Hard drive mods

[Pentium]

So, this is pretty much aimed at LongRunner. You're the local HD guru

Since I don't know a whole lot about hard drives, I'm curious as to what mods you have done to hard drives to improve their reliability? It really is amazing how much longer the older drives seem to last, with of course the 7200.7 being my favorite (They never die!)

[LongRunner]

I haven't actually done such a thing, but I can think of two ideas for the Seagates:

• Enhanced power filtering – for the 7200.7 (PATA and original, bridged SATA versions). Each of the +5V and +12V rails have a tiny ceramic capacitor (if I was to guess, those would be C0G types filtering noise in the MHz) right next to the connector, and some more substantial caps (filtering at around PSU frequencies I imagine) a bit further away. [Looking at my ST3120026A] The 7200.7s I've seen are factory fitted with a size 3216 (metric) ceramic cap on the +5V (two on the SATA version) rail and a large(-ish) tantalum capacitor on +12V, though there is provision for an extra 3216 or 3225 (metric) ceramic capacitor on the +12V and (PATA version) +5V rails - and that's where the opportunity is.
  The current catalog for TDK's series of standard MLCCs (multi-layer ceramic chip capacitors, the full name for the type already mentioned) shows 16V X7R types in 10µF for 3216 metric or 22µF for 3225 metric. (Yes, they are capable of making electrolytics look bulky.) So you can have 20µF or 44µF, total, on the +5V input, no ghetto mods necessary. You could also remove the original tantalum capacitor (either 6.8µF or 15µF 35V - the voltage is de-rated as they can be damaged by inrush current) and replace it with a 10µF ceramic, relieving yourself of the explosive potential of a failing tantalum. (I don't advise getting 6.3V ceramics for this purpose, as they have no markings and you probably won't gain much from the higher capacitance density, any more than you want to avoid blowing them up by installing them in the wrong position.)
  And it's more than likely that X7R is overkill and X5R or X6S (which are more compact) would be fine. (No ordinary HDD can withstand 85°C even when powered down.)
• SeaSink (made up name) – for Barracuda ATA V and earlier (and also some early 7200.7s which still had the holes for the SeaShield, before they stopped drilling them), an aluminium sheet cut, drilled and formed to shape, with an area stamped inward to press on top of the MCU (with thermal compound at the junction), cooling it down (it's quite a hot chip). Especially of interest, IMO, for the Barracuda ATA IV (and III, II but those were less popular), where the MCU is in BGA (but at least with leaded solder). Or, of course, there's the ghetto version – using thermal epoxy to stick a small finned heatsink directly to the MCU. But I don't think anyone wants to permanently modify their HDD in a way that restricts its usage.

If you're going to solder on the PCB, it's best to uninstall it from the HDA (head/disk assembly) and set the HDA aside, as you don't want to break the good drives you have. (For the Seagates in question, this requires a size 8 Torx screwdriver.)

I too love the 7200.7, but the Western Digital drives of 2005~2006 are almost as good.

[LongRunner]

Attached is a photo of the PCB on my ST3120026A with coloured rectangles indicating which supply rail a part works with:
Red and yellow are obvious
Orange – +3.3V (regulated on-board, used by the SDRAM, flash and part of the MCU)
White – −5V (generated on-board by a buck-boost converter for the head preamplifier)
Pink – +1.5V (regulated on-board, core voltage of the MCU)
Cyan – +24V (gate supply for high-side MOSFETs in the SH6950? The capacitor on that rail is referenced to +12V, not ground)

The buck-boost converter is made up of the large inductor, the 8-pin chip near it (which contains a P-channel MOSFET and a Schottky diode) and the surrounding MLCCs. The small filtering inductors on +3.3V were (no doubt to save a few cents, Seagate having ended their 5400RPM U Series) replaced by 0Ω links in later production (I've pinned that change down to late 2003 or early 2004; the drive shown was manufactured 2003-10-12, while I have an ST340014A made 2004-02-25 with them jumpered out).
My drive doesn't have a discrete flash chip, but when one was used, a tiny MLCC was installed on +3.3V adjacent to it.

The large black components across the +5V and +12V rails are transient voltage suppression diodes. Those provide some degree of protection from voltage spikes, but they can fail shorted if overstressed (or defective). In data recovery they are often simply removed but for continued use, they should be replaced.

And as it turns out, you can install not one, not two, but four good-sized MLCCs on +5V – talk about overengineering.

If you need to, when setting the drive as "slave" you can park the jumper across pins 5 and 7, which are both ground (look closely and you can see that pins 1, 5 and 7 of the jumper block have thicker traces – the same goes for the ground pins on the data connector).

There are different versions of the PCB so make sure you know what you're dealing with.

[LongRunner]

A bit late, but another thing I'd suggest (this one is at no cost, and with minimal risk) is to clean the contact pads on the back of the PCB. (They are generally plated with silver or on pre-RoHS drives, tin/lead; silver in particular is prone to tarnishing which makes it a highly questionable choice of material. You can use an eraser to remove the oxide/sulfide, restoring the contact surfaces to like-new condition. However, don't use erasers on gold contacts, as it wears them out; use isopropyl alcohol instead.) In the case of Barracuda IV/V/7200.7 there are 22 contacts (two rows of 11) for the head assembly and 3 for the spindle motor (as is obvious from the photo). (You can also apply solder to them and wick away the excess – and clean up the flux after – making them as good as on the pre-RoHS models.)

[LongRunner]

And now for something a bit more adventurous:

If you look again at my ST3120026A, you can see two prominent parallel groups of 5 resistors each near the motor driver (SH6950 made by Texas Instruments). The group of 1.58Ω resistors (directly above the SH6950 in the photo, and with a small parallel capacitor) is used to sense the spindle current, while the group of 1.00Ω resistors (to the right of them, and oriented 90° differently) is used to sense the current in the head actuator (VCM). It follows that altering their resistance will change the current limit.

So this is a somewhat accidental finding on my part; I had the charger for a Nintendo Wii U Gamepad (which I plan to examine in my mini-review thread, together with the PSU for the console itself) die and figured I could use one of the 1.58Ω resistors – not from my ST3120026A, but from an ST340014A (which also had one of the ceramic caps on +3.3V already busted off, though it still worked) – to replace a blown parallel pair of current sensing resistors in the primary circuit. All that happened then was that it blew up the resistor I scavenged (plus another of the original resistors), but I did find that the drive still worked (as far as I can test it; I don't have a working way to connect to PATA drives at the moment) only taking slightly longer to spin up. (I verified the resistor assignment against this close-up photo.) I then attempted to take another one of the 1.58Ω resistors (in part to see if I could get it off undamaged), but only managed to break it . Fortunately, the PCB itself survives, so restoring the board to the original configuration isn't out of the question. And the drive is apparently still operational, only taking significantly longer to spin up than originally.

So I may have found a workable modification; if the resistance is increased (either by removing 1 or 2 of the original resistors, or by installing higher-value replacements), the drive will take longer to spin up but draw less peak current, reducing stress on the motor driver and the PSU feeding the drive. Alternatively, you could put in lower value resistors to increase the current limit, making the drive spin up faster at the expense of greater stress. (It might also help a drive with a mild case of stiction to get going, although I wouldn't do it for any vaguely critical purpose.)

This is at your own risk, of course (such tampering definitely wouldn't be covered by any warranty – though all of these drives are long past that anyway), and I wouldn't mess with the resistor bank sensing the VCM. Decreasing the current limit too far may cause the drive to fail to even start spinning, and increasing it too far may fry the motor driver (and possibly the motor coils themselves).