Effects of grill patterns on fan performance/noise (2011) | Hacker News Reader

Effects of grill patterns on fan performance/noise (2011)

pugetsystems.com

360 pointsyread121 comments

Love this:

> Always look at the date when you read a hardware article. Some of the content in this article is most likely out of date, as it was written on September 19, 2011.

Not only do they have a date on the article (and many posts in recent years simply don't), but they draw your attention to it because it's long ago.

It's such a small but important detail. Instantly increases my trust for the company.

I actually came to complain about that warning. Why does it exist? These are just experimental observations. They are valid forever. Many of the thermodynamics texts I needed at university traced their first editions to the 19th century. They didn’t come with weird disclaimers.

It's not weird, and it's not a disclaimer! Many articles I come across online don't have a date on them.

I suspect it's the feature of the website engine. I have seen it on other websites as well - it just appends this disclaimer, after a preset amount of time.

Yes few clicks looking at other articles on that website confirm that they show that warning for all posts older than ~1 year. Makes sense for a blog focusing mostly on latest computer hardware and software benchmarks which can get outdated quit quickly.

While this is broadly true it does ignore some interesting developments in CPU fans. Ten years ago fan sizes were smaller and blade designs less differentiated. Now, there are fans optimized for air flow for air cooled systems and fans optimized for static pressure for cooling radiators in water cooled systems. This opens up a whole bunch of questions which may obsolete these results.

Does the change and differentiation in fan blade design have implications for grill interference noise with regards to this data? Do fans that are optimized for flow versus those for pressure behave the same? Do 80mm, 120mm, and 140mm all have the same grill noise characteristics?

Anecdotal, but even with the new fans, stamped out grills still perform poorly in terms of noise. They're popular because they're cheap.

I'd gladly pay $10 more for a PC case with a less noisy grill, but I can't find any that has this, not even ones that are advertised as "silent".

It's a pretty easy mod if you really want to swap to wire grills. 10 minutes with a dremel and 10 more to make it look ok again and you can have a completely open fan mount.

I did it for honeycomb grill with metal scissors, and replaced to cheap metal wire ring grille. It significantly reduces noise.

Well, for one thing, the once-excellent model of fan they picked aged as well as an avocado green Chevy Nova with a rusted out coffee can muffler:

https://graphicscardhub.com/best-silent-pc-fans/

Weird, I remember Noctua fans being really good ten years ago too. Choosing Antec feels like a deliberately mid-range choice, but I don't remember how they were thought of back then.

That list is terrible. They clearly didn't do any testing and just copied off the spec sheet, and ignored what enthusiasts actually recommend as good fans in 2022 (Corsair fans are terrible, Arctic F12 instead of P12, etc).

You're comparing fundamental laws of physics to consumer product interactions. PC fan design is not static on the quiet end and neither are grill designs though they're less fluid. The interaction between them matters though for the kinds of changes and noise they're trying to measure.

Okay, but also keep proper perspective. 2011 is not some kind of ancient history in the domain of fluid mechanics and fan design.

Reminds me of the time some Rust advocates insisted 2016 was an era before programmers were aware of the concepts of abstraction and separation of concerns:

https://news.ycombinator.com/item?id=19109418

The link seems to be you just constructing that straw man, with no one insisting on those things at all.

I said the problem was that the function calls didn't obey well-understood practices for abstraction.[1]

Responses defend it on the grounds that it was a long time ago, i.e. 2016.

Where's the straw man?

[1] https://news.ycombinator.com/item?id=19109093

Where does anyone insist that "2016 was an era before programmers were aware of the concepts of abstraction and separation of concerns"?

Like I just said, the part where they insist my expectations of that library where too high because it was 2016, and my expectation was that it obeyed proper abstraction.

I don't see anyone insisting on that, either.

>>they insist my expectations of that library where too high

>I don't see anyone insisting on that, either.

Okay, I'm really lost -- if they weren't disputing that my expectations were too high, then their comments ("it was way back in 2016, man!") were not responsive at all. Is that really the pillar you want to lean on?

In any case, let's review:

Steve Klabnik dismissing my criticism because the library is abandoned (as if that's a defense of a 2016 library not having abstraction):

https://news.ycombinator.com/item?id=18943056

User cetra dismissing my points because the library "had not been updated" in a number of years (as if you can't expect a 2016 library to have to proper abstraction until it gets updates):

https://news.ycombinator.com/item?id=18943056

fpgaminer dismissing my points because "Static, perfect code is rare" (as if failure to abstract irrelevant details is okay because no one gets the code perfect with no need to ever change):

https://news.ycombinator.com/item?id=19109529

kaoD insisting it was a dark age of rust, justifying the poor abstraction:

https://news.ycombinator.com/item?id=19110116

Are we looking at the same links?

Yes. I don't see a single one of those comments implying that Rust developers didn't know about proper abstraction in 2016.

The closest might be kaoD, who says:

> I'm sure they were more concerned with making the code work, making it secure, etc.

i.e. proper abstraction was not their priority at that time.

I think you are reading into these comments and finding meaning that doesn't exist.

So you agree they weren’t saying anything that substantively refuted my claim that proper use of abstraction was a reasonable expectation to have of Rust devs in 2016? Then what exactly are you disagreeing with?

Edit: So, yes, I agree that Rust devs knew about a abstraction. When the replies tell me that, they’re not disagreeing with anything I’ve said. My point is that it therefore follows that it was a reasonable expectation to have of the code, that it obeys abstraction. They were giving excuses for why it might not have kept up with eg newly discovered bugs, but not why it would fail to get such basic stuff right … besides, of course, it being the dark ages of 2016.

I don't think linking to that comment favours your point, honestly.

> Rust 1.0 was just released and the ecosystem was mostly maturing at that point. You're talking about version 0.2.36 of a library that had been in development for less than two years during a quite tempestuous time in Rust.

And like I said at the time, that would excuse failure on some edge case. It wouldn't excuse the thing I was actually criticizing, that "the calls for common functions require you to specify low-level implementation details that don't matter to you". That's the kind of thing the typical software dev gets right when designing the function, because abstraction was very well understood in 2016.

It's not something they would only figure out after fixing the thousandth bug.

Linked thread shows you getting schooled. You used a crappy library that died before it ever made it to version 1.0, a mistake that everyone makes at least once.

On topic: This article was written by an SI (system integrator). 2011 is ancient history as far as PC hardware goes. Eg today's fans are a lot more efficient at the same price point and most enthusiasts need a lot more static pressure for radiators.

> 2011 is not some kind of ancient history in the domain of fluid mechanics and fan design.

But it is for most of the other articles this site publishes about consumer computer hardware, like the two that were published surrounding this one ([1], [2]). This article just happens to be an exception.

[1]: https://www.pugetsystems.com/labs/articles/Product-Qualifica...

[2]: https://www.pugetsystems.com/labs/articles/Product-Qualifica...

Is Rust on topic here?

many posts in recent years simply don't

Blame Google. It favors the newest content, regardless of quality.

Lots of valuable information on the web is just not searchable on Google, even with exact phrasing.

Shakespeare is next.

Acoustics consultant here. Its too bad there's no mention of using a windscreen on the sound level meter, especially at 2 inch distance. It would make sense if the mesh performed best, while airflow was restricted nearly most if air rushing by the microphone element was contributing to the level measured, and not the noise of the fan itself.

What a cool and fascinating job you must have.

Hey thanks. Yes, esoteric enough to keep it interesting day to day. I mostly work with architects and design engineers to coordinate quiet building systems and good interior acoustics--projects anywhere from K-12 to major performing arts centers to commercial offices, including some of your US offices...

On that point, I feel everyone's collective pain on the situation with open offices. I don't have the power to avoid them, so the best I can do is advocate for getting the signal-to-noise problem solved right.

Thank you so much for what you do.

> signal-to-noise problem solved right.

One evening, I was sitting alone in the eating area of a very high end office, and noticed a whooshing sound coming from the top edge of the room. I thought it was maybe some air vent, but there was no wind outside. I stuck my phone camera up behind the front lip of a shelf and saw a set of speakers. They were playing something close to brown noise!

And, I learned about sound masking [1]. So, apparently there's two ends to be avoided, in the signal-to-noise problem!

1. https://cambridgesound.com/learn/sound-masking-101/

There are various factors:

1. Partitions around workspaces create a modest barrier effect, not so much realizable for your nearest neighbors but those more distant. No one installs partitions up to 48" or 52" but you have to at least break line of sight to the noisemaker to realize any improvement. This reduces signal.

2. Acoustically absorptive ceilings avoid the overhead reflection that would be the next cue to an occupant. The partition comes first, but this is second. Another signal reducer.

3. Background noise, whether a consistent HVAC system or sound masking system raises the noise floor of the environment. We have a pretty good sense of what level is acceptable to most people, but there will always be those with sensitivities. The noise is usually pink noise with some EQ to sound like HVAC air distribution. Unfortunately there has to be some treble in the noise signal to reduce the consonants of speech, which can be more annoying.

You won't make nearby co-workers inaudible, but the hope is that those 20 ft or so further will be less problematic. For inaudibility you have to get S/N to around -10 dB, that's a noise floor of 10 dB higher than the source, which is only realizable with walls at least to the ceiling.

I adore how diverse the HN audience is.

A couple of decades back I was very involved with work that required paying attention to acoustics.

I remember building this room with walls lined with 1 foot thick Helmholtz resonators tuned to a range of frequencies. The same with portions of the ceiling. And, of course, there were broken-up and angled surfaces to help diffuse sound, avoid creating standing waves and stimulation of room modes. Back then I even wrote a bunch of software to run acoustics analysis and evaluate room characteristics based on a range of parameters, including construction materials, etc.

Here are a couple of screen shots:

https://i.imgur.com/sXQmc8d.png

https://i.imgur.com/q0kHVbQ.png

Since then I have always been very aware of acoustics in every environment. Restaurants are particularly horrendous. Nearly all of them seem to be reverberation chambers designed to destroy sound. Most would benefit greatly by deploying a few very simple tools to control sound and allow patrons to have conversations in a reasonable environment.

They measured the intake side, though, so it was more of a vacuum effect. Would a windscreen still make a difference in that scenario?

I can’t think of why not. The velocity might be lower, but moving air is moving air.

Yes, I agree. Inlet side is quieter than discharge for most fan types (propeller, centrifugal, etc), but noise due to air movement are on both sides.

I would have expected that you'd _want_ to factor in air noise because humans don't only hear fan noise in their computers. Or am I misunderstanding?

You know how someone blowing in your ear makes more sound than when you blow in someone else's ear? The airflow itself can make sound from eddies and whatnot.

Most people don't put their ear next to a case fan to see if it's annoying or not. You experience its sound from several feet away, and the measurement should reflect that.

I think they mean the excess noise of a volume of air moving past your ear, which is due to the turbulence created while air is passing over your ear and not the soundwaves being transmitted by said volume of air.

Practically speaking no you wouldn't want to factor the turbulence generated at the microphone because the user of the computer wouldn't be hearing that.

"Wire" is effectively the air passing around a cylinder.

There is a lot of information about airflow around cylinders. Eg [1]. The top left diagram should be the one that applies with small diameters and the kind of airflow rates in a PC fan.

I would guess that all the other shapes have micro-scale sharp edges that leave vortexes which both slow the flow down and make noise.

[1]: https://i.ytimg.com/vi/pW0JfEBE9h8/maxresdefault.jpg

I agree, it feels like a missed opportunity to not explore this. I'd love to see an analysis of the swirl design (or other commonly used manufacturer patterns), with the edges rounded or beveled on the intake face, exhaust face, or both.

Like, even on cut or stamped grills, how much could performance be improved by e.g. sandblasting the finished piece from one or both sides and taking down the edges a bit?

The wire is only just laminar flow... I found a random PC fan and assumed a 1mm wire diameter, 2.3m/s airflow, and the reynolds number comes out at 36.

That tells me if you were to get a 1000x really powerful fan, then the grille would start giving turbulent flow, and might no longer perform the best.

Page with lots of details: http://labman.phys.utk.edu/phys221core/modules/m8/turbulence....

This is 10 years ago article, coolers tech changed a lot since that.

I even remember, when in mac clones used ultra-high-speed coolers, which now used only in rack devices.

And yes, ultra-high-speed coolers, has much higher speed and make turbulence.

I mean that's the main difference between the wire and all the others, really; the wire is round metal, the others are shapes stamped from flat metal.

That said, my current PC has plastic "angled slats" at the back (outlet) and a honeycomb and mesh (filter) at the front, so it could be quieter. I'll keep it in mind if I ever buy a new one. I believe the wire one was the most common on older PCs (90's / 2000's).

I seem to recall the wire covers being on the IBM XT power supply so I think they are the OG.

And 'stamped from flat metal' has a cross section of a cuboid... And a cuboid is also the shape of a brick... And there is that old expression... "As aerodynamic as a brick...".

Ah yes, just like the space shuttle. ;)

"They took a high-performance business jet, added extra drag, and ran the engines in reverse to simulate a flying manhole cover."

https://youtu.be/pfNQW4jToHE?t=340

The problem is you have very messy flow originating from the fan with a ton of swirl, so it is likely not like the top left diagram and more like the bottom two. It's totally unsteady flow, like in this CFD: https://www.automotivetestingtechnologyinternational.com/wp-...

Wind tunnels use honeycomb flow straighteners to take care of this.

A full blown analysis might use computational aeroacoustics software that can calculate the noise generated from solid geometry in a given flow. This field has advanced considerably since when the article was published due to newer methods like LBM. It'd be beyond the scope of journalism like this, but it is no doubt done by companies with resources like Apple.

I remember back in 2000 ish I worked in R&D of a PC manufacturer and had to check the new PSUs and cases with a 'Test Finger' to make sure all the holes were small enough. The test finger was really expensive IIRC.

I also had to strap a full CRT PC to a table that rotated through various axies to measure the EMI/RFI coming from the complete setup while it was running and every now and then the straps would slip and the whole thing would crash down and make a big bang and a mess. Good times.

We often use stepped "reach gauges" for machine guards used in manufacturing and such: https://www.mcmaster.com/gauges/machine-guard-safety-gauges/ Not too expensive though.

...

Why was the "test finger" so expensive? What was it made of?

Why not just use a $3 wooden dowel rod from Home Depot?

Very likely they just wanted or needed certification-level accuracy. A wooden dowel would be very inaccurate for a bunch of reasons, including that its dimensions will change with humidity.

Test fingers and the like are very expensive because of the serious precision they have to be machined to, and are made of expensive materials to minimize dimensional changes the result from environmental variations and use.

I imagine they're small manufacturing runs as well. How many ISO-certified test fingers do they sell per year? Adds to the cost.

Maybe they just told me that to stop me loosing it all the time!

All that cost savings goes out the window the first time you get sued ;)

But also keep in mind that test fingers are used for more than just seeing if it can poke through a screen.

Way way way back when I worked in the auto supplier industry, we needed to test to make sure the auto-up windows and auto-closing minivan doors would actually not cut your finger/hand/leg off. We always used pencils as a first try. They break a lot easier than a dowel rod of the same diameter. If we were happy with the results, we'd get out the test finger. If it looked a little iffy, we'd suggest that the engineers should use their own fingers if they were really confident in their work.

You can buy small stainless steel cylindrical gauges accurate to 0.0001" for not much more than that.

BSI standard for test finger appears to have been withdraw [0] in favour of EIC standards. You can still get test fingers on Amazon though [1].

[0] https://knowledge.bsigroup.com/products/standard-test-finger...

[1] https://www.amazon.co.uk/GOWE-IEC61032-IEC60529-Probe-Finger...

Going from 36 to 50.2 dBa is not a 41.4% increase, you can't use percentages with a relative scale.

Technically it's a 2530% increase, but that's silly because it doesn't correspond to how people experience sound.

Sorry if this is a nitpick, but by design, sound doubles in perceived loudness (roughly) every 10dB, and dBa is a calibrated version of dB which is intended to correspond to how people experience sound, so that increase is only around a tenth of the figure you gave (although you're absolutely right that it's a lot higher than 41%).

Yes, but the zero point is arbitrary, so a percentage of dBa is meaningless. You can say it's 14.2 dB louder, or 10^(1.42) more energy, neither of which is very meaningful to the average person.

So then wouldn't it basically be 142% louder?

I don't mind being nitpicked when I'm being pedantic. Indeed you can (roughly) approximate perceived loudness that way, but only the absolute difference in dBa is meaningful, looking at a relative difference in dB is meaningless.

Rather than using percentages under the noise readings, it would be much more useful to show the delta.

If the unguarded fan is at 35.5 dBa, and the turbine guard is at 50.2 dBa, the delta is 14.7 dB. This is the same increase in perceived noise as an unguarded 60 dBa fan and a guard that raised the noise level to 74.7 dBa, not a relative difference of 41% and 24.5%.

In that case 36 to 50 is well over twice as loud, right? So while 2500% is wrong, 42% is clearly wrong as well. It's closer to 130% (or 230% if you're not doing additive percentages).

https://www.youtube.com/watch?v=1L2ef1CP-yw

Matthias has an interesting video that is tangentially related talking about the distance of a fan to a window and air flow. While watching I was wondering how this interacts with PC case fans and if they are losing a lot of performance by being directly against the case.

I think grills on plastic bladed fans are an anachronism. Most fan blades used to be metal (some still are) and those can likely hurt you badly if you stick your finger in the blades. Plastic fan blades came later, but by that time people expected fans to come with blade guards so they do, even though they now serve no practical purpose. I can stop my plastic-bladed box fan by flicking my pinky finger between the blades and it doesn't even leave a bruise. But I leave the fan guard on my metal bladed lasko, I think that fan might cut my finger off if I tried the same trick. Those metal fan blades are fairly sharp and have a lot more momentum.

Sharp blades, hot radiator, equipped for cooking.

I think it's always a good practice to exercise caution and respect for mechanical and electrical devices, no matter how low the perceived risk may be. While standards exist for e.g. voltage or current under which the relative risk is reduced, careless operation of even low-energy systems may at best damage the equipment, or at worst result in injury. It also tends to lead to the normalization of dangerous practices, which transfer to dangerous situations.

PC fans are surprisingly dangerous.

So much so, that Dell and HP are forced to put a little triangle "warning!! fan!" sticker next to the fan in a laptop.

Okay, maybe not. In the case of the blower fans, you are more of a danger to the fan than it is to you. Touching the thin blades while operating is often enough to snap some or all of them off the hub.

On normal axial fans, DIY PC market fans are often underpowered, at 0.1-0.2A at 12 volts. This isn't a lot of power, and the rotation speed is low and as you have found, not very dangerous.

However, outside of the DIY PC market, even in any PSU, fans are often rated at 0.3 A, and the average graphics card fan is rated 0.6 A. Once you get into the several watts territory, the fan is capable of operating at 3000+ RPM. On OEM PCs, 80-90mm CPU fans are rated at like 0.9-2.5A. In servers, fans up to 4A are common. These operate at 10,000 RPM or more.

The reason you do not hear about this is because most systems oversize the fan and run it at a low speed to reduce noise, but keep that machine that is choking in dust running at full speed so the factory floor doesn't halt or something. I have seen Dells in these kinds of situations where nobody will clean the PC in 20 years, but it has to, and does, keep trucking. These kinds of fans produce robot vacuum cleaner territory of static pressure, up to many inches of water column at 250 CFM.

These kinds of fans are extremely dangerous. China/AliExpress calls them "high speed violence fans" for a reason. The motor hubs are usually made of steel in higher powered ones, and the stall torque is high. The blades have a swept, sharpened tip made of glass fiber reinforced plastic, and the blade assembly carries several hundred grams to kilograms of inertial force. These fans will mangle fingers. When I was very young and inexperienced, [warning graphic], I caught my finger on the sharp tip of a 90mm fan from a Dell while it was spinning down. In a separate incident, I also managed to touch a Intel stock cooler on full speed, and the blade got caught under my fingernail.

Always use a suitable fan guard or safety equipment when testing cooling fans or working around them. The more powerful fans will take off chunks of your fingers, and even a tame seeming fan can quickly speed up to dangerous power levels without notice when the system controller senses a case open, fan failure, or high ambient situation.

>I can stop my plastic-bladed box fan by flicking my pinky finger between the blades and it doesn't even leave a bruise.

I might be misreading this. Are you saying you can stop a running box fan, powered on, force being applied to the impeller by the electric motor, with your pinky?

Easily. It's a bit painful but does no damage. And once the blade stops the motor gives almost no torque.

With a plastic blade? Yah, you can. The plastic is very light and has very little momentum behind it, despite the speed.

I calculated that a 20" box, 1000rpm fan blade moves at around 50 mph. Those large ones might hit pretty hard but will not cause serious injury.

I think he's talking about a small 8 inch desktop model, going maybe 500rpm - that's only 7mph.

The difference of 5 to 15dbA is quite a big deal, as 10dBA higher values are roughly perceived as "twice as loud".

Interesting that "no grill" is always the most quiet, is there no configuration possible (or has never been explored) where the air flows in such a way that sound waves cancel each other out? I.e., even holes emit noise 1/2 phase changed from the uneven ones? Would be a nice area of research. Perhaps it exists, I'm lazy (some 5 mins of searching gave me nothing).

Not what you describe, but that reminds me. There are panels that 'cancel' (some of) the noise by guiding the air through foam channels. Had never heard of such thing before, personally.

[0]: https://youtube.com/watch?v=tMLIzedVvH8&t=1031 (timestamp included)

I'm aware of some research from my department, though not the lab I work at:

https://www.ingentaconnect.com/content/ince/ncej/2013/000000...

https://www.sciencedirect.com/science/article/abs/pii/S00224...

They place specially shaped "obstructors" in fan ducts which create destructive interference with some of the prominent noise tones.

I'm just speculating here, but the grill is the source of the additional noise, rather than something that attenuates noise already present in the airflow. Therefore, to get two holes to emit sound out of phase, the mechanism by which they generate sound must be synchronized somehow.

One way that such synchronization can occur is by the blades passing the holes, and I suppose that the high noise of the 'turbine' grill is caused or exacerbated by the blades alternately aligning with the holes and the ribs. I recently learned that tire treads are made with a pseudo-randomized block size, as with a same size all around the sounds each makes as it contacts the road would be periodic, producing a siren-like sound with a definite pitch.

The swirl pattern presumably mitigates this effect by having little variation in the overall blade/rib alignment through one revolution. There are also tire treads like this.

The difficulties of using interference in reducing fan noise are that it becomes less effective the whiter the noise is, and that destructive interference somewhere usually creates constructive interference elsewhere else.

I was wondering if a honeycomb laminar flow grid would make it quieter.

For passive noise cancellation you would need to add material that absorbs/dampens sound, or you would create a baffle/muffler of some sort. The only alternative I know of is ANC.

This PC case fan (https://noctua.at/en/noctua_anc_project) is designed for ANC. It generates the inverted signal inside the fan.

Interesting to see the classic wire grill at the top. All PC PSUs used to have them, but I guess cost reduction took over.

At least the slightly upmarket ones still use a wire grill on the intake, where the fan resides.

As a household with two Puget Systems workstations, i can attest that this article is very representative of their attention to detail. The workstation built in 2015ish was their Serenity series (nod to Firefly), designed to be as quiet as possible. It does so amazingly well by using large fans and lots of acoustic filtering. My 2022 gaming machine is bigger than that but still very quiet in almost all circumstances. Great place to have a custom build. Not the cheapest though.

I just realized that my Synology DS218play NAS is so loud because of the fan grill at the back. Too bad I can't easily remove it...

Looks like it's nothing some pliers or a dremel can't handle. Also if it's the one I'm seeing on an image search, it looks like the grill at the back is of the mesh variant - it's plastic, molded in a rounded-off fashion. There's probably some tweaks you can do, if it stays cool enough, lower the fan speed or make it dynamic depending on temperature.

Additionally, replacing the fan itself with a quieter one can also help. Image search for "synology noctua" shows this is a thing (and looks like some of them are just removing the grills altogether).

Not quite sure you'd have to get down and dirty like that. Here is a video of a (seemingly) similar model where a guy simply unscrews the grill without even opening the case:

https://youtube.com/watch?v=lBaeK5ry-aM&t=30s

I put together a silent pc earlier this year and I'll never go back. I had a Dell "new" XPS with 16 cores and 32GB on it, and the fan for the laptop itself runs 80% of the time, and even the docking station for it has a fan that runs pretty often. This thing in contrast never makes a single sound.

Related: "How do different fan mesh patterns affect fan and chassis airflow?"

https://www.silverstonetek.com/en/tech-talk/wh_chessis

The author doesn’t consider % void volume / aka ‘open-ness’ of the grill. (Or if they did, I missed it.)

I’d predict the most open to Have the least resistance, which would give the least noise.

I predict that is a large impact of laminar vs turbulent flow, which is not really an intuitive thing to reason about. But it could be that a more closed but weirdly shaped grill is better than a more open configuration? I'm not an expert, just have some experience with microfluidics, back in the day.

An engineer wishing to design a quieter or more efficient fan may be interested in developing a model in which 'open-ness' is a factor, but as a user, I am primarily interested in firstly creating the required airflow quietly, and secondly efficiently. For my concerns, direct measurements are more valuable than indirect ones.

The author did not compare the grills at specific airflows, but the information is still useful.

This is really cool!

Measuring something as that nobody even thinks about or realizes is there, and finding an actual big difference in performance is always a delightful discovery to me.

If close eyes for article date, I seen at least one defect.

20 years ago where known, that place instead of grille deflector tube (like lens hood), will make cooler significantly quieter, just because tube absorb some sound and vibrations. I don't know, why this variant does not shown in this research.

The point of a grill is to block dust and debris, which presumably I want. Otherwise I would just not use one and my fans would be the quietest and cool most effectively. What is the trade-off between the two metrics sited and effectiveness of each type of grill at its intended purpose?

I think the purpose of the grill on a computer is not so much about dust and smaller debris as it is about (a) protecting user fingers against fan bites, and (b) protecting the fan against bigger pokey things that could damage the fan blades or jam the fan.

hmm, I always assumed there was some other benefit. I'm taking all mine off. Seems pretty easy to not ram my finger in it.

No roaming pets or children I take it :p.

Debris are usually covered by a foam filter in front of the fans (if at all). I’ll usually take that off though and just clean it every once in a while.

No, I comment on posts about pc fan grills at 6am...I'm single :)

That's exactly the kind of thing I do in the morning over coffee in those quiet few minutes before the children are up. Rest assured that you can still continue to comment on obscure tech minutiae at odd hours of the morning.

I have a large fan at home that I remove the front protector from because it gets dirty often and is quieter.

My young nephew shoves his hand it in all the time for fun with zero harm.

"It doesn't affect me, so it cannot be an issue"

On the flip side, a few of my hobbies include devices where the fans are usually exposed - and I've cut myself a handful of times because of it.

It's very size/shape/rpm dependent.

zero harm to him or to the blade?

Big fans are at a relatively low RPM, and household fans might be made with kids sticking their fingers in them in mind. I've got a little noisy desk fan with metal blades in it that I would prefer not to take my chances with. I've also cut myself on my RC Plane's props a few times.

In super small form factor PCs, sometimes you use a fan grille on the inside, to make sure that wires, etc on the interior don't get caught in the fan.

The real secret to a quiet PC is not running the fans until absolutely necessary. My PC is silent 95% of the the time, because of massive heatsinks, good venting, and a lack of silicon sympathy, I can tune the fans to only come on at 90th percentile temps.

The other main source of PC noise nowadays is coil whine. That one is harder to deal with, because it's often down to just luck (and that's because companies apparently don't care enough to solve this problem during production).

I hate coil whine so much! It really is luck of the draw. Even if you RMA you could just get another one thst whines.

I think a more interesting question would be whether the design actually makes any difference or if the airflow just depends on the free/unobstructed area.

Didn’t think it would make such a big difference. I am also surprised the mesh was so quiet, I would have assumed it would have been the loudest.

As someone with three long-haired cats, Mesh is my only option unless I want to take on a monthly task of removing fur from all the fans.

What can explain the case where there's more air moving out the exhaust side than being sucked up the intake side?

Coandă effects might drag air from somewhere else, not just from the intake.

https://en.wikipedia.org/wiki/Coand%C4%83_effect

Edit: I see the tester used an anemometer, which measures just the air moving THROUGH the anemometer.

If the exhaust is smaller than the intake, then air will move at a higher speed through the exhaust. Thus the anemometer will show a higher reading.

It has to come from somewhere, by conservation of mass.

Indeed. I was thinking it's from some other holes of a non-airtight case. But actually, the mass conservation happens through air around the anemometer (i.e. air not getting measured, but still moving).

Évaporation and gas formation of the electrolytic capacitor goop comes to mind :)

The airflow into a fan is in response to the low pressure created by the blades, an so will approach over a wide angle. The outflow, however, is initially strongly biased to a direction perpendicular to the plane of the blades (together with some swirl imparted by the blades.) Because of this, the anemometer shown is likely to intercept a larger fraction of outflow than the inflow.

Maybe the swirl will also have an effect, depending on whether the fan blades and anemometer blades rotate in the same or opposite directions.

Hm, but does this explain why some fans show higher inflow than outflow?

Try this with a desk fan: turn it on and stand a meter away on the outflow side - you will feel the flow. Stand a meter away on the inflow side, and the flow is much weaker and probably undetectable. That's because it is flowing in over a whole hemisphere upstream of the fan, but flowing out in essentially one direction.

Here, as with the anemometer, you are not measuring the total mass flow; what's being measured is the velocity of part of that flow, and you are intercepting more of that flow on the outflow side. If you were to measure and integrate the velocity over two hemispheres centered on the fan and divided by the plane of the blades, with a barrier in that plane to prevent recirculation, and calculate the mass flow from that (you can assume constant density as the pressure change is tiny), you will find that the total outflow downstream equals the total inflow upstream.†

† To actually do this experiment, you will need a sensitive anemometer which measures over a small area and does not unduly interfere with the flow, such as a hot-wire probe type. There may be additional difficulties in averaging turbulence.

https://electricalworkbook.com/hot-wire-anemometer/

Reply on news.ycombinator.com