Anadtech has a nice db chart on today's large hds

[philipms]

http://www.anandtech.com/storage/showdo ... i=3061&p=5

Anadtech has a nice db chart on today's large hds (750gb & 1tb) in their review of the..
Western Digital SE16 750GB: "Quiet" a Performer"

People here wanting a quiet big drive might want to check out their review.
Cheers,
Philip

[MikeC]

The measurements are taken at a distance of 5mm from the rear and front of the drive being tested in order to minimize surrounding environmental noise.

Not very useful; it's way too close to ensure accuracy and freedom from distorting recording artifacts. Even the relative comparisons cannot be reliable; ie, even if a drive measures lower than another w/their test setup, it's no assurance that it actually IS quieter.

[miahallen]

Mike, I find your comment very interesting due to the fact the results seem to be accurate according to what I've read and seen here & elsewhere. Especially because I trust SPCR's testing methodology above all else, so why would the appear to be at least somewhat accurate (comparatively)?

BTW - did SPCR get out of the business of writing reviews?

[JoeWPgh]

Not very useful; it's way too close to ensure accuracy and freedom from distorting recording artifacts. Even the relative comparisons cannot be reliable; ie, even if a drive measures lower than another w/their test setup, it's no assurance that it actually IS quieter.

I'm not sure I understand this. Not that I'm challenging your assessment, but wouldn't the choice of microphone have an enormous bearing this close?

If the setups were all the same, what recording artifacts would shift the results from one drive to another?
Aside from the Hitachi placing so well, the results seemed to show the usual suspects at the top of the list and once again, pretty close to each other.

I'm not going to run out any buy a Hitachi based on this review, but it's piqued my interest.

[fresh]

BTW - did SPCR get out of the business of writing reviews?

I was wondering the same thing. I remember how I used to read at least once a week reports about new coolers or any other component for that matter. Now, I only turn to silentpcreview when I have to see the usability of older components, whereas new, I found reviews of those on a random google search sites.... Pitty. I hope you get back on track with testing upcoming components.

Even though results of HDD acoustic testing isn't as precise as here on SPCR, it's good to see results being similar to those on SPCR.

[jaganath]

a 1 terabyte drive measuring 26.8dBA at 5mm? that would be a good result... from 1 meter!

[fresh]

Hitachi's new terrabyte HDD has been mentioned as the quitest drive on many reviews. Later on they discovered that it might be due to IDLE state, where its spinning slows down. I cannot confirm that, it's just what I heard. Beside that drive, samsung HD501LJ prooved the quietest among other drives, which seems right

Well results at least seem similar from the view of comparison, whereas numbers are, I agree, pretty low considering it's measured 5mm away from HDD. That happens when you don't test inside an acoustic laboratory.

[tonyw]

BTW - did SPCR get out of the business of writing reviews?

I was wondering the same thing. I remember how I used to read at least once a week reports about new coolers or any other component for that matter. Now, I only turn to silentpcreview when I have to see the usability of older components, whereas new, I found reviews of those on a random google search sites.... Pitty. I hope you get back on track with testing upcoming components.

Even though results of HDD acoustic testing isn't as precise as here on SPCR, it's good to see results being similar to those on SPCR.

I'll second (or third) this request of seeing more reviews on SPCR.

[Das_Saunamies]

I'm sure more reviews can be arranged... now if you gentlemen would kindly ship SPCR the free samples and arrange for paid time off for the reviewers so we can get this show on the road.

[Max Slowik]

I don't trust Anandtech's noise-testing methods.

http://www.anandtech.com/casecoolingpsu ... i=3022&p=7

If the dBA bottoms out 38, which is approximately someone humming, then their office is way too loud, their equipment amateurish, or both. Given their tendency to run falsified "leaks" and break NDA on to of that, I just don't trust them in general.

Also, measuring sound at under a foot? What are they thinking? I'll tell you:

Whether or not they know that their benchmarking process is flawed, they want to have the largest and most comprehensive lists of compared hardware. Rather than change their methods and re-review old hardware (or just break clean and start empty) they hang on to their worthless, but numerous results.

[DaveLessnau]

Anandtech is one of my regular reads. But, I'm beginning to wonder about them. Recently, they've made several weird decisions about how they test that render their results invalid. For instance, 1) they test their power supplies inside a sealed case with no thought for heat evacuation, and 2) they test their heatsinks in a computer without a case fan. I'm not smart enough to know if their disk-drive sound measurements are valid. But, in general, it looks like they make measurements that might seem technically valid at the component level even though the results are meaningless (or even counter-productive) at the system level.

[nwavguy]

I'm not defending Anandtech in general, but I think some are being a bit unfair in judging their drive noise testing methodology. As an engineer who has done some acoustic testing work, I can say there are trade-offs with any acoustic measurement technique. There are many benefits to the "close mic" technique they're using. In many ways, it puts the drives on the most level playing field for comparison even if it doesn't necessarily represent the noise outside the PC case.

The more additional components (i.e. drive mounting method, case, etc.) you include in the measurement, and the more background noise included, the less repeatable the results will be. In other words, the quietest drive in the test case may well be one of the noisier drives in YOUR case. By measuring the noise right next to the drive, you remove these other variables and essentially measure JUST the drive noise.

I can also point out Anandtech's results closely parallel the anecdotal reports and reviews of subjective noise from a variety of sources. Most people agree the Samsung's are the quietest followed by Western Digital and/or Hitachi. So this helps confirm Anandtech's testing is valid for a variety of mounting schemes, etc.

I would also like to see more current reviews being performed on SPCR. PC components tend to change rapidly and a manufacture's replacement for a previous component can often be far noisier (as was the case a while back for Seagate) or far quieter (i.e. Western Digital) than their earlier versions. So outdated noise tests are basically useless as the models tested are often no longer available and the replacements are often very different when it comes to noise. It was sad to see Seagate go backwards on the noise front.

So I welcome reviews like the recent Anandtech review. At least it includes some of the latest drives and uses a valid objective measurement technique. It also may help give some of the noisier manufactures some incentive to make quieter drives. Some similar reviews here on SPCR would only help things further.

[whiic]

"The more additional components (i.e. drive mounting method, case, etc.) you include in the measurement, and the more background noise included, the less repeatable the results will be. In other words, the quietest drive in the test case may well be one of the noisier drives in YOUR case. By measuring the noise right next to the drive, you remove these other variables and essentially measure JUST the drive noise."

Yes, but that's the point: Anandtech uses a system that have many noisy components in it, while SPCR uses a system that is primarily intended to be quiet. Anandtech measures HDD noise from a very short distance because they have to. If they had a silent testing system they could measure from a longer distance.

Measuring up close isn't a solution to better evaluation of HDD noise, it's just a bad workaround of the problem. (Without the workaround, all HDDs would appear the same, because CPU fan, PSU fan, etc. would completely mask all HDD noise.)

There's two real solutions that would eliminate majority of background noise:
- silent test system (expensive)
- passively cooled stand-alone PSU for just the HDD (cheap).

If I wanted to compare HDD idle noise, I would use my 230V to 12V (2A) + 5V (2A) (single molex connector) power supply (and use molex to SATA power connector adapter if required). Something like that would probably cost around 10 bucks. (I got mine with a cheap 20 eur USB enclosure. Yes, the enclosure really uses molex connector, meaning hot-plugging the connector to the USB enclosure can toast the drive permanently. Quite a bad design, but hey: I got a passively cooled molex power supply with no fan!)

Of course for seek noise measurements you need a system to connect the drive to. Silent system is good (and expensive) but not necessarily something Anandtech would invest in. So, running the HDD as far away from the computer would decrease the influence of the system fans. This way the system doesn't need to be silent and "decently quiet" might be enough. The HDD would have to be connected to a computer either through some external method (like USB/FW/eSATA) or by using very long internal IDE or SATA cable.

I think Anandtech's measurements are not only low on informational value but they may be in fact misleading, thus we might be better of completely without them. Of course this discussion isn't quite creative. It'd be better to give feedback to Anandtech instead or discuss the matter on Anandtech's board, in hope that they'd improve the noise measuring method. I understand benchmarking HDDs either through USB or through IDE cables that are longer than ATA standard permits can negatively affect performance, but I'm talking about noise measurements only. Performance should be measured in a consistent, different method. Likewise, noise measurements should be consistent with other noise measurements but not necessarily with performance measurements.

I do agree that products change a lot between generations. I'm looking forward for a 1TB round-up here on SPCR. I understand it might take a while as many of the interesting products are still awaiting widescale availability. Also, 1TB drives may remain a niché for a while (due to their price) so reviewing some recent generation 500-giggers could be prioritized. T166, WD5000AAKS and T7K500 await reviewing. Personally, I'm very curious to see how T7K500 would rate as I've read so many positive forum-user's reviews on it, and my personal experience on Hitachis of older generations aren't that positive (noisewise, otherwise there's no problems). But, like nwavguy said, it's an ever-evolving business.

[nwavguy]

Yes, but that's the point: Anandtech uses a system that have many noisy components in it, while SPCR uses a system that is primarily intended to be quiet. Anandtech measures HDD noise from a very short distance because they have to. If they had a silent testing system they could measure from a longer distance.

Measuring up close isn't a solution to better evaluation of HDD noise, it's just a bad workaround of the problem.

We disagree on that point. It's easy to come to the conclusion you did, but even in a quiet system or testing a drive outside of a system, there's a strong argument for the close mic technique. It doesn't matter that you can't compare close microphone measurement to far measurements. You can almost NEVER compare two different sets of acoustical measurements unless they are taken under very strict anechoic conditions with equipment properly calibrated in that environment, etc. Such a test setup is cost prohibitive for either SPCR or Anandtech.

What's important is simply this: How do the drives tested compare to EACH OTHER in a variety of systems? To answer that question, the most practical solution is to take the measurements with a close mic technique to eliminate as many other variables as possible. You're only looking for a relative measurement, not an absolute measurement (which, as I pointed out above, would be very difficult to meaningfully compare to any other absolute measurement).

Sound pressure levels drop off exponentially as the square of the distance. So when you start to get very close to the device being measured, you will automatically exclude virtually all other sources. The other noises in the Anandtech test system should not be a factor in how they test. That's apparent because they did measure a rather wide spread among the different drives.

When you measure from further away, even if the drive is the only thing making noise in the system, you're as much measuring how the drive is mounted, how the case/mounting assembly is damped (or not), etc. rather than the actual drive noise. For example, one case/mounting set up may be well damped at a particular drive's dominant noise frequencies, and even if that drive is noisy in many systems, it would measure well from a distance in the test set up giving a very misleading result.

Also, as you move away from the drive, the measurement is subjected to reflections and standing waves. Standing waves cause cancellation at some frequencies and make for highly innacurate measurements. If you throw one rock in a pond you get a nice set of clean ripples in the water. If you throw two rocks in, the ripples collide with one another and some are cancelled out and never reach the edge of the pond (i.e. the microphone). That's much like what happens with reflected sound waves as you move away from a noise source. The only solution is to test in a free field anechoic environment which isn't practical for what we're discussing here. So the next best solution is to measure very close to the source (where the rock hits the water) so the reflections and other sources of noise don't influence your result.

So, again, I think close mic measurements are the most practical solution, and have the fewest negative side effects for comparing drive noise among different drives that will be used in a thousands of different systems and configurations not just the test system.

[MikeC]

I really wasn't going to say anything here, but I feel compelled to do so. First, I almost completely disagree with nwavguy about the merits of close mic SPL measurement. The only good things about them are...

1) you can use a poor SPL meter in an inadequate, noisy setting
2) it's cheap

The problems with Anandtech's methods are much greater than this, however. The HDDs are not even isloated from other noise sources which appear to be far louder than the HDDs being measured. That's totally illogical and careless if you're seeking any kind of accuracy in acoustical measurements.

The results are not useful even to compare noise between HDDs, never mind absolute loudness values. Why? Because of the compression of data due to boundary effects. From 3' and greater distances, the frequency response of a microphone is close to constant. However, the response changes as we get closer to the mic. This is the "proximity effect" that causes a bass boost in the frequency response of the microphone. This boost changes very rapidly in the last few inches before the sound source touches the mic. For these reasons it is very hard to accurately determine this effect. In serious testing, complex calculations can be made to compensate for proximity effects based on specific mic characteristic, the exact distance, and other factors.

It's well known that the fundamental frequency of the noise of a 7200rpm HDD noise is at 120Hz. There is no question that the proximity effect will boost this resonance dramatically for each and every HDD measured. In some cases, it will probably far exceed the noise produced by the drive at all other frequencies, due to the boost. So whether quieter or noisier, most of the measurements will tend to fall in a relatively tight range, dictated mostly by the boost in the bass. But HDD airborne noise is most not due to the 120Hz fundamental; rather it's the mix of harmonics, tonal and broadband noise through the midband and higher that tends to give each drive its distinctive acoustic signature. Jam the mic right up to the HDDs, and the bass boost basically drowns out all the other differences.

I repeat myself: even if a drive measures lower than another w/their test setup, it's no assurance that it actually IS quieter. Furthermore, the relative differences between them are likey to be greater than their measurements show. (Aside -- I have no doubt that their 4-platter Hitachi 1tb drive measured lower with their test than the 500gb Samsung, but I seriously question whether the reading is representative of reality as you and I would hear it.)

In contrast, SPCR....

1) uses a room where the ambient is lower than that of the HDD being measured, usually by a few dBA.
2) uses a mic that can measure down to those levels
3) isolates the drive so that no other noises can interfere with those readings. (we also mechanically decouple the drive so that it's more or less in free air; the vibration is assessed separately so that users can determine just how much benefit damping will make. )
4) makes the SPL measurements at the correct 1m distance with the HDD and mic well away from room boundaries so that no proximity effects can occur.

.........

Finally, as for frequency of reviews, yes, it's been a slow couple of months, & there have been many reasons. SPCR is certainly not abandoning hw reviews, but there's been serious attrition in staff, and it's very difficult to find people who are knowledgeable enough, conscientious enough, methodical enough, good enough writers and can access the SPCR lab easily to be reviewers for SPCR. The buildup of resources, specialized tools, and consistent methodology for reviews makes SPCR unique in many ways, but unlike most other sites where reviewers are scattered across the globe working in a virtual community, we can't do this. People have to come to the SPCR lab to do the testing in order to keep consistent with previous, well-establish testing routines. Naturally, we have to pull from a much smaller talent pool than most other hw sites, who can draw pretty much across the globe.

If you or anyone you know in Vancouver is a good candidate to be a reviewer for SPCR, let him/her and me know!

[JoeWPgh]

Thanks MikeC. I hadn't thought about the bass boost. Makes perfect sense and you explained it very clearly. Thanks again.

[whiic]

nwavguy: "Sound pressure levels drop off exponentially as the square of the distance."

Here's where your theory collapses. Sound pressure level has an exponential relation when the source of the sound and point of measurement are point-like object (like singulaties of a black hole). In general, the relation applies into practice with normal sound-emitting objects and microphones, when distance is great enough (i.e compared to the distance between the source and the mic, the objects emitting and measuring sound waves are small in relation).

When you measure sound pressure emitted by a HDD that has 3.5 inch form factor at a distance of 3mm, the HDD cannot be simplified as a pointlike object but more like a wall. Distance to nearest atom belonging to the HDD is 3mm away but the corners of the HDD vibrate as well, emitting sound further away (approximately 90mm if the point of measurement is over the center of top cover, 3mm above it). Sound travels only at 300 meters per second so it will take 0.3 milliseconds longer to reach the mic from edges of the HDD than from the closes point. If half of the duration of the pulse matches this 0.3 ms (i.e duration is 0.6 ms), and thus frequency around 1700 Hz, the sound waves from center of the HDD and corner cancel each other out perfectly.

But that's just from the center and the very tips of the HDD and there's sound emitted from all around the HDD top cover. To make matters even worse, the microphone ain't pointlike object either...

I think this is the boundary effect: when you can't assume the source and microphone as a pointlike objects, there's differences in time required for soundwaves to travel, and higher frequencies are distorted (while low frequency continues to amplify no matter how close you put the mic). This no doubt is closely related to properties of the microphone but might also be extremely sensitive to moving the microphone. If you place the microphone 1 meter away or 0.999 meter away (1 mm displacement) you can't notice the difference, but if you place the mic 3mm and 2mm away from the surface, there will be one, no doubt. Not only would it affect the sound pressure but due to the proximity effect, it would also affect noise spectrum.

Thus, my opinion on close mic technique stands. My opinion might not be correct but based on my limited understanding it feels justified.

MikeC: "with the "A" weighting curve, which applies a further big boost to lower frequencies to compensate for human hearing deficiency at those freqs."

Wouldn't A-weighting try to simulate human hearing, thus ignore frequencies that humans are deficient in hearing, instead of "boosting" the importance of frequencies that are difficult to hear?

MikeC: "I have no doubt that their 4-platter Hitachi 1tb drive measured lower with their test than the 500gb Samsung"

7K1000 has 5 platters. Anyway, it's not that hard to believe, especially if lower frequencies are amplified in relation to higher ones. Samsungs have above average vibration and might appear noisy when measured at 3mm.

[MikeC]

MikeC: "with the "A" weighting curve, which applies a further big boost to lower frequencies to compensate for human hearing deficiency at those freqs."

Wouldn't A-weighting try to simulate human hearing, thus ignore frequencies that humans are deficient in hearing, instead of "boosting" the importance of frequencies that are difficult to hear?

Yes, you're right. I was obviously thinking inversely... The incorrect comment is removed.

I think your comments actually do describe the actual mechanics of the boundary effect.

[nwavguy]

OK... this is getting way more complicated than I intended but acoustic measurements are complex. For those not versed in the engineering principals, common sense doesn't always lead to the correct answer. There are many sources of errors including such things as other noise sources, standing waves, reflections, boundary effects, the frequency response and limitations of the measuring equipment, etc.

I'm well aware of boundary effects and they are indeed a source of error. But I would argue they're a less significant source of error than the other effects I mentioned in my previous post. As I said originally, it's a matter of choosing a practical measurement method with the fewest side effects. No measurement method is perfect.

As an engineer, I've done a lot of loudspeaker testing for high-end audio applications. Unless one has access to an anechoic chamber, it's common practice to use nearfield (close mic) techniques to measure the low frequency output up to about about 500 - 2000hz. The frequency response of the test microphones is well understood in these close mic applications and the microphone calibrators we used actually fit over the end of the microphone positioning the sound source within a few millimeters of the microphone element. See:

http://www.claudionegro.com/swacoustic/ ... field.html

If you try to measure the output and frequency response of speaker from a distance of anywhere from a few inches to several few feet away, you're more measuring the effects of the room and the standing waves at the microphone location. You get VERY unpredictable and relatively useless test results. Just moving the microphone a few inches produces a radically different frequency response curve. I realize a hard drive is not a speaker, but we're interested in similar frequencies and the same acoustic effects of room interaction apply.

One thing I was not clear on was that SPCR measures with the drive suspended in free air. I was under the impression some were advocating testing the drive mounted in a quiet system being used much as it would be in real life--that is generally a bad idea for testing. If the effects of the system chassis are effectively minimized, however, that removes a significant source of error.

I also agree it MAY be worthwhile to minimize the noise from everything else in the test environment but only if that noise significantly contributes to the measurements being made. With a good close mic technique, it likely does not. If it did, it would create a "noise floor" and no drive would test below that level. For this application, it doesn't matter what the exact exponential nature is of dB measured versus the microphone distance (i.e. point sources, etc.). Typically as you get very close to a source, the source will dominate the measurements even if there are other louder sources nearby. It's an easy thing to verify. You just unplug the drive and see what the rest of the system measures. If its at least several dB below your quietest drive measurement, it's a minimal source of error. I would expect Anandtech did just that but I don't know?

We may just have to agree to disagree here. I have no affiliation with Anandtech but, at least on the surface, their technique seems to be reasonably valid. And the fact their measurements coincide well with many other (usually subjective) reports only adds to the validity of their test results. If you go to a site like Newegg.com you'll find the user comments about drive noise roughly coincide with Anandtech's findings. Yes, SPCR or others could certainly improve on Anandtech's testing, but given the lack of other current reports, it would be a mistake to dismiss one of the few current objective drive noise tests just because they test differently than this site does.

[Tephras]

I read an article at StorageReview.com comparing two of their own testbeds. What's interesting in the context of this discussion is the page where they talk about environmental measurements, it is of interest because they measured the idle noise of some drives from a distance of 18mm (testbed3) and 3mm (testbed4) and compared the results:



The Deskstar 7K400, for instance, is noisier than Caviar WD2500KS when measured from 18mm but it is the other way around when they measure from 3mm. There is a noise difference between Caviar WD3200JD and the MaxLineIII when measured from 18mm but not when measured from 3mm.

[nwavguy]

Both the 3mm and 18mm Storage Review test distances are less than one inch and would generally be considered "close mic" or nearfield measurements relative to what's being discussed here. As the link in my earlier post explains, the closer you get the microphone in nearfield measurements, the higher the frequency you can accurately measure with minimal interference from the test surroundings and standing waves.

I suspect the differences noted might be partly due to high frequencies being less accurately measured at 18mm and partly due to where the noise emanates from in the drive. For example, seek noise likely comes from a somewhat different area than the spindle (idle) noise and this may vary from manufacture to manufacture. This is one of the negative trade-offs of close measurements and may require some experimentation to find the best microphone location. A more ideal solution would be to take several close mic measurements for each drive and average the results.

In reality, any cost effective measurement technique involves some significant compromises. I don't think it's reasonable to dismiss the Anandtech test results just because they used a close mic technique (as does Storage Review). In a perfect world we'd have several independent reviewers conducting carefully controlled standardized anechoic testing. But in reality, we have to take what we can get.

[continuum]

IIRC, SR is trying to find the point of highest noise on the drive and using that measurement. Not sure about AT.

[whiic]

"Both the 3mm and 18mm Storage Review test distances are less than one inch and would generally be considered "close mic" or nearfield measurements relative to what's being discussed here. As the link in my earlier post explains, the closer you get the microphone in nearfield measurements, the higher the frequency you can accurately measure with minimal interference from the test surroundings and standing waves."

Yeah, as noise to be measured get closer, the ambient noise is less significant in proportion to measured noise. But especially the higher frequencies are distorted by the surface of the sound emitting object you are measuring. You seem to be obsessed about eliminating the influence of ambient noise (and ambient reflections, and ambient blahblah, ambient...) alone. Don't you care about the quality of the measurement from the sound emitted by the macroscopic, non pointlike object to be measured? Don't you really care about the fact that while you make ambient noise less significant by using close mic, you distort higher frequencies as measured eminating from the source? If you measure from a distance like 1 feet or 1 meter, the distortion to higher frequencies is less significant because the distance travelled from each point is almost the same, difference being much smaller than half of the wavelength human is able to hear.

For example at the high end of human hearing (20 kHz) the wavelenght is 1/20000*(300 m/s) = 0.015 m. And when you measure noise from 1 meter, the sound from center of top cover travels exactly 1 meter, while the sound from the corners travel sqrt(1^2+0.09^2) meters = 1.00404 meters. Difference is 0.00404 meters, being much smaller than the wavelenght of 20 kHz whine => no significant distortion even at the very edge of human hearing. (Of course frequency responce of the mic will affect measurement, but if there was distortion to merging of sound waves that were in different phase that distortion would be added to inaccuracy of the mic, making bad even worse.)

And do remember that from 1 cm, the distances to the object's center is 0.01 meters and sqrt(0.01^2+0.09^2)=0.090554 meters, the difference is about 0.08 meters, which is more than 5 times the wavelength of 20 kHz whine and 10 times the half of wavelength! With measurement point 1 cm from the surface, 1900 Hz will be the first frequency to be completely canceled out. Also frequencies 5600 Hz, 9400 Hz will be canceled out. Some distortion would be noticeable probably at even as low as 1000 Hz and the only non-distorted frequencies would be 0 Hz, 3750, 7500, 11250 and so on. (With 1 meter measuring distance the first frequency to be completely canceled out would be 37 kHz, well beyond human hearing.)

You might probably say that measuring even closer will reduce the effect of ambient noise, and while that is technically true, it has nothing to do with what I described. What I described is an unevitable drawback of making close mic recording on a macroscopic, non-pointlike object. Nothing to do with ambient noise, ambient noise reflections, ambient... anything.

What you are suggesting is pretty much against the very principle of SPCR: shout louder, you'll drown out the ambient noise. I know that anechoic chamber is well beyond SPCR's budget, but building an "anechoic" wall isn't necessarily. Like you might have noticed, a room with no furniture echoes badly, but a room with, doesn't. Simply put some unused matresses against the walls during sound measurement would probably handle wast majority of echoes even though it'd fall short of industrial anechoic chambers. It's still far better than using close mic technique. Close mic is probably a priceless tool in diagnostics (like locating a rattling valve or diagnosing bearing condition without opening or even stopping an engine), but I don't think it's a solution to all the problems.

Special microphones to be used in close mic technique do exist and they probably specialize on eliminating sound waves that come from the edges of the object (and in different phase). Using them (like that one in those pictures on the page you linked) probably gives a good measurement... well, at least in range 500 to 2000 Hz, which isn't necessarily enough for all purposes. Also, it's easy to find the spot where to measure the responsiveness of a speaker (right in the center), but where to measure the noise produced by HDD? You can measure from one spot or several but no matter how many spots you use, the noise source could even be a singular spot on the PCB. Unless the close range mic just happens to be over it: it's ignored. But a user of the computer probably don't ignore a whining HDD. You can locate the source of whine, but doing so, you compromise the ability to compare the results and cannot compare the drives to one another. Well, there's not much sense in comparing them against each other if you knew you missed the spot that eminates an annoying whine.

HDDs make noise in numerous ways while you may have accustomed to measuring speakers that can be simplified to produce sound with each cone installed and very little resonant noise from the sides and the back... and those resonances aren't probably published in data sheets so "who cares" attitude would probably take place even here. Well, I would care if my speaker cases start vibrating in some harmonic frequencies at higher volumes. I happen to like my old 70's ASA speakers with massive sound damping wooden structure more than those cheap plastic multicone speakers with no doubt a better frequency response. But I like PA more than HiFi so that explains a lot... and probably makes me a minority here in SPCR forums. I still would like my computer to be silent because:
- I'm not deaf (yet) and
- I don't listen to "music" (that includes noizecore) all the time.

[nwavguy]

Like I said earlier whiic, we may have to agree to disagree. Most of the idle noise coming from a HD is under 500hz while the seek noise is a bit higher but the very high frequency components are very low in level. If I have time someday, I'll take a spectrum measurement from a variety of distances in my lab and post the results.

You don't have to believe me, but unless you're testing in free field in an anechoic chamber, if you try to measure from 1 foot or 1 meter away, you're as much measuring the room, table top, chassis, etc. as you are the hard drive. And there will be standing waves at virtually ALL frequencies causing lots of cancellation effects significantly distorting the measurement. My argument is it's better to at least accurately measure the lower frequencies than not accurately measure ANY of the frequencies.

And nearfield measurements DO NOT require special microphones as you suggest. The exact same mics are used for near and far field techniques. See the link I provided earlier in this thread if you want more information. When the manufacture, or a testing lab, calibrates a microphone they do it by slipping a calibrator over the end of the microphone that places the sound source within a few millimeters of the microphone element. This is as nearfield as you can get. And that same technique is generally used for ALL measurement microphones regardless of the distance they're used at. They calibrate mics that way because it's the easiest way to eliminate all the other effects I've been talking about.

I love silent PCs. I have two inaudible ones under my desk as I type this. I'm as interested as you are in finding the quietest components. That's why I support objective testing of CURRENT drives that I can actually buy such as the recent test by Anandtech. Most of the other published tests--including the ones on this website--are for drives that are no longer available and as such are nearly worthless unless you buy your hard drives at garage sales and flea markets.

[whiic]

"Most of the idle noise coming from a HD is under 500hz while the seek noise is a bit higher but the very high frequency components are very low in level."

But higher frequencies than 500 Hz are more easily heard due to sensitivity of the ear. Also, they may be considered more annoying because the don't resemble natural ambient noise (like wind for example) and may be harder to "tone out" phychologically. For example I have no problem forgetting wind sound or whether it's raining outside or not. Those sounds can be considered phychologically comforting ... at least for some. Whine, pulsating humming, sharp clicking, chirping on the other hand may be harder to ignore. They may be too pure tone (sine wave), too periodic and too predictable. It's very hard to tone out a sound that repeats itself at constant interval... heck, even if you leave the room after listening to it for a long time, you'll probably hear it still. While white noise is difficult to absorb completely, at least it's relatively "natural" sound, like a heavy raining.

"And there will be standing waves at virtually ALL frequencies causing lots of cancellation effects significantly distorting the measurement."

If there was cancellation at all frequencies, you could call that attenuation instead of distortion, which is what you'll face when doing close mic recordings.

"My argument is it's better to at least accurately measure the lower frequencies than not accurately measure ANY of the frequencies."

That's purely a matter of opinion. You say it's better to measure frequencies below 500 Hz only, and completely ignore anything that goes above it (even though human hearing is most sensitive at 1000-3500 Hz), than to measure the whole range of perceivable sound wave frequencies with a not-so-perfect accuracy.

I'd prefer the latter.

"And nearfield measurements DO NOT require special microphones as you suggest. The exact same mics are used for near and far field techniques."

Yeah, they don't require them (and I don't know if such mics are even available). But if no special microphones are used to limit which sound waves may enter and which may not, the frequency range is severely limited due to cancellation. In my example calculation the first complete cancellation would be at 1900. And of course it depends on shape and size of the surface too. 2.5" laptop drives would cause less cancellation but there is differences even between same form factor drives. To completely ignore the surface shape and size, I'd probably consider measurements reliable only on that 500 Hz you happened to mention. And I don't think that's enough. Many old ball-bearing drives would probably be quite tolerable if we could just ignore all that goes beyond 500 Hz. They might even be quieter, because I've heard that many fluid-bearing equipped drives vibrate more than BB counterparts and below 500 Hz... that could be called vibration.

"That's why I support objective testing of CURRENT drives that I can actually buy such as the recent test by Anandtech. Most of the other published tests--including the ones on this website--are for drives that are no longer available and as such are nearly worthless unless you buy your hard drives at garage sales and flea markets."

I agree on that, but to measure objectively doesn't necessarily mean to measure close mic. Everyone else measures close mic so what extra value would it have if SPCR would also measure the same 7K1000 using the same close mic recording technique as Anandtech, StorageReview and all other reviewers did? You suggesting that SPCR should review like everyone else does pisses me off. It's kinda like those housewives who want to inform that [insert your favourite sports car here] is an impractical car and there's no room for children and for groceries. It's as if [insert your favourite sports car here] was intended for housewives... But they just can't keep their mouths shut because they are the majority and they want the minority to obide. It's not like they'd by a [insert your favourite sports car here] even if it had rear seat but they just have to dictate the rules.

[jaganath]

Most of the other published tests--including the ones on this website--are for drives that are no longer available and as such are nearly worthless unless you buy your hard drives at garage sales and flea markets."

This is true for only one or two drives on the Recommended HDD page. If you followed the Silent Storage forum you would know we keep up with the most recent developments in HDDs, and we trust members of the SPCR community to flag up which ones have potential in the quiet stakes. An unenclosed 3.5" HDD is always going to set the noise floor for an otherwise well-constructed PC, and I don't see that changing, platters spinning at 7.2krpm make at least 20dBA.

Most of the idle noise coming from a HD is under 500hz while the seek noise is a bit higher but the very high frequency components are very low in level."

Maybe that's the case on a pure spectral level, but they are certainly the most annoying from a subjective point of view. The soft whoosh of a soft-mounted HDD is nowhere near as distracting as the banshee-like wail of hard drives now thankfully gone EOL.

[nwavguy]

You suggesting that SPCR should review like everyone else does pisses me off.

If you look at my posts, I've NEVER suggested SPCR change their testing methodology. I'd welcome a variety of objective tests. All I've said all along is I think it's unfair to dismiss the Anandtech review as inacurate or worthless when, in fact, it's based on relatively sound measurement techniques--especially when others use a similar methodology and the Andantech results correspond well with the latest subjective reviews.

If there was cancellation at all frequencies, you could call that attenuation instead of distortion, which is what you'll face when doing close mic recordings.

Standing waves DO cause distortion, not just attenuation as you suggest. The cancellation happens only at some frequencies, not all frequencies equally. So, for example, a drive that has an especially dominant acoustic output at a frequency that's canceled by the particular surroundings and microphone position would measure as being much quieter than it really is. The simple fact is this: The closer you get the microphone to the noise source, the less inaccuracies you have due to cancellation and other outside influences. If you don't agree with that fine, but it's a well documented fact.

Acoustic measurements in the real world, as I've said all along, are all about compromises. You have to pick the lesser of the evils if you want the most accurate results. As you suggest, there's some value in using a variety of methods and comparing the results. That's all the more reason to not dismiss the Anandtech numbers.

[nwavguy]

Most of the other published tests--including the ones on this website--are for drives that are no longer available and as such are nearly worthless unless you buy your hard drives at garage sales and flea markets."

This is true for only one or two drives on the Recommended HDD page. If you followed the Silent Storage forum you would know we keep up with the most recent developments in HDDs, and we trust members of the SPCR community to flag up which ones have potential in the quiet stakes.

I was specifically talking about OBJECTIVE measurements--the subject of this thread--not subjective reports from users. Unless I haven't found it, I don't think SPCR has objectively tested (measured) the current crop of drives? Drive models change often enough that it's hard to find multiple objective noise tests of current models. That's all the more reason to not dismiss the Anandtech findings.

As I pointed out earlier, subjective reports without any measurements are easy to find (i.e. on Newegg, here and in many other places). But, unfortunately, few people run out and buy 5 different current drives to compare. They just buy one. So their subjective point of reference is usually the much older drive they're replacing. For example, if I replaced a Seagate 7200.9 with a current Hitachi I'd be thrilled to report the Hitachi was much quieter. But I might be completely unaware a Western Digital would have been an even quieter choice. So, while useful, subjective reports only tell part of the story.

[jaganath]

The simple fact is this: The closer you get the microphone to the noise source, the less inaccuracies you have due to cancellation and other outside influences. If you don't agree with that fine, but it's a well documented fact.

if this is so, then why do all the accepted noise standards (ISO 7779/9295/9296) measure from 0.5m or greater? if what you say is true shouldn't they be measuring arbitrarily close to the noise source?

[nwavguy]

The simple fact is this: The closer you get the microphone to the noise source, the less inaccuracies you have due to cancellation and other outside influences. If you don't agree with that fine, but it's a well documented fact.

if this is so, then why do all the accepted noise standards (ISO 7779/9295/9296) measure from 0.5m or greater? if what you say is true shouldn't they be measuring arbitrarily close to the noise source?

I believe it's because the testing is done in an expensive anechoic chamber that prevents reflections, cancellation, outside sound sources, etc. from influencing the measurment. To my knowledge none of the independent websites measuring hard drive noise, including SPRC, have one. Even the drive manufactures likely contract with someone to make the measurements for them.

If you have such a chamber available, it's better to measure further away so the localized sources of noise in the drive (as I previously mentioned) are averaged out. The chamber prevents reflections that cause cancellation at the measurement microphone so you can safely measure from further away. If you don't have an anechoic chamber, however, a close mic technique will likely provide the most accurate results. If you're curious about anechoic chambers, there's some info here:

http://en.wikipedia.org/wiki/Anechoic_chamber