18 Jan
2011
18 Jan
'11
8:44 p.m.
Chuck writes:
On 18 Jan 2011 at 14:18, Shoppa, Tim wrote:
> Isn't it wonderful the way the pendulum
swings? First parallel
> cylinder select (e.g. SMD and other bus and tag type address
> interfaces), then step in/out cylinder select (e.g.
> SA4000/SA1000/ST-506) and serial cylinder select (e.g. ESDI), then
> parallel cylinder select (e.g. IDE/ATA/SCSI), then serial cylinder
> select (e.g. SATA).
It would seem to be more a matter of "smarts on
the disk drive" than
anything. IDE/SCSI and SATA all have controllers on the drive and so
have no dedicated signal lines for positioning. In particular
IDE/SATA LBA and SCSI don't require any knowledge of the drive
geometry, other than the sector size and the total number of sectors.
with ST506/412 and ESDI, the "smarts" were still external to the
drive.
There's a difference between what the interface doesn't require a drive
To do, and what the most common drive implementations actually do.
(I do agree that IDE LBA and SCSI, allowing a disk to be addressed as a
"big bunch of blocks", are substantially more sophisticated than purely
Physical interfaces.)
You might be surprised at how smart many drives were even though they were
Using the "not so smart" ST1000, ST-506 or floppy interfaces.
e.g. the cabling may only have "Step in" and "Step out" but on the
drive
the drive itself is counting which cylinder its in to apply appropriate
precompensation. E.g. LSI Shugart floppy drives.
Any ST-506 non-stepper-motor drive knew what cylinder it was on thanks
To internal logic.
The SA1000 explicitly defined "buffered step mode" whereby the step
Commands on the interface could come in way faster than the physical stepper
Could move, and a counter on the drive played out the steps later. So on
The very first implementation, the drive was already smarter than the interface
Would imply.
Tim.