Fixing a loud PSU fan without dying | Hacker News Reader

Fixing a loud PSU fan without dying

chameth.com

45 pointssprawl_62 comments

An important side note is that Cooler Master’s “10-year Warranty” is garbage:

> I looked at Cooler Master’s warranty, and for issues within the first two years you have to deal with the retailer. That would be Amazon in my case. So I looked at Amazon’s information on warranty issues. Their policy is that if it’s more than 30 days since purchase, you have to send it off to a third-party repair center and wait for them to diagnose and try to repair it. Here’s the kicker: Usually repairs take up to 20 business days (including delivery time), but could take slightly longer

Everyone knows electronic devices tend to die early or last forever. Cooler Master and Amazon are working a shitty dodge here, and I bet they avoid most DOA warranty claims because of it.

I ran into this same setup with a microwave that failed in a few months after purchase and Walmart.

Tried to deal with the manufacturer, but they couldn’t help and sent me to the retailer.

Went to the store, popped it up on the counter, had a short conversation and got the expected “you have to deal with the manufacturer”.

Is there anything else I can help you with today?

Actually yes. Can you throw that out for me?

The confusion on the guy’s face was great.

Spent more than enough time on the $100 microwave. Their problem now.

My experience with cheap modern microwave failures has been the door sensor failing, which for safety ofc prevents the magnetron from running. I had one fail in about 2 months, thankfully fixable with a $10 sensor and 15 minute of work. Same goes for a lot of appliances, repaired a dryer that had its door sensor fail (in fact, they all tend to use identical door sensors as far as I've seen, dryers and washing machines and microwaves).

My personal experience with a modern microwave (they mostly seem to be the same design internally, coming from the same chassis with the same electronics just a different button panel) was that the internal light bulb blowing generated a surge (it was a mains voltage bulb) that wasn't fused so the next nearest thing in the mains circuit was a trace on the motherboard that vaporized.

There is no way of easily changing this bulb (inside the main casing with no access panel for the bulb) so for want of a single in-line fuse, the entire microwave was rendered scrap[0] by the lifetime of a light bulb.

[0] - Except for the fact that I care not for electrical safety "DO NOT OPEN" warnings of doom due to being actually competent with handling high voltage equipment and being able to do a board level repair on the burned out trace without touching the very large capacitors associated with the very high voltage side controlling the magnetron...

In Europe, it actually matches the law, it is actually better.

- You have a 14 days retraction period, where you can send back the item and get your money back, defect or not. Amazon gives you 30 days and free return shipping.

- Legal warranty is 2 years and is provided by the seller, not the manufacturer. It is Amazon in this case, but it could be a brick and mortar shop or any other seller, including independent sellers. And they have to solve your problem within 30 days (repair, refund, replacement, ...). How to deal with the manufacturer is their problem. Amazon here is perfectly in line with the law.

- Manufacturer warranty and commercial warranty comes in addition to the legal warranty is whatever is written in the contract.

So here, the manufacturer doesn't need to provide warranty directly to the consumer for the first 2 years as the legal warranty covers it. And I am pretty sure that the 8 extra years offered by Cooler Master are even worse than what Amazon offers. For example, they may require you to pay for return shipping and wait even longer. There may be some fine print there too.

Similar in Norway, though not it's only for production defects and not expected wear and tear from regular use, or misuse etc.

However it's on the seller to prove it was not a production defect if they don't want to fix or refund.

Back when CD burners kept getting faster and faster my buddy bought a 4x writer and ended up with a 52x writer as they kept breaking down after 12-18 months. IIRC he went through four or five before getting the 52x which lasted.

Same in the UK, 2 years is the legal minimum for all electronic goods and no contract can be written to override that. If it uses electricity in any way, it must last 2 years.

It helps prevent ewaste.

Noctua fans are usually a good choice (though I think their industrial line is better, especially if you can get a deal on the splash proof IP66ish ones). That all said, other than having extremely good lifespans for ostensibly a retail oriented fan, the other thing they tout (the low RPM and quietness) sometimes doesn't work well when you're replacing something much different like a high RPM Delta fan. The Noctuas may not push enough air, especially if there's a controller involved. I've had to build add on fan controllers before where whatever it is won't spin the fans fast enough until they heat up too much, which makes things even more fun. Especially for a PSU it's probably worth looking to see if there's a major difference in airflow (and at what RPMs), as it would suck to have your fixed up PSU cook itself because it's not flowing enough air.

Other less prominent vendors often do make easier drop in fans, that often have very good lifespans that might be easier to adapt than Noctua. Sanyo Denki fans easily will long outlive even Noctua (but don't typically play in the retail space); Delta and Dynatron also usually sell at least decent parts that are much better than whatever random chinesium fans lots of stuff come with. All of the above have much more variety in the profile of fan sizes and configurations and I would probably be looking through their parts catalogs to see if they don't have something that's a better fit (they often do).

The only other interesting thing is that I have picked up is soldering seems like a good idea in cases like this, but at least in bigger setups I've seen solder joints fail because of vibration. Maybe not an issue here, but crimping is usually a much better solution for connecting things that need to have a long life and might be vibrating like something attached to a fan.

I've "fixed" countless noisy computer fans over the years with simple lubrication. Peel off the label, remove the bung, pry out the circlip and pull the rotor + blade assembly out, clean off any remaining old grease/oil, and then pack the bushing with petroleum jelly before reinserting the shaft. Reassembly is the reverse of disassembly.

doing this on a fan that used a dry lubricant can burn a wire lead from increased startup torque requirements to get it spinning.

it's not bad advice but it's worth noting that petroleum jelly is pretty thick compared to a machinists' oil or dry graphite.

Then there are the fans with magnetic bearings. Had a few that started making noise but never figured out from where, given there are no touching parts.

I think magnetic bearings necessarily also got touching parts.

The leverage the motor exerts is huge relative to the shaft, and chances are that when the bearing is chattering audibly, it's already been running dry with much higher friction for a while, and worn enough that you'll need the extra viscosity of a grease to fill the clearance.

I have a personal rule that I only follow instructions when I understand more than 10% of the words in them, so I'm buying a new fan.

That's a good personal rule to copy

I would glue an additional fan onto the psu hoping it would stay cool enough for the master not to notice the heat build up. If that doesn't work open it up and cut the wires or unplug the fan.

Never had much success doing this. At best, it worked for a few days, maybe weeks before buzzing again. I tried a variety of lubricants, stuff that I had on hand (WD-40, 3-in-1, lithium grease...) but not petroleum jelly though.

So unless it is a weird fan I can't find elsewhere, now, I just buy a new one.

WD-40 isn't a lubricant, 3 in 1 is really a cutting oil and too thin, and I can't imagine lithium grease doing well in a high speed bearing...it's really only good because I think it's less toxic. These are all real thin and not appropriate in a bearing. I'm not sure about petroleum jelly, but as crazy as it sounds I've packed all sorts of bearings including fan bearings with basic synthetic marine grease (the pink shit that comes in tubs and grease cartridges), or occasionally heavy weight synthetic gear oil (think 75w90), and profoundly improved the life of things and reduced drag. Only real drawback is god forbid you get real grease on your clothes or something you care about cause it's not coming out.

Also if it came packed with graphite or something else exotic none of the above applies and doing the above can burn a motor out.

> WD-40 isn't a lubricant

WD-40 is a lubricant. The WD-40 company says it is [1], the Wikipedia article says it is, attempts to reverse-engineer the (secret) formula have found compounds typically used as lubricants and it is rather obvious to anyone who has used it.

It is the best lubricant for you application? Probably not, and that includes fans. It is probably not the best at anything unless you have an Atlas missile lying around. It is not the best lubricant, penetrating oil, water displacer, rust protector, etc... but it does all of it to some extent, and it is cheap and convenient.

Anyways, thank you for the suggestion of using gear oil, I didn't try this one and it sounds like a good candidate. I don't have marine grease but what I understand is that it is similar to basic lithium grease but formulated to resist the harsh marine environment (so maybe with some corrosion inhibitors).

[1] https://www.wd40.com/how-to/faqs/

I mean I'm sure water could be considered a lubricant too, but it's almost never a good one. WD-40 I'm sure in some technical sense can function as a lubricant but it's a really crappy one and if nothing else tends to dry out quick. Even as a penetrating fluid it's pretty shit even compared to liquid wrench or ATF and acetone. I don't know why people bother using it other than as a water displacer.

Marine bearing and chassis grease is really different than white lithium grease. Lithium grease in general probably isn't advisable on a fan the more I think about it as it tends to destroy plastic and rubber. On a real bearing like a wheel bearing even a thin coat of high pressure grease like synthetic marine grease is going to outperform basically everything else (unless you've got an oil pump involved), and again would be the first thing I'd try and pack a bearing of rotating assembly with hands down. t's probably cheaper or about the same if you go pick up something like a tub of Supertech synthetic marine. I'm sure if you go ask somewhere like BobIsTheOilGuy forums someone there will be able to point to an optimal solution.

My method has been to just give the fan a good whack until the annoying noise goes away.

Check your credit card agreement. Some cards have a "purchase return protection" for when a retailer won't take back a product.

It's kindof a gimmick because it's easy to forget you have it, but if you have it, use it.

I see they have the exact same frustration with the results of trying to research how to safely work on a PSU as I've had. I know there's plenty of info out there about working on CRTs; are the caps in PSUs really more dangerous than that?

Modern computer power supplies are far less dangerous than CRTs were.

CRTs were dangerous because they used extremely high voltages (often as high as 25 kV), and that voltage was often present on the CRT itself, which acted as a capacitor. The capacitance of the CRT was relatively low, but it still stored enough energy to be very hazardous. There's ways to safely discharge them, but you absolutely need to know what you're doing and work with one hand behind your back.

The voltages in a modern switching power supply are much lower (typically no more than ~1.5x line voltage ≈ 180V), and the capacitors usually have "bleeder" resistors which will discharge them to safe voltages within a few seconds. I'd still give one a few minutes before touching it - and maybe check one or two of the bigger capacitors with a multimeter - but it's still much less dangerous.

The main danger with CRT HV is not the HV itself but what the shock will make you do, such as jump and cut your hand on something sharp in the chassis, or even crack the neck of the tube, or fall and hit yourself. There is very little current available, especially if the set has been off, but it's comparable to a very painful static shock. Instead, the "low" voltage B+ supply is far more dangerous when the set is on, as that is basically rectified mains.

Beware. I took a poke from the focus anode of a CRT when I had a job repairing mainframe terminals in the early 80s. I thought I had discharged it, but apparently I didn't.

I had to sit down in a quiet room for the rest of the work day, before I could function again. It was probably 13 kV, which was typical for monochrome displays back in the day.

Also, anything in a CRT television can kill you because TV's tended to have a "hot" chassis. The repairmen always plugged a TV into an isolation transformer so they could ground its chassis while working on it. There's a lot of counterintuitive things about those old electronics due to clever cost reductions.

The focus anode is lower voltage (~1kV) but its supply may be capable of higher current.

I might have gotten my terms mixed up, but it's the suction cup thingy on the side of the tube. And it was disconnected, so the poke I got was from the charge built up on the capacitance.

We had a procedure for discharging it, which I must have bungled.

It hurts if you poke the wrong part with a finger while it's live. I'm still here, ymmv.

I was once told about a chap who got his ring finger blown off because the spanner he was holding inadvertently made contact with a car battery terminal and the engine. Ouch.

I am pretty certain that the G2 Noctua fans will start at 5V and they also in the 4 pin model will turn off at 0-19% and turn on at 20%. Its a bit of a shame Noctua don't make 2 pin fans anymore but it is possible on the latest most efficient models to get the low start voltage.

The G1 Noctua fans don't, I have a 120mm still in my case that is a decade or so old and it starts at a bit under 7V, but all the G2 140mm will start at 5V IIRC the Noctua presentations on this. Wish it was laid out in the specifications however.

I learned about fans many yea^H^H^H decades ago when I had a sun external scsi disk box.

One day at Fry's Electronics I found they sold external scsi cases. So I bought a case and a cheaper bare scsi drive and hooked it to my sun.

The fan was quieter too!

and after a few months, the fan started getting louder. and louder. and then one day when the drive seemed to stop responding, I found it had failed.

sun fans were annoyingly loud, but they had one interesting property: THEY ALWAYS SPUN.

I've replaced at least 2 PSU fans, all without incident. The last time, I bought a pack of fans with that specific tiny connector, making the repair simpler.

Get PSU from original manufacturers instead, someone like Seasonic. Most common brands simply repackage stuff and not really the best quality.

Seasonic makes (made?) really excellent PSUs, but they've been real thin on new inventory over the last year or so. Maybe they're getting out of direct retail? Even a second hand older Seasonic platinum PSU is still hands down better than a lot of the new garbage that's getting turned out.

I noticed some of their models are often out of stock. I really hope they aren't quitting.

Interesting article and great results.

PSUs are probably the only piece of Hardware, i DO NOT fiddle around with, not even for saving lots of money or time.

PSU Capacitors are no joke, especially in (older) TVs. However, I expected more like an instruction manual why PSUs are dangerous and how to properly discharge capacitors before taking them apart less than how to replace a fan although it's dangerous and getting away with it with a simple hint to not touch the capacitors (no offense).

BTW I would recommend to use electrical tape and shrink tube instead of simple adhesive tape (as it looks in the pictures) to prevent lose cables touching the spinning fan.

A new semi-passive 850-watt fully modular PSU is around EUR 130, the Noctua fan around EUR 30.

I guess if you know electronics and how to safely handle the PSU internals, the risk of injury is low, but I personally would not risk it for EUR 100.

Also, if the only problem was the noisy fan, I guess selling it used would have returned most of the investment, leaving him with like EUR 50 in added cost. Compared to the price of a modern gaming PC, that's nothing (also avoiding not risking your life).

C grade high school physics understanding makes the risk exactly zero. Selling defective stuff without declaring is bad mojo, with declaring it just delegates the pain in the bum to someone else..

I’ve been tearing things open after ignoring the “dangerous if opened” stickers since I was 8 years old. I’m in my 30s, you’d think something would have caused me harm by now, but no.

There's a fair bit of survivorship bias in this. That said, modern power supplies use "bleeder resistors" to discharge the capacitors once powered off.

> There's a fair bit of survivorship bias in this.

Is there? This kind of statement has the potential to exhibit survivorship bias, but I feel like the opposite—"12-year-old dies replacing a power supply fan"—would make headlines. Definitely haven't seen that.

Capacitors should be respected for sure, but people don't routinely die in DIY electronics tasks.

Electrical injuries claim some 1000 lives per year in the states, and 20% of all electrical injuries are sustained by children (lethal and non-lethal). I don't think every electrocuted child makes even local headlines.

Sobering for sure, but I think vanishingly few of those are from residual energy in capacitors in unplugged appliances during disassembly, [1] and certainly far from the "this forum is missing x people just like doubled112 because of tragedies" that I imagine when you say "there's a fair bit of survivorship bias in this".

[1] a quick search mentions things like damaged cords, the classic metal object into the outlet, etc. I installed tamper-resistant outlets everywhere in my house to prevent the latter as part of child-proofing. I think they're mandatory now in new construction. I also may have gone a bit overboard trying to instill caution in my children about this this particular risk; my son tonight asked me if he was okay after his sleeve brushed against the metal part of a USB-C connector while he was plugging in his tablet.

I rather meant statements like "I did X and I didn't die" are open to survivorship bias, than asserting many must have died from this particular cause. I believe bleeder resistors are written in blood though, I vaguely remember electrocution tales from the times when people were expected to replace vacuum tubes in radios.

That's more than an order of magnitude less than fatalities from cars. In my book, that puts it into "be careful" territory but far from "don't even think about it".

Nearly everyone deals with roads every single day. Very few people disassemble electric devices per day. The risk base is not comparable.

Also https://xkcd.com/795/

As a child, I remember being up in the attic with my grandfather. He touched some wires, swore (which was unusual for him), and said 'that was 220'. I still have a healthy respect for power.

A lot of the risk is simply overstated. Yes, bad things can happen but that also goes for crossing the road yet we don't limit that to certified professionals. Just have some basic understanding of electricity, common sense, don't attempt the work while tired and you'll be fine.

I will put hundreds of dollars in Noctua fans into a second-hand chassis without thinking twice.

It's sometimes uneconomical from a cost-ratio perspective, but it is crucial to making datacenter-grade equipment actually useable at home.

If you need pretty good fans for cheap as dirt, there is also Arctic Cooling.

They just never last. Arctic fans perform really well especially for their price but they all seem to develop problems. I have probably bought about 15 different Arctic fans from the F and P range and none of them survived 5 years, most were dead or developed noise within 1-2 years. Noctua on the other hand the old 80mm fan from the early 2000s still works just fine and remains quiet. Noctua fans are crazy reliable, they cost more too but I would suspect over the life of the fan they end up similar priced or cheaper.

Arctic fans seem to last 10+ years for me. I haven't bought any recently though and I run them at 500 rpm or so, that's my strategy: many and large fans, running slowly.

while arctic make fairly good high performing fans for cheap their bearings/acoustics are kind of bad. i bought 5 of the original p12 max fans with the ball bearings. they all made a chirping noise at various rpms (even after getting 5 new ones from arctic). the new p12 pro fans perform great but around 1200 rpms the motor is very loud and annoying

really wish arctic would now focus on the acoustics of their fans because its there weakest point. i wouldnt mind spending a bit more per fan if the sound profile wasnt just straight terrible which personally theirs are to me at the moment

I have a 12U 19 inch rack built into my computer desk, and I have a couple of NASes in it (2x HPC-8316SA-55RB1). The 40mm fans in the included CRPS PSUs are loud, whiny, and rattly, all at the same time.

I replaced all 4 of them with Noctua NF-A4x20s, wired to run at full speed all the time. They still report their speed so the IPMI management interface doesn't consider the power supply fan to have failed, but the PSU can no longer control the fan's speed.

The PSUs don't run any hotter and I can't hear them now.

I have a used Eaton PW9130 UPS in the bottom of the rack. The 80mm (exhaust) and 60mm (inverter heatsink) fans were likewise louder than I'd like. I replaced them with Noctuas too, again wired to run at full speed all the time, and the UPS' Web/SNMP card confirms it's still no hotter than 30'C internally. I can't hear that now either.

Hilariously, the most critical fan, the original inverter heatsink fan, is a 2-pin fan, so it probably can't even detect when it has failed (unless it's detecting fan failure by monitoring current consumption). The original rear exhaust fan uses a locked rotor sensor rather than a tachometer, which required a bit of bodging to convince the UPS that it has not failed. Oh well.

I’ve also done a noctua fan replacement on my ups. My worry is that they are rated for lower airflow than the original fans they replaced. Have you checked whether it stays cool when running on battery?

It's a permanent double conversion UPS; it's always inverting. You could activate the "high efficiency" bypass mode to directly connect the input to the output in the presence of mains AC input, but this would also pass through disturbances like fluctuations, harmonics, and surges, so I don't have it enabled. This wastes about 80W in my setup but whatever. I'm not worried about the inverter temperature is the point I'm trying to make; but I was considering this when I did it, which is why they're wired to run at full speed all the time.

Hilarious to read this on "hacker news" - replacing a standard fan is too dangerous! This is not styropyro.

PSU capacitors can kill you, even after being disconnected from mains power.

I think perhaps you might have not read the entire article? =)

The danger isn't so much in the fan, but in that the fan is INSIDE OF A COMPUTER PSU. There are mains AC voltages (220V, or 110V) here, and even if unplugged (which is should be) there are also capacitors in there, which you should definitely be cautious of.

I've worked in DC (datacenters) before - and I've seen people accidentally drop screws into power supplies...and well, electrical arc, boom, you can guess the rest. And in a domestic situation, a 4" cooling fan (yes, I know, larger) stopped suddenly due to a motor issue, and send flying bits of plastic shrapnel around (always wear eye protection!).

This isn't quite like tinkering with your little Arduino board, or Raspberry Pi.

If you did read the article =(.... I think that's a bit arrogant and disingenuous to make fun of people saying you should be cautious around things that are connected to AC mains, or that involve capacitors.

I appreciate that most folks talking about power on HN are talking about computers, but I decided to not become an electrician in an earlier life.

Taking standard precautions doesn't mean you suddenly shy away from doing basic maintenance. You can use iso alcohol to clean things even if its flammable, just don't use an open flame at the same time.

> I've seen people accidentally drop screws into power supplies...and well, electric boom, you can guess the rest.

Are we talking like, bad electrical shock, hospitalization, death, etc?

Reply on news.ycombinator.com