expected, no more resonant coupling into the chassis. Even at maximum fan speed it now sounds expensive.

I wish I'd tackled this before, it's been annoying me for years!