Vulnerability of HDD
External drives: fall
When 1tb external drive or of more capacity of 3,5" form factor falls from the table being switched on, the spindle is often seized. A set of 3-4 aluminum platters 1.6 mm thick has serious inertia moment, and usually there is no corresponding amortization mechanism except thin spongy rubber - enough to make a fly laugh. The edge of the bearing becomes deformed and seizes after fall. Manual wedging often increases plays and vibration up to derangement of positioning on the track, and often the transplantation of entire set is required for donor spindle. Well, if the spindle bents, the transplantation is just inevitable. The concentricity is measured with the help of laser pointer, reflecting on the ceiling - look like mirror galvanometer. However even if the data can be read after these operations, it goes slowly and sadly. In case if the drive is full, I'm pretty sure to have red eyes for a couple of weeks. You can't even imagine how many sectors in 1-2 Tb and how is it to rewrite them in half-manual mode. Hence is the price for the work - from 300-400 $ without spares.
As a source for latterthe following donor is used -a drive of the same model as the patient, coinciding in "Blood group" (smaller structural features). It is taken to pieces which are used once and discarded after dataremoval. Due to the fact that the donor is a consumable material, the customer makes advance payment for it, and disregarding the work outcome (data self-recovery is usually paid on actual basis). The average donor price was at 70-100$ level for pretty long time due to thesearch overheads, but the price has recently increased - after well-known situation on HDDmarket.
Here we should make a bit of a tangent. Donors - a sore point for any company or professional of my specialization. Fully stocked donor fund shall, according to estimates, include 15-20 thousand drives released in the past ten years. Although HDD overall stock listcontains ten times smaller number of models, each of them was assembled at different times at different factories. Hence are the differences in the packaging - boards modification (processor, "spinner" and other chips), mechanics options (heads and commutators are purchased from several manufacturers), heads' maps and firmware versions. It is advisable to keep them all on the shelves waiting for "insurance case" - an order for recovery with a need of a donor. Special fertility goes to WD as our market loser (or course we speak about flooding in Thailand). It is estimated that WD had produced about 3,000 drive modifications for over last 5-6 years! Can you imagine how hard it is to look for donors for it?!
It is clear that such a fund is an expensive and time-consuming thing. Only large companies can afford it, because they have to keep track of all released HDD modifications and buy several items from each batch. Donor drive selection has grown into a separate business long ago. In 2004, one donor had been taken to me via DHL from London, and it had exact specifications I needed-the case was about a laptop model, which was out of production since 2002.
Now such services are starting to appear here. Of course, they are not cheap, so customers should be prepared to show some generosity. An option for economic ones is to wait for indefinite time until repairman will find a donor in the secondary market, or look for it by his own. Smaller firms are trying to establish mutual collaboration, and with this purpose special resources are created in the world web (exchange database and bulletin board).
But let's get back to external drives. People, put them away from the edge of the table! Cables shall not hung for any awkward movement/child/dog/poltergeist not to be able to bring the drive down on the floor. Simply saying, when the hard drive is switched on, deal with it as with boiling kettle. If someone does not know, drive's network cable is made short purposefully (80 cm), and surpluses are even laid in the stand. The standard is adopted by the European Union after several tragic incidents. Meanwhile, widely spread vertical drive arrangement is a mere evil: the support often slips on the table, and besides, it does not always prevent overturn. And it is separate and considerable risk factor. A fall from its own height for modern HDDs of 2 Tb or larger can become fatal.
External Hard Drives on Amazon
2.5" form factor hard drives
These drives are less capricious: their platter's mass and dimensions are significantly lower, external body usually has amortization items, and models with built-in accelerometer have time to park heads already during first 30 cm of vertical "flight". Even in the worst scenario, everything is limited by stuck heads (though, their release shall be carried out by an expert) or "bads" at the contact place, and often the fall goes without any consequences.
External drive models with enhanced mechanical protection are very practical in this matter, so I recommend to use them. For example, some 2.5-inch Transcend drives has been tested for impact resistance in accordance with U.S. military standards. This choice of words is widely used in marketing nowadays, so I need to clarify: it is a document named MIL-STD-810F Method 516.5 Procedure IV, resolving 26 free falls from 122 cm height (4 ft) onto concrete floor, and every face, edge and corner of the device is checked. However, all operations are carried out in switched-off mode. If the drive will stand all these tortures and subsequently will operate normally, then one shall not doubt in its impact resistance properties.
Construction which was certified this way, has three-layer structure of amortization: HDD lies in of foam rubber cover, followed by a chamfered plastic body, distributing impact loads and finally "gumshoe" - resilient silicone external covering. It also provides moisture and wear resistance, as well as increased friction for the drive not to fall down from the table at the slightest cable tug or inclination. A price for better protection is a bit larger size and weight (4-6 mm in each dimension and 20-25 g), and somehow brutal design.
In contrast, many glamour models have a resemblance with a remnant made of slicky plastic. Sometimes it looks nice, but is not practical at all. Such drives are often equipped with velour or leather bags, but it does not add amortization, but simply protects gloss from scratches during transportation. I highly recommend putting the cover off before operation: it gives out warmth a lot. If active positioning of heads is initiated for several hours (for example, to run defrag), you can face serious problems. This is typical especially for high-capacity models like WD, Seagate, Hitachi and Toshiba. Products of Fujitsu and Samsung, according to my own impressions are more heat-resistant.
External drives: interface
Besides low impact resistance, an interface adds problems to external drives (and their users). The vast majority of models have connection with PC via USB 2.0. FireWire remains exotic in our area and eSATA is mainly used in desktops. eSATA is presented only in high-end models of notebooks, but now it is rapidly supplants by USB 3.0. The second USB version is widely spread, even some tablets have it. This thing is quite reliable and proven: the appropriate standard has just celebrated 10 year anniversary lately.
In relation to external drives, USB 2.0 disadvantages appeared long ago: low exchange speed (actually 32-35 MBps), and lack of supply on bus (500 mA according to the standard, but in many laptops the actual figures are 250-300 mA). The latter prevents drive from the run and initialization by the system, so Y-shaped cables with two connectors for USB-ports are manufactured (second one is two-pinned, only on power). This solution is forced and not very convenient, especially given the frequent lack of ports in laptops. Besides, the reliability of USB-interface is not absolute. When moving large data arrays via USB, there are distortions and one uncorrected error on 30-50 GB in average. Most often it occurs in adverse conditions - bad USB-cable, intense inducing (for example, from power cables) and so on. Such volumes are common for usual users who copy media files which are not critical to failures. But transmission errors are dangerous enough for more serious content (so, my experience remembers corporate database "crash" because of this). For this reason, and to save time, experts prefer to connect it directly through SATA / IDE. Physically, the external drive is most often connected to a computer via USB-A -> mini-USB cable. Often mini-USB port built-in the chassis appears to be a weak point: it becomes loose after one or two years of active use. The weak metal is to blame, and it is peculiar for Chinese. From numerous "plug-ins - plug-outs" the slot bandage becomes slightly wider, springing edges weakens, and plug attach becomes unstable. The force for unplugging drops significantly, cable stops "working" or simply fall out of any random load, even under its own weight. Surely it makes routine operation extremely complicated, and slot replacement requires soldering.
Against expectations, there was little improvement with the introduction of a USB 3.0 standard: prolonged micro-USB connector of B type, built in the drive appeared to be as short-lived as previous one. It seems the thing is in metal and soldering economy. Thick and rather rigid USB 3.0 cable (it is really good by its own) transmits big effort to the connector in case of bending. Thin-walled slot bandage become loose of it and after a few months the cable loses tight fit. Moreover, slot mount to the board is not always reliable; there are cases of cold solder and other defects. As a result we get contact tinkling and drive fallout from the system. Loose-fitting connection works worse day after day. One can achieve some stability unless pressed that ill-fated connector with a finger, but it can't be called normal operation. It is risky to take device with dozens of gigabytes of confidential data to warranty repair, and for removing the data securely and cleaning the drive one should open the body and lose warranty. Both options are equally bad way outs.
The optimum solution for external drives is the removable USB-cable "appendix" 10-15 cm long, which lays into a special notch on the chassis. The cable ends with USB connector of A type: it is enough to have it for connecting to a laptop, and stationary systems are serviced by the extender (it is included or sold separately; 60-80 cm section is enough). Such practical designs appear on the market and their number increases - manufactures could hear requests, finally. So that is the example of how one of them (A-Data) describes its model: "Integrated and wrapped-around USB-cable provides convenience and secure portability for any mobile user, protecting you from problems with storage or cable loss"
External drives: supply
As we know, 3.5" form factor hard drives require two voltage types - 5V and 12 V. Therefore, "full-sized" external drives almost always require power supply. The power supply unit can be built-in (more common in multi-disk storages) or external one. In the first case power cable slot is standardized (computer IEC 320 or "eight" usual for radio equipment) and does not cause any problems, but the external power supply can be connected to the drive differently. In my opinion, the most practical option - supply of 12V value to the case, which is converted into 5V by the internal circuitry.
Of course DC-DC converters take some space and give out warmth, but they have reliable two-pin connector. The best scenario is when it is made in the form of a tube, as it is being done in laptops for a long time. This type of connector cannot be connected improperly, it is difficult to break it, and in case of a fault (the contact in the slot sometimes comes off), it is easy to repair or replace it. And the whole power supply unit can be changed, if necessary for cheap equivalents, there are plenty of then on sale, because the characteristics are the most commonly used. Just make sure that the output current is not less than 2A.
Typical external box from ST Lab. Complete usual is available: easy-dismountable aluminum case, two interfaces, correct supply and even a small fan on the far edge. There is also a support for vertical installation. The supply of two nominal requires four pins already and connectors here are less standardized.
Often they are round with four pins, like enlarged PS/2. To me, it is not really good solution: the resistance of the slot to side and pulling efforts are insufficient, and the key is only a narrow "cavity" in the bandage, and even an arrow on the case indicates the correct orientation. Being in a hurry or having bad lighting one can plug the connector inverted by 180Â°. Experience has proven that even trapezoidal arrangement of pins does not prevent you from mistake.
Therefore, I encountered exactly such a case couple of years ago. LaCieBigDisk external drive had "cracked up" on two HDDs located in the same surface (there were such drives before 2 Tb drives era and now the need for such bulky solutions has passed away, if only it is not network storage).
Being in a lather, the user has not looked at the arrow and pushed the connector in wrong way and then with horror on the face he heard cracking attempts of drives to unwind. I did not face it as well and suggested that the internal controller has burnt because of 5V and 12V pins mix up and maybe drives themselves. But luckily not: fortunately, both HDDs were normally recognized at the benchmark computer and the whole case started to work with "supply" from other drive (there were two of them in the department).
After studying the power routing, I realized that the inverted connector short-circuited both voltages to the "ground" and stabilizers has just burned in PSU. Now it produces 6V and 9V voltage, and the drive cannot run because of it. The moral is: always check connectors' orientation, and do it visually if possible, and when you connect, do not put a lot of effort. If something goes tight or crooked, you're mistaken.
I have seen sad consequences of sticking inverted PS/2, IDE, SATA, Molex, USB, FireWire connectors. It seems everything is done for foolproof there: beveled edges of socket connector, asymmetrical contact location, the key in the form of filled socket or L-shaped plate. Nevertheless, our people are tough-skinned - they will break anything. It happens worse than this: I had to repair burned drives after the attempts to plug 4-pinned Molex tried into 40-pin IDE interface (say hello to the processor on the motherboard) or to stick floppy drive power connector to spread on SATA-drive test leads (like, what's wrong - there are four pins here and there...).
Repairmen sometimes see a lot...
Do not use Cut-Paste operations in Windows file explorer, use Copy-Paste only. In other words, necessary folders should not be moved
(followed by automatic, uncontrolled removal from the old location), but only copied. Only after successful completion of the process and, preferably, comparison of folders by at least total volume and number of files, you can manually clean the source drive.
External drives: copying
The process of data transfer from PC to an external storage device or from it to PC requires attention. First of all, do not use Cut-Paste operations in Windows file explorer, use Copy-Paste only. In other words, necessary folders should not be moved (followed by automatic, uncontrolled removal from the old location), but only copied. Only after successful completion of the process and, preferably, comparison of folders by at least total volume and number of files, you can manually clean the source drive. The same is for file managers, even proven FAR or Total Commander. I always try to replace the movement of files with their copying with subsequent manual deletion (the exception is work within one section where movement is equivalent to renaming). The fact is when having high-volume copying (and its speed via USB 2.0 is low, respectively, the process takes much time) various troubles are not excluded - from blinking light and unexpectedly encountered drive defects to OS deadlock or accident USB-cable unplugging by awkward motion. In all of these cases, the external hard drive "falls off", and transfer data is surely terminated. Only file manager knows what is left in the source. Moreover, if the destination disk is damaged because of incident (physically or logically) then you will lose your data at the drop of a hat.
It is not always an easy task to recover files for source disk even if it has just been deleted. As Windows always writes into the system partition, erasing vacant clusters. The external drive can be formatted for FAT32 file system, which is generally hard to recover. Meanwhile, these drives are usually sold having FAT32 on board. The reason for using the outdated system is simple (single file size limit of 4 GB it is something!): only FAT32 is reliably recognized by all user platforms - Windows (from 95 to "SevenÂ»)/Mac/Linux, as well as different network storages. NTFS - closed product of Microsoft, its specifications have not been fully published, and not every third party OS supports it. Moreover, there is no wide "native" support from exFAT, Ext2/3, and other trendy file systems.
So manufacturers do not make themselves act: to insure against users claims ("Your drive is not working!"), they get to 15 years-old file system (released in August 1996 together with Windows 95 OSR2 release). It is fair to say that FAT32 provides higher performance of external drive than NTFS; especially it is seen with flash drives and slow HDDs. The reason is smaller volume of service data requiring update under reading and especially writing of files.
Seagate: handle with care
The history of 7200.11 series resounding failure is well known, and it is, perhaps, worth writing separate article. The blow to the reputation and heavy financial losses in the midst of the global crisis - it is very serious. The corporation has tried to make amends for sins, releasing popular and inexpensive twelfth series in 2009. The market took it with a bang.
And year and a half after the start of sales 7200.12 faced massive return to the repair. Couple of thousands of "bads" appears first and drive disappears from the OS because of serious file system damage. Another problem is that the "fall" occurs rapidly - sometimes just for a few hours. And there may not be any visible symptoms the day before that. Firmware is trying to cope with arising defects to the last breath, and when there is no reserve left, "funeral with dancing" begins. Users often do not notice this moment. Well, the drive is chirring with heads more than usual, so one can not stand it out of the herd. Well, it "brakes" slightly, so one can pay no attention to it. Well, S.M.A.R.T. subsided, so it is somewhat relative on Seagate. There are new attributes we must look at (for example, HighFlyes), but programs-analyzers mostly do not know about it.
I assume problems appear because of off-design load. Twelfth "home" series is the result of total economy and it is assembled from low-end components. Firmware is also simplified - particularly there is no adaptive channel recording settings such as slower positioning in case of heating. At high density recording (platters of 500 GB volume) it reduces the reliability during continuous operation. The drive load where passport characteristics are valid is 2,400 working hours per year, or 200 hours per month. That is 9-10 hours a day 5 days a week, that's a schedule of a typical office. All that is higher - at your own risk.
And these models are used pretty often in 24/7 mode, even for torrents under aggressive settings, filling broad link in both directions, or FreeBSD with writing millions of small files at the same time. And the cooling is as luck would have it - often arranged by ineffective case cooler overgrown with dust. But the drive experiences very high load and obviously off-design one. Definitely here should be enterprise-class drives (EnterpriseStorage family, one and a half to two times more expensive), but who is explain it to the office manager, who buys 50 items for the office at once. By the way, Seagate reduced warranty to two years for the consumer models...
Seagate: control heating
Not everyone knows that in modern Seagate drives SMART attributes give a lot of detailed information about heating. Just a few programs can correctly decipher its code, and HDDScan in this case is a pleasant exception. Here is an extract from its report, from which we can learn not only current but also minimum and maximum temperatures reached.
- AirflowTemperature 190 66 64 34 C 45 190 66 64 23
- AirflowTemperatureMinimum C AirflowTemperatureMaximum 45 190 66 64 34 C 45 194
- HDA Temperature 34 40 34 C 0194
- HDA TemperatureMinimum 34 40 23 C 0194
- HDA TemperatureMaximum 34 40 60 C 0
Here: HDA is HeadandDiskAssembly (in colloquial speech - a can);
AirflowTemperature - heating of lower (system) reading head, it is approximately equal to the air temperature inside the can.
The magnetic resistor of the head is the thermo-resistor as well; so separate sensor is excessive. In other words, both heating of the hard disk drive (this is important for spindle bearings and electronics) and magnetic heads are controlled; durability and efficiency are directly affected by the temperature.
The drive cover and upper head heats up by 5-15 degrees more due to the friction of air on the cover (it is dragged by the platters, so there is a real storm inside with a speed of 20-30 m/s). That's why "by touch" drive test often causes user's anxiety, although according to thermal attributes the heat is not so high. Also the necessity of lid blow-off is obvious or, at least, presence of free space above it (3-5 cm) for passive convection.
Seagate: how to know the date of release
Seagate drives are not dated in a way understandable to everybody - instead there is DateCode on labels of five, rarely four digits. Their format is YYWD or YYWWD. The code is rather tricky and connected to corporate reporting.
The first two digits - YY - means fiscal year, which begins in the Seagate on the first Saturday in July of the previous year. It means all drives released from August to December of 2011 are marked with 2012, and it often makes customers surprised. The following figures - W (1-9) or WW (10-52) - is the week of the financial year. Finally, the last digit - D - is the number of days passed since the beginning of the week, and the week lasts from Saturday to Friday inclusive.
Therefore, in order to calculate the exact release date you will need a calendar. If you do not want to bother, I propose to use online calculator.
There are more cases when recording channel of the hard drive goes bad and the user knows about it only when the drive is no longer recognized at the next startup. The table was incorrectly recorded in the service area (usually it is SMART area) - and bye bye. “Service area” can be fixed within processing method, the drive runs up normally, but there is a few joy in it: the user data is already corrupted. You can rearrange head back and forth, but the data cannot be read - it is practically erased from the platters. Well, it is not actually so, but there is too large percentage of uncorrectable errors. All files (except the smallest ones) are corrupted.
And the most unpleasant thing is that you will not able to catch this bug at the moment of its appearance – there are no control mechanisms. And there may be huge screw up in case of massive file record. 3-4 years ago the reading mode after recording could be met in more expensive models of drives (separate heads, reading head goes along the path after the writing head, and the commutator is able to switch quickly), but in order to simplify the firmware and reduce energy consumption this way was rejected. So the problem source is saving.
Earlier, few years ago, when hammering drive has usually appeared before me, my first action was to look for a donor. In 80% of cases the hammering sound was pointing out to the death of the head or a commutator; so magnetic head assembly interchange helped to retrieve data. Now the procedure has changed: the first step is to remove the cover and inspect platters carefully for the presence of scratches and notches (a notch is a circular scratch). In some cases they are present - and not always they are easily visible to the eye. Alas, a quarter of a millimeter scratch is fatal, and after finding it the work is over. It is useless to exchange donor head with damaged platters - they will be also damaged in a few minutes, not enough time to get any data off the drive. While touching "hillocks" having grown up on the edges of scratches, the head is heated and quickly breaks down.
If the scratch is located at the edge of platter, it is possible to do something with it. Sometimes it is possible to rearrange magnetic head assembly limiter to prevent heads from getting on a dangerous section. The hot-swap technology is implemented as there is service area placed there often, which is read at the start of the drive. The board with the firmware initialized into RAM is rearranged without power off from the donor drive to the patient drive (spindle is stopped programmatically beforehand).
Why is this happening? The fault is on recording density of modern drives. High spatial resolution (EXHAUSTION?!) of recording and reading elements are required. Therefore, the maximum head altitude is lowered and contains up to 10 nm in some models.
Actually it is a pure form of nanotechnology (nano-object is conventionally that object having at least one dimension of less than 100 nm, or 0.1 microns).
In case of any push, say nothing of the hit, the head strikes over the platter, and the consequences are unpredictable. Quite often things go smooth; current platters are covered with a solid lubricant like Teflon reducing friction and mitigating damages. But sometimes worse scenario takes place the scratch appears and start to grow. The process of its "maturation" may take from a couple of seconds to a few hours and user here is usually the passive observer. He begins to act when there is a knock and the situation is already irreversible, and the data is already lost. An immediate power shutdown and expert assistance may help in this situation. But as I explained above there are little chances for recovery.
All these horrors, however, relate to high-speed (7200 rpm) and high capacity 3.5" form-factor drives (750 GB and above). "Green" low-speed family (5400-5900 rpm) and drives of previous generation are much more stable. Laptop models are needed to hit very hard to get scratches, the mass of moving system and linear speed is much less there. Also there is often an accelerometer in the laptop or in the drive itself which drives head into park zone beyond platter limits in case of any speed-up. In case of free fall of switched on drive it is enough for him to fall from 20-30 cm height to park, so if laptop flies down from the table, platters and data are not actually endangered.
Scratches and especially notches do not occur often on 2.5" drives. There's another problem there: in case of frontal hit onto the cover or board, heads may stuck, and thin spindle axis bends in case of side hits with the risk of vibration. It is practically impossible to correct the axis and the data can be saved only by rearranging platter package (with heads) on donor spindle. The procedure is not easy and requires special equipment and considerable experience.
Have you ever lost data on a laptop, PC or RAID?
When it is beyond repair
I was brought two disks to repair and there was no need to recover data. One of them, 500 GB, had side hit - it seems the system unit has just collapsed. Now the bearing "sings" and the disk itself has serious vibration (frequency 120 Hz, the sound is similar to massager). The spindle axis shifted from the hit. Drive is still working, but surface reading is unstable - there are areas of extreme slowdowns. I tried to adjust read/write parameters, expand defect list but everything was in vain. Stable zones diminished, but did not disappear. It is impossible to use this drive; it is predestined to become a donor for spare parts. I bought it from the client for a symbolic amount.
By the way, the hit was recorded in system log file, which is stored in the service area on the platters. This log is often checked by service people, if you try to refund the drive under the warranty. The corporate software from the manufacturer is used here (in some cases to access the log some undocumented commands are required). So I'm going to disclose a secret to you, if you do not like the behavior of the drive and you dream about exchanging it, you should hit switched off drive, and wrapped into a towel in order not to leave external traces.
The second drive had been struck to the bottom with a screwdriver - pretty usual thing in case of careless installation. Normally some details should be removed from the board in this scenario, but here the hit took place to flat ribbon cable going from the board to the engine. It is only a couple of centimeters long, has 4 cores, and two of them were broken. Well I cleared the surrounding area and soldered shunts. It works, but looks not good, and cannot be refunded under warranty.
The client asked to solder a new ribbon cable - I have it, but I can't really solder them to the engine's pins. There's a good heat sink, a small soldering iron will fail, and heat it with greater one is dangerous - hydrodynamic bearing does not like overheating: the oil inside will lose viscosity, and sometimes it may even spilled outside of the can. Shortly saying the principle is "it works - do not touch". And I did not promise to solve warranty issue.