Anyone tried resolving PSU coil whine?

[RejZoR]

Recently bought RTX 3080 and upon playing games I noticed horrendous buzzing. I thought, great, I got a buzzer card. But upon inspection, the buzzing doesn't seem to come from the graphic card which is actually surprisingly quiet even though it has one of the slimmest heatsinks with only 3x 85mm fans. Instead, it's coming from the PSU. It wasn't buzzing like this when I had GTX 1080Ti in the system, but does with RTX 3080. The power delivery shouldn't be an issue as I'm running a high end Platinum PSU with 750W from BeQuiet. I'll check the power loads of the system, but I don't think it's an issue really.

 

Apparently specific load level creates some sort of coil whine inside my PSU which wasn't present at lower load level before. And I'm wondering, there are several open induction coils inside. What would be a preferred method to try and silence them down assuming one of them is responsible for the buzzing? I know hot glue is one of methods, but I'm wondering if anyone ha used any other method and what was the efficiency of the approach, maybe one that's less permanent. I don't want to fill entire PSU with hot glue just to notice I haven't really fixed anything.

[Morrie Sells Wigs]

There's a certain irony to the "be quiet - silent wings" sticker.

[Ralphred]

Back in the old days, coils were coated in wax to help combat this. If you know which coil is the culprit you can remove and dip it, if not some kind of conformal coating (available from electronics stores) may help if directed at the coils. 

Hot glue does sound like a good compromise, non-invasive to the point it doesn't really matter if it's there or not.

The sound comes from "loose" turns in the coil, the idea is to get some medium in to the gaps to prevent the movement and stop the noise.

 

Personally, there are some nasty voltages in those psu's tho, so would probe each coil with some plastic chopsticks I have, to keep a good distance. After identifying the offending coil remove it from the board and dip it in bees wax for a while (so the coil reaches the melting temperature of the wax, you want it to infuse in to the coil, not just sit on the surface).

There is a caveat with this though, the wax melts about 60-65 degrees C, as my CPU/GPU never go above this temperature I'm confident my PSU never does either, if I thought it did I'd have to use hot glue, which melts a lot higher (150-200) and would obviously require some slightly more elaborate (and less safe) heating rig than a bath of boiling water.

[Elisis]

1 hour ago, RejZoR said:

What would be a preferred method to try and silence them

There is no preferred method a consumer could carry out. Either learn to live with it, or buy a new PSU.

[RejZoR]

18 minutes ago, Elisis said:

There is no preferred method a consumer could carry out. Either learn to live with it, or buy a new PSU.

That applies to enclosed induction coils on graphic cards...

 

I've checked power usage and it's below 500W when playing Shadow of Tomb Raider with all maxed out and using ray tracing. So, power clearly not an issue, just the load apparently.

[mdk777]

Big slabs of silicone caulk is the method PSU makers apply.

But yeah, there is no certainty it will work.

 

 

[Guest]

Unfortunately, higher efficiency requirements over a wider range of loads and the cost of copper has required us to use lower switching frequencies that are within the range of human hearing. 

 

[Elisis]

5 hours ago, RejZoR said:

power clearly not an issue

I never said so? But okay. 

[RejZoR]

1 hour ago, Elisis said:

I never said so? But okay. 

I just added that so I don't have to type 2 separate posts. But okay.

[Kilrah]

3 hours ago, jonnyGURU said:

Unfortunately, higher efficiency requirements over a wider range of loads and the cost of copper has required us to use lower switching frequencies that are within the range of human hearing. 

Not at all, quite the opposite actually, frequency keeps going up for that reason.

 

What usually causes coil whine is not the switching frequency, it's when one of the consumer devices changes their draw significantly periodically within audio frequencies, which is common with GPUs (load spikes up while computing a frame, then drops when done until it gets a new one to render etc, by hundreds of watts).

 

So you're not hearing the (small and fast) current variations when the switch turns on/off, you're hearing the (big and slow) load current variations between the high load and low load states.

[RejZoR]

Unless RTX 3080 cards have massive power draw fluctuations on millisecond scale, I don't think that's the case here. Need to inspect further by taking PSU out of the case and listen to it more closely because within the case it makes it seem like noise is coming from somewhere else, but can't possibly be graphic card because I put my head right next to it and there was none of this distinct buzzing sound. Need to repeat with actual PSU to see what's up. Then take it apart and try to locate the offending coil inside of it. Hopefully it's one that can be tinkered with...

[RejZoR]

No luck. Opened up the sucker and after whine spotting using a funnel shaped roll, the biggest offender is the winding thingy between two big capacitors and the central coil. And the damn thing is enclosed into a casing and wrapped into insulating tape. Filled all the coils with hot glue just for good measure and while I might have slightly toned down the whine, but couldn't do much with the encased coil. Filled some holes with the glue, but it has done very little or nothing. Oh well. At least I tried.

[akio123008]

8 hours ago, Kilrah said:

you're not hearing the (small and fast) current variations when the switch turns on/off, you're hearing the (big and slow) load current variations between the high load and low load states.

Given this is causing the coil whine, you could try to add an extra (big) output capacitor in that PSU to smooth out the spikes. 

[Guest]

On 1/2/2021 at 12:38 AM, Kilrah said:

Not at all, quite the opposite actually, frequency keeps going up for that reason.

 

What usually causes coil whine is not the switching frequency, it's when one of the consumer devices changes their draw significantly periodically within audio frequencies, which is common with GPUs (load spikes up while computing a frame, then drops when done until it gets a new one to render etc, by hundreds of watts).

 

So you're not hearing the (small and fast) current variations when the switch turns on/off, you're hearing the (big and slow) load current variations between the high load and low load states.

Umm.. no.

 

Not in the PSU, at least.  I believe you're looking at this strictly from a component point where DC to DC conversion is done, and not at the PSU where the AC to DC conversion is done.  But I can't think of any modern parts that don't use MOSFETs (using Schottky diodes instead?) for switching and with MOSFETs, the higher the switching frequency the more times per second you have switching losses.

 

Or maybe you're referring to the noises coming from the transformer when a double forward PSU is "hard switching", which doesn't actually have to do with switching frequency, but the way the voltage is switched.

 

If you want to increase the efficiency, you lower the switching frequency.  If you're willing to sacrifice efficiency you can improve dynamic performance and you won't have audible noise, but that's not where the industry is trending to.

[Kilrah]

2 hours ago, jonnyGURU said:

If you want to increase the efficiency, you lower the switching frequency.

If you reduce the frequency you improve MOSFET efficiency, but you increase transformer size/weight/losses (you initially mentioned increased cost of copper, and that's what I was answering as lowering frequency means more copper).

 

MOSFET losses depend on the type and new parts keep becoming more efficient at higher frequencies, so frequencies keep going up as a result.

[RejZoR]

Hot glue maybe cut off tiny bit of buzzing, I've now inverted my PSU so it's facing inside of the case. I now got it to such low level it can be heard, but as soon as there is any ambient sound in the game, I can't really hear it anymore. So, I guess mission failed successfully. I had to reroute the cables from the PSU so it's getting more direct fresh air within the bottom tunnel where it's hidden. It's getting half of the bottom fan airflow so it should be fine cooling wise, but instead of releasing buzzing noise directly out of the bottom of the case it's now pointing inside and only small amount of noise comes out of the back of it on the exhaust side. Not ideal, but far better than I had initially where it was so loud I was seriously considering replacing it. Which is weird given it was the highest end model from BeQuiet few years ago and then fails at acoustics. From company called "be quiet". A bit disappointed given 500W load seems like something it would experience most commonly (while having some reserve left so it's not running at 100% load...

[Guest]

12 hours ago, Kilrah said:

If you reduce the frequency you improve MOSFET efficiency, but you increase transformer size/weight/losses (you initially mentioned increased cost of copper, and that's what I was answering as lowering frequency means more copper).

 

MOSFET losses depend on the type and new parts keep becoming more efficient at higher frequencies, so frequencies keep going up as a result.

I'm telling you, you're incorrect.  At least in practical application.

 

There is a "balance".  You have to find the fine line between switching losses and losses in copper.  Currently, the balance requires a switching frequency just above audible noise level and slightly larger magnetics to obtain current efficiency requirements.  If you increase the switching frequency, you can reduce the size of magnetics, but the less copper will not make up for the higher switching losses.  Furthermore, losses in the copper can be made up for by using heavier gauge copper.  For example:  One of the primary differences between a Corsair VS and the Corsair CV is that the output choke simply uses heavier gauge copper.  

[artuc]

@jonnyGURU

 

Any reason they can't dip or pot chokes and transformers before assembly?

 

I know often they silicone goop after they're on the board.

[YoMz]

28 minutes ago, artuc said:

@jonnyGURU

 

Any reason they can't dip or pot chokes and transformers before assembly?

 

I know often they silicone goop after they're on th board.

from the pictures provided by OP, it doesn't look like BeQuiet! does this for their PSUs... I always thought it was common practice to dip them in some kind of coating before assembly as well...

[Guest]

12 hours ago, artuc said:

@jonnyGURU

 

Any reason they can't dip or pot chokes and transformers before assembly?

 

I know often they silicone goop after they're on the board.

 

12 hours ago, YoMz said:

from the pictures provided by OP, it doesn't look like BeQuiet! does this for their PSUs... I always thought it was common practice to dip them in some kind of coating before assembly as well...

 

It is SOP in most factories.  Problem is, one part RTV that is non-electrically conductive yet not a thermal insulator is HELLA EXPENSIVE!!!!!

 

If you're paying a tier 1 to build you a $500 PSU, then the cost of RTV is a drop in the bucket:

 

 

If you're trying to build a $100 PSU at a China owned and operated factory, well.... not so much.

[akio123008]

On 1/4/2021 at 6:43 AM, jonnyGURU said:

I can't think of any modern parts that don't use MOSFETs (using Schottky diodes instead?) for switching

You can't use a Schottky diode for switching (as far as I know) Perhaps you meant a thyristor or something alike? (similar symbol?)

[Guest]

50 minutes ago, akio123008 said:

You can't use a Schottky diode for switching (as far as I know) Perhaps you meant a thyristor or something alike? (similar symbol?)

You're absolutely correct.  You always use a transistor.  I was thinking of the SR, using SD's instead of FETs and being less efficient.

[Mark Kaine]

On 1/2/2021 at 6:08 AM, jonnyGURU said:

Unfortunately, higher efficiency requirements over a wider range of loads and the cost of copper has required us to use lower switching frequencies that are within the range of human hearing. 

 

This reads like satire honestly. How much copper are we talking per psu? 5 bucks or what?

 

 

 

 

[Guest]

7 hours ago, Mark Kaine said:

This reads like satire honestly. How much copper are we talking per psu? 5 bucks or what?

 

 

 

 

"5 bucks"

 

There's no fixed number as the cost is increased based on the amount of copper used (PCB layers, trace size, copper windings, etc.), which varies from unit to unit.

 

The cost increase I've seen is average 7%.  This equates to about a 3% cost increase on average per unit.

 

It's not satire.  This is what I Googled for you:

 

https://tradingeconomics.com/commodity/copper

 

https://www.reuters.com/article/global-metals/metals-china-data-and-vaccine-optimism-lift-copper-price-idUSL1N2IV0JW

 

 

[Tristerin]

As for precious, semi, and non - the metals industry boomed Q4 of 2020 (for myself at least) but the supply chains are....fooked going forward atm.  Makes sense copper went up.