Date: Mon, 25 Nov 1996 11:19:33 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: msmith@atrad.adelaide.edu.au (Michael Smith) Cc: terry@lambert.org, joelh@gnu.ai.mit.edu, grog@lemis.de, chat@FreeBSD.org Subject: Re: ATAPI (was: Who needs Perl? We do!) Message-ID: <199611251819.LAA22823@phaeton.artisoft.com> In-Reply-To: <199611240424.OAA24031@genesis.atrad.adelaide.edu.au> from "Michael Smith" at Nov 24, 96 02:54:53 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > They are standard drives which do not have an off cycl;e for thermal > > recalibration. > > This is not strictly true. If they are actively being used, it's true. > > This makes them faster to dump an incoming stream of "A/V" data > > and turn around for more data, but it makes them much more sensitive > > to thermal variance. > > An incorrect conclusion, a conjunction and a falshehood. I'm just quoting the Seagate literature for the one I bought... 8-(. > > If you have a machine you leave on all the time, and you scsiformat > > after it has reached thermal equilibrium, and never remount after a > > crash until it is, again, at thermal equilibrium, and you maintain > > a standard thermal profile with consistent ventilation to a controlled > > environment in which the machine is placed, you can use them all day > > with no difference, except they are slightly faster over a bursty short > > haul. > > This is derived from dream-delirium facts, and should be tossed out > before either baby or bathwater. If I have equipment sensitive to thermal variance, and I limit the amount of thermal variance, then the equipment will not react. One way to do that is to not change power conditions (ie: leave the thing on all the time). Another is to not vary the amount of thermal dissipation the device gets over any given time period. Ie: keep a constant thermal conductivity in the air by controlling the humidity, and providing a constant temperature thermal sink into which the device can radiate (constant ventilation rate, constant temperature for incoming air, etc.). > > If you don't do any of these things, they are slightly faster over a > > bursty short haul, but they have a *significantly* decreased MTBF. > > This was probably derived from the above observations, but may be > true regardless. It is certainly speculation. Just like inadequately ventilated hawks don't go belly-up. 8-). > Some theory : > - harddisk platters are made of aluminium. > - aluminium expands and contracts corresponding to temperature. > - harddisks generate quite a lot of heat. > - disk control logic uses a variety of techniques for getting the head > to the right place on the disk as quickly as possible. One of > these techniques calls for guesstimating "about where" to sling > the head before actually looking at the disk to see where it is. > - in order to compensate for the expansion/contraction of the > platters, the drive logic performs a periodic operation known as > "thermal recalibration", where it hops the head across the disk > comparing where it thinks the head should have landed with where > it actually _did_ land, and updating it's idea of what is where > accordingly. > - in 'conventional' disk drives, this process is uninterruptible, > and can take hundreds of milliseconds. > > So, to deal with the "AV" crowd, whose hardware often can't handle > being starved of data for several hundred ms, drive manufacturers made > the recalibration process interruptible, so that data operations > continue and recalibration occurs in the "background". > > Even with the use of a servo surface, it is not practical to abandon > thermal recalibration at all. Sorry if "don't have an off cycle" implied this for you... it was not my intent. Obviously, you could not use the drive at all immediately after power on (before heat up -- before thermal equilibrium is established) if the driver were never calibrated and only correctly indexed at equilibrium. But the problem with doing writes during thermal slew is very real; there is always a window, especially if the drive is being actively used, where it will be "hotter than you are calibrated for". There is also the problem that, if the calibration is interrupted, the thermal slew is not radially constant, and therefore the calibration would be expotentially off, with a curve related to the difference between the format temperature and the temperature at the time of the operation. Since you don't format at a fixed temperature, the drive logic can not estimate this curve accurately in the case it has been interrupted. One potential fix would be to use a thermister to know the temperature at the time of the format, and calculate relative to that. I don't know any drives that have this. This still assumes that the heat dissipation is constant and invariant over time, or there is an increasing window for failure. So ventilation and constant temperature of incoming air flow for cooling are required to ensure best operation. The Seagate pamphlet suggested an external case, in fact. Regards, Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199611251819.LAA22823>