It is funny that sometimes your expectations are not high prior to the teardown. But under the microscope you see things you have never seen before. This was the certainly the situation here.
This is a 160Gb hard drive with a firewire (IEEE1394) interface.
On the rear are two firewire connectors, a DIN (low voltage) power connection and an on/off switch
But this is not just any hard drive but a LaCie model designed by F.A. Porsche! 😉 I Googled it and its the same Porsche who designed the iconic 911. It has always made me smile – why would a famous designer want to design a box? And why would somebody at LaCie think its was worth paying for a famous designer. Did they really think it would sell more drives?
A little anecdote, back in the 80’s we used to call hard drives Winchester Hard Disks. It was common knowledge that they were called that after the town in Hampshire, England where IBM had a plant that developed and manufactured the first drives. Apparently the Wikipedia history of hard drives disagrees and claims it was a code name chosen based on the winchester rifle.
Four screws and a bit of prying you can pull the standard Western Digital IDE hard drive out of the steel frame
Removing the connectors and pulling the interface board out on one side just connectors, passives and a 24.57Mhz crystalOn the rear you can see three chips. One (FW911PLUS) by Oxford Semiconductor who I have never heard of, a bit of research they were established in 1992 and made interface chips (This one is for IEEE1394) and were acquired by PLX Technology in 2009. This is beside an Agere FW802C which is a two cable IEEE1834 transceiver/arbiter. The only other device on the board is the 3 pin AK33174G5 which is likely a voltage regulator.
Turning to the WD1600 drive the board has a 4 terminal connection to the spindle motor and is screwed into the case.
The board is set into foam, and the connection between read/write actuator and the board is just a 20 pin press fit. Hence the need for the screws and the foam.
- Samsung K4S641632H 64Mb DRAM – buffer
- WDC WD70C26 controller – and above this the small 8 pin ic is
- ST 25P10AV a 1Mb serial Flash
- Marvell 88C6590-LFE is the read-write channel IC
- ST L6283 SMOOTH is the spindle motor controller
Turning the board over, you need to remove the five screws, and the large label, and the small sticker hiding the sixth screw.
When you take the lid off you see the drive as it is commonly seen (But with my camera reflection visible in the highly reflective magnetic disk)What you see here is the top of the actuator arm that in this case holds four read/write heads, as there are two disc on the platter that are read on both sides. This actuator arm is scanned across the disk using a frictionless voice coil actuator. There is a coil in the actuator arm between two strong magnets, as current in the coil is varied the Lorentz force swings the arm around the pivot. Above the magnet you can see back of the connector to the PCB.
You can see the coils mounted into the plastic of the actuator arm
For completeness this is the spindle motor that is attached to the base by three more torx screws. I will keep this motor as I am sure there is something I can use a 12V 5400rpm motor for.
On the side of the actuator attached to and obscured by a flex ribbon cable (Which I have cut off here) you can see a single flip chip ic. This must be the hard drive pre-amp that I will look at later on.
Here is the read/rite heads looking at the side of the actuator arm. When you take them off of the platter they lightly spring together.
Under my microscope imaged with a 5x objective (This is a focused stack of ~60 images). I do not know what the purpose of the ‘solder’ balls on the end are. At first I thought this was the read/write head, but now realize that is wrong. They might be a counter weight to help keep the head close to the disk?
Separating them and taking one apart this is looking down onto the read head
You can see the four connections on a micro flex ribbon cable that attaches to the pre-amp die. The black rectangle is called a slider. The slider enables the head to travel across the disk with just a few nanometers of separation. They work on the same principle as a hovercraft, the metal pattern you can see on the underside here, causes an air cushion to be created as the arm swings across the disk that is spinning at 5400rpm. The air cushion is self correcting such that the buoyancy of the head keeps it ~5nm-10nm off of the smooth disk surface – quite amazing really.
Now I was puzzled at this point – I know the head is basically a miniature electromagnetic coil that creates and reads magnetic dipoles on the cobalt coated disk but I cannot see the coil. In this picture (and pretty much any picture I have seen on the web) you cannot see the actual read/write head. I thought the coil might be built on the underside of the slider, I tried a silicon ic de-cap process but it showed nothing more about the slider. On a hunch I wanted to see how the gold bonds were connected up so I tried mounting one vertically and looking at the end.
So thats the actual head, searching for pictures of the read/write head I did not find any examples of the head imaged this way – as I said in my opening sentence something I have never seen before.
The read/write head is patterned onto the end of the slider which is just 0.3mm tall by 1mm wide. Looking carefully at this it is apparent that the coil is actually patterned onto a slight bevel at the top of the slider. This was very difficult to image – I tried adjusting the angle of the die (This made the lighting of image cause some glare) where you can sort of make out the bevelled edge (This is a composite focus stack of 133 images btw)
This is a small think film circuit of 3/4 layers. You can make out the two connections on the left come in and form a coil and a base, the two on the right connect through the green squares, I guess they must sense the induced current in the coil.
The coil is just 40 μm wide with 4 turns and a line/space pitch of just 3 μm. So they have patterned a metal track of 1.5 μm width that has one third of it on a ~45° angle. It also looks like there is a piece of ferrite at the centre of the coil. (This would significantly enhance the electromagnetic affect on the coil.) I am impressed how they fabricate this small head, and wish I knew more about the process. Do they form them on flat substrates like ics then singulate and attach to the end of the slider?, I could see no evidence that this was attached (Or not an integral part of the slider). Or do they pattern them on a thick block then singulate and then pattern the slider base?, or are they individually patterned on the end of the already formed slider base?
I decapped the pre-amp that is on the actuator arm and expected to see a fairly simple (4 channel) amplifier. Well this is not a simple amplifier, its a 4.77 mm x 3.12mm made on what looks like a 0.8μm BiCMOS process. There is still some residue of the underfill material present despite this having repeat extended exposure to hot sulphuric acid. Since you can make out the main structure of the die, I did not want to risk damaging the die by another acid soak.The chip is made by Texas Instruments. Searching for SA1774 or any derivatives does not turn anything up which is quite unusual for TI. A couple of years prior to 2001 there is a press release about differential pre-amplifiers for HDD. But it does not look like TI are still in this business segment.
The layout of the die is quite relaxed with distinct blocks of circuitry and lots of grey space between them. I guess even in 2004 the throughput of this amplifier is over 1Gb/sec hence the need for the complex circuitry.
On inspection one thing intriguing about this die is the metallization. Mostly chips are using copper metal layers (Modern chips) sometimes with Aluminum top metal to bond to. Or they are using just Aluminum (Or actually Aluminum with a little bit of silicon alloyed) mostly in older chips. Occasionally Tungsten has been used for metal interconnect. This chip appears to have three different metals for metals 1,2, &3
The top metal (M3) the white tracks are clearly Aluminum. Metal 2 is the dark tracks vertical tracks, given there appearance I think they are likely Tungsten. Metal 1 is the yellow horizontal tracks. They are also Aluminum but appear different due to the dielectric insulator covering them. So the chip has an Al/W/Al metallization scheme which is pretty unusual.
Hard disk drives are an exquisite masterpiece pushing the boundary of mechanical design as well as high speed digital processing to read and write out the huge data streams. Although I have seen many images and teardowns before, it was fascinating to actually see and image the drive head.