In one of my earlier posts, I salvaged two IBM Deskstar 75GXP DTLA-307030 30.7Gb hard drives. One of them was sick – to the point that recovery just wasn’t economical timewise, so I decided to take the opportunity to take it apart and photograph it. Of course, I wasn’t expecting clear glass platters, as this drive didn’t seem too sick, but I wanted to have a chance to show exactly what is inside a modern hard drive.
Lovingly emblazoned with ‘deathstar’ in permanent marker to distinguish it from its still-functioning brother.
Hard disk disassembly has been covered to the death online, this is just one more to add to the pile, I guess. The main difference is that this one is written by yours truly, and I’ll do my best to add some commentary along the way. I’ve also used a more decent camera (my Nikon D3200 with Tamron 17-50mm f/2.8) and lighting (fluorescent for focusing, and three flashguns on bounce for the shot).
Drive labelling back then was quite a detailed affair – lots of information there, but most importantly is the November 2000 manufacture date which gives us a reference for its age. While it can be said that it is hardly modern, IBM was at the forefront of hard drive technology back then, and some of their elements will be evident as we take a tour of its innards. The drive itself is made in Hungary, according to its label …
The rear of the drive shows a very complicated jumpering diagram. These Deskstars featured the widest compatibility jumpering system I had ever met – first of all, there is an option for 16 head and 15 head geometry to get around BIOSes with 16 head bugs. Then there is the 2Gb Clip feature which causes the drive to report itself as a 2Gb drive to get around capacity limitations. Finally, the Auto-Spin Disable is a feature used with arrays of these drives and RAID controllers to enable staggered spin-up. There is the obligatory Master, Slave and Cable-Select settings, but there is also a “Forcing Dev 1 Present” feature which is useful when dealing with slightly incompatible ATA/ATAPI slave devices. Lots of settings. The drives shipped with white jumper shunts – someone had probably lost one in the case of this drive and replaced it with a blue shunt. Of course, this is an IDE drive, and in 2013, IDE is well and truly almost irrelevant.
The underside shows the drive with a reduced length PCB. There is a cache RAM, a motor drive and a head driver. One of the improvements it seems is the use of a minimal flash RAM (I’m not sure if that’s a serial flash chip next to the main controller), as these drives stored many of their firmware elements in a service cylinder on the actual platters themselves, eliminating the need for a large flash chip for the firmware and reducing cost. Unfortunately, this meant some of the buffer memory was “consumed” by the firmware – so an advertised 2Mb buffer on IBM Deskstars had less which was usable. There is a three-phase spindle motor from PM-DM (Precision Motors Deutsche Minebea) connected by flat flex. Around the spindle motor is a silver plate of material – this is a type of diamagnetic material which resists intrusion of magnetic fields into the platter area and is adhered to the black aluminium drive frame. Interestingly, the main IC is covered with a black plastic surround – this may be to prevent accidental contact damage or ESD – it is attached simply by adhesive tape and can be easily removed like so.
The assembly is accordingly made in Romania. The other side of the PCB has something to tell us …
Apparently the PCB is made in Japan! The two rows of solder-surfaced contacts are used to connect to the head assembly inside the drive through a springy bed-of-nails like connector on the drive frame.
The connector is visible in the lower left corner. There is a plastic-foam layer to prevent PCB contact and shorting to the frame, as well as providing some static resistance possibly. Underneath this layer, we can see a set of screws hiding!
It is quite common in many hard disks to have screws buried in many different places. It is essential to have a good Torx screwdriver set covering multiple sizes, as well as a bit of experience in disassembling drives (that’s if you wish to do so). I’ve dismantled many (45+ drives) so far, so this doesn’t surprise me. But these guys at IBM have taken additional care to cover their screw heads with an adhesive plastic cover – this may be to prevent air flow through these screws which may contaminate the disk chamber, or to prevent shavings from coming out and shorting electronics. Or just to present an even surface to the foam.
Hidden screws abound in the front label as well. In this case, there were two partially covered by the label, and one completely covered. These were covered with plastic adhesive circles – probably to even the surface for better label adhesion. I’ll take a moment also to say that this drive has a screw right on top of the spindle – so the spindle core itself is attached to the cover providing additional rigidity and support. This idea was also “rediscovered” by Western Digital, later on in the life of hard disks. I would also take a bit of time to comment on the “arc” shaped cutout covered by adhesive aluminium foil – the one near the word Deskstar on the label.
And also this aluminium foil piece covering the side of the drive. You would have seen many of these around on your drives – these cover cutouts which are used during the manufacture of hard drives. These holes allow instruments such as the Servo Track Writer (STW) to move the heads, measure the radial distance, and record the initial servo data which defines each and every track on the platter surface. This data remains on the drive for life and can only be properly produced using very expensive equipment which can define track widths into the um range.
Removing the Torx screws around the cover, and giving it a pry, and we’re in. Note that other drives may have nuts securing the head bearing to the cover, and some drives (Fujitsu, Quantum) may use many Philips on the outside and Torx on the inside. Some older Seagate Medalists didn’t use any screws for the cover, merely metallic adhesive tape.
The platter is shiny as a mirror, but dust already begins to settle on it as soon as it’s open. The drives themselves are built in a cleanroom to ensure the highest level of cleanliness, but are not sealed. The drives have a breather hole, which is backed with an air filter, responsible for keeping the atmosphere inside the drive clean-enough for operation over the lifetime of the drive.
That’s a much nicer picture. So what can we see here? On the lid, you can see the cut out for the servo track writer’s arm positioning equipment. You can also see a clear plastic round device which is the air filter for air coming in via the breather hole.
On the bottom part of the drive, the large circular disks are the platters which store your data. A number of these (in this drive, two) are secured to the spindle using the platter clamp (this one containing six Torx screws) and a number of spacers (grey, just underneath the platter clamp but also normally made out of aluminium).
There is a small cutout on the bottom left with a white “pillow” inside – this is a recirculating filter channel which takes off some of the airflow flowing around the drive platter and passes it through a filter thus continually cleaning the air.
On the right side of the drive, in the bottom corner is the magnet assembly which is responsible for head positioning along with the voice coil attached to the pivoting head arm. This forms the voice coil motor which is used to position the heads radially on the disk. These magnets are one of the big prizes (aside from the shiny platters) which you get from disassembling the hard drive. These magnets are very powerful and can come in quite handy for a variety of things. To the bottom right, there’s a back square cutout which a rubber piece is fitted to – this piece also has a small square magnet on it (IBM specialty) which is used to hold the armature in the parked position for transport.
In the photo above, the head arm pivot bearing is on the right, with the arms being rigid up to about 3/4 length. After that is a thin aluminium suspension which applies pressure towards the disk, and it has a suspension which holds the heads themselves. The heads are connected by four, very thin wires, to the flat flex which (sometimes, but not in this case) has a head preamp and leads the signal to the PCB on the outside rear of the drive.
IBM showed their prowess in these drives by producing media that was capable of much finer magnetization domains than their competitors, and producing media based on glass platters which were supposed to be more stable and less prone to expansion/contraction and warpage. Their head technology was also progressing quite quickly with GMR heads being one of their specialties. Finally, they pioneered the mass usage of ramp load and unload technology. This drive shows this clearly – the heads are parked safely in a plastic structure, off the disks, separated from one another and secured against vibration.
This was considered an expensive solution, and many drives of the time relied on Contact Start Stop (CSS) which had the heads rest in a landing zone in the centre of the platter. This posed numerous problems as head damage would accrue over time due to the heads “dragging” against the media until sufficient speed and airflow were developed so the heads could fly on a film of air. It also resulted in lost capacity as some of the platter was designated a “landing zone” where no data would be stored. It also resulted in drive fragility – the G force handling of the drives were reduced as head-slap could damage heads and roughen the media causing accelerated head wear on start-up. Ramp-load and unload was a superior technology which increased the start-stop numbers quite significantly by keeping the heads from ever contacting anything! They’d be safely held apart on the ramp until the disks were ready to receive them, and when the drive was idle, the head ramp provided a safe area to park the heads temporarily to reduce power consumption (less air friction) and reduced risk of media contact and damage due to vibration.
This technology took a while to be adopted, but was initially adopted on laptop drives where shock resistance was an issue, but is almost universally seen on all drives nowadays. The head load-unload sequence does make a slight quiet clunk/click sound and some people object to that on laptops – but the benefits are numerous! Depending on the idle timer setting, there could be adverse reactions sometimes where the heads load and unload rapidly causing accelerated ramp wear and that may result in plastic particles contaminating the drive (the main failure mode problem with ramp-load technology).
Here you can see the heads are kept apart, the bottom head facing up is visible in the image. These things are small, and so are the wires – this required extension tubes to get a good shot of!
I even folded back a head to get a shot of the surface. The geometry is very carefully designed to maintain a consistent flying height in a variety of situations (temperature, air pressure and density). You can see the wires very carefully bonded to the head slider itself at the top, and a very special multiple level construction of the suspension arm.
So there we are – a hard disk, taken apart. Hope you enjoyed the photos!