I’ve expressed no scarcity of admiration in previous writeups (stretching again…oh…1 / 4 century or so…) for the flash reminiscence business’s incremental value, capability, efficiency and reliability accomplishments over time. I suppose I’m (greater than) a bit biased as a result of my post-college (not together with during-college-co-op) engineering “origin story” entails eight years with Intel as an functions engineer, throughout the embryonic days of that trendsetting firm’s solid-stage storage trajectory.
Fact be advised, nonetheless (and if you happen to look again at my protection, what I’m about to say right here shouldn’t be a shock, both) I discover the electromechanical marvels which can be laborious disk drives (HDDs) to be much more wonderful engineering accomplishments:
- One or (often) a number of platters, spinning at speeds as much as 15,000 RPM. Every platter mated to at least one or (often) two learn/write heads, hovering over one or either side of the quickly rotating platter just a few nanometers away, and tasked with rapidly accessing the specified track- and sector-stored particulars.
- Low-as-possible energy consumption and high-as-possible ruggedness and reliability, in distinction to different contending design issues.
- And ever-more knowledge squeezed onto every platter, due to PRML (partial-response maximum-likelihood) sensing and decoding and now-mainstream PMR (perpendicular magnetic recording), next-generation SMR (shingled magnetic recording) and rising successor HAMR (heat-assisted magnetic recording) storage strategies.
I’ve watched loads of movies and seen loads of images of the insides of a HDD, however I’ve by no means taken one aside for myself. Normally when a drive dies (I’ve solely had a number of achieve this over time, and never at all times via an inherent fault of their very own), I simply both apply a powerful magnet or a heavy sledgehammer to it, in each instances to render the data saved on it inaccessible to others, then toss it. However current occasions gave me one other inspection alternative.
These days, when outfitting a NAS (network-attached storage system), as an alternative of going with model new HDDs at full retail costs, I as an alternative decide up refurbished Hitachi International Storage Applied sciences (HGST) Ultrastar drives generally obtainable on the market on Amazon, eBay, Newegg and elsewhere. HGST, a Hitachi subsidiary fashioned when the corporate acquired IBM’s disk drive enterprise in 2003 (subsequently explaining the “Ultrastar” model that different old-timers would possibly nonetheless bear in mind), now not truly exists. Western Digital acquired HGST in 2012 and the HGST model was phased out by 2018. This means simply how outdated these drives are, though I’ve so far had nice luck with them; they typically additionally embrace multi-year warranties from refurbishers.
I used to be poking round within the software program person interface of my QNAP TS-328 the opposite day:
after I seen a “warning” indication for the three TByte drive in NAS bay 3:
Diving additional into the UI indicated that sixteen sectors on the HDD have been reporting S.M.A.R.T. ID 197 “Present Pending Sector” attributes:
Right here’s the definition of attribute 197, courtesy of Wikipedia:
Depend of “unstable” sectors (ready to be remapped, due to unrecoverable learn errors). If an unstable sector is subsequently learn efficiently, the sector is remapped and this worth is decreased. Learn errors on a sector is not going to remap the sector instantly (for the reason that appropriate worth can’t be learn and so the worth to remap isn’t identified, and in addition it would grow to be readable later); as an alternative, the drive firmware remembers that the sector must be remapped, and can remap it the following time it has been efficiently learn.
Nevertheless, some drives is not going to instantly remap such sectors when efficiently learn; as an alternative the drive will first try to jot down to the issue sector, and if the write operation is profitable the sector will then be marked pretty much as good (on this case, the “Reallocation Occasion Depend” (0xC4) is not going to be elevated). It is a critical shortcoming, for if such a drive accommodates marginal sectors that persistently fail solely after a while has handed following a profitable write operation, then the drive won’t ever remap these downside sectors.
QNAP’s personal steerage additionally rightly indicated that I didn’t essentially must do something instantly. Indicative of the low urgency, I hadn’t even gotten a notification e-mail from the NAS:
Nonetheless, I as-always had a spare 3 TB readily available, and replacements are lower than $40, so I went forward and did the swap. A number of hours later, post-RAID rebuild, I used to be good to go once more:
Leaving us with in the present day’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 dimension comparability functions. Word, per my earlier age feedback, the April 2013 manufacturing date stamped on the label:
Not a lot to see on the 2 sides, other than mounting-screw websites:
Nor on one of many two ends:
The opposite finish is extra fascinating: left-to-right are the 15-pin energy connector and 8-pin SATA connector (the opposite two websites to the precise are nonfunctional):
Lastly, let’s take a look on the backside:
That four-trace flex cable coming from the PCB presumably powers (and manages) the motor that rotates the platters…however we’ll must dive inside to substantiate:
Talking of diving inside…if you happen to revisit the topside overview shot, you’ll see six hex head screws across the perimeter, plus three extra beneath the tamper-detecting clear plastic. By the way in which, earlier than I neglect, the opposite gap seen in that very same shot, to the precise of the three plastic-swathed screw heads and within the decrease proper nook of the HDD, is the “breather gap”. It’s typically explicitly labeled with “don’t cowl” or comparable warning messages, though not on this case, and its objective is two-fold: to equalize the atmospheric stress inside and outdoors the HDD, and to offer an exhaust path for any moisture that may in any other case accumulate throughout the drive. Anyway, let’s get these screws out:
After eradicating all 9 seen ones, I nonetheless wasn’t capable of get the “lid” off the HDD. A peek beneath the label confirmed my suspicion as to why; yet one more was lurking there:
With it additionally extricated, I efficiently achieved liftoff:
Word within the picture, the white “air filter” beneath the aforementioned “breather gap”. Its key perform is to catch any particulates that may in any other case end in a “head crash” (recall how tiny the hole between a learn/write head and the platter is in regular operation) or different calamity. There are two extra, “puffier” particulate filters on this design, each of that are seen in these chassis overview photographs. One is within the decrease left fringe of the platter stack, and the opposite (tinier, though each are small; search for the white within the photographs) is within the decrease proper platter-stack edge. You’ll see closeups of each later. Talking of closeups, right here’s our first alternative for visible inspection of the actuator meeting and skim/write head stack:
Word, too, the dividers separating the heads from one another, in addition to the platters. Once more, you’ll have the ability to view them in standalone perspective shortly.
And right here’s one of many two earlier talked about inside air filters in additional element:
Inside disassembly step 1: take away the highest of the spindle. These six Torx screws have been (understandably) fairly tightly hooked up, and your complete platter meeting would rotate each time I attempted to take away one, thereby explaining the ensuing top-of-platter fingerprints:
This magnet is extremely highly effective; prying the highest of the actuator away from the arm meeting took a little bit of muscle and fairly a bit extra persistence:
That is pretty much as good a time as any to share a superb video that I discovered post-teardown on the Exhausting Disk Drive Wikipedia web page, which clearly and concisely explains how the actuator meeting (together with different HDD components) operates:
Had I identified prematurely, as proven each on this video and one other from that very same Wikipedia web page, that I might have operated the HDD with the highest off, I might have tried that too!
Oh nicely. I used to be now capable of pry the topmost platter over the spindle and off the stack:
Right here it’s standalone, in each prime aspect (once more, with fingerprints added by yours actually):
In making an attempt to get a second platter out, I turned the meeting the other way up. Out flew the around-spindle round spacer between the highest two platters, together with one other unknown-function piece that you simply’ll see within the earlier images proper subsequent to the actuator:
To get extra platters out, I used to be going to wish to get the pinnacle stack out of the way in which first. Eradicating two extra Torx screws didn’t get me far, however uncovered to clearer view the interconnect between the actuator and the electronics on the underside aspect of the HDD:
Equally, I used to be capable of carry off the inter-head spacer meeting by eradicating yet one more screw:
What was revealed was surprising, a minimum of to me:
It is a four-platter 3 TB HDD; you’ll see the opposite platters shortly. Reflective of this reality, there are eight complete learn/write heads: a pair per platter, one per platter aspect. Doing the maths will get you to 750 GBytes of per-platter storage. However the finish of the actuator arm has sufficient websites for 10 learn/write heads; as you may see, the underside two are unpopulated on this configuration. My guess is that this similar actuator-and-arm assemblage additionally finds use in 4 TByte HDDs, the place every platter (5 complete) holds 800 GBytes of knowledge. Agree or disagree, readers?
Talking of the actuator-and-arm assemblage, regardless that I’d eliminated all seen screws holding it in place, it nonetheless wouldn’t budge. I subsequently suspected it was additionally restrained from additional attachment to its underside, so it was time to show the HDD over and proceed my disassembly job there. Six eliminated screws later, I had my reply:
Placing the PCB apart for a second, I’ll draw your consideration to a number of notable-to-me particulars:
First off, within the decrease left quadrant, you’ll see a number of extra Torx screws, that are certainly nonetheless holding the actuator-and-arm assemblage on the opposite aspect in place. Second, bear in mind the flex cable that I beforehand talked about led from the controlling electronics to the actuator? You’ll see its connector to the PCB within the decrease left quadrant. And at last, within the center are 4 terminals which press-connect to the PCB and, by way of one other flex cable middleman, permit those self same electronics to manage the platter spindle motor. Oh, about that PCB…
The again aspect is pretty unmemorable, except you’re into check factors and vias, that’s. I admit that I’ve at all times discovered it a bit odd that HDD producers go away the possibly short-circuitable (or a minimum of ESD-shockable) PCB uncovered, versus masking it with non-conductive tape or different materials previous to delivery the HDD to the tip buyer. Talking of non-conductive materials:
Nonconductive foam, plus a little bit of thermal tape atop the primary IC, initially obscures the majority of the PCB prime aspect’s contents from view. Let’s do one thing about that:
Sadly, the remnants from the thermal tape rendered the primary IC’s product markings partially illegible, so that you’re going to must take my (and my illuminated magnifier’s) phrases as gospel that that is what’s stamped atop the TQFP:
LSI *
6045
TNNFU52220
EAQ0501315 TH
Sadly, I can’t discover any significant info on-line on this chip (and LSI Logic was acquired by Avago-now-Broadcom in 2014), however I’m guessing that it’s the first drive controller chip, managing the interfaces to the SATA bus, the actuator circuitry, and (not directly) the spindle motor. One other writeup I got here throughout referred to it because the LSI MEL-B1B1 (a minimum of that’s what the second product mark line of their picture says), however I can’t discover any info related to that IC title, both.
Happily, the opposite ICs’ identities are simpler to decode. Beneath and to the left of the LSI 6045 is an IC with the product code OA73080, which seems to be the spindle motor controller, though I can’t determine who the chip’s producer is. To the precise 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 right here was from Samsung. Beneath and to the precise of the SDRAM is a 4 Mbit serial flash reminiscence (marked 25FS406, multi-supplier sourced), which probably shops the drive firmware. Bear in mind the sooner talked about “4 terminals which press-connect to the PCB and, by way of one other flex cable middleman, permit those self same electronics to manage the platter spindle motor”? Their quad-spring mate is on the backside of the PCB, under the LSI 6045. And talking of pressed-together connections, the diagonally oriented mate to the beforehand famous flex cable routing to the actuator is within the higher left nook of the PCB.
Once I turned the HDD again over, I spotted that the different air filter had fallen out. Fortunately, I used to be now capable of cleanly extract the actuator-and-arm assemblage from its confines. Listed below are 4 overview views from varied vantage factors:
And right here’s one ultimate closeup of the learn/write head stack, full with the lacking fifth pair:
Partially by swiveling out of place after which utterly eradicating the beforehand famous inter-platter spacer plastic piece, I used to be capable of get all of the platters out:
And since I now not bear in mind what order they initially stacked within the drive, to not point out the truth that I’ve manhandled all of them, I don’t suppose anybody’s going to get any knowledge off them:
With that, one other teardown concludes. I’ll now flip the keyboard over to you to your questions and different ideas within the feedback!
—Brian Dipert is Editor-in-Chief of the Edge AI and Imaginative and prescient Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the corporate’s on-line publication.
Associated Content material
