Notes from a repairman: hard drives as you never knew them

By Ilya Seidel

3DNews

14 min

In my first articles, we looked in great detail at solid-state drives - flash drives, memory cards and SSDs. This type of drive came into widespread use recently, only 5-6 years ago, and many users, fortunately, who have not yet encountered any breakdowns, have a rather vague idea of the weak points and precautions. This, by the way, was shown by the responses to the articles.

But the most common, irreplaceable and well-deserved storage devices are, of course, hard drives (aka HDD, aka hard drives). For more than twenty years now - approximately since 1988, when mass production of HDDs began, not a single PC can do without this component. Alas, the most unreliable of all. The only thing worse than a hard drive in this regard is floppy disks, but, fortunately, they are almost out of use. There is hardly any experienced user who has not suffered from HDD failures or failures. Therefore, repair and data recovery from this type of media is an established and respected activity.

I started working on hard drives back in 2002. At that time, Fujitsu drives of the notorious MPG series failed en masse: due to an unsuccessful, overly active flux that corroded the processor on the board, they failed almost all of them. Collapse occurred after 6-9 months of work. The repairmen who were the first to master the technology of “frying” boards and editing service area modules (typical prices are $15-25 per disk) were then on a high horse. Patients were brought to them in batches, and over the summer they could earn enough to buy a car, and in a year – to buy an apartment (these are not stories, I know such people personally).

The same Fujitsu MPG3204AH drive, released in 2001. After repairs, it has worked to this day. Capacity 20 GB, designed by today's standards is very primitive. Now it’s even strange to remember that people suffered with it (the editors apologize for the quality of this and some other illustrations in this material and express the hope that this is compensated by their informativeness)

I followed in their footsteps: I mastered the repair technology, bought the PC-3000 complex, which also runs on the ISA bus and under DOS, gave several advertisements in the press and on the Internet, notified my friends - and things went well. The Fujikis turned out to be a good training base, and they also provided income. The main contingent is students, scientists, doctors, musicians and journalists.

Regularly, once every 2-3 weeks, concerned men called with the same question: “ Do you repair alloy wheels?” I answered: “ I fix it, but only up to 5 inches in size.” All HDDs have a case - a “can” of aluminum casting, and at that time there were still Quantum BigFoot drives with a five-inch form factor. The bewilderment of the interlocutor (what kind of disks are these, for toy cars, or what?) quickly dissipated...

Discs of that generation have long since disappeared from the scene. New times - new songs. The capacity has increased a hundred times (from 10-20 GB to 2-3 TB), new design solutions, interfaces and areas of HDD application have given repairmen a lot of experience and, as usual, posed many problems. I will give my notes about some of them.

From left to right: wheels manufactured in 1993, 2002, 2007 and 2010. The electronic board was constantly reduced in size, and the number of parts on it decreased. All this is in the name of economy: with fierce competition there is no other way to survive. Alas, by the end of 2011 the number of HDD manufacturers seems to be reduced to a minimum

And this, for contrast, is a modern disk board with a SCSI interface. In such products they do not skimp on electronics: the profit from the server segment is already quite good, you can claim a five-year guarantee

⇡ # Repairman and SATA loops

Let's start not with the drives themselves, but with what is connected to them. Regularly, about once a month, I come across defective cables on drives with a Serial ATA interface. This leads to data transfer errors, computer freezes and inability to boot. After replacing the cables with new branded ones, everything disappears. Several years ago, when I was collecting material on handling hard drives, this was not observed, and I noted the reliability of the SATA cable, contrasting it with a parallel “comb”.

Alas, since then the quality of the cables that are included in boxes with motherboards (which are what PC builders usually use) has dropped noticeably: someone once again decided to save money. Don’t feed the Chinese rice - let them simplify the technology somewhere and reduce the cost of products by half. They strive to reduce the cost of those components that you cannot immediately check - the composition of the solder or flux, the cross-section of the wires, the coating of the contacts. It’s obvious that they screwed up the last one: the contacts in the cable are buried and are practically invisible, it costs nothing to install brass lamellas, avoiding the gold plating required by the standard. After six months, brass, of course, oxidizes (the connection is not gas-tight) and the contact is broken. During transmission, data is corrupted with all the ensuing consequences.

It is more difficult to save money on the disk itself: the contact comb is visible there and all computer scientists know what a gold-plated contact looks like (smooth, slightly matte shine). And the control at factories is serious. So they piled on the trains, fortunately few people pay attention to their “branding”. Externally, all the trains are difficult to distinguish, the accessory is massive and cheap, the thought of marriage does not occur to anyone.

Now you have to remember this: electronics is the science of contacts. A competent computer technician should always have a new branded cable (or preferably several, of different lengths) in stock. In case of incomprehensible “glitches” of the drive that appear out of nowhere, the first thing you need to do is change the cable.

“Shoals” are possible not only with contacts, but also with wires. Colleagues shared an observation: after removing the insulation from a non-working SATA cable, they discovered that the grounding conductor had oxidized and moved away from the shield of the twisted pair (there are two of them in the cable, each with its own shield). This sharply reduced noise immunity and led to transmission errors. After cleaning and resoldering everything was fixed. Although, of course, if possible, it is better to simply replace the cable.

There is another problem - no longer a “Chinese” one, but related to a change in standards. SATA cables of early releases (2003-2006) were held on the contact plugs by friction alone. The developers considered that this was not reliable enough (the threat of accidental disconnections remained), which is why the second version of the cables (starting in 2007) received spring latches at both ends. It would seem great - another reason for failures has been eliminated. But it's not that simple.

This is what the old (left) and new versions of SATA cables look like

On many drives of the previous generation, including actively used ones (2008), the SATA connector does not have a protrusion for the latch, which is why the cables of the new version fit weakly on it and do not lock - the latch does not work. The tip can slip due to anything - even from the vibration of the disk basket, or from the elasticity of a coiled cable. It is clear that this sharply reduces the reliability of the connection and is therefore unacceptable. Only an “old” cable without latches with its tight fit will do here (I don’t consider the option of fixing the connection with hot-melt adhesive - in the jargon “with snot” - although it is quite popular among assemblers). By the way, repairmen in their stands use cables of the first version, as they are the most universal (and sometimes there is no time to tinker with latches).

The old version of the cable (on the right) has an internal protrusion on the connector, which is responsible for a tight fit. In the new version, this protrusion has been removed, and fixation is carried out by a latch that clings to the protrusion on the disk itself

I remember a case where a client computer crashed due to similar problems. The included yellow cable from the motherboard (with latches, of course) was weakly pressed against the drive plate, which is why soft bads began to grow on it (the sector is written with an incorrect checksum and when read it gives a UNC error, although the data itself is correct). As luck would have it, the defects occurred in the registry, and Windows stopped loading with a BSOD - blue screen of death.

I deployed a “field hospital”, read out all the soft bads with a long read and wrote them back. Everything worked, the disk is like new. The yellow cable, of course, had to be replaced with another one - red and without latches. Hard drives are nowhere without tight contact in the interface lines. Electronics engineers call such a contact “dry” and value it very much: there are no transient processes and, therefore, the signal practically does not degrade.

When assembling or repairing a computer, I advise you to check all the cables - they should fit onto the connector plugs quite tightly, with noticeable force. I do the joining 2-3 times from each end to wipe off random dirt and oxide film from the lamellas (who knows, it’s gold plating, titanium nitride or even bare brass - the Chinese love such jokes). Hence the need to have in stock reliable trains of different versions and lengths (20-30-50-80-100 cm).

The best train will always be the minimum length . It’s not without reason that branded workstations (HP, Dell) are usually assembled on custom, very short SATA cables, sometimes as long as 15 cm. By the way, according to the standard, the internal SATA connector must withstand only 50 docking and undocking cycles, so its switching life is relatively small (the external eSATA connector is another matter, its durability is as much as 10 thousand cycles).

“Thick” and “thin” trains. There are even subtler ones: there is no such thing that the Chinese could not make worse

In addition to length, SATA flat cables also differ in thickness . It ranges from 5 to 10 mm, which is associated with the cross-section of the conductors (from 30AWG to 26AWG - the caliber marking is usually present on the cable), as well as with the density of the shielding braid (lowering it is a favorite trick of the Chinese, who save copper in all ways). Of course, you should always use the thickest cable - this increases the signal level and reduces interference. On a thin long cable, another disk may not be recognized or will work intermittently - this is due to the low load capacity of the interface chips.

Marking of SATA cables. Pay attention to the numbers in front of AWG: the smaller they are, the better - the conductors are thicker

SATA cables included with motherboards often have an angled connector at one end. Connected to the disk, it reduces the likelihood of accidental disconnection, saves space in the system unit and improves installation. However, the corner connector does not like tampering: if you accidentally pull it, you can break the contact strip on the disk, and this is a non-warranty case and a difficult repair.

Consequences of careless handling of the SATA cable. The bar is destroyed, the contact slats are literally hanging in the air. Repairing the connector is impractical

You can identify a low-quality cable using SMART. Unreliable contact generates transmission errors, causing attribute #199 UltraDMA CRC Error Count to increase. It is also worth paying attention to attributes #5, #197, #198 - their growth often indicates degradation of the disk itself ( for more information about the SMART attributes, see below - editor's note).

⇡ # Repairman and PATA loops

The scope of the parallel interface is constantly narrowing, but it is still far from dying out. For example, PATA 2.5″ hard drives are still produced - after all, you can’t install a SATA controller in an old laptop. And there are still plenty of PATA DVD drives. So you often have to work with 80-wire loops. Here is a case from recent practice.

A regular customer called - the system won’t boot, he writes something about “invalid disk”, urgently needs help. According to my inventory, this computer has an old PATA drive from Hitachi, DLAT series. They are quite simple and can be repaired even on the road. Moreover, it’s time to renew the contract...

Had arrived. I look - the disk is recognized in the BIOS, but with distortions in the model name. Naturally, the download doesn’t work either. This is typical for the loss of a bit in a word transmitted via PATA. The culprit is usually a damaged cable or a broken (bent, dented) pin in the contact comb on the disk. The latter happens during careless assembly, when the block is inserted into the connector at an angle or completely upside down (our fellow assemblers don’t care about anything - even the mismatch between the key and the slot in the frame).

No one has climbed into the system unit for two years, so the pins are excluded. This means there is a problem with the cable: one of the conductors has broken or the fit of the connectors on the cable has become loose (there are banal knife contacts that cut through the insulation; if you pull the cable “well”, they can come off). I replaced the cable with a new one (you should always have it with you) - everything worked. No repairs required, everyone is happy. But how could the PATA cable spontaneously deteriorate? All the computers in the office are from the same company, assembled in the same way. The train is folded into an envelope and tightly secured with a nylon tie. So, this screed has become stiff over time (and maybe from the heat), and its rigidity has increased. In an effort to restore its natural round shape, the tie pushed through the outermost wiring of the cable. Elementary Watson.

Well-built PATA cable. It is optimally suited for a compact system unit, where the hard drive and the connector on the motherboard are separated by just a couple of centimeters

Conclusion: There are no small details in assembling a computer if you want long, trouble-free operation. In particular, PATA cables are best secured with soft wrapping wire in plastic insulation. I don’t see any alternatives to it: I’ve already said about the tie (besides, it’s not removable, you’ll have to snack if anything happens, and this is also a risk of damaging the train - there have been cases), the elastic bands dry quickly and fall apart, the tape comes off. Branded computers (for example, HP) use special flat clamps with a latch, but I have not seen them on sale.

According to the standard, the PATA cable must have a length of 18 inches, or 46 cm (all other options, from 15 to 90 cm, are amateur performances of the manufacturers and do not guarantee quality). For most system units, this length is excessive, and the excess should be collected into an accordion, bending the cable at an angle of 90° or 180°. Make sure that it does not touch the fans and does not interfere with general air circulation. This is an important aspect of cooling the system unit: every motherboard has heating components without individual cooling, such as memory modules and some controllers, and “shielding” with a cable does not benefit them.

Already at the end of PATA’s “career”, cables appeared that were well protected from kinks and did not complicate ventilation inside the case. True, they cost almost an order of magnitude more.

Well, one last thing about the cable that is going out of use: avoid sharp bends, do not allow dents, or tension near the connectors. The conductors in the PATA cable are very thin and easily break if handled carelessly. Often the defect is not externally noticeable (elastic insulation hides the gap), and the behavior of the disc can be very diverse. This was demonstrated by the incident described above. In such cases, the first thing to do is replace the cable. You should always have a spare new cable on hand, since it costs a few rubles.

⇡ # Five smart things to do with a SATA drive

Has your favorite hard drive suddenly started behaving strangely, slowing down or freezing? There were no shocks, no overheating, the food was of high quality, and the SMART readings were normal? Let's see what a competent and careful user can do before running to the warranty office or to a repairman?

1. Replace the SATA cable with a new one, preferably branded and thick. The conductors must be AWG26 caliber - this is usually written on the braid; the width of such a cable is 8-10 mm. AWG30 cables with a width of 5-6 mm will NOT work. If you have a choice in length, take the shortest one (as a rule, 20-30 cm is enough, although 50 cm are more often on sale). Connect the cable to another port on the motherboard or external SATA controller. After this, parameter SMART #199 (C7) UltraDMA CRC Error Count should not increase!

2. Clean the SATA connector on the drive itself (7 flat contacts, of which two pairs of signal contacts and three ground contacts are longer) from dirt and oxides. Use isopropyl anhydrous alcohol and a microfiber cloth. Do the same with the adjacent power connector (15 pins).

3. Unscrew the electronics board from the disk (a Torx T9 or T6 torx screwdriver may be required in new models), find the silver-plated contact pads on the back of the board. There are two of them: 14-20 contacts for data, 3-4 contacts for the spindle motor. All areas should be light, if darkened (red, brown, dark gray) - use a soft eraser to wipe off the oxides until they shine, wipe with a cloth with alcohol. Carefully screw the board to the disk. The screw tightening torque is small, up to 30 N*cm (hold the screwdriver with three fingers). Otherwise, the edges of the slot will become wrinkled, which the warranty department may later notice - “signs of repair by unauthorized persons,” and hello.

This problem occurs even with new, just purchased disks. Oxidation of contacts on the board is facilitated by changes in temperature and humidity during long-term transportation, mainly by sea. Storage in poorly heated economy-class warehouses and polluted air in our cities also have an impact (sulfur exhaust from bad gasoline and coal smoke are especially harmful).

Oxidized contact pads (bottom right) on the board of a Seagate 7200.10 drive that has been in storage for over a year. If the disk had not become a head donor, the board would clearly need cleaning

From time to time I sell extra discs at Molotok and other flea markets, and pre-sale preparation, in addition to thorough tests, includes the procedure described above. In one of 15-20 cases, there are particularly picky buyers: having seen by the screws that the board has been removed, they believe that they were given “second-fresh sturgeon” and demand a refund. Well, the customer is always right.

Disk board after the “pioneer” repairman. Enthusiasm plus the lack of a proper tool, and the result is twisted screws

4. If freezes persist or the computer reboots, check the north and south bridges on the motherboard. Perhaps something is overheating, then you need to improve the cooling (change the thermal paste under the radiator, increase the airflow, etc.). Of course, you should also check the power supply for voltage stability under load. Change the power supply branch suitable for the problem drive, choosing the connector closest to the power supply. Disable all other consumers from this branch. Increase the disk start delay in the BIOS to 3-4 seconds - this will smooth out the surge in load on the power supply and help equalize the voltage, especially along the 12 V line.

5. If the problem persists (in particular, entries like “a controller error was detected” appear in the OS event log), then the next steps are to update the SATA controller drivers and flash the BIOS to the latest version. On nForce chipsets, disabling the NCQ command queue can help; to do this, uncheck Enable Command Queuing in the properties of the SATA controller on the channel to which the problem drive is connected.

The second part of the material will be published tomorrow, 10/26/2011.

⇣ Contents

Comment ( )

If you notice an error, select it with the mouse and press CTRL+ENTER.

Related materials