Removing electrical noise introduced by fan

[tony359]

Hi all,

I am working on a sound processor, it tends to overheat so I fitted a fan inside. The fan is connected to the +15V of the PSU via a DC-DC converter based on the LM2596 to reduce the voltage to around 10.5V.

Unfortunately the analogue circuitry is picking up the noise from the fan in two ways:
1. The fan is close to the analogue section, the motor itself gets into the outputs so I'll have to move it somewhere else
2. Even with the fan running outside of the case, there is still some noise which is clearly going back to the processor via power line. It disappears if I power the fan with an external PSU.

I'll take care of point 1, but what about point 2? Any suggestions on how I can shield the noise coming from the fan?

Attached a spectrum analysis of the output, you can see the 271Hz noise peak - and the scope is showing some noise with a 271Hz period on the DC line going to the fan.

Is there a simple solution to this? Thanks!

Edit: I tried replacing the LM2596 with a resistor but I still see noise going back. I've also tried with a linear regulator and it's the same. I need to find a way to stop the noise FROM the fan to go back via power line.

[tony359]

I've been experimenting with low pass filters. The best outcome - complete removal of noise - was with using the coil of a primary transformer and a 1000uF capacitor. I don't have inductors handy so that was a test. However I've also achieved a very good result with a 4.8Ohm resistor and a 1000uF capacitor, the noise is now much manageable. I'm not skilled enough to be able to calculate what inductor I might need for a low pass filter so feel free to chime in! Oh, the filter shall be small and not very expensive

Outcome is attached.


Fun fact: another fan I have delivers a much higher noise on the line and over many harmonics. Attached.

[redwire]

How big is the fan? It's 12V? 1,000uF is huge. An inductor is only useful at high frequencies, not the BLDC motor pitch 171Hz.
I find the chinese LM2596 switcher boards are junk - counterfeit IC's and very noisy, need a lot of capacitance added to the input and output and even then they are low efficiency.
Maybe look for other power supply rails you can take power from, like the VFD filaments that are not sensitive to noise.
Or the relay power rail.

[tony359]

I have +5, -15 and +15. I could try 5V fans and the 5V rail. However it's not the LM2596, the same happens with a resistor in its place.

I've resolved with a resistor and a capacitor. 10Ohm resistor, 1000uF capacitor. The noise has disappeared completely.

But you might be right in suggesting a different rail. Maybe the 5V one is less critical as relays are powered by it. Though, they're muting relays, when they're activated, it means sound is muted
The original design of the PCB allowed for a fan which was never implemented and it was off the 15V rail though.

Fan is 60mm 12V.

[прямо]

If a small 5V DC fan less than 1W is good enough to help dissipate the heat, you could use a 5V ground isolator BS0505S-1W, because it removes Bluetooth and USB VBUS noise completely.