Oscillating fan noise when in curve mode, but not in fixed % (case)

[RejZoR]

So, I've noticed recently with more fans in my new Kublai case, although I remember having it in old case as well. When I have a custom curve set in my MSI mobo, fans produce this repeating oscillating noise at certain RPM range. Imagine it sounding like "wroom wroom wroom wroom" noise every second, imagine pressing gas pedal in a car a bit more than it's a speed limit and then releasing it to slow down a bit instead of running at the exact speed you want pressed just right constantly. It's not super load this, just loud enough I hear it and it's pissing me off. What is more funny is if I use a fixed speed percentage outputting same RPM instead of a fan curve doing same RPM, the oscillating noise goes away. I'm using SilentWings 3 fans for all slots in my case, that's 6x premium fans that don't have the slightest vibration when you're holding them in a hand while running and it's doing this.

So, what I'm wonder if this is a common thing with PWM fans on MSI mobos or even in general? It sounds to me like something isn't right with the PWM circuitry for it to be pulsating like this on fans (I'm aware PWM being just that, but you'd think they were not doing it like this). Also, I don't think the curve is the issue since it's attached to a System temperature which fluctuates far less than CPU and is way more constant, so it's not temperature changes that are causing the RPM changing and resulting in that oscillating noise.

 

[Bill_Bright]

If the noise is oscillating, then it would seem the RPM is not maintaining a consistent speed. Plant your finger on the hub of each fan. This will change the RPM of that fan. See if you can then tell if the noise if coming from one or all. If you have a multimeter, you can see if the voltages are fluctuating too.

 

[RejZoR]

The thing is, RPM is fluctuating like with any fan where you never have a dead on 1000 RPM, but it's always 1005, 1003, 1000, 999 etc. It does this even when using 50% fixed speed. But without this noise. And all fans do this in what seems to be a perfect sync which is why it's so amplified and why I'm kinda blaming the mobo PWM power delivery. If fans were doing it, they'd be doing it in totally different intervals each.

 

[TheoneandonlyMrK]

The thing is, RPM is fluctuating like with any fan where you never have a dead on 1000 RPM, but it's always 1005, 1003, 1000, 999 etc. It does this even when using 50% fixed speed. But without this noise. And all fans do this in what seems to be a perfect sync which is why it's so amplified and why I'm kinda blaming the mobo PWM power delivery. If fans were doing it, they'd be doing it in totally different intervals each.

Sounds like it's to sensitive to changes, is there a setting for hysteresis ,i use a fan controller that did this if set at a very low hysteresis so with constantly monitoring temps and speeds it was making continuous adjustments.
That or possibly monitor one component not all but again you might not have a setting for that.

 

[lilhasselhoffer]

Natural Frequency.
Beat Frequency.
Spring mass damper system.


For the non engineers, the fans all oscillate at slightly different frequencies. Said oscillation may be very small alone, but you're running multiple fans with large distance between them. The net effect is that at certain frequencies the beat frequency (read: sum of all frequencies) adds to a multiple of the natural frequency of the system. Without any damping, the natural frequency multiple self propogates, creating excessive vibrations.

This will explain inconsistent noise, the response at specific fan settings, and why you should invest in some damping material. As a side note, the previous case probably had different construction, aligning with your observation that it did not occur as bad there.

Fixes are:
1) Damp the fans to convert mechanical energy into heat.
2) Change natural frequencies by reinforcing the case structure or item placement in the case.
3) Change fan curves, such that the interference of the frequencies is destructive rather than constructive (frequency fans counteract one another).


For a somewhat basic, but pretty thorough, first look I'd recommend the following:
Physics Classroom Lesson on Interference and beats
The topic goes much deeper, but KISS is a powerful message.

 

[Frick]

Hey Hoff! Stick around.

 

[RejZoR]

All fans have integrated rubber dampeners and some are even additionally suspended on rubber holders instead of screws. And these are one of the smoothest fans I've ever used. Most others always have some sort of vibration or sound going on, but not these. And like I've said, only when on a curve. With fixed RPM, no such issue. Hysteresis seems very loose in general, when temperatures or even % changes, it takes few seconds for fans to even catch up so I don't think that's an issue.

Isn't PWM basically a constantly pulsating power delivery, to me it seems like it only seems to be doing this when RPM falls at a certain range and it doesn't seem to be able to keep up those RPM smoothly just in that mode. When you use fixed, all is fine. Running at 50% which is about the same as at what they are at desktop idle and one produces this oscillation and another doesn't.

 

[lilhasselhoffer]

Hysteresis is the loss of energy through variations in magnetic field potential and magnetic field angular change. If there were hysteresis losses in the system you'd experience a dramatic increase in power pull, not a mechancial issue as described.

Pull up the spring mass damper system information. A Mass (the fan blades) is connected to a Spring (the mechanical deformation of the material and input force from the motor), with Damping provided by either elastic deformation in the material or rubber pads.


Without math...because people seem to hate it when that gets too real...every fan has a frequency. Perhaps the natural frequency is very low, to the point where we squishy humans can't detect it. That's easy, just overdamp the thing with squishy materials and run it at low frequencies. That doesn't change the system, it just makes the response different. If I was to explain all of this, it'd be a semester of math, so you'll have to confirm yourself or trust.


Now, one fan isn't an issue. Multiply it several times over. Again, overdamping will make the response more tame if the beat (summation of frequencies) is low, but if the frequencies are constructive and a multiple of the natural frequency the overdamping is like wearing a ponch in a flood [too little, too late].


Why isn't it an issue with a set frequency versus a curve? Curves aren't really a curve. It would take a huge calculation pool, which the fans don't respond to. So most curve controllers use a sampling plan and refresh only with sufficient deltas in temperature. This means that a consistent temperature will often lock fans to a frequency over a period of time, allowing the accumulation of energy (again, complicated interactions amongst oscillating bodies). The fixed points do cycle through the constructive interference, but don't stay there long enough to build up energy. Momentary constructive interference is easily dissipated in an overdamped system.



But hysteresis....fine. If you think it's the fan controller going wonky, replace it. You've described a mechanical oscillation (sound), and ascribed it to an energy loss method for electronics. An energy loss mode which would generate heat, not mechanical waves. If you've seen something else my explanation may be inaccurate, but I've outlined why your description matches physical properties. Without more input data, this is an exercise in trying to solve a jigsaw with half the pieces missing.

PBS did a fantastic job on destroying a wine glass via the natural frequency and the sound waves (mechanical waves) of a singer. If you feel so inclined, it's worth a watch (and demonstrates how something relatively weak can induce destructive forces:
PBS video on Youtube, wine glass shatter

Hey Hoff! Stick around.

Much obliged, but no.

I've been thoroughly disappointed by how this place often becomes more about spreading insane theories and fanboy hate. I lost respect when a long answer with all the math laid out is disregarded. I'll gladly help or respond to those who have shown some spark in the past, but can't muster the energy to slog through it any more. After one too many heated discussions with people here that got...less civil and more dangerous than I'd like...I was given the opportunity to review input versus reward and the equation never balanced.

 

[INSTG8R]

My Fury used to do that at exactly 41% drove me mental and forced me to use AB to adjust the curve...

 

[Caring1]

The thing is, RPM is fluctuating like with any fan where you never have a dead on 1000 RPM, but it's always 1005, 1003, 1000, 999 etc. It does this even when using 50% fixed speed. But without this noise. And all fans do this in what seems to be a perfect sync which is why it's so amplified and why I'm kinda blaming the mobo PWM power delivery. If fans were doing it, they'd be doing it in totally different intervals each.

Are all the fans on one header or spread around the Motherboard fan headers?
Have you tried disconnecting one fan at a time to see if power draw is relevant?

 

[RejZoR]

Are all the fans on one header or spread around the Motherboard fan headers?
Have you tried disconnecting one fan at a time to see if power draw is relevant?

Hm, this got me thinking, I have most of the fans split with a Y cable because motherboard doesn't have anywhere near enough fan headers. I don't think power delivery itself is a problem, but it might affect PWM where only one fan is sending back PWM signal, but the other one is hooked to the same power source. So, it's reporting like one fan, but it has a load of two. Could this be the reason?

Coz it sucks that motherboard can't take 7 devices on fan headers (1x AiO pump and 6 fans). I have like 5 physical headers only...

 

[Bill_Bright]

The thing is, RPM is fluctuating like with any fan where you never have a dead on 1000 RPM, but it's always 1005, 1003, 1000, 999 etc

But those fluctuations are insignificant and I doubt you could hear it. I am talking about wide swings like 1000RPM down to 600RPM then quickly back to 1300RPM with no change in taskings.

You said in your first opening sentence, "I've noticed recently...". But later reading suggests it is not all that recent.

Did this problem just recently appear? That is, when your current fan power connections, did it used to work correctly?

 

[RejZoR]

I listen to music non stop. Recently, I wasn't and I noticed this oscillating noise from fans. That's what I meant.

 

[Bill_Bright]

Hmmm, so it could have been happening since day one.

If me, I would use one fan per motherboard header. Power the extra fans through a manual speed controller connected directly to the PSU and see what happens.

 

[RejZoR]

The 5820K and GTX 1080Ti need a lot of wind to be comfy...

 

[Bill_Bright]

That's fine, but you can satisfy that temporarily by opening the side panel and blasting a [hopefully quiet] desk fan in there - at least while testing with one fan per mobo header.

 

[Steevo]

Hysteresis in the PWM control loop. The gain control setting is too high, or the sampling rate calculation is wrong.

I have to adjust these same settings in PWM and servo controllers for my job.

 

[Frick]

Much obliged, but no.

I've been thoroughly disappointed by how this place often becomes more about spreading insane theories and fanboy hate. I lost respect when a long answer with all the math laid out is disregarded. I'll gladly help or respond to those who have shown some spark in the past, but can't muster the energy to slog through it any more. After one too many heated discussions with people here that got...less civil and more dangerous than I'd like...I was given the opportunity to review input versus reward and the equation never balanced.

You're not wrong ... and such equations rarely balance out to be honest. At least not online.