Hard drive reliability

[Pentium]

What is more important when you want a hard drive to have maximum reliability, low ripple from the PSU or keeping the temperature down? I know having both is ideal but what is more of a factor?

[shovenose]

Temperature matters less than you think. I'd say ripple I guess.

[c_hegge]

Yeah, I would agree. I'd say the PSU is probably more likely to be a factor than the temperature.

[Pentium]

I figured. As a rule I like to keep my HDD's below 40C if possible but that's hard sometimes with laptop drives

[Wester547]

I disagree with Shovenose. The temperature is extremely important and heat is one of the main causes of electronics failure. The whole "10*C difference doubling or halving the lifespan" rule applies to both HDDs and electrolytic capacitors.

What is more important when you want a hard drive to have maximum reliability, low ripple from the PSU or keeping the temperature down? I know having both is ideal but what is more of a factor?

They both go hand-in-hand. Higher ripple from the PSU automatically means higher HDD temperatures, at least in my experience. Ripple is essentially heat... and ICs don't like noise in the provided DC power.

What's more important than keeping the HDD temperatures down (though below 40*C is a good rule of thumb and it definitely helps) is keeping the power cycles down. Temperature swings can be a problem as the jump from hot to cold can cause expansion and contraction in the platters (rising temperatures cause expansion and falling temperatures cause contraction, exacerbating the effect in question) and result in premature wear. Too much vibration does this as well. This is how you see HDDs last longer with 24/7 use (IE; very minimal power cycling if any). As for laptop drives, you could always buy a well-designed cooling station for them, but then adding more fans adds more vibration to the laptop chassis and that can eventually be a problem too, not just for the HDD but for the BGA solder joints and the lead-free solder.

That being said, I'd keep the HDD temperature above 30*C as if a HDD isn't warm enough that can cause quicker failure than being on the warm side (the HDD's lubricant can freeze or seize to some extent if it's cold enough).

[Pentium]

Thanks for the reply Wester. That makes sense....the higher ripple equaling higher temperatures. I also do leave all my computers on 24/7. I was just worried about my hard drives failing, but the ones in my main rig are always on, always between 32-37C, and when my PSU is idle, the ripple is <15mV. They have around 10,000 running hours so I will expect them to last much longer they're both Seagate, one is a regular Barracuda and the other is a Constellation ES.

I have a spare internet surfing laptop and no matter what I do, the thing is never colder than 45C and usually hangs around 49C which really bothers me. But there is no important data on it thankfully

[Wester547]

If there's an option to enable head parking on the laptop HDD (APM) that could reduce temperatures by as much as 20%. People leave it off because the occasional "clicking" noise and load cycle count increase in S.M.A.R.T. vexes them, but that's all of merit, IMO, for the lower temperatures. Some laptop HDDs don't see much of a difference temperature wise by enabling or disabling head parking, however, so YMMV.

[LongRunner]

The whole "10*C difference doubling or halving the lifespan" rule applies to both HDDs and electrolytic capacitors.

Who said that applies to HDDs??? As for electrolytic capacitors, it's still an approximation. (I think the rule of thumb for polys is a 10x difference in lifetime for every 20°C change.)

So to say polys are more heat-tolerant than conventional electrolytics isn't totally correct. Let's say you compare Panasonic FR with Chemi-con PSA (admittedly quite an old polymer series). Sure you need three 2200µF 6.3V FRs to match a single 470µF 6.3V PSA, but PSA is only rated for 2,000 hours at 105°C and FR from 10D x 20L on up is rated for 10,000 hours at the same temperature. So if you did, following the rules of thumb would give you this:

At 85°C: FR at 40,000 hours/PSA at 20,000 hours
At 65°C: FR at 160,000 hours/PSA at 200,000 hours

The cross-over point is a bit under 75°C. But as stated, it isn't at all a fair comparison, as when Chemi-con PSA was new it wouldn't have been competing with Panasonic FR over motherboard VRMs - it would have been competing with the ultra-low-ESR electrolytics (Nichicon HM/HN, Rubycon MBZ/MCZ, Panasonic FJ/FL, Suncon (originally branded Sanyo) WG), which were rated for only 2,000 hours at 105°C. (Now that polys are no longer the expense they once were, it's not even worth making those. No one is interested in electrolytics with a low endurance rating and few other redeeming features.)

For HDDs, these are the age-and-heat-related issues I'm aware of:

• Gradual demagnetisation of the recording layers (definitely affected by heat, but to what degree???)
• Evapouration of bearing lubricant (but I'm not aware of this actually becoming a problem to date)
• Uneven thermal expansion and contraction eventually leading to micro-fracturing of inflexible materials (pretty sure the media structure would be most vulnerable to this)
• Also as an effect of thermal expansion and contraction, uneven preloading of ball bearings (resulting in minuscule, albeit not insignificant, flattened contact points) (fans for example use springs to provide a constant preload, independent of temperature changes, but HDDs can't do that)
• Off-gassing of internal components (materials used inside HDDs have to be carefully selected to minimise it)
• Eventual wear-out of the protective layers on the media surface due to the proximity of the flying heads (how long does this take???) (what effect does temperature have on this???)

I don't know what effect temperature would have on the heads, short of them becoming so hot they can't operate until they cool down (as sometimes happens in minor head crashes that don't outright destroy the heads).

[Wester547]

So to say polys are more heat-tolerant than conventional electrolytics isn't totally correct. Let's say you compare Panasonic FR with Chemi-con PSA (admittedly quite an old polymer series). Sure you need three 2200µF 6.3V FRs to match a single 470µF 6.3V PSA, but PSA is only rated for 2,000 hours at 105°C and FR from 10D x 20L on up is rated for 10,000 hours at the same temperature. So if you did, following the rules of thumb would give you this:

The endurance test isn't a wear out test. It's just an industry standard test. It is not the be all end all of capacitor lifetimes. Manufactures don't actually publish those. What -can- be said is aqueuous electrolyte expands much quicker and dries up much faster, especially with added heat, than non-aqueuous electrolyte does (electrolyte that still uses organic solvent of sorts in the form or quaternary ammomium salt compounds but that is much more resistant to heat, even if not quite as resistant as truly conductive and solid polymers).

The cross-over point is a bit under 75°C. But as stated, it isn't at all a fair comparison, as when Chemi-con PSA was new it wouldn't have been competing with Panasonic FR over motherboard VRMs - it would have been competing with the ultra-low-ESR electrolytics (Nichicon HM/HN, Rubycon MBZ/MCZ, Panasonic FJ/FL, Suncon (originally branded Sanyo) WG), which were rated for only 2,000 hours at 105°C. (Now that polys are no longer the expense they once were, it's not even worth making those. No one is interested in electrolytics with a low endurance rating and few other redeeming features.)

Again, those are only industry standard tests that expressly state the capacitor series in question will last -at least- this long at a certain temperature, at maximum ripple rating, and at 105*C, at a switching frequency of either 100KHz (105*C and low ESR capacitors) or 120Hz (85*C or general purpose capacitors), with the ESR rating stated at either 20*C or 25*C and -10*C respectively as those would be the temperatures the capacitors were restored back down to. It is never stated that they fail after those amount of hours. True conductive solid polymers at least don't have any liquid to dry out in them at all. Hybrid/functional polymers have very little water in them. Organic polymers are more of a cross between electrolytics and true polymers, but can still vent and bulge just like electrolytics with liquid in them, hence the safety vents. Also interesting about the 10*C drop = 2x the life is that in the KZG datasheet, Chemi-con state that for every 5*C the life is doubled... that also sets in stone, at least IMO, despite the fact that KZG is a crap series, that aqueous electrolyte is definitely more sensitive to heat. Also, the test conditions and prerequisites the endurance lifetimes have to conform to are rather loose - 200% or less of the dissipation value (ESR).

[*]Gradual demagnetisation of the recording layers (definitely affected by heat, but to what degree???)

That's a huge one. Hard drives a very fragile devices and are not meant to be abused, though admittedly at this point are somewhat dated technology. The platters expand and contract with even somewhat rapid temperature increases and drops, and that leads to their wear out quicker than not.

I don't know what effect temperature would have on the heads, short of them becoming so hot they can't operate until they cool down (as sometimes happens in minor head crashes that don't outright destroy the heads).

Head crashes are definitely another potential problem.

[LongRunner]

The endurance test isn't a wear out test. It's just an industry standard test.

I know that.

...or 120Hz (85*C or general purpose capacitors)...

I don't think there's any such thing as an 85°C low ESR capacitor.

Also interesting about the 10*C drop = 2x the life is that in the KZG datasheet, Chemi-con state that for every 5*C the life is doubled... that also sets in stone, at least IMO, despite the fact that KZG is a crap series, that aqueous electrolyte is definitely more sensitive to heat.

Not really - they say the same thing for their non-aqueous series (LZA, LXZ, LXY, LXV).

As for HDDs, given how variable their lifespan is to begin with, I think it's misleading to state any rule-of-thumb about how operating conditions affect it. Run it in a PC with inadequate RAM (thus accessing the swapfile excessively) and no case fans (resulting in overheating), slide it across the floor, power it on and off five times a day, hook it up to a PSU that's gone so bad it can barely turn on, and if it's a quality HDD it could still outlast one that's run non-stop in an air-conditioned room, in a PC with eight powerful fans, powered from an X-series Seasonic, but badly made. Of course, the keyword being could, this won't necessarily happen. But it's surely possible.