[net.micro] 5.25 inch hard disks

BillW@sri-kl.arpa (05/12/83)

From:  William "Chops" Westfield <BillW@sri-kl.arpa>

I just read in "Systems and Software" that MAXCOR Inc will anounce
an 8 platter 380 Mbyte 5.25 in hard disk drive at NCC.  For comparison,
DECs new RA81s, which are 14 inch drives, hold 476 Mb.  Offhand, I
would say that its a good time to sell your stock in companies making
14 inch drives....

BillW

dmmartindale (05/19/83)

For someone suggesting that 380Mbyte 5.25 inch disks will make the
14 inch drives obsolete:  Hang on a minute.
Does this little drive have the same sort of seek times as a large drive?
The large Winchesters (Fujitsu, Ampex, DEC) have access times that are
equal to or shorter than the best of the removable-pack large disks,
and higher data transfer rates.  (The RA81 is 28ms average seek, for
example).  All of the small winchesters I've seen are slower.
The number of seeks that can be done per second is very important to
the performance of the multi-user systems that disks like the large
Winchesters are installed on, and unless the little disks can match them
in they they won't replace the large disks.
Besides, if you have the technology to build a 5 inch drive with 380 Mb,
why not use it to build a 1 Gb 14 inch drive?

tjt@masscomp.UUCP (05/24/83)

1) Most currently available (i.e. shipping now in commercial
   quantities) 5 1/4" hard disk drives use stepping motors and
   lead screws for seek positioning, and the typical seek times
   are ~85 msec (e.g. Ampex Pyxis).  In contrast, seek positioning
   for larger disk drives is done using linear or rotary voice
   coils.

2) Disk drive manufacturers are already announcing 5 1/4" hard disks
   using rotary voice coil actuators for their heads.  Average seek
   times for these disks should equal or better those for current
   8" or larger hard disks.

3) The linear bit density will also be improved on these drives by
   roughly a factor of 2 (e.g. ~32 512- byte sectors/track instead of
   current ~16-18).  However, drives like the Fujitsu Eagle have a
   bit rate some 50% higher than current SMD compatible drives.

4) Track-to-track seek times are dominated by settling time, and tends
   to be in the 3-10 msec time.  For most disk controllers, you will
   miss a disk revolution while seeking to the next track, so all that
   matters is that you can seek track-to-track in less than one
   revolution (controllers that implement a "spiral format" can do
   better -- i.e. offset sector 0 of successive tracks by just enough
   time to do a track-to-track seek).

5) Longer seeks (and therefore the average seek time spec's) are
   determined by how fast and how far the head moves.  The speed at
   which the head moves for voice-coil actuators ought to be faster
   for smaller disks (the head assembly can be shorter and therefore
   lighter), plus the heads don't have to move as far on a small disk.

6) If you have more disk surfaces (heads), you can transfer more
   data without having to do a seek.  You can approximate one
   N-head disk with K N/K-head disks, although most disk
   controllers would lose a revolution during the drive select.
   To calculate the improvement gain from an extra head:
   You can now continuously transfer N+1 tracks of data in N+2
   revolutions compared to N tracks of data in N+1 revolutions
   (the extra revolution is for the track-to-track seek).

	(N+1)/(N+2)	(N+1)(N+1)   N^2 + 2N + 1 	 1
	-----------  =	---------- = ------------ ~ 1 + ---
	  N/(N+1)	  N(N+2)	N^2+2N		N^2

   Normally you add a platter (i.e. two heads), but the effect is
   still not that important once you get ~6 heads.

Of course, other considerations of designing disks are overall size,
manufacturing costs, reliability, etc., etc., etc.  These
"intangibles" are the reason the current 5 1/4" hard disks don't have
faster seek times: a whiz-bang 5 1/4" disk would have been too
expensive for micro-computers, and too small for mini-computers.
Now that most of the development has been done for the larger disks
(e.g. better head positioners, higher bit densities), AND that
micro's are getting bigger and mini's getting smaller (you can't tell
the players without a scorecard), the technology will trickle down to
5 1/4" disks quickly.


		Tom Teixeira
		Massachusetts Computer Corporation (MASSCOMP)
		...!decvax!genrad!masscom!tjt