I just remembered, I still have a dead 4TB WD RED drive sitting on my desk (it is from the same series as yours and is a couple of months newer), it failed on me last year.... I should pull it apart too (that is why it is sitting on my desk, I was going to do a video on it one day!), my drive is from 6th March 2018 and it is also an EFRX version.

In my experience, actual physical head crashes are extremely rare in modern hard drives (especially with 3.5" form factor drives) and the failure mode is almost always down to corruption of the embedded servo data, resulting in the drive being unable to track properly, or a problem with the drive electronics - the hub motor / head actuator drive chips often get very hot and are known to fail. The motors themselves very rarely give problems, so if a hard drive won't spin, it's usually going to be an issue on the PCB. Chattering heads (the actuator arm banging back and forth between its travel limits) is often caused by bad servo data, in which case there's not usually much that can be done to correct it. Total failure of the drive to be detected by the operating system is invariably the main onboard microcontroller/DSP chip and intermittent detection failure is usually just bad cables/connectors. The little "bag" which you thought was a dessicant pack is actually a filter designed to trap any particles that get thrown off the surface of the platters - thses things are generally spinning at between 4000 and 8000 rpm and thus any dust and debris that lands on them will be hurled off at high speed due to centrifugal force - as a result, opening the hard drive in a non clean room environment is nowhere near as damaging as popular myth would have us believe - I've done it many times, particularly on certain laptop drives that have a habit of sticking (I'm looking at you, Samsung) without any problems, and many of those drives are still working years later (9 years and counting for one of my own drives). I wouldn't recommend it of course, but if you absolutely have no other options available to get a drive fixed, then it's not something you should be totally paranoid about, as it is entirely possible to get good and reliable results from this procedure. Obviously you should work in as clean an area as possible, using minimal air circulation (turn the desk fan off), keep your fingers and any tools off the platter surface and try to get the job done and the drive closed up again as quickly as possible. (and try not to let your inevitable sweat drip onto the platters - that's never going to end well :) )

27:19 And that's where he discovered he had pulled out the wrong drive, and the defective drive is still in the NAS :P

15:06 That window is used for writing the servo track during production. PS: I have been working for 4 years at Quantum Corp.

12:17 I’m pretty sure that’s just the recirculating air filter. Any little bits flying around will be carried along the outside of the platters and get trapped there

The high capacity drives are filled with helium, that is in order to be able to hover the heads closer to the surface of the platters than atmospheric air allows. The breather hole is needed on non-helium drives because the turbulence when the drives spin up, compresses some of the air, which causes lower pressure on top of the platters, causing the heads to hover too close, so when spinning up, they'll suck in air to compensate (over simplification), the helium drives have a slight over pressure to compensate for this. The white pad is a filter, designed to collect contaminations thrown off the platters. however, activated charcoal filters, and desiccant is common to find also. Regarding the screws, this is very common, i think they use custom sized torx, as i have experienced this many times. The trick is to loosen the screws one by one, and re-tighten them before going on to the next, if you loosen all screws at once, the last screw will carry so much load, that it can be hard to remove. The various aluminum parts are nickel plated, that is why they have a texture that could look like cast some places. Actuator arms have been both machined and cast over the years, but cast is more common in older drives. The head parking feature is used on all HDDs today, because of the flatness of the platters and heads, they will actually stick if they ever contact the platters, so the heads can't rest on stationary platters like they did years ago. Another fun fact, to ensure the heads are parked in case of power failure, it actually uses the platter motor as a generator, to generate just enough power to return the actuator home, before it is spun all the way down.

Coil doesn't need feedback as they use servo info written on the surface of 1 platter

*17:06* _The coil is an _*_orthocyclic_*_ (non-helical) winding, onto a sector-shaped bobbin._ _You can see the uneven step-over zone of the coil, nearest to the heads, where each loop of wire jogs to the next adjacent winding plane._

For a bigger hard drive Teardown and much younger Dave. Eevblog # 395.

It is amazing, given how much tech goes into these things, how really inexpensive they are.

The heads are created in layers using the same processes used for semiconductor chips. They go through thousands of operations to build a three dimensional structure using exotic materials (platinum, tantalum, gold, copper, diamond-like-carbon, silicon carbide, etc.). They are oriented vertically (standing on edge), so that thousands can fit onto a single wafer. The crazy part is that they have machines with enough precision to mechanically separate them from each other, with the air bearing surface still in tact (it is not simply a smooth surface).

Just looked up their Dutch address printed on the case and yep it's one of those tax dodging buildings. You can look the street name up Taurusavenue 5 Hoofddorp. That whole building has thousands of companies registered in The Netherlands so they pay like 0.1% Tax while they virtue signal on Twitter how much they care.

I think your diagnosis that it has crashed is incorrect. The head positioning is done via feedback from the read signals from the heads, either a single servo head or (in this case) embedded servo information on all heads. There has been a failure in the read amplifier and thus the head servo has lost control, and that is what you see and hear. When you view videos from people who know how to recover these, you will see them replace head assemblies and make them work again (at least to read the data).

If your re-assemble the magnets and drop one of the aluminium rings between them, you can see Lenz's law in action as the ring moves much slower than you would expect.

Used to love taking apart hard drives as a kid. It really felt miraculous, even in the 90s, to have such a precisely made machine in your hands. It's actually a little surprising how superficially similar a modern unit looks to one of those old ones, though I'm sure the engineering has only gotten more advanced.

Just want to say to everyone, that drives today from any manufacturer are pretty damn robust. They are machined to really tight tolerances and some have the ability to park the heads before physical damage can occur by detecting the G forces applied to the drive. So the damage is less likely to be physical scrapes on the platters unless you've really dropped the drive or given it a jolt. Desktop drives are more easily physically damaged from movement but desktops are less likely to be moved. But really, the most likely failure point for any non-physically damaged drive are the pressure pads that let the signals travel between the board and the motor and heads. At least that is what I've found to be the case. The pressure contacts eventually begin to put less and less pressure on the pads over time and they also eventually start to oxidize, both problems create higher resistances in the connections and that eventually causes issues spinning the drive, moving the heads and reading/writing the data. If you are placing a drive in a freezer to get data off and it's working then those pressure contacts are likely the reason why your drive is failing. I've seen the pads for the head connectors oxidize causing read/write issues and I've seen the pressure connection pins for the motor connector loosen enough over time that the platters won't spin up properly or even at all. Physically cleaning oxidation and bending the pins for the motor connector to make more pressure on the pads is usually enough to give a drive some life support but if you want reliability then it might be necessary to replace any aged pogo style connectors that are going to the head assembly. [... _or just replace the pressure connectors with some carefully made direct wired connections and live a happy life with an _*_almost immortal_*_ drive_ ...] And I've never had to, but if you use something like Deoxit then only use it on the board side and not the mechanical side, or at least be careful and don't spray anything at it. Just use a q-tip or cotton swab and don't get anything inside the mechanical part... I've also seen shorted protection diodes on boards but that seems more likely an issue on external drives and less likely on internal drives unless the power supply is going out or you've got some fancy pants removable drive system. And they really need to add some protection diodes on all power rails right over the motor driver and maybe even the DSP chip, too. Faulty power connections occasionally destroy those things.

It's actually not just a desiccant packet. That is the micro particle catch. If something were to get into the drive, it would get flung into that catch.

WAY cooler video than I expected. The microscope shots were just amazing to pause and just soak up all the incredible design and manufacture. There's lots of voodoo in those devices...

I'm sure others have pointed out that with modern drives just changing the torque on the screws on the top plate is enough to cause them to fail. Then you get into the helium filled drives...another story all together.

The platters make nice wind chimes.

Holes near the read amp have small traces going to them - may be for the gold plating process - connecting all the copper together, then punched out after plating

The yellow stuff is urethane. Those platters are actually glass with a vacuum deposited material. Not a desiccant pouch, it's an air filter to catch any flying particles. Eventually all of the air in the drive will be pushed through that filter from the spinning platters.

I remember making a mobile out of old platters in high school. I've repaired computer bits and bobs before, straightened out pins, soldered wires back on, but the tolerances are so tight on hard drives, they're just beyond me haha

This where you come to find that a capacitor in the circuit went open and was causing undesired oscillations causing the heads to miss align and not boot!!!!

I've dissected a lot of hard drives in the past, but they still amaze me, at how precise they are put together. Really blows my mind!

5:00 That red potting compound or whatever is telling "I'm sorry Dave, I'm afraid I can't do that." 5:30 Traces used for the electroplating that were then cut, I assume? 26:40 Gauge platters, I reckon. Perfectly wrung together.

Dead Red Revolver? I'll see myself out.

The first ever EEVblog video I saw was the teardown of the old big hard disk. Kept on watching the channel ever since :D

What amazes me is how they can produce such high precision so cheaply. They can print silicon, but they can't print hard drives!

I'm glad you found the grease port after removing the silver sticker! unfortunately, this one was damaged already as you clearly could hear the dry bearing. but there is still hope for other ones.

Dave thanks for keeping Andy @PhotonicInduction inspired and his spirits up. A lot of people including myself were quite worried. God bless.

The 2 notches at 16:31 I believe is to balance the spindle when the screws are attached. The noise should be coming from the seeking motion when the drive starts and it's looking for the index ring or something. I've not seen any more scratches on the disk like it used to when drive fails nowadays.

It's almost always the header that fails first. Anyway it really is a thing of beauty. Thank you for the teardown.

I'm impressed that at @8:26 just after taking off the top that you've already got visible dust specks on it (upper right corner, and upper left corner near the middle of the platters). Just goes to show how important clean room facilities are due to how quickly it can become contaminated.

The head is probably made via pulver metallurgy (e.g. sintering) and then machined to final dimensions. This process makes sense for high volume high precision parts

A few years ago, I bought 4 6TB WD red drives thinking they'd be great for my new NAS not knowing they were the SMR drives. I moved ALL of my data onto these drives before I discovered the difference between CMR and SMR and was very upset that they'd put this in their red series. I contacted WD support, and they replaced all 4 of my FAX drives with the FRX drives for free. Not only that, but they sent me the drives first so I could copy my data over and requested I sent my FAX drives back after I received the FRX. There is a nearly $100 difference between the two drives, so WD essentially gave me $400 because I was a poorly informed customer... gotta say I am a WD customer for life now.

I did enjoy that, sure, thanks! Now I hope you always keep this drive as something you can continue to show other family members and new friends who haven't seen the inside of one in person yet.

It's highly unlikely that part is cast. The draft angles would be an absolute nightmare.

That connector looked like a grinning face. Love it!

20:25 Those aren't test pads. Those are Piezo dual-stage actuators for fine positioning a head onto a track or jogging to an adjacent track. They work in such a way that they cause a twist in the arm that is converted into a side-to-side motion at head. Read about it in the whitepapers published by various drive makers.

Servo data is written on every surface. The slot under the large sticker is where an arm from the servo writing machine connects to the head assembly to write the servo tracks on the platter during production. The head assembly whacking back and forth means the firmware can’t find the servo tracks and calibrate. It tries a bunch of times then gives up and packs itself.

I remember the first time I took 1 of my old dead HDD's apart and scavenged the magnets etc out of the thing

As you saw, The hard drive platters are so smooth and flat that you can ring them together like Guage blocks (assuming you don't finger print them up). I recycle these and take them apart all the time. I like to ring several platters together end to end.

I always take these apart to use the magnets, the aluminum bushings are pretty useful too.

Almost like Casimir effect when you put those platters together -- but it's just atmospheric pressure at work :-)

Hey Dave, those screws may be Torx Plus not standard Torx.

It's so amazing to me that the magnetism from the linear motor is so strong but still so localized that it doesn't spread over areas of the disk and harm data while the disk spins!

Yup, pulled apart a few of these things, and *could* tell you what steps to go through - but IMHO if you don't need to successfully get it back together again after, figuring it out is half the fun!

That resistance you're feeling in the spindle motor is just back-EMF from the brushless motor acting on the permanent magnets in there. Fun fact; you can connect the terminals of that motor directly to a standard hobby aircraft/drone brushless DC motor ESC to get things spinning real quick!

The first video I watched on your channel was the teardown of a HDD that was probably 1 millionth of the capacity of this one

Dave, the so-called death clicking is not usually caused by problems on the surface of the platters. Mostly, it comes from either head assembly problems or platter motor problems. As you mentioned, tolerances are extremely small, so if heads can't move properly or platter rpm isn't spot-on, the drive's controller enters a loop trying to seek valid signal(or data - I'm not exactly sure whther it's the actual onboard controller of the HDD or the motherboard). At least that's what I remember from university :) There is another type of click, which is more rhythmic, and very loud - that's if there's issues in the signal path of the heads so they don't get any feedback off the tracks. The servo then launches them to the middle and they hit hard and click. But that's not what you hear here. By the way, if u listen carefully, when you spin the drive open, there is a specific sound when the heads passed over the scratch you made - that totally destroyed the head :) Oh, one more thing - those notches on the platter shaft I think are weight balancing holes so that the center of mass is exactly in the middle.

The small square bag in the corner is a air filter, the shape of the chamber gives you a clue, the air movement from the platters draws a little air through it as the drive spins. The separation plates between the platters are there to reduce the amount of air that moves across the faster moving edges of the platters, to reduce drag. Quite often in smaller drives the platters are made from glass with a coating applied by vacuum deposition.

Oh no, you didn't take apart the spindle motor. It would have been interesting to see the engineering that goes into spinning all those platters.

Seal looks like some kind of silicone. In the old days they used the spongey/sticky material that tend to turn into dust over time- I guess that's why they switched to silicone based seals :P

Now I don't have to teardown my dead hdd thanks to this video, it satisfied my curiosity quite well.

The seals on all the screws help to clean the drive before opening it, or they prevent gunk that is difficult to clean out around the screws falling inside the drive while opening it. Obsolete since welded shut drives wit gas filling.

The two machined semi circles on the platter hub are almost certainly not for a holding tool but are a balancing operation done post assy

Platter screw was probably an ip7, torx plus #7. We used those in cell phones, they are able to take higher torque without stripping the head. Normal torx can be used, but is not the preferred tool for them. Torx plus drivers will not fit torx screws at all.

I never cease to be impressed by the precision of the machining in hard drives, especially at the price they're sold for. With the lid off and powered up you can't even tell that the platters are spinning. The head carrier arm has to be machined as there's no way to hold any decent precision with a casting.

I pulled a drive apart a few years back, the second I took the cover off I knew it was game over - a fine coating of a silvery dust was covering everything, took the platers out and found one of them was scored really badly - all that fine dust used to be my data. But I totally agree with you Dave, the engineering that goes into a hard drive is just amazing.

I doubt “most complex mechanical product” applies. But “Most precisely engineered mechanical device” Definitly applies.

Dave, you should know that those metal flanges are being connected by a non-reversible tremie pipe to the differential girdle spring on the “up” end of the grammeters. Also, I wonder if you can "lift" finger prints from that glorious 4K video?

I used to dismantle lots of server harddisks to get the magnets out, they are extremely useful.

You should get an impact driver, if you don't already have one, to help with getting stubborn screws out. They work well on screws attached to things that spin because the rotational impact action is faster and can apply more instant torque than you can with your hand. I've used them to take apart dead hard drives.

the sound looks more like it was the bearings failing

Didn't inspect the high resolution video, but i hope you don't use the biometrics of your forefinger for something important. ;)

Dave! Didn't realise you were so young. As mikeselectricalstuff says, this is a "voice coil" drive. It was the major step forward when we moved from stepper motor drives into positional feedback. One of the platters is usually reserved for initialisation information so that the head gets to the right place regardless of temperature. I remember one Bathurst winter having to sit there for half an hour with a hair dryer to get an old stepper drive warm enough for the heads to line up and read enough to boot.

I think the white object removed at 12:13 is a filter that uses the multiple-disk centrifugal pump effect to pull the air pre-filtered across the disk heads in a smooth, clean laminar flow.

Sounded like bearing failure to me, would explain the dependency on orientation for the noises. Then the motor would not reach its operating speed and the data looks garbled to the controller, hence it stops.

The pads are airfilters for any minuscule particles that sneak in during production.

I still have some platters from an old IBM drive I took apart. Out of the thousands of HD's I have taken apart for my job, this drive was the only one with glass platters. Glass. Fascinating.

I believe the slot on the underside of the HDA is for the servo writing machine to place the embedded servo information (for proper actuator tracking) onto the platters before the controller goes in

You should do a factory walk-through at one of the big HDD producers.

Actually basic design of hdd kept the same for 20 years. Just endless polishing of technology.

The last time I saw something visibly wrong inside a hard drive, it wasn't just a head crash, but the head actually dangling by a copper trace. Was the motor still electrically connected to anything so that maybe it was acting like a generator? They usually make nice bearings. The platters also make nice first surface mirrors, or xmas tree decorations.

I use the platters as first surface mirrors for playing around with cheap lasers they work quite well

Last 3 weeks 4 out of the 5 wd 6TB HDs failed in my nas. Had to replace all of them. All cmr versions with 30-40k hours on them. Replace 1, test, rebuild, fail another, backup all with raid in fail mode and 1hd with bad blocks. Create new raid, fail another disk after successful test. Hell of a week.

and now assemble everything back ! :D :D

Hard drives are wonders… I started with a 5.25 inch, 40 meg voice coil actuated Miniscribe 3053 back in ‘88, still have it, that actuator was never raucous but made great sounds :-)

As a guess I think they would have a tool similar to the angle grinder tool for removing discs to remove the platter screws.

Whenever I have dead HDD that is unreadable, unformattable, and unusable, I always disassemble those, so I can get magnet inside.

Dave, maybe consider running different manufacturers or batches of drives in NAS(unless RAID0). I got WD+Seagate combo, probably not matched in performance, but will not fail all at the same time.

I once had an HP laptop power supply fail exactly 1 day out of warranty. Sometimes I wonder just how close these warranty periods are to the MTBF.

Fun fact when they were stuck together you wrung them and shoved all of the air out to almost friction weld them together like you would gage blocks for percision measuring

Last time I teared down an HDD I try to unstick the magnets using a screwdriver and a hammer. One of them come out, the other broke into pieces and some spark come out as the screw driver hit it. That was impressive.

For those who still own old, really old VHS, or especially Betamax video-recorders, yes, the precision is not as great as a modern HD, but they are a far more 'mechanically' complex device than this HD by a mile. As for precision, until the day that higher density hard-drives become the norm in homes, actually the VHS & Video-8 / Hi-8 video heads were the most accurate precision machined device consumers ever owned.

Acrylic lacquer on the platters... to mske them into interesting coasters. Maybe try some etching to show up the magnetic recording pattern. One of the surfaces is likely to be a servo surface, recorded at manufacture and used subsequently for finding the travks.

The three large half-circle brackets are air baffles, to shape the airflow and control turbulence. The three white pads on the inner side of the lid are for noise dampening.

Contaminations and imperfections on the disk platters might not be visible with normal light. If you care enough you can use "green light" to check for damage. See kzfaq.info/get/bejne/or94mKl7y5eVfaM.html for more information.

Another dead red, red for danger as they say! All the 4tb greens I have (12 in total) made it past 60k hours and are still in use! Just bought a load more 4tb greens, might as well back a winner!

For those who are interested on the tech that goes into making and using those heads, these heads use something called "Giant Magnetoresistance" to read magnetic fields. The trick here is they are injecting electrons of different spins between the magnetic layers separated by non-magnetic materials and depending on the field it can align those electrons to be the same spin and it manages to achieve lower resistances or higher resistances depending on what is written on the platter It is very sensetive to those fields and a lot of spin offs (pun not intended) of that promise an alternative to solid state storage. It is absolutely an amazing field i am glad i got introduced to it with my physics degree. Could write a whole conference on it and still not cover everything that we know, much less what we don't know.

Just took apart a Western Digital Enterprise SATA drive. Interesting to compare between them. For example, instead of the plastic pads on the lid, they actually glued a piece of metal to the top to stabilize the lid. Drive had nearly 50k hours 24/7 on it. Apparently you do get something for the extra coin spent. I've been saving all the drive platters; maybe they'll make a neat wind chime one day. They make a good "ding" noise. 👍

A thing of beauty. All done by robotic milling of cast parts.

Would have been nice to see what SMART detected if anything.

I always remove the magnets from hard drives before tossing dead ones - they are crazy useful. Have a friend who is a veteran and he has shrapnel still surfacing from an old wound - gave him one of those neodymium magnets, advised him to put it in the ziplock freezer back to keep it clean and it worked a treat - as more pieces of shrapnel start to surface underneath his skin, the magnets really speed up getting it out. He's told his buddies and now they are all going around looking for more hard drives. So if you know of anyone with any injuries with shrapnel these magnets work a treat. So many uses!

These are always so fascinating. It's mind boggling how far storage tech in general has come. Back in the day you needed a cargo plane to move 1MB of data around lol. Going to guess the failure mode of this drive was a head related issue. Head was not reading correctly which caused it to keep seeking, and it would crash at the end zones because it did not know where it was. At least that is my guess.

10 years from now: Technology found a way to recover TBs of data from platters shown in a 4k video. :-)

and we are all paranoid about getting a magnet close to a hard drive,,, and yet the disks spin past the servo magnet cage at 5400 7200 10000 15000 rpm....

I had the exact same size/model fail as well (made 2015) and it's been sitting on the shelf waiting for me to extract the cool magnets from it.

At 20:13 I believe those rectangular copper pieces where the head flexes are piezo elements when energized lift the heads up to float on top of the hard driver surface, If those fail the head crashes and starts to scrape the disc surface.