I'm not sure the point-of-origin problem is completely solved (or solvable.)

The directional microphone takes noise from a small area. For a large object, the object may be larger than the volume to which the microphone is sensitive. (In radio astronomy terms, I'd say the beam of the antenna is smaller than the object.) This means that changes in the pointing of the microphone will affect the recorded noise. (E.g. is it centered on the PSU fan, or on the side of the case?)

Less importantly, there is also a question of where you measure the distance from - centre of the object, or nearest face/corner. Again, this won't be significant for a hard drive or CPU fan, but can be for a complete system.

Excellent stuff.

Just listened to the new samples and while yes, the quantitative difference between P80 and 500KS is not as great as in the original recordings even at 1m I'd still say 500KS is noticably louder than P80. However, having four platters against just one doesn't bode well for comparison. I'd really love to hear two platter seagate and wd drives.

In any case, based on the new recordings I'd position the drives as P80>WD74Raptor>WD500KS. I'm not sure about barraIV as it seem to have the lowest idle noise of all, but it also has annoying high frequency whine, very much like one I own.

BTW that's one expensive microphone. I was thinking about buying one for myself to do my own sound recording, but at $379 msrp that's pretty much out of my range.

Great move guys ! Thanks a lot once again.

Could it be possible to offer an unreduced quality 16-bit (flac) version of the new reference recordings ? I easily understand you have to use mp3 encoding in order to limit internet bandwidth usage (as well as providing a format most people already know of) but it would be nice to have reference files available in uncompressed format as well, because 128kbit/sec constant bitrate mp3 encoding in its nature isn't quite optimized to minimize compression artefacts when encoding noise-type sounds as the codec's focus is on musical-type sounds. Using flac compression level 8, this would give file sizes roughly double of the current mp3 files.

The distance information should be provided in centimeters, not in feet and inches.

The metric system is the one to be used in a scientific context and is pretty established on this planet.
http://en.wikipedia.org/wiki/Metric_system

You have to remember, this is NOISE, not music. We've done extensive critical listening; the quality of the MP3s is indistinguishable from the originals for all practical purpose. They do what they are meant to do perfectly well.

I agree, but I wouldn't put it so bluntly. Though SPCR aspires to a scientific standard of rigour, we are not really a "scientific" website. Metric would be useful for Europeans, who pretty much all use metric; AFAIUI imperial units are still more common in North America?

Only $379? I paid over $700 after tax. Are you sure you've included the K6 module that is needed to supply power?

RE: Metric system: Does is matter how we measure distance so long as we measure the same distance every time? Or are you merely picking at semantic differences? The measurement distances were chosen because they are easy for us to remember and reproduce. It makes absolutely no difference to me what unit is used to count the distance. What matters is that I can easily find the 1m and 1ft markings on my meterstick, but I have to hunt around for 30cm. </grump>

What, if we mention feet & inches we're somehow not scientific?! You mean 1ft is not 30cm? And 1 meter is not 3.3' or 39"? That's pedantic, dull and silly. Who gives a hoot!? We're using the same distance whether you measure it out in feet or meters or li.

Maybe you'd like to curl up with a few pages of conversion factors?
http://www.onlineconversion.com/length_common.htm
http://www.wsdot.wa.gov/reference/metri ... htm#length
http://scphillips.com/units/convfact.html

Officially the UK stopped using imperial units (with the exception of pints) a while ago - there were several cases of people being charged for refusing to sell bananas in kg at one point... However, pretty much every Brit I know will give you their height in feet and inches, and their weight in stone.

I just wanted to state that using the metric system would be beneficial for an international forum since it is so wide spread and at the same time accepted by the scientific community.

Using a measurement system that is used by 95 percent of the worlds population (as is the metric system, see http://ts.nist.gov/ts/htdocs/200/202/lc1136q.htm) would just make the forum easier to read and understand for many readers of this forum.

It depends on how many international readers you have, though.

rofl, no I did not include the module, I know next to nothing about mics so when I googled for prices I thought K6 module was optional... Well, now it's for sure out of my range.

The comments about point-of-origin issues reminded me of some loudspeaker measurements I've seen. A good example can be found in Stereophile's Dunlavy SC-IV/A review. Check out how the measured response changes with distance from a large speaker.

But to simply state that English units, which are widely used and known in the US, Canada & the UK, should simply not be used is draconian & rigid. We've always used 1 meter as our SPL measurement distance, and for our mic position, 3" was used because originally, I had a little piece of wood that length which was used as a template to set the distance. It's 7.5cm -- what's so hard about that?

Also, the new 1' distance is clearly stated to be 30cm w/in the article.

Conclusion: You're nitpicking for no good reason.

BTW, US, UK & Canadian visitors account for ~60% of traffic to SPCR. EU=30%, Aus/Nz ~5%. The rest mostly from Asia.

Metres, Feet! Bring back the cubit.
And while you're at it, can you release the recordings on 8 track cartridge or at the very least a vinyl picture disc with the SPCR logo.

It's not that relevant, actually, because despite our comments about how a system sounds different from different positions, the basic character of a PC doesn't usually change much, and most people will hear a mid-tower PC from the front, and from above -- ie, it's usually placed on the floor under the desk where the user sits.

Also, a speaker system is asked to reproduce a huge range of sounds that vary in intensity, duration, frequency, attack, decay etc. The linearity of the speaker is extremely important, so changes in measured linearity caused by mic positioning are certainly causes for concern, as this can dramatically affect the perceived "goodness" of the speaker.

For SPCR's recordings, all we need is a reasonable approximation of the sound. We're not trying to capture every aspect of a PC's noise -- just its essence so that you can hear what we're writing about.

I'm not as interested in the methods used to record as I am about the fact that SPCR is has revised their system so much.

It's really difficult to get 'accurate' recordings no matter what method you use. I just think about how impossible it is to get my guitar to sound the same through a microphone as it does in real life. There's just no chance.

I like the fact that SPCR supplies recordings of the products that can be compared to one another on a 'level' field. But it's the comparison that matters the most. Without listening to many recordings back to back, it's nearly impossible to make sense of what you're hearing. The recorded sound is never going to be a reproduction of the product, so we can only compare products recorded in the same way and make our own judgements.

This change comes at the expense of the comparability of all of the SPCR recordings to date. Since it's become the new standard, this won't be an issue in a year, but it is now. I really hope that the accuracy has improved considerably. Otherwise I'd wonder if it is necessary.

I'm not complaining. If anything, I'm just astounded that they're concerned this much about improving the 'accuracy' of an inherently innacurate process. If they think it's that much better, I'll trust them. But for now, I will weigh the subjective reviews much more heavily.

Candor,

That was my biggest concern about changing the system, too. But thinking through, although we select products for review that have a good chance of being quiet, probably you'd really want to buy only 10-20% of the products we've reviewed. So the number of products affected is much smaller, and you know product life cycles are so short these days, it's already shrinking naturally.

I guess the bottom line question would be: If a new product recording sounds really nice, better than a recording of an old quiet champ, and we tell you that the new one is good but not as good as the old one, would you believe our words or our recordings? No different than before, right?

It sounds to me that since you already know how limited the recordings are, you probably never gave them that much weight anyway. I'm in wholehearted agreement with you about the recordings -- this is why we describe tham as snapshots. It's not the same as hearing the real sound for yourself; it's an attempt to get you as close to this as we can.

<aside>The closest you could get to hearing what we hear is for you to use the same sound card & headphones as we do. The recordings would have to be done in stereo, using a binaural stereo mic inside a really good dummy head. Then you'd be pretty close. Of course, we'd need to get another mic preamp -- $400 -- (our current one is mono) and invest in a binaural stereo mic -- maybe like Neumann's $8000 model. </aside>

Well, sorta. They are helpful and interesting. Plus, like you said before, they provide insight into the character of the sound. (Essence) And when heard side by side, worthwhile info about volume, which, as you know, is what my previous post was concerning.

I'm just clarifying because I do find the recordings totally worthwhile, if only as further detail to a thoughtful review. And yes of course, your written reviews are naturally the final word on the product. (Bows. Walks backwards, quietly out of the room)

MP3 was designed for compression of music, not noise. Anyways, there is another argument against using MP3: since a company owns a software idea patent on it, software that supports legal playback of it requires permission of that company, again at least in the US.

http://www.gnu.org/philosophy/savingeurope.html

Ogg Vorbis may fit your needs the best, and it sounds better than mp3 at the same file size. So using Ogg Vorbis could reduce your bandwith costs.

Unfortunately, unless you follow the controversies that you mention or are passionately into computer audio, you're unlikely to have a Vorbis decoder installed. Media players with free (as in beer) MP3 decoders are not hard to come by, even in the US ... Winamp, iTunes, etc., and they are a standard that just about everyone can play without special software. I might add that Vorbis is also designed for compressing music, so it has no advantage over MP3 in that regard.

The MP3 samples, which are encoded as 128 kbps mono tracks, are transparent for me when I listen to them side by side with the original PCM files. Because the files are mono, they are effectively the same bitrate as a stereo song encoded at 256 kbps. I believe HydrogenAudio when they say that even the experts have a hard time hearing compression artefacts above 192 kbps using LAME.

Oh, the nitpicking lol. But hey, it lead to improvement before.

Sure, ogg is great. Though I prefer lossless myself, ogg is right up there in my book. I listen to classical music and hear, personally the disadvantages of lower bitrate mp3s...but 128bit mp3 for _noise_ is certainly sufficent. And yeah, I'm a bit of a GNU/Linux fanboy. This post is coming to you from my OS/distro of choice, VectorLinux. But suffice it to say that 99.9% of WinOS users don't even have an ogg codec installed...


Meters/Feet, bah . I'm a chemistry undergraduate, and I can certainly appreciate the metric system...but this is an english forum! Since when was a well known english unit a problem in an english forum? This isn't spcr.de or something.


Anyway, sorry, had to get that out of my system lol. I considered the former recording setup easily the best online, and now you guys have made it better. A big thumbs-up to you guys!

DrCR

________________

Hello all,

A caveat to start. I am fully aware that everything I will be saying will probably be overkill for the application at hand. I fully acknowledge that while I do know some things, my knowledge is limited and hope that this post doesn't come off in the wrong way. I'm merely an insomniac who needs something to do at the moment. My attempt is to be educational, and not a nag. And though it may seem like it, I am not criticizing the current setup in so far as it pertains to what it is being used to accomplish. I would personally do a few things differently, but I think that concept of overkill is going to appear quite often. If any of this comes across in a negative fashion, then please feel free to pull it down and I will adjust as necessary.

I will start by simply following the signal flow.

Sennheisers are fantastic mics. I use a variety of them at school and on the job when I'm with my bosses at the symphony. If you are wanting to measure the noise of that the given pc part makes, then using a super cardioid is a great way to zero in on it. Although, if you are making a recording to compare the sound produced by item and the ambient noise to the room, then an omni-directional mic would be more in order. This will address two of the issues previously mentioned. Having the super cardioid at a greater distance from the sound source does reduce "proximity effect." In other words, directional microphones have an increased low end response as they get nearer to the sound source. Simply put, they get boomy. Having an omni-directional microphone eliminates proximity effect since it theoretically picks up sounds equally in all directions. Having a cardioid microphone means you are rejecting the room reflections that are emanating from behind the microphone, which does have a profound affect on a sound. (Yes, I know I'm knee-deep in overkill).

A binaural Neumann mic would be superb, but your other options would be to set up a two cardioid mics in XY, ORTF, or set up two spaced omni's. XY setup invovles two microphone elements set as close to each other as possible, one pointing left and one pointing right with one microphone element on top of the other. This is called a coincident pair. It will tend to create a wide stereo image which might not be the most accurate for this purpose. An ORTF stereo pair is when the elements of the two microphones are set up on the same horizontal plane at about 6-7 inches apart at about a 110 degree angle or so. The technique was developed by French radio and television stations back in the day. It was designed to model the way humans hear. You do have to be careful about phasing/comb filtering issues (I will spare you that description) which can cause serious misrepresentations about the sound in a room.

One must also be aware of a microphone's frequency response. The me66 has two different frequency responses. The pdf on Sennheiser's site is hard to make out, but it looks like there is a bump in the high end frequency starting at about 5kHz, peaking at about 9kHz, and then rolling off after 10-12kHz. What this means is that sounds residing in the sibilant range (any hisses, the 'whoosh' of air) will be amplified between 3-6dB. To us, this means that it will sound about twice as loud as it really is. There is apparently another frequency response that the me66 can have that has a cut in the high frequencies. If I were approaching it, I would find a mic that possesses as flat a response as possible.

In other situations, cable would matter. This is more of an even geekier side-note, but CAT-5 cable is likely to become the next standard of audio cable and I predict that in less than 10 years XLR cable will no longer be the mainstream. I won't get into all that here, but basically they have transmitted 4 channels of analog audio up to a 1/3 of a mile without any signal degradation on a single CAT-5 run! Try running four XLR cables 1/3 of a mile and watch what inductive and capacitive coupling do to your signal. Anyway, since the cables you are using in this rig are "electrically short," then this doesn't matter.

We use M-Audio products on many of our ProTools HD rigs at school. My experience with them in comparison to other audio interfaces (RME or Apogee) is that the M-Audio stuff tends to sound a little more brittle in nature. But, we are also talking the difference between a $200 sound card and a $500+ sound card. 24-bit and above (although 64-bit is overkill, even for audio folk) is ideal for any and every application with the exception of storage based needs. 192kHz would be the ideal sampling rate as the Nyquist theorem shows that there is a "brick wall" filter at 1/2 the sampling rate. By sampling at 96kHz, we loose all frequencies above 48kHz. But, you say, humans don't hear above 18-22kHz on a good day. Well, we don't "hear" it, but our brain does process it in some way that does affect the harmonic content of a sound and therefore can play a major effect on our aural response. Psychoacoustic studies have shown that humans can detect (not necessarily hear) changes in frequencies up to 80kHz. Again, I realize that this is somewhat superfluous to recording computer noise, but I figure if you weren't interested in audio you would have stopped reading a long time ago. Please oblige my insomnia.

I honestly know nothing about the power amp being used, so I can't really comment on that.

I also do not know the monitoring system being used. Take the same theory discussed about microphones, flip the signal flow in reverse, and you get the same thing for speakers/headphones. Ideally, you want as flat a response as possible. Otherwise you will have an inaccurate representation of the sound. What's interesting is that with this equipment that tends to have a "brighter" response (more high frequency content) you get the image of hearing more noise than is actually in the room. On the other side, we are drawn to this sound because typically we've lost a lot of our high frequency hearing by our early 20's. I have tinnitus at 27 and its no fun. I have to be careful about adding in the frequencies I want to hear because they might actually already be there, I just can't hear them.

When it comes to codecs for posting the sound online, storage-wise, you cannot beat a compressed (lossy) codec. As far as accuracy is concerned, you can throw that out the window. Any codec that is lossy (Ogg, AAC, WMA, MP3) can have compression ratios upwards or exceeding 40:1! To minimize the audio information these codec do the equivalent of taking the bass and treble knobs on your stereo and turning them almost all the way down. If you are ever curious about this, take your favorite classical cd and import it as an mp3 and listen then A/B them. The reason why mp3's don't sound all that different from the radio is because modern music generally tends to have about 4dBs of dynamic range (the difference between the loudest and softest sounds in piece). Whereas classical music can attain a dynamic range of 24dB very easily. So, we may think mp3's aren't all that bad (and they are convenient), but unless you EQ the recording to compensate for that compression, then all of your hard work can't be fully appreciated. Also, the Fletcher-Munsen curve plays a big role in how we hear. What these guys said was basically that our hearing changes at different amplitudes. At low amplitudes, we hear less lows and highs, but retain our hearing in the mids. The reason for this is left to the evolutionists among us. So, not only is the mp3 getting rid of the highs and lows, we are trying to hear low-level sound at low levels. I would recommend that if you are going to use a compressed codec that 320kbps be your choice rather than 128kbps. If you think that it's worth the extra space and download time, then a lossless codec (Wav, FLAC, WavPack, Shorten, Monkey's Audio, OptimFROG) might be a more accurate choice. Ideally, 24-bit/192kHz lossless would be optimal.

Like I said, overkill. But, in case someone wanted a little bit of audio theory, there it is. I won't rant about how mp3's are destroying the world and that ipods are killing everyone's hearing, but please, please watch your listening levels when gaming or anything else. Being kept up at night because your ears ring too loudly is not a happy thing. If you are an avid gamer and like to have your face blown off, please invest in a decibel meter and follow the OSHA standards chart for listening levels. At least until they invent nano-machines to restore all the cilia in our ears, then its all fair game.

Hope this hasn't seemed pompous or heavy-handed, neither of those things were my intention. I hope it contains some relevant information for someone. If any of the other knowledgeable folk see any mistakes, then please point them out. I would be much obliged.

~S

Just one of those facts for fun, quality microphones can easily jump up $10-14,000 no problem. Between my professors and our collection at school I'd say we have about $200,000 worth of microphones. It's a weird feeling when you are holding a pair of microphones in your hands and your realize that you could buy a car for what they would sell for.

souagua --

Thanks for your thoughtful comments.

I think you're right that our Sennheiser mic is a touch bright. I can hear it when I record ambient noise and sometimes conversation. But in the context of the noise we record, it's not really a big deal; there's actually very little content up that high in any of the stuff we record. Except maybe with smaller fans at full speed.... and HDDs have become so much more muted in the time that SPCR's been active.

The fact about mic prices is sobering. I was looking at the "high end" version of the $380 KP-6M T.H.E.mic we originally started off with. The TT-3M has been described as utter neutral and transparent, with zero noise or distortion. Tempting, eh? Its price is $1500.

But when everything is considered, what would help SPCR acoustic measurements and recordings the most is an anechoic chamber. I was trying to raise $$ to build one underground -- under my garage/backyard... but haven't been making much effort in the past year.

souagua,

You've done an excellent job of picking our setup apart. I appreciate the thoughts. I think you summed up my response in your first paragraph: What you suggest is "overkill for the task at hand".

I think I can agree with you on most points, but I'm not sure why you suggest a stereo recording. What is the advantage for a point source?

I should also point out that *our* monitoring system makes very little difference. What matter is the monitoring systems of our readers — and these are highly variable and (most likely) generally poor quality. Every hard drive I do I wonder how many people actually have a system that will reproduce the 120 Hz hum accurately without either accentuating it (through a subwoofer) or attenuating it to the point that it's no longer audible (most computer speakers).

Ultimately, the best we can do is give users a rough idea of what the gear we test sounds like. Accuracy is important, but not as important as you might think because by the time the hardware is installed in an actual system in an environment outside the lab, the sound changes anyway. It's only the major components of the noise that matter in a review — inviting readers to expect that a piece of hardware sounds "exactly like the recording" is misleading no matter how accurate our recording setup is.

Got the building permit? It would be sad to raise all that money and then find out you can't get a permit!

Thank you for your gracious responses

Yeah, overkill. Like I said, I was tired and I couldn't sleep because I was like a kid before xmas because my computer parts were in town, I just had to wait for the fedex store to open!!!

First off, I am completely biased towards stereo recordings. As far as the stereo techniques, ORTF or spaced omnis, would give a more accurate stereo image of the sound in a room. Since we hear binaurally, and mics are a mono source, it would help give you a sense of space in the room. Having a mono source makes it difficult to differentiate the room noise and the component noise. Again, it all depends on the approach: do you want to record the component noise, or the difference between the component and the ambient noise in a room. If you are able to try it, I think you might be surprised at the difference. Although two super- or hyper-cardioid mics would not be optimal for that setup. We'll have to do some recordings in our studio during the student summit using both high-end and low-end mics and see what we end up with.

Concerning the anechoic chamber, for precise test measurements (and coolness factor) it would be exactly what you need. Although, I think the room you have set up is perfect in the sense that it is closer to the room in which our quiet pcs will reside. What I might suggest, after building that nifty floating room, is to test in both. Get the measurements both in and out of the lab since that will provide theoretical and real world test results. This is how every manufacturer of audio gear gets their specs looking great. They only test in the lab and very precisely. But (visiting the audio world for a moment), you get a pair of speakers at an outside venue at 3,000 ft above sea level with 30mph gusts of wind at a cool temperature, then those specs go right out the window. Yet, real world specs are the type of specs audio engineers need to make intelligent choices about the sound system they are putting in.

Personally, I think measuring the type of noise generated from the components and the interaction of the components in the room in addition to chamber testing would be ideal. The reason is because different frequencies need to be dealt with in a different manor. If a component makes a high-pitched noise, then you could build a noise trap that is a multiple of the 1/4 wavelength of the frequency and that should help clear up that hiss or chirp or whatever. If you are dealing with lower frequencies, then sound proofing the corners of the case would help absorb those frequencies since bass waves get trapped in corners. The type of noise will dictate the solution.

Alright, I need to finish reading the manuals for the fifth time and start assembling. Pictures and AAR will follow.