Hacking the Seagate hard drive

As I've explored how to tease the secretes of the password out of the xbox hard drives I've learned some very interesting (even if not terribly useful) things about the Seagate 10.2 gig (white label) drive. I figure that rather than keep all this information horded up and kept to myself I'll share it instead. I'd really appreciate some comments on this piece as it represents quite an investment in time and effort and I'd like to know whether stuff like this is wanted (or liked) by you, the modding community.

Thanks for the read :-)

-nB

UPDATED: 25JULY2004 added 2 more pics (the missing ones ;-)



Exploring the Seagate 10.2 gig hard disk drive

Well, here we are with another puzzle.
What is the current 10 gig Seagate hard drive in the Xbox?

Let me start by saying that I have never seen a drive made so in-expensively. I'm guessing that Seagate underbid Western Digital and initially provided fairly nice drives (the rubber coated ones) based on their belief that they could make the drive cheaper and thus increase the margin on the drive to the point it becomes worth it. This while under enormous price pressure from Microsoft. The result is the white label 10.2 gig OEM drive ST310014ACE.

Why do I say this drive is built cheap? Because it feels like it!
After peeling off the white plastic label stuck to the foil backing we see what those 4 bumps are: Something resembling hot melt adhesive. They are there for one of two reasons *(that I can think of, if you've got better leave it in the comments ;). Reason number one? harmonic dampening. What's that you ask? Well the drive could have some vibrations that would cause the stamped cover to vibrate. By adding these dots those vibrations would be dampened. The only other idea I can come up with is that they are there to increase the drives thickness.
I already mentioned the thin stamped cover, what I didn't yet mention is that one of the magnets for the voice coil actuator is on the outside of the case. I have never seen a drive built this way and I questioned it at first. After some thought it became apparent why they went this route. If you don't need the increased force by available from having the magnets closer together, you can avoid building a support structure for them by integrating it into the casing.		 CHASSIS_bump.JPG (123018 bytes)
Next observation:
As I remove the cover to the drive and reveal the mirror finish platter (yup, only one of ‘em), I see that the central hub and retainer clip is held in place by nothing more than a group of four clips which look initially to be holding a central (smaller) hub in compression (much like a diamond is held in a men’s ring). As it turns out these prongs are actually not holding anything. They are there for balance. In the manufacturing process these prongs are bent up or down in relation to each other to change the balance of the overall stack-up of the spindle, drive motor, platter, retainer hub, and compression ring. The retainer hub is thus held down by the retainer ring, which is in turn not held by the clips but rather is simply pressed onto the center spindle.		 HDA_Spindle.jpg (111034 bytes)
The drive actually could hold 20 gigs of data but only one head and armature is installed, thus only half the capacity of the platter is available. The GMR (head) is controlled by a TI chip (1761CD/136A456TJ); thus far I have been unable to find anything (at all) about this chip.

The armature its self is not retained by anything once the cover is off. This is another indication of the design being a (relatively) heavy cost reduction. Most drives I've come across have a retainer mechanism for the head armature.		 HDA_TI_chip.JPG (109924 bytes)HDA_armature.JPG (118896 bytes)
Onto the controller board.
The Seagate drive logic is run by one of two chips (that I have found yet). One is by ST micro and the other by Agere. Both these chips are proprietary for Seagate (with the Agere even bearing the name Seagate on it). This chip is likely a DSP with some custom logic or other relatively minor (but important) changes to it. There is a memory close by (my sample part being a Winbond industrial spec SDRAM (1Mx16 k 3.3V 166MHz 50 TSOPII). It is organized as 2 banks, each 512K deep and 16 bits wide (for a total of 2MB (that's a big 'B' so megabytes not megabits). The entire datasheet for this particular ram is available HERE. LOGIC_dsp.JPG (126508 bytes)LOGIC_dsp_agere.JPG (129699 bytes)
Finally the other major device on the board is positioned very close to the spindle motor contacts and as such likely is the motor controller. It bears the name SMOOTH and is built by ST micro. My device part number is 100222354 / B992W0317 (again a search on ST micro's website yields nothing . . . I think they have a sucky site). The only other active components are a pair of semiconductors by Fairchild, a voltage regulator and an 8 pin SOIC that is an "Integrated P-Channel PowerTrench MOSFET and Schottky Diode" (more info HERE). That's it for the active components on the logic board. LOGIC_smooth.JPG (126260 bytes)
The board its self is manufactured by topsearch, an Asia based board fabrication company.
The HDA (Head disk Assembly) connects to the Logic board through two connectors. J7, which consists of three pads and connects the "Smooth" chip to the drive's spindle motor, and J1, a connector consisting of 22 pads which has a mating connector of compression pins which connects to the GMR head and the voice coil actuator. The rest of the pins connect to power, ground, or signal (obviously) most (all?) of which connect to the proprietary DSP.
The pin configuration uses a zig-zag approach as:
1-3-5-7-9
2-4-6-8-etc.
The pinout (updated as I find out new information) is as follows:		 LOGIC_spindal_motor_contact1.JPG (115758 bytes)LOGIC_HDA_signal_contact.JPG (120005 bytes) 
+-----+-----------------------------+
|PIN# | Signal                      |
+-----+-----------------------------+
|1    | Voice Coil actuator 1(P/N?) |
|2    | Ground                      |
|3    | Voice Coil actuator 2(P/N?) |
|4    | Ground                      |
|5    | dT compensation 1           |
|6    | Ground                      |
|7    | dT compensation 2 (gnd)     |
|8    | Ground                      |
|9    | Signal pair A 1(Vcc?)       |
|10   | Ground                      |
|11   | Signal pair A 2             |
|12   | Vcc?                        |
|13   | Signal Trio C 1             |
|14   | Ground                      |
|15   | Signal Trio C 2             |
|16   | Vcc?                        |
|17   | Signal Trio C 3             |
|18   | Signal                      |
|19   | Signal Data 1? (P/N?)       |
|20   | Signal pair B 1             |
|21   | Signal Data 2? (P/N?)       |
|22   | Signal pair B 2             |
+-----+-----------------------------+
Notes on the table above:
All of the signals are guesses (some better than others) here is how I made the guesses:
Pins 1 and 3: I opened the drive and followed the trace to the Voice Coil actuator.
Pins 5 and 7: disassembly of the HDA yields a small thin flex-PCB with the connector and this component on it. It looks to be a resistor and the only reason I can think to put it in the drive case it to measure temperature (Thermistor then rather than a simple resistor )
Pins 9 and 11: These two pins go to paired traces inside the HDA and connect to pins of the mysterious TI chip. On the main logic board it looks as if one of these two is tied to VCC
Pins 13, 15, 17: This trio is routed together inside the HDA, but appears unrelated on the logic board.
Pins 19 and 21: These two signal lines made a very direct route to the DSP, were surrounded by a ground guard ring, and were given obvious routing priority on the PCB. This makes it fairly obvious the lines are important and what's more important than low signal level data?
Pins 20 and 22: These signals are paired traces inside the HDA and also appear to be such on the logic PCB.
As to the accuracy of the signals? Someone get me a datasheet on the TI part and I'll get it all puzzled out for you ;)		 LOGIC_Data_lines.JPG (110759 bytes)
In this pic you can see why I picked out 19&21 as data signals. Note the guard trace circling them (it's tied to ground).

There is a forum thread on XIC dedicated to this topic HERE. Come on over and say hello ;)