The HDD testing system underwent a major series of upgrades that reflects the growing sophistication of SPCR readers — and our own increasing awareness of hard drive acoustics issues. The acoustic testing tools of careful listening and SPL measurements remain; they are enhanced with a new vibration testing methodology, as well as measurements of power draw in an effort to predict thermals.
May 11, 2005 by Devon Cooke and Mike Chin
Most people who work with computers know that hard drives make noise. So do the hard drive manufacturers, who are usually good about providing at least some data about their acoustics. Often, the data is very comprehensive, detailing noise at idle as well as in seek. However, anyone who has tried to make a truly quiet computer soon discovers that the specified acoustics don’t really predict the noise produced by a hard drive in their own system. We have long been aware of the role of the HDD in PC acoustics, and suggested many ways of subduing their noise, including some commercial add-on products.
Meanwhile, we have been working to create a system of HDD noise assessment that combines traditional “hard” data measurements along with subjective listening and our own unique silencentric methods. This article documents our new HDD testing methodology.
Before you can understand how we assess HDD noise, you need to know something about HDD noise itself, how it is created, what forms it takes and why we often describe it as insidious. The following section about HDD noise is partly a summary of information contained in the article Recommended Hard Drives.
THE NATURE OF HDD NOISE
The noise of a disc drive mounted in a case comes in two forms:
The acoustic noise that a drive produces is a result of mechanical movement
that occurs inside the drive enclosure. There are two sources of acoustic noise inside a drive: The motor and the actuator.
The bulk of the noise produced by a hard drive comes from the electric motor
that spins the discs. This is the noise you hear when the drive is idling.
Except when first turned on and spinning up to speed, the drive is at a constant
speed, so the noise at idle is always present. The primary frequency of the
HDD noise is easy to predict: Simply divide the motor speed by 60. This is
because RPM is the spin rate per minute, while frequency or
Hertz is the rate of cyclical movement per second; there are
60 seconds in a minute.
Primary Frequency of HDD Vibrations / Noise RPM Typical Drive Interface Frequency (Hz) 4,200 Notebook 70 5,400 Notebook, PATA 90 7,200 PATA, SATA 120 10,000 SCSI, SATA 167 15,000 SCSI 250
Keep in mind that HDD noise is never composed of only this fundamental noise,
the first harmonic. Like many sources of noise, there are many harmonics (multiples
of the fundamental frequency) at higher frequencies. There is also broadband
noise, a hissing sound that seems related to the spinning of the discs themselves,
and this noise varies in overall frequency balance and loudness level from
drive to drive. Some appear to have hardly any of this noise at all, while
others exhibit a fair amount. Generally, the higher the number of platters,
the higher the idle broadband noise.
Although it is less constant than motor noise, the seek noise produced by
the actuator is louder and usually more irritating. Its irregular character
tends to draw attention so it can’t be
tuned out easily. Seek noise can be reduced by adjusting the speed that the actuator
moves across the discs, although this acoustic benefit often comes at the
cost of drive latency. Many drives provide this feature as an option in the
form of Automatic Acoustic Management (AAM). Generally, the higher the number of platters, the higher the seek noise. Again seek noise varies quite a lot, with each drive exhibiting a unique combination of sounds.
These are the reasons for the unique acoustic signatures of different HDDs of the same speed and number of platters; they have the same fundamental tone, but each has a unique set of harmonics and broadband noise, affected not only by design aspects such as internal damping and bearing type but also production variances including balance. Add the variance of seek noise and you have quite an acoustic range.
Vibration Induced Noise
a hard drive the way it is designed to be mounted — screwing it directly
to a computer case chassis — conducts the vibration to the case. This is referred to by Seagate, the world’s dominant HDD company, as structure-borne acoustics.
A case that is vibrating due to conduction of mechanical energy from a HDD produces noise by itself. The large thin metal panels of the typical case act as sounding boards which convert the vibrations into airbone noise. Aluminum cases are especially
susceptible — so much so
that the issue gets special mention in our
article on Computer Case Basics and Recommendations. This is because aluminum has about 1/3 the density of steel, and yet the sheet metal used for aluminum cases is no thicker than in typical steel cases.
A vibrating case can also create additional mechanical noise. A loose screw, a badly fitted side panel, or a cable
resting against the side of the case — all these are prone to rattling when the case
The sources of vibration in a hard drive are identical to the sources of
acoustic noise: Once again it is the moving parts, the motor and the actuator.The primary frequency of the HDD vibrations are 70, 90, 120, 167 or 250 Hz, as mentioned before, along with harmonics, broadband noise, and the complexities of the seek vibrations. Add whatever harmonics the case itself might add to this rich mix of sounds, and you can begin to understand that the total noise of a hard drive depends not only on how it is mounted but what kind of case it is mounted into.
The vibration produced by a HDD in idle does not necessarily correlate with its airborne
noise. While some drives are quiet and vibrate very little, there are also many drives with low acoustic noise and high vibration, and
vice versa. Seek vibration, by contrast, usually correlateswith seek acoustic
noise. A drive with loud seeks in free air usually has even louder seek noise when installed
in a case. The AAM feature discussed above is just as effective for reducing
seek vibration as it is for direct acoustic seek noise.
SPCR HDD TESTING GOALS
There are literally dozens of computer hardware review web sites that run performance
benchmarks on hard drives. Some of them do a very credible job. Few measure
noise, and even fewer have the expertise to fully profile the noise produced
by a hard drive. At SPCR, we are interested in testing how a hard drive sounds;
its performance is of secondary concern.
Our basic point of view is that while there are perceivable performance differences between similarly rated drives, they generally tend to be subtle and small, with a 10% difference being considered huge. Noise differences, however, are far more dramatic and obvious. In both measurements and subjective listening, the noise difference between two otherwise similar HDDs can be as large as 100% (10 dBA). This can mean the difference between a computer that’s pleasant to work around and one that’s difficult to tolerate for any length of time.
Our methodology for testing hard drives highlights acoustics. Our benchmarking tools are hardware-based —
an SPL meter, for example — rather than software applications. The metrics
by which we compare drives are airborne and vibration noise, rather than throughput
OUR HDD TEST METHODOLOGY
All acoustic testing is performed in two states: Idle and Seek.
If the drive being tested supports AAM (most do), separate seek tests are run
with AAM enabled and disabled. These two (or three) possible activity states
are reproduced twice: Once for airborne acoustics and once for vibration noise.
SPCR’s hard drive test bench.
For all noise measurements, the hard drive is attached to one
of our open bench test systems. The system employs a quiet notebook drive in
a Smart Drive enclosure for minimal noise, and any fans are stopped temporarily
during testing. Extreme head actuator movement (seek/write) noise is produced
by using the AAM TEST feature of Hitachi’s
HDD Feature Tool. The resulting noise is about the worst a HDD will
ever sound. (NOTE: Hitachi’s downloadable Feature Tool (v1.97)
utility works with most modern hard drives, including SATA interface drives.
It contains an option to enable the Automatic Acoustic Management feature
that reduces the seek noise in exchange for a small decrease in seek performance.)
If the Hitachi Feature Tool does not work with a particular drive, then seek noise is created by defragmenting the drive (if the drive has previously been used),
or by copying a large chunk of data to the drive and copying it repeatedly within
the disc. This operation ensures maximum seek noise by forcing the drive to read and write
different parts of the disc simultaneously.
Assessing Acoustic Noise
Each drive is measured for SPL one meter away from the top of the hard drive.
Hard drive noise tends to be directional, the loudest position being directly
over the top. SPL readings typically drop by 2~3 dBA/1m when measured from
the side of the drive. The drive is placed on a soft foam to ensure that no
vibration noise is produced during testing.
SPCR’S SOUND LEVEL METER
The B&K model 2203 sound level meter is a professional caliber SLM that dates
back ~20 years, weighs over 10 pounds, and is completely analog in design.
It has a dynamic range that spans over 140 dBA. The unit’s absolute sensitivity
reaches below 0 dBA, although its real limit may be the internal noise of the mic and electronics, which are said to be at ~15 dB. A quiet environment is a prerequisite to low noise
testing; the lab has been measured down to ~17 dBA at night, and a 14
dBA adjacent room is also available for any HDDs that are quieter.
Audio recordings of the direct acoustic noise are made using SPCR’s
high resolution digital recording system. All recordings are made
with the microphone centered 7.5 cm (3″) above the top panel of the drive.
The HDD itself is placed on soft damping foam atop a tall stool. This setup
ensures that only the direct airborne noise from the drive is captured in
Audio recordings are made with the omnidirectional mic 7.5 cm above the top center
of the drive.
If you wonder why the recording microphone is positioned just
3″ away when the SPL readings are taken a meter away, it is because some
of the drives emit too little noise for it to be audible over the background
noise of the recording equipment itself. The 3″ recording distance has
been the same for every sound recording posted in MP3 format at SPCR, so it
is a reasonable de facto reference.
In addition to measuring and recording the acoustic noise, we
also evaluate how a drive sounds subjectively. Even though subjective
impressions cannot be reduced to simple numbers, we consider this evaluation
the most important part of a hard drive review. Measuring the volume
of a drive from a particular angle and distance cannot capture the distinct
quality of noise that it produces. Often, the MP3 recordings can give a rough
impression of how a drive sounds, but even these do not fully capture the
complete noise quality of a drive.
Some examples of things we might comment on include differences
in noise when listened to from a different angle, head reset noise (typically
a “chirping” sound that occurs every few minutes), or harmonics
that are especially pronounced. We may also comment if there is an especially
large difference between seek and idle noise, or if AAM is especially effective
or ineffective for a particular drive.
Assessing Vibration Noise
Measuring vibration is a complex task. Since vibration is essentially motion,
a proper measurement of vibration must measure motion and acceleration along
each of the three spatial axes. Additionally, because vibration is oscillating
motion, frequencies must be specified to make the measurement
complete. Our lab is not equipped to make measurements of this
complexity, which require investments in expensive specilized test equipment that will have no other usefulness.
After much thought and experimentation, we realized that it is not necessary
to measure the vibration directly. Instead, we can gauge the effects
of vibration, the structure-borne acoustics. To this end, an
old aluminum electronics project box (43cm x 25cm x 10cm) was pressed into
Aluminum box, open side up on test bench. One of the PC systems used for testing is in the background.
Placing a powered drive on the sound box immediately makes its vibrations
audible. The box acts as a sounding board and resonating chamber for the HDD
vibrations. It exaggerates the vibration-induced noise to make it easier for
us to hear and evaluate. The air inside the box resonates along with the
thin aluminum panels at the primary spin frequency of the drive (as well as
its harmonics). We experimented with placing the HDD at different positions
on the board and found that the greatest noise was produced near one edge.
We applied duct tape on that spot to prevent short circuits and extraneous
mechanical noise, but the box is otherwise unadorned. The box actually has
only five sides; the sixth is a cover that we left off. The box is placed
with its open side down on a somewhat resonant test bench.
Musical drives: SPCR’s highly resonant aluminum sound box, set up to amplify hard drive noise.
It is possible to measure the SPL of a HDD on the box. But this is a measure
of sound, not vibration. Our sound box is an artificial
means of producing sound from HDD vibrations; it does not produce noise that
is directly comparable to how the vibration will sound when the drive is used
in a real situation. Furthermore, any SPL measured in this way would include
the direct acoustic noise of the drive as well as the vibration noise. Thus,
we cannot equate drive vibration with the SPL of our test box.
So, how do we assess vibration? The only way we can: Subjectively, by listening carefully. We may
not be able to report vibration according to standard units of measurement,
but we have many HDDs on hand that we can compare. And, if we can tell which drives vibrate more than
others, we can put them on a scale. This is exactly what we did.
We listened to and compared over a dozen drives, and ranked them on
a 10-point scale. 10 represents no vibration at all; this will probably not
be reached with any drive that has moving parts. The ranks of 1 is
reserved in case a drive reaches new vibration high in future. All
the notebook drives we tested scored either 8 or 9; we expect most notebook
drives to be in this range. The lowest vibration 3.5″ drive received
a 7, and it was barely audible when placed on the sound box.
In practical use, the audible difference in vibration noise between 8 and
10 should be minimal. Most of the time, the drive’s airborne noise will
overpower its vibration noise at this level, but there will be exceptions if
the drive produces very little airborne noise or if it installed in
a particularly resonant case. Although our test setup amplifies the vibration
noise enough to discern between low vibration drives, differences in this
range are unlikely to be audible under ordinary circumstances. Most users
will be happy with anything above 7.
MP3s of vibration noise are recorded 7.5 cm (3″) from the side of the sound box.
We also tried recording our vibration tests. We experimented with many different mic positions before settling
on 3″ from the sound box, centered on the side farthest from
the HDD. At this position,
- the resonant vibration noise is most emphasized, and
- the direct acoustic noise of the drive is more reduced than in other positions, allowing the vibration noise to be heard more clearly.
Our intention was to make these vibration sound recording available for download, but after a couple of weeks of experimentation, we decided against this idea. The reasons are many and complex, but in a nutshell:
These HDD vibration recordings do not
represent the exact acoustic characteristics of the drives; the vibration induced noise is exaggerated and the acoustic noise deemphasized.
These recordings are meant to be compared
against each other. As explained previously these sounds will be
at the same frequencies: 70, 90, 120, 167 or 250 Hz depending on spindle
The problem is that these lower frequencies are not accurately reproduced by most PC audio playback systems. What you hear from these vibration noise recordings will be more seriously affected by the fidelity of your audio system than with the broadband noise recorded and posted at SPCR till now. We ran into trouble confirming our own vibration assessments (based on the live sound) when listening to the vibration recordings. This experience was enough to convince us that the vibration recordings are useless by themselves for those who cannot actually feel the HDD vibrations in their hands and listen to the sound that the HDD induces on the box.
But to satisfy your curiosity and perhaps entertain you a bit, here are a few HDD vibration box recording samples. Note the difference in tone (pitch, frequency) of the 5400 rpm (90 Hz) notebook drive compared to the 7200 rpm (120 Hz) 3.5″ drives.
Vibration Recording Samples
Ambient: This recording was made with no noise sources in the room
HDD Vibration Noise Level 9: The lowest vibration notebook drive
HDD Vibration Noise Level 6: “Average” vibration 7200 rpm 3.5″ HDD
HDD Vibration Noise Level 2: The highest vibration 3.5″ HDD tested
For many complex reasons discussed later in this article, we will not be taking temperature measurements of the drives. However, power consumption testing is done. The power drawn by a HDD essentially determines the amount of heat it must dissipate. The exact temperature seen at various points on a drive will depend on its mechanical design, the materials used in the drive, and other factors, especially how it is mounted in a case and where it is positioned. (That is a hint to why temp measurements are not being done.) However, total power draw of the drive is the starting point for any serious thermal considerations about a hard drive.
Standard 3.5″ hard drives are powered from both the 12V and 5V lines. Standard notebook drives draw on only the 5V line. To measure the power drawn by a drive, a simple circuit interrupt was devised. It inserts 0.25 ohms of resistance in the +12V line and 0.2 ohms of resistance in the +5V line. The basic concept is the same as the one used in A $5 DIY Power Meter, but without the complications of AC current; this is DC, to which Ohm’s laws can be simply applied.
HDD current measurement rig.
Current measurement rig at work.
Step 1: Determine the current for each voltage line. The voltage drops across the small resistances are neglible, well under 0.3V for the +12V line and under 0.2V for the + 5V line when a 7200 RPM 3.5″ drive is in idle. The amount of voltage drop allows us to determine the current drawn on that line by using simple Ohm’s law:
I (current) = V (voltage) ÷ R (resistance)
The current is calculated by dividing the voltage drop (measured across the resistance) with the resistance. This formula is used to determine the current draw on each of the two voltage lines.
Step 2: Calculate the power. Once the current is known, then we can apply another simple variant of Ohm’s law:
P (power in Watts) = I (current) x V (voltage)
The current was determined in Step 1. The voltage here is the voltage seen at at the terminals of the HDD for each of the lines. Because of the voltage drop caused by the resistance, the voltage at the HDD is usually a bit lower than 12V or 5V, but for the sake of simplicity, we’re just going to use 12V and 5V. Any error introduced here will be consistent for all the drives we compare, and it will be no more than 0.25W in the worst case.
The original plan was to measure power draw at startup, idle and seek/write. The latter two tests were retained for the final test methodology, but we gave up trying to measure power draw at startup because it is simply too dynamic and dependent on the response time of our multimeters for the results to be reliable. For what it is worth, most 7200 RPM 3.5″ drives showed a spin-up peak power draw of ~20W. None were significantly lower; a couple were slightly higher. This figure has little bearing for heat / noise in normal PC operation and is only really useful if you are trying to perfectly match the power needs of the system with the power supply.
REFERENCE DRIVE RE-TESTED
The drives that have been low noise benchmarks for SPCR were tested using this methodology. These drives are the Seagate Barracuda IV and the
Samsung Spinpoint P80. There are two known versions of the P80 series; one
uses a Nidec branded motor, and the other, a JVC. We tested one sample of
each version. The model number of both versions was SP0802N.
Because the noise characteristics of the Seagate and Samsung drives have already
been well documented, our examination of these drives will be brief. A more
complete review of these two drives was done in May 2003, and the Samsung
Spinpoint was re-reviewed with a sample of a later generation Barracuda line
a year later.
SPCR REFERENCE 3.5″ DRIVES NOISE EVALUATION Drive Model
Mfg date – firmware
(10 = no vibration)
Seagate Barracuda IV
ST340016A – firmware 3.10
Samsung Spinpoint P80 (Nidec motor)
June 04 – firmware TK100-24
Samsung Spinpoint P80 (JVC motor)
Feb 05 – firmware TK200-04
n / a
Seagate Barracuda IV
Even though it was released four years ago, the Barracuda IV still ranks
as the one of the quietest drives we know of. Its high frequency noise is
minimal and it is characterized by a smooth whoosh that sounds
more like airflow than motor noise.
Its seeks are sharp and quite loud relative to the quiet idle noise, but
they measure in line with those of other drives. They are intrusive because
they are so much louder than the idle noise, not because they are loud in
absolute terms. With AAM enabled, the sharp clatter of the seeks is dulled
and sound less intrusive.
Vibration is quite low for a 3.5″ drive, although not the lowest we’ve
seen. We know from experience that this drive benefits considerably
from a suspension mounting to reduce vibration.
Power draw is modest, but anecdotal evidence suggests it runs hot. The extra foam damping on the bottom side of the drive (abandoned since the B-IV) may play a part in trapping a bit more heat than other drives.
Samsung Spinpoint P80, Nidec Motor
The Nidec motor Spinpoint was the version that earned our original recommendation.
Idle noise is rougher and slightly louder than the Barracuda,
but inside a case it is unlikely to be audible. Compared to the Barracuda,
the character of the sound is more metallic, but the overall volume is not
enough to be intrusive. Any fan run at stock voltage would easily drown
out the noise of the hard drive.
Seek noise is softer and more broadband than the sharp seeks of the Barracuda,
although there is a considerable amount of low-frequency rumble. The reduction
in noise with AAM enabled is spread across all frequencies, making the reduction
in noise more a change in volume than a change in character.
We were surprised at the amount of vibration produced by this drive. While
it is not the worst we encountered, it is far from a low-vibration drive.
Our recommendation of this drive in the past has been based on the assumption
that it would be suspended to reduce noise; there are definitely better
choices if suspension is not used.
The power draw is slightly lower than the Seagate Barracuda IV; it may run a touch cooler. This difference does not account for the anecdotal evidence above how much cooler the Samsung runs compared to the B-IV. The aluminum top over of the Samsung may help cool the unit better.
Samsung Spinpoint P80, JVC Motor
The JVC motor version of the Spinpoint has a reputation for having a high-frequency
whine that makes it less suitable for use in a quiet system. However, our
sample (the only one we’ve examined personally) exhibited only a trace of
this whine. Perhaps the difference in motor is more pronounced in the higher
platter models. As with the Nidec version, the idle noise is likely to disappear
inside most cases.
Surprisingly, the seek noise produced by the JVC Spinpoint was markedly
different in character from the Nidec’s seek noise. The character of the
noise is midway between the Barracuda IV and the Nidec Spinpoint. Higher
frequencies are more pronounced than in the Nidec version, but they still
lack the sharpness of the Barracuda seeks. I would characterize the JVC
seek noise as a chatter, the Nidec as a rattle, and the Barracuda as a clatter.
As with the Nidec version, AAM reduced the seek noise, but did not appreciably
change its character.
Vibration for this version was much better than the Nidec version and was
roughly on par with the Barracuda. In spite of the louder seeks of the JVC
Spinpoint, the seek vibration was slightly less than
either of our other reference drives.
The measured power dissipation differences compared with the Nidec version are within the margin of error for the test system and thus insignificant.
Noise Recordings in MP3 Format
contains 10 seconds of idle noise, followed by 10 seconds of seek noise with
AAM enabled and 10 seconds more with AAM disabled. Keep in mind that the audio recordings
paint only part of the acoustic picture; vibration noise is not recorded, and
drives often sound different depending on the angle from which they are heard.
HOW TO LISTEN & COMPARE
These recordings were made
To set the volume to a realistic level (similar to the
Although our methodology provides us with a very thorough means of testing
for drive noise, our methodology by no means tests every characteristic
of hard drives. Among the characteristics we do not test:
We’ve already mentioned why we don’t run performance tests: Generally we don’t believe they are all that significant; many other sites do such testing extensively. Properly testing the performance of a hard drive is a complex and labor-intensive
task. This article
on the methodology for testing HDD performance by StorageReview.com details
the many factors that must be considered when evaluating a hard drive’s performance,
including the intended use, interface, throughput, and seek times. Running performance tests in addition to noise tests simply demands too much of our time and effort. We’re interested in providing acoustic information not available elsewhere.
While designing our test methodology, we seriously considered devising a test
for drive heat dissipation. Ultimately, we decided against including such testing
in our reviews for four reasons:
Ultimately, there are only two ways in which we consider HDD heat to be relevant:
1. Effect on Case Cooling: The amount of heat produced by a hard drive is roughly proportional to the
amount of power it consumes. Power consumption varies very little between drives,
perhaps by four watts at most across all models of a given form factor and spindle
speed. Even the difference between a 2.5″ notebook drive and a 10K RPM 3.5″ drive is only in the range of 10-15W,
a negligible amount when the CPU and the VGA card can each draw >100W.
There is not enough difference between different hard drives
to justify testing power consumption. In practical terms, the total amount of
heat that a hard drive adds to a case varies only minimally between different
2. Effect on Reliability: This is something even certified product testing enterprises
such as UL or CSA have trouble with. It requires a huge number of samples and
extensive testing tools, neither of which are practical for SPCR. In fact, the
best reliability info is probably obtainaible only from large distributors’
historical records after a product has been sold for a while. Everything else
is conjecture and informed guesswork.
In general, a drive’s operational temperature is affected by
spindle spin speed and number of platters: The higher they are, the
hotter the drive will tend to run. Most modern HDDs are rated for safe operation
up to 60°C. This seems a very high temperatrure, and surely, drive reliability
must be adversely affected by running too close to the maximum temperature for
any length of time. We would suggest 45~50°C as a safer maximum temp for end
users who have access to the same temp monitoring tools that we have. If you are more cautious, set a lower maximum HDD temp target, monitor the drive temps, and use manual or thermally controlled active cooling to keep the drive temps down.
TESTING NOISE-REDUCTION PRODUCTS
In addition to testing hard drives, SPCR also reviews products that are designed
to reduce hard drive noise. No matter what technique these products use to reduce
drive noise, our approach to reviewing them is the same: Use the noise reducer with a moderately
noisy drive to conduct our standard
acoustic and vibration tests, then compare the results with and without the noise reducer. In some cases we may
try more than one drive, perhaps choosing one with very high vibration and another
with high noise. The obvious thing we’re looking for is the difference the noise
Most noise reduction products tend to enclose a hard drive or mount them to the case in a different way. Often, the operating temperature of the drive may be affected because of reduced heat conduction to the case compared with standard 4-screw mounting. For this reason, we will also record temperatures of the drives with and without the noise reducing product. The temperatures will be read from S.M.A.R.T. We are interested in the relative change, not the absolute value reported.
* * *