Conference hydra::amiga_v1

I think this is a somewhat controversial issue...

Any electronic device undergoes stress when powered up/down.

when a disk drive (at least RD type class) is "off" it's heads
are resting on the media. When it powers up, the heads ride on the media until
sufficient air flow allows them to "fly". So there is some additional wear
and tear here. Also, some disk's use whats called "thin film media" which
has a funny lubricant. This media is succeptable to "stiction" - the heads
get stuck to the media! Although modern technology is curing this, it exists
over some small population and is usually aggravated by loooong periods
of being "off" (ie heads on media). RD54 has this stuff and we have
done both long term shelfing and start-stop testing (I believe 40-50,000
start-stops and keeps on ticking...)

In general, I guess its up to you. Way the possible effects agains noise/power
consumption and then take a good guess :-)

All disk devices (and other things as well) have an MTBF (Mean Time Before
Failure) associated with them. This means that after some time (typically 15 -
30K hours) a drive is expected to fail (although we've seen 40-50K in reality)


Hope this helps!

John

    FYI 40K hours MTBF is approximately 4 1/2 years.

    It is in general a bad idea to spin down winchester drives of any
    kind.  This is due to seeral factors:
    
    1) When drives spin down the heads land on the surface.  There is
    a measurable probablility (albeit small) that the heads will land
    too hard - basically a head-crash (contact event as its known in
    the trade).
    
    2) There is lubricant on the surface of the platter.  In some cases,
    the heads may stick in the lubricant (stickision) and have bad
    consequences as the drive spins up.
    
    3) Because of the centripital forces, the lubricant will migrate
    to the outer edges of the platter.  After a long time, there may
    not be sufficient libricants to prevent the heads from doing damage
    on landing or takeoff.
    
    4) The lubricant may glob causing a bumpy ride down for the heads
    - essentially not good for the heads.
    
    
    In the for what its worth department:
    
    MTBF is the MEAN time to failure, not how long it will take for
    YOUR unit to fail.  It is a statistic that is derived from the
    behaviour of the population of units.  40K hours says that you have
    a probability of one failure in an average of 40K drive hours.  Some people
    may see lots of failures early in the life of the products.  Other
    owners will not see a failure ever.  Put another way, if you have
    4 drives with a 40K Hr MTBF, you can EXPECT a failure on the AVERAGE
    of once every 10K hours.

    BobW

I understand the merits of leaving the HD running, but how does one trade that
against the risks of damage from a nasty power failure?  Is it possible to get
power conditioners (cheaply) which shut down and wait for manual reset after the
first power failure?  I think that would be the best strategy for a home
computer, at least the way I use it.

	Paul

I agree with this 100%.

When a drive is powered down, it's heads are parked WHERE NO USER DATA IS
LOCATED.

When a drive is left spinning, the heads can be ANYWHERE. If the
drive/controller don't properly handle power failure (and bad glitches etc...)
then there is a MUCH greater potential for data corruption than if the drive
is powered down.

John

    With the SupraDrive they recommend "parking" the drive head before
    moving the drive.  Would it be worth it to run "PARK" everytime
    a shut the power off?

Most drives park themselves at power down

John

    re.: -1
    	Oh, really?  Which ones do?  Do you mean most types of drives,
    or most of the installed base of drives?

                     
Winchester drives in general. 

Of the several RDxx drives DEC sells (RD52,RD53,RD54,RD31,RD32,RD33) 
only one (RD31) must be manually parked.


The basic method used is to re-direct the back-emf generated by the
spindle motor into the voice coil actuator to drive the heads
into a "landing zone"


John

I've watched the head motor as I've removed power from an RD50 and RD51 hard
drive...neither one seemed to automatically "park"; however, both did move the
heads to an intial position (track 0?) when I *applied* power to them.  Note
that I did not have any controller connected to either when I did this test.  I
would have tested an RD52 but am not sure where to look to detect head movement.

Right, RD50 and RD51 will NOT automatically park. These drives are (relatively)
old - especially as far as technology is now.

RD50 and RD51 have stepper motor actuator's. That's why you can "see the heads
move". This method is slooow.

RD52 was one of the first (5�") drives to incorporate a Linear Voice Coil
Actuator. This is very similar to the operation of a speaker. There is
a permanant magnet fixed to the HDA (head and disk assembly), and a voice coil
(coil of wire to form an electro-magnet) fixed to the actuator. (BTW - an
actuator is the framework which holds the heads (and flextures) and the voice
coil). 

This actuator is mounted on a pivotal axis, such that interaction
between the voice coil and the permanant magnet cause a  "seek".

You can not see the heads on this type of drive since everything is sealed
inside the HDA.

All hard drives will, upon power UP, recalibrate themselves. This is where
the drive finds track zero, and stays there till the host tells it otherwise.

A neat "trick" used by drive manufacturers these days is to re-direct the back
EMF generated by the spindle motor (the motor that spins the disk platters)
into the voice coil to allow the drive to automatically park it's heads.
The Landing Zone (as it's called) is typically at the inner most section
of the disk platters - away from user data!

Disk drive technology is amazing. The vendors are doing some truely unique
stuff. How about a 765 megabyte, 5�" drive, ESDI interface at 15 megabit
per second transfer rate (1.875 megaBYTES!). This is top of the line stuff.
Now, SCSI has potential to reach 1.2 - 1.5 megabytes in asynchronous mode, 
or 4 (some vendors claim 5) megabytes per second in synchronous mode!
And all that on an 8 bit wide data path. Wait till SCSI-2 hits the market - 
32 bit data path, with built in (to the architecture anyways) caching commands!

Not only  that, but these transfer rates allow packing mucho data in a small
space (hows a 200 Megabyte 3�" drive grab ya?, you can just about fit
4 of these little buggers in a 5�" slot - thats 800 megabytes and on 4
different spindles!)


John
John