I’ve expressed no shortage of admiration in past writeups (stretching back…oh…a quarter century or so…) for the flash memory industry’s incremental cost, capacity, performance and reliability accomplishments over the years. I suppose I’m (more than) a bit biased because my post-college (not including during-college-co-op) engineering “origin story” involves eight years with Intel as an applications engineer, during the embryonic days of that trendsetting company’s solid-stage storage trajectory.
Truth be told, however (and if you look back at my coverage, what I’m about to say here shouldn’t be a surprise, either) I find the electromechanical marvels that are hard disk drives (HDDs) to be even more amazing engineering accomplishments:
- One or (usually) multiple platters, spinning at speeds up to 15,000 RPM. Each platter mated to one or (usually) two read/write heads, hovering over one or both sides of the rapidly rotating platter only a few nanometers away, and tasked with quickly accessing the desired track- and sector-stored details.
- Low-as-possible power consumption and high-as-possible ruggedness and reliability, in contrast to other contending design considerations.
- And ever-more data squeezed onto each platter, thanks to PRML (partial-response maximum-likelihood) sensing and decoding and now-mainstream PMR (perpendicular magnetic recording), next-generation SMR (shingled magnetic recording) and emerging successor HAMR (heat-assisted magnetic recording) storage techniques.
I’ve watched plenty of videos and seen plenty of photos of the insides of a HDD, but I’ve never taken one apart for myself. Usually when a drive dies (I’ve only had a few do so over the years, and not always through an inherent fault of their own), I just either apply a strong magnet or a heavy sledgehammer to it, in both cases to render the information stored on it inaccessible to others, then toss it. But recent events gave me another inspection opportunity.
Nowadays, when outfitting a NAS (network-attached storage device), instead of going with brand new HDDs at full retail prices, I instead pick up refurbished Hitachi Global Storage Technologies (HGST) Ultrastar drives commonly available for sale on Amazon, eBay, Newegg and elsewhere. HGST, a Hitachi subsidiary formed when the company acquired IBM’s disk drive business in 2003 (therefore explaining the “Ultrastar” brand that other old-timers might still remember), no longer actually exists. Western Digital acquired HGST in 2012 and the HGST brand was phased out by 2018. This indicates just how old these drives are, although I’ve to date had great luck with them; they often also include multi-year warranties from refurbishers.
I was poking around in the software user interface of my QNAP TS-328 the other day:
when I noticed a “warning” indication for the 3 TByte drive in NAS bay 3:
Diving further into the UI indicated that sixteen sectors on the HDD were reporting S.M.A.R.T. ID 197 “Current Pending Sector” attributes:
Here’s the definition of attribute 197, courtesy of Wikipedia:
Count of “unstable” sectors (waiting to be remapped, because of unrecoverable read errors). If an unstable sector is subsequently read successfully, the sector is remapped and this value is decreased. Read errors on a sector will not remap the sector immediately (since the correct value cannot be read and so the value to remap is not known, and also it might become readable later); instead, the drive firmware remembers that the sector needs to be remapped, and will remap it the next time it has been successfully read.
However, some drives will not immediately remap such sectors when successfully read; instead the drive will first attempt to write to the problem sector, and if the write operation is successful the sector will then be marked as good (in this case, the “Reallocation Event Count” (0xC4) will not be increased). This is a serious shortcoming, for if such a drive contains marginal sectors that consistently fail only after some time has passed following a successful write operation, then the drive will never remap these problem sectors.
QNAP’s own guidance also rightly indicated that I didn’t necessarily need to do anything right away. Indicative of the low urgency, I hadn’t even gotten a notification email from the NAS:
Nonetheless, I as-always had a spare 3 TB on hand, and replacements are less than $40, so I went ahead and did the swap. A few hours later, post-RAID rebuild, I was good to go again:
Leaving us with today’s dissection candidate, the Ultrastar 7K3000 3TB (PDF spec sheet), as-usual accompanied by a 0.75″ (19.1 mm) diameter U.S. penny for size comparison purposes. Note, per my earlier age comments, the April 2013 manufacturing date stamped on the label:
Not much to see on the two sides, aside from mounting-screw sites:
Nor on one of the two ends:
The other end is more interesting: left-to-right are the 15-pin power connector and 8-pin SATA connector (the other two sites to the right are nonfunctional):
Finally, let’s have a look at the bottom:
That four-trace flex cable coming from the PCB presumably powers (and manages) the motor that rotates the platters…but we’ll have to dive inside to confirm:
Speaking of diving inside…if you revisit the topside overview shot, you’ll see six hex head screws around the perimeter, plus three more underneath the tamper-detecting clear plastic. By the way, before I forget, the other hole visible in that same shot, to the right of the three plastic-swathed screw heads and in the lower right corner of the HDD, is the “breather hole”. It’s sometimes explicitly labeled with “do not cover” or similar warning messages, although not in this case, and its purpose is two-fold: to equalize the atmospheric pressure inside and outside the HDD, and to provide an exhaust path for any moisture that might otherwise accumulate within the drive. Anyway, let’s get those screws out:
After removing all nine visible ones, I still wasn’t able to get the “lid” off the HDD. A peek underneath the label confirmed my suspicion as to why; one more was lurking there:
With it also extricated, I successfully achieved liftoff:
Note in the image, the white “air filter” underneath the aforementioned “breather hole”. Its key function is to catch any particulates that might otherwise result in a “head crash” (recall how tiny the gap between a read/write head and the platter is in normal operation) or other calamity. There are two additional, “puffier” particulate filters in this design, both of which are visible in these chassis overview shots. One is in the lower left edge of the platter stack, and the other (tinier, although both are small; look for the white in the shots) is in the lower right platter-stack edge. You’ll see closeups of both later. Speaking of closeups, here’s our first opportunity for visual inspection of the actuator assembly and read/write head stack:
Note, too, the dividers separating the heads from each other, as well as the platters. Again, you’ll be able to view them in standalone perspective shortly.
And here’s one of the two earlier mentioned internal air filters in more detail:
Internal disassembly step 1: remove the top of the spindle. Those six Torx screws were (understandably) quite tightly attached, and the entire platter assembly would rotate every time I tried to remove one, thereby explaining the resulting top-of-platter fingerprints:
This magnet is incredibly powerful; prying the top of the actuator away from the arm assembly took a bit of muscle and quite a bit more patience:
This is as good a time as any to share an excellent video that I found post-teardown on the Hard Disk Drive Wikipedia page, which clearly and concisely explains how the actuator assembly (along with other HDD elements) operates:
Had I known in advance, as shown both in this video and another from that same Wikipedia page, that I could have operated the HDD with the top off, I would have tried that too!
Oh well. I was now able to pry the topmost platter over the spindle and off the stack:
Here it is standalone, in both top side (again, with fingerprints added by yours truly):
In trying to get a second platter out, I turned the assembly upside down. Out flew the around-spindle circular spacer between the top two platters, along with another unknown-function piece that you’ll see in the earlier photos right next to the actuator:
To get more platters out, I was going to need to get the head stack out of the way first. Removing two more Torx screws didn’t get me far, but exposed to clearer view the interconnect between the actuator and the electronics on the bottom side of the HDD:
Similarly, I was able to lift off the inter-head spacer assembly by removing one more screw:
What was revealed was unexpected, at least to me:
This is a four-platter 3 TB HDD; you’ll see the other platters shortly. Reflective of this fact, there are eight total read/write heads: a pair per platter, one per platter side. Doing the math gets you to 750 GBytes of per-platter storage. But the end of the actuator arm has enough sites for 10 read/write heads; as you can see, the bottom two are unpopulated in this configuration. My guess is that this same actuator-and-arm assemblage also finds use in 4 TByte HDDs, where each platter (five total) holds 800 GBytes of data. Agree or disagree, readers?
Speaking of the actuator-and-arm assemblage, even though I’d removed all visible screws holding it in place, it still wouldn’t budge. I therefore suspected it was also restrained from further attachment to its underside, so it was time to turn the HDD over and continue my disassembly task there. Six removed screws later, I had my answer:
Putting the PCB aside for a moment, I’ll draw your attention to several notable-to-me details:
First off, in the lower left quadrant, you’ll see several more Torx screws, which are indeed still holding the actuator-and-arm assemblage on the other side in place. Second, remember the flex cable that I previously mentioned led from the controlling electronics to the actuator? You’ll see its connector to the PCB in the lower left quadrant. And finally, in the middle are four terminals which press-connect to the PCB and, via another flex cable intermediary, allow those same electronics to control the platter spindle motor. Oh, about that PCB…
The back side is fairly unmemorable, unless you’re into test points and vias, that is. I admit that I’ve always found it a bit odd that HDD manufacturers leave the potentially short-circuitable (or at least ESD-shockable) PCB exposed, versus covering it with non-conductive tape or other material prior to shipping the HDD to the end customer. Speaking of non-conductive material:
Nonconductive foam, plus a bit of thermal tape atop the main IC, initially obscures the bulk of the PCB top side’s contents from view. Let’s do something about that:
Unfortunately, the remnants from the thermal tape rendered the main IC’s product markings partially illegible, so you’re going to have to take my (and my illuminated magnifier’s) words as gospel that this is what’s stamped atop the TQFP:
Unfortunately, I can’t find any meaningful information online on this chip (and LSI Logic was acquired by Avago-now-Broadcom in 2014), but I’m guessing that it’s the primary drive controller chip, managing the interfaces to the SATA bus, the actuator circuitry, and (indirectly) the spindle motor. Another writeup I came across referred to it as the LSI MEL-B1B1 (at least that’s what the second product mark line in their photo says), but I can’t find any information associated with that IC name, either.
Fortunately, the other ICs’ identities are easier to decode. Below and to the left of the LSI 6045 is an IC with the product code OA73080, which appears to be the spindle motor controller, although I can’t figure out who the chip’s manufacturer is. To the right of the MEL-B1B1 is a 512 Mbit DDR2 SDRAM; the one in my drive is from Winbond (the W9751G6JB-25), whereas the one documented here was from Samsung. Below and to the right of the SDRAM is a 4 Mbit serial flash memory (marked 25FS406, multi-supplier sourced), which likely stores the drive firmware. Remember the earlier mentioned “four terminals which press-connect to the PCB and, via another flex cable intermediary, allow those same electronics to control the platter spindle motor”? Their quad-spring mate is at the bottom of the PCB, below the LSI 6045. And speaking of pressed-together connections, the diagonally oriented mate to the previously noted flex cable routing to the actuator is in the upper left corner of the PCB.
When I turned the HDD back over, I realized that the other air filter had fallen out. Thankfully, I was now able to cleanly extract the actuator-and-arm assemblage from its confines. Here are four overview perspectives from various vantage points:
And here’s one final closeup of the read/write head stack, complete with the missing fifth pair:
In part by swiveling out of position and then completely removing the previously noted inter-platter spacer plastic piece, I was able to get all the platters out:
And since I no longer remember what order they originally stacked in the drive, not to mention the fact that I’ve manhandled all of them, I don’t think anyone’s going to get any data off them:
With that, another teardown concludes. I’ll now turn the keyboard over to you for your questions and other thoughts in the comments!
—Brian Dipert is Editor-in-Chief of the Edge AI and Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.