Yeah, but that's PC folk for you. You see similar behaviour in benchmarking all the time "OMFG! 193021 GruntMarks??? D00dZ, your box is teh SuX0r, I get 193025 GruntMarks with the same setup. You need to tweak, and overclock, and align it with the poles of teh compass, and stuff".

The problem with this aproach is that "inexpensive SPL meters" tend to have minimum measurable noice close to 30db (some even higher) and we are worried about measurements in the 20db vicinity(sp?) =>

ruprag --

That's one issue. I guess bigred is thinking about using these less sensitive SLMs up closer so you can get a reading even with very quiet stuff. But then 2 other issues crop up:

1) proximity effects -- the closer you get the mic to the sound source, the less accurately it reads & the more differences between sound levels get "compressed". No simple way to adjust for this, especially if using the Radio Shack SLM, which reads down only to 50 dBA. This means you really don't want to read any lower than maybe 55 dBA to keep the measurement even readable. It means getting the mic real close. Not good.

2) Sample variance -- the cheaper the SLM, the poorer the QA, and thus the more virance between different units. So you get 55 dBA with one meter; who's to say another sample of that same meter doesn't read 60 dBA? And then which one is more accurate? Even the pro SLM require periodic calibration; they do go out.

No, I am afraid this is not a viable solution for comparative sound analysis.

I do have a solution that has been languishing for lack of attention from me -- involving a high resolution recording system to record all reviewed noise makers using a single standardized method and creating high-enough quality MP3s so people can hear & compare the recordings of noise for themselves. I have all the gear: High accuracy pro test mic, pro mic preamp, 24/96 sound card, good audio software. Need to find the time to get it all working the way I want!

Ah well... Yeah I was thinking about using it up closer so that it could detect the sound. Isn't there some crazy calculations that can be used to sort out proximity effects?
Also I was just using RadioShack as an example. I wouldn't mind spending $100-$200 for an SPL if it could do the job.

I like your idea Mike but as far as I see it would only be useful for comparing the type and "sound" of the noise. But maybe thats all you intend it for?
From person to person there is no way of controlling how loud they play the file. Also someone listening to the files on a SoundBlaster16 soundcard and $25 speakers will hear something different than the person using an M-Audio Audiophile 2496 and $2k speakers.

bigred --

Your concerns are valid. My idea is to record several reference fans at 12V and maybe 7V -- at either 1m (or 1/2 meter if 1m does not work -- for being too low in level). These would be Panaflo 80L, the Nexus 80... maybe a couple others. They would be used to calibrate the level of your playbakc system. Here's how it would work:

1) You must have one of the reference fans.

2) Download the appropriate recording of the reference fan you have.

3) Play it back on ONE speaker, with your reference fan next to it.

4) Set the level of the playback so that your reference fan (at whatever voltage) and the recording of the ref fan sound about the same loudness.

5) Don't touch any of the playback level settings.

6) Listen to recordings of whatever fans, PSUs, HDDs, etc that are in the SPCR database. Relatively speaking, they should have at leas some semblance of the original, and you will get some sense of the REAL loudness level of these components without actuially buying and trying them. Sure, if you have a better sound system, you're hear something closer to the original, but even if it is not great, you'll still get a better idea than NOT hearing the sound files at a calibrated playback level.

That's my project for the fans -- and eventually all noise making components we review. It is an ambitious project.

MikeC's suggestion seems great. Both doable and should give rather accurate results. I guess the sound files could be one of the outcomes of the great overtaking of fan testing that MikeC is doing.

Until we have that the best we can do with very simple things is to compare the noise level with a fairly well known reference. The 80mm L1A @12V or 7V or 5V is what is quite commonly used as reference.

Both versions has the drawback that they are a bit sensitive to the sample of reference fan you are using. Still this is a lot better than no reference at all.

MikeC, your idea appeals to me. However, I foresee a significant problem: truly quiet components will be difficult to audition this way on all but the quietest PCs.

True. I am not sure you you could get around a noisy PC. Maybe you could rip the sound files to CD and try playback on a stereo...

I'm new here and have wondered the same thing. How many here have a Playstation2? Every PS2 probably has the same little cooling fan inside.

When you veterans of SPCR say your PC sitting right beside your desk is a quiet PC (but not a silent PC), would that be less noisy than how a PS2 sounds when it is say, right under the T.V. and your sitting on the couch maybe 6 feet away? I can hear that little fan in my PS2 during moments of in-game silence.

segask -- definitely quieter than that. No one uses a little fan around here -- 80mm is min, generally, and they're spun really slow -- usually under 1000rpm for me.

Rather than trying to determine exactly what is quiet, it could be easier and just as effective to find out what the folk of SPCR interpret as quiet.
Instead of trying to standardise recording equiment, environment etc. why not just concentrate on relative measurements - take a recording with whatever microphone you like, with whatever recording peculiarities, from a set distance from the case. Then take an identical recording with the PC either switched off, or taken somewhere far away (if it happens to be part of the recording equipment)
Do a frequency analysis of the recordings, subtract the former from the latter and have a look at the power spectrum - and there you have the additional sound power added by your system.
Doesn't tell you much about the equipment, but it would be useful for finding out, as big red requested, 'exactly what someone means when they say "really quiet" or whatever.'
Of course it'd be better if SPCR could review every fan on the planet so we can just sit around on our comfy chairs...

nannygoat --

Your "difference" data makes perfect sense if all you are trying to do is "show the noise" in an abstract way. But there is a serious problem with this. Generally people have no idea what "a 20 dBA difference" sounds like -- or what a particular freq curve sounds like. In fact, a 20 dBA deifference is a purely abstract idea, because the subjective experience of hearing a 20 dBA difference can vary tremendously depending on conditions.

80dBA vs 100dBA -- that's a 20dBA difference, right? What do these sound like? Loud and louder.

OK, how about 50dBA vs. 30dBA? For me, loud and quiet. That's very different from the 1st difference, isn't it? But it's still 20dBA.

How about 30dBA vs. 10dBA? It's still 20dBA, but in this case, the difference is HUGE -- one is fairly quiet but plainly audible in a home environment; the other is inaudible. Period. You can't hear it because it is below the ambient noise level.

Even people who work daily with sound measurements don't always have a precise grasp of what something actually sounds like from a measurement or graph.

Sound is NOT what we measure; it is what we hear. By definition, this means what you hear is not the same as what I hear or anyone else hears. In fact, we simply don't know exactly what it sounds like to anyone but ourselves!

This is the whole point of my long term goal of making precise audio recordings of the noises -- so people can hear and judge for themselves! Of course, the point brought up above of noisy playback PCs has thrown a minor wrench into the works...

*chuckles*

Nutball - nice one - you hit the nail on the head with this one. Very well done *chuckles*

Gruntmarks ... it might as well be, might it not? .

*still chuckling*

MikeC - if you'll do these audio recordings, be sure to include some kind of "reference" volume segment, A silly joke, or the SPCR slogan or somesuch - something uniform that'll be included on EVERY measurement.

Why?

Because my sound settings are likely different to yours are different to (pick someone). That way, I can set the volume to what I interpret as "normal conversational level", and THEN compare as to how quiet/loud the audio-sample in question is.

Hope this makes sense? .

*trying to help the cause*

shathal --

You have not read my earlier post -- or not read it thoroughly enough. Please read it again -- the one addressed to Bigred Posted: 11 May 2004 18:16.

This borders on going off-topic, but why not take some recordings of classic problems, if only so that we're all close to the same page? Like a good example of bearing noise vs. motor noise vs Airflow turbulence vs.... etc.

This would focus on the quality of the noise rather than the actual volume, and it would help the rest of us better understand the noise descriptions that you and the other reviewers give.

Yup, that's in the plans, too.

MikeC I've read the post, but its logic somewhat escapes me.

Your concept works, but assumes that if I want to know about fan ABC, I'll go out, buy it, and compare it's volume against your recording.

Whereas my suggestion is of the "unform" volume sample type, so that once you've set the volume right, and it's always the same, you can caompare the different fans against each other - without having to buy them.

Or am I being daft and am missing something (i.e. - the "cannot see the forest for the trees" case)?

shathal -- Nope.

Because all the recording are made the same way at the same level, all you need to do is establish the correct playback level for any ONE fan. You do this by matching the playback level of the recording to the real sound level of the actual fan which you have. Now you have calibrated your playback system to the correct level -- every recording will be at the correct level as long as you don't change the playback level now.

Not sure how else to explain this. Does everyone else get it?

Ahhh - it's the any ONE fan that dropped the penny.

I.e. - Match my Plebby-fan with your plebby-fan at 12V (or whatever, I guess 12V is the easiest), and we're on to the standard that you'll record at.

NOW it makes sense .

My understanding before was "I compare MY evercool to yours, but if I want a Panaflo, I need to buy a Panaflo before comparing that to yours..." (a mistake in my logic) .

Ta.

Yeah I love the idea.

There are only a few potential problems that could pop-up, but yeah I'm excited to benchmark my system against other's out there.
The potential problems I was thinking about are people's computers being louder than the fan/recording (which you've addressed), or variations in the loudness of fans, and the potential problem of getting power all the way up to a speaker.

MikeC - your ida how to bypass the problem with hard to get somethign that measure 10dB noise is very interesthing, however have (as everything) flaws. First flaw is, that your machine have to be fanless, no pump hum, no monitor hum and even the drive has to be into acustic box
THEN and only THEN you can create something referencing

The second thing is, that in case of noise, you have not only to standartize the distance mic/fan, but also the fan holding! Fan sound louder when resonation on desk and much quietier in my hands - and vice versa.

And next problem is the acusting noise background, when we talking about 10dB, then we getting into problems.
Acustic chanber is a goood thing If not this one, then all the references have to be taken using one person room (differences in rooms/furniture could mean a lot) and during short time, at about there hours after midnight (things like that...)

And the last problem finaly mentioned bigred - most ppls have VERY noisy computers, so they can't even hear the fan sound you recorded, when set the volume to the desired "reference" level...

Sorry for bringing up so much problems I would love to get reasonable readings, like my fanless server was capable of true comparing of HDDs noise (because it don't generating any noise at all, not even some electrical hum fmah mentioned on his great fanless AMD XP cooling), however the lack of proper measuring tool make this impossible...
And it just happens that I have Seagate, Maxtor and brother WD drive... eh... Just to get Samsung and ... but...

trodas --

None of the problems you mention are that significant except for the one Bigred brought up: Users' loud computers.

10 dBA: There is no way to get any audio recordings of anything at 10 dBA -- this is below the electronic noise level of ALL recording gear I know of. (Certainly nothing I can afford, for sure...) The mics, the preamps, etc -- they all make at least around this much noise, so there's no point trying. I know that my min noise level will be not much better than ~15 dBA. So for very quiet stuff, I will try to measure from 1/2 meter or 30cm... which will increase the record level up 3~4 dBA -- this might be enough to bring super quiet stuff above the electronic noise. I have to experiment more with recording really quiet stuff and the resultant sound files before deciding. But 10 dBA is not an audio recording target that is reachable for anyone.

I may end up with 2 recording distances: 1 meter for fans at 12V and 50 or 30cm for fans at 7V or lower.

Why?

Because the most important aspect of fan noise at reduced voltage is not noise level per se, but the quality of the noise: clicking, ticking, swishing, chattering, etc etc. It's more important at the low speed levels to hear the quality of the noise than to get an accurate level comparison to the 12V level. Everyone knows reducing voltage makes the fan MUCH quieter... and all the fans can still be compared at the 7V level.

Obviously these details still have to be worked out -- with experimentation.

Background acoustic noise... I now have a 20x10x8' room that is as quiet as ~12 dBA. I have successfully measured 14~15 dBA devices in this room at 1 meter. Whether I can record sound files at this level still remains to be seen. I have fanless humless PSUs that can drive components and the recording system may actually be in an adjacent room so that its noise is not an issue at all.

Fan mount is simple, I already have a standard soft mount & it is easy to duplicate at home when trying to calibrate your playback level.

Vibration testing is part of this project BTW. I have not got around to it yet, but will be working on a vibrometer device and a standard hard mounting system for fans in order to measure and rfeport of vibration from the fans.

What about using some form of spectrum analyzer software on the recordings? Something like Baudline which is a free Linux program designed to (among other things) analyze inputs from sound card mikes? This might allow better characterizations of noises and their qualities in addtion to getting some really good readings on sound levels at different frequencies. That might also be a useful factor, because I suspect that many of us might find a low pitched 30dB hum more tolerable than a high pitched 30dB whine...

I'll admit I haven't tried playing with this program myself, but it looks very powerful, and as if the only real performance limiter is the quality of the sound input recording. It might also be a way to approach BigRed's basic idea of creating a common measurment platform - if one recorded a known reference device, then the test unit under the same defined conditions it should be possible to scale both recordings so that comparisons would be possible.

Gooserider

Gooserider, you're obviously not reading my posts very thoroughly either. I don't understand why people are having so much trouble understanding my proposed project. It's really quite simple: It does not matter how accurate the recordings are if you don't play it back at the original loudness level! You will get the quality of the noise but NOT an accurate loudness level unless your playback system is calibrated.

If you set the volume too high, the difference between softer and louder noises will be exaggerated. If you set it too low, the differences will be compressed.

There is no need for any spectrum analyzer software in recordings. That is only for graphical and abstract analysis. I am recording the sounds so you can hear them for yourselves. I can assure you that the fidelity of the recordings will be very high and very consistently done. They will be high enough quality that with a high quality playback system, you will get very close facimiles of the original sounds -- if played back at the original level. The recordings will all be made in the same room under the same conditions with the same equipment.

So how will you know where to set the volume level? By comparing a recording to the real thing under the same conditions and setting the volume control of your audio system so that the recording is at the same subjective loudness as the real thing.

You can actually try this with the recordings on page 3 of the AcoustiPack Noise Damping Kit, about 2/3 of the way down.

There is a recording of a Panaflo 80L at 12V from 12" away. Also a recording of a Thermaltake Volcano 7+ fan, same conditions. (also a few others...) Pull both files down. Listen to them at the same volume setting --ie, don't touch the volume control as you swithc from one to the other. The difference is easy to hear, and you can also hear the quality of the sound.

But to hear both recordings at the correct level, the volume control must be at the right setting. You need either a Thermaltake Volcano 7+ fan or a Panaflo 80L fan on hand to calibrate your audio system volume level.

Hopefully you have a Panaflo. Hook it up to 12V and place it next to one speaker on something soft (so it does not vibrate). Turn the other speaker off. Now play the Panaflo fan sound file. Adjust the volume control so that the sound from the speaker is at the same level as the fan itself as you listen from 12" away.

Once you have matched the levels, your audio volume level is calibrated -- for all the recordings on that page. (There are several others) When you play any of those files, you will hear the sounds at about the same level as they were when I recorded them.

This assumes decent loudness linearity in the playback system. Which is why I mention that the higher your playback system, the more accurate and realistic the sounds will be, not only by itself but in comparison.

Now does anyone/everyone get it? Please speak up those of you who do get it! I am feeling like maybe this project should be shelved... I mean what's the point if no one understands how to make use of the results....

There is a HUGE difference between Silent (what this site strives for) and Quiet (what I strive for). I have been in an anechoic chamber before and .. my gawd .. the silence was deafening. I, like many, couldn't take more than about 10 minutes in that room. It was too blasted quiet.

I am VERY happy if my systems are 30 dBA C-weighted at 1m or less. I suspect that there are many here who consider those too farking loud. Sure, the spectrum does have an impact on human perception of sound, but specifying the sound pressure level at a specified distance and a specific frequency weighting DOES give a pretty good indication of how loud the system is.

fractal

Silence per se is not what we seek, really; the goal is a computerthat contributes no additional noise to our environment. Big difference there.

Anechoic chambers are weird, aren't they? They don't exist in nature. Pink noise is the gist of "natural peace & quiet". It's what we are used to, it's what most of us need a bit of to feel comfortable.

All this notwithstanding, and putting aside your apparent understanding of dBA (although you can't have dBA and C-weighting at the same time; the A in dBA refers to A-weighting) -- do you understand how to get realistic reproduction of PC component noises using the system I propose?

Would someone please try it & report how it worked for them?

I like this idea.

Once the sound files become available (assuming they do), I more than likely won't bother calibrating my sound levels. I will use the sound files back-to-back to compare two or more products I'm considering purchasing.

For instance, if I can't decide between a Panaflo L1A and an AcoustiFan, I'll listen to them both, and decide for myself which one is more pleasant to my ear (ie. which is is quieter, roughly how much quieter it is, and what the noise quality is like). I'll weigh that against the difference in CFM, and make my decision.

The calibration process is definitely a good idea as well, as it will benefit those with already-very-quiet computers. But for those of us with louder computers, or who just don't care exactly how quiet something is, provided it's the quietest option available, calibration won't be necessary, alleviating the loud-computer problem.

MikeC, I'll put my hand up as one who understands and thinks your project has a great deal of merit, in principle. The devil, of course, lives in the details of implementation. I think the goal of obtaining and delivering accurately reproducible recordings of *very* quiet fans (e.g., Panaflo at 5.75V) poses an exceptional challenge, as you've pointed out. There are other challenges, like the frequency response of the microphone and any non-linearities or noise introduced by the electronics. Alas, there is no control over the speakers/headphones used by the end-user--both of which can be notoriously inaccurate even if "calibrated" to the same loudness/level.

Nonetheless, I support this heroic effort. I think the advantages far outweigh the pitfalls. The basic concept seems quite simple. As Putz mentioned, a side-by-side [i]comparison[\i] of sounds allows for judging both quality and relative perceived loudness--even if "absolute loudness" may not be accurately reproduced.

For the record, on my lab's system, I calibrate a full frequency spectrum in a sound-attenuated booth (not quite an anechoic) to a pre-defined microphone voltage, taking into account the electronic noise floor. It's quite a finicky undertaking with, as MikeC points out, rather expensive gear.