22 Dec
2006
22 Dec
'06
7:59 a.m.
Out of interest, which is the correct terminology when defining a single point
on a disk's surface - is it better to talk in terms of cylinders, or tracks?
People often seem to talk about floppy drives in terms of tracks, heads and
sectors (e.g. for an 80 cylinder floppy with two sides they'll still talk in
terms of tracks 1-80 in conjunction with a side number, even though the media
has 160 tracks in total)
Talking in terms of cylinders seems to be generally the norm with hard disks
though, and perhaps seems a bit more sensible - but remember that hard disk
manufacturers are the people who introduced decimal megabytes into common use :-)
Personally I prefer to talk in terms of 'surface number' rather than 'head
number', as a given surface could conceivably have more than one head in order
to improve media latency - but I'm unsure as to whether to use
cylinder/surface/sector or track/surface/sector. It's all just semantics, but
for the stuff I'm currently working on I'd rather go with the 'correct'
version even if that happens not to be the one most commonly used...
cheers
Jules
22 Dec
22 Dec
10:23 a.m.
Out of interest, which is the correct terminology when
defining a
single point on a disk's surface - is it better to talk in terms of
cylinders, or tracks?
Depends on whether you want two numbers or three, I'd say. :-)
People often seem to talk about floppy drives in terms
of tracks,
heads and sectors (e.g. for an 80 cylinder floppy with two sides
they'll still talk in terms of tracks 1-80 in conjunction with a side
number, even though the media has 160 tracks in total)
The medium has thousands of sectors, too, but it doesn't seem to bother
you to speak of sector number within the context of a given track; I
don't understand what bothers you about speaking of track number within
the context of a given cylinder.
Talking in terms of cylinders seems to be generally
the norm with
hard disks though, and perhaps seems a bit more sensible - but
remember that hard disk manufacturers are the people who introduced
decimal megabytes into common use :-)
They also brought us ZBR; unless you're going to the trouble to get
hold of the notch table, or you're talking about pre-ZBR disks, using
cylinder/track-within-cylinder/sector-within-track terminology, while
physically accurate, is not terribly useful - regardless of which words
you use for the numbers.
Personally I prefer to talk in terms of 'surface
number' rather than
'head number', as a given surface could conceivably have more than
one head in order to improve media latency - but I'm unsure as to
whether to use cylinder/surface/sector or track/surface/sector. It's
all just semantics, but for the stuff I'm currently working on I'd
rather go with the 'correct' version even if that happens not to be
the one most commonly used...
I'm not aware of any standard beyond that of common use, on this point
(which of course does not mean there is none). It does seem to me that
using "track" for cylinder number is likely to be confusing, because
when I've seen it used, it's been referring to track-within-cylinder,
which is your "surface" - except for floppies, confusingly enough.
/~\ The ASCII der Mouse
\ / Ribbon Campaign
X Against HTML mouse at rodents.montreal.qc.ca
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
10:46 a.m.
Out of interest, which is the correct terminology when defining a single point
on a disk's surface - is it better to talk in terms of cylinders, or tracks?
People often seem to talk about floppy drives in terms of tracks, heads and
sectors (e.g. for an 80 cylinder floppy with two sides they'll still talk in
terms of tracks 1-80 in conjunction with a side number, even though the media
has 160 tracks in total)
The problem is that "Tracks" is ambiguous - it could be "track number the
start
of the surface" or "track number from the start of the entire media" both
uses of the
term have become common enough that the meaning has to be explicitly stated.
My vote is for Cylinders to refer to the position over any surface (ie: from
Cylinder 0 - funny - many referefnces refer to this as Track-0 even if they
use the Cylinders convention otherwise), and for Tracks to refer to the position
from the beginning of the media (entire disk) - But I
know I've used Tracks
in both ways - best to explicitly state which convention
you are using.
btw - with floppies, it can be worse - NorthStar floppies are logically treated
as a single range of 35, 40, 80 or 160 tracks (depending on the type of drive)
however when you reach the innermost Track/Cylinder of Side-0, the next
logical track is the innermost track/cylinder of Side-1 - in other words, it steps
in for increasing track numbers when they exist on Side-0 and out for
increasing track numbers when they exist on Side-1 - So absolute track and
sector numbers (absolute from start of media) will be different than you might
expect if you are used to a system which advances C/H/S though it's
logical progression. This is yet another ambiguity with "Tracks".
--
dave06a (at) Dave Dunfield
dunfield (dot) Firmware development services & tools: www.dunfield.com
com Collector of vintage computing equipment:
http://www.classiccmp.org/dunfield/index.html
11:13 a.m.
Dave Dunfield wrote:
Yes, I'd be naturally inclined to use cylinders - I just wondered if there was
any good reason not to, given that (particularly in the floppy world) 'tracks'
often gets thrown around to mean 'cylinders'. As Brian says, it's probably a
hang-up from the days when floppy disks were pretty much single sided and so
the terms did mean the same thing...
Oh, that's another thing - is there a correct convention for track numbering?
There seem to be equal numbers of apps out there which treat the first track
as track zero as there are which treat it as track one :-) (Personal
preference would be to number from zero, just because it likely makes some
programming calculations easier, but I'm not sure if it's technically wrong to
present that visually to the user...)
Out of
interest, which is the correct terminology when defining a single point
on a disk's surface - is it better to talk in terms of cylinders, or tracks?
People often seem to talk about floppy drives in terms of tracks, heads and
sectors (e.g. for an 80 cylinder floppy with two sides they'll still talk in
terms of tracks 1-80 in conjunction with a side number, even though the media
has 160 tracks in total)
The problem is that "Tracks" is ambiguous - it could be "track number the
start
of the surface" or "track number from the start of the entire media" both
uses of the
term have become common enough that the meaning has to be explicitly stated.
My vote is for Cylinders to refer to the position over any surface (ie: from
Cylinder 0 - funny - many referefnces refer to this as Track-0 even if they
use the Cylinders convention otherwise), and for Tracks to refer to the position
from the beginning of the media (entire disk) - But I know I've used Tracks
in both ways - best to explicitly state which convention you are using. btw - with floppies, it can be worse - NorthStar
floppies are logically treated
as a single range of 35, 40, 80 or 160 tracks (depending on the type of drive)
however when you reach the innermost Track/Cylinder of Side-0, the next
logical track is the innermost track/cylinder of Side-1 - in other words, it steps
in for increasing track numbers when they exist on Side-0 and out for
increasing track numbers when they exist on Side-1 - So absolute track and
sector numbers (absolute from start of media) will be different than you might
expect if you are used to a system which advances C/H/S though it's
logical progression. This is yet another ambiguity with "Tracks".
Hmm, that's interesting. I'm messing around writing a floppy image manipulator
[1] that'll work at the filesystem level. NorthStar's very low on the priority
list though, so I'll just keep that in mind - when it comes to it though it
should only be an issue at the interface between the filesystem driver and
disk image driver, so it's not something that the user needs to worry about.
[1] Actually, I'm trying to avoid the word 'floppy' as there's no reason
that
it shouldn't work with hard disk images too, and probably even tapes.
cheers
Jules
2:23 p.m.
On 22 Dec 2006 at 10:46, Dave Dunfield wrote:
btw - with floppies, it can be worse...
Much worse. In fact, when talking about floppy addressing, one needs
to take into account the following:
a. The physical position of the sector/side/track on the disk.
b. The address in the header for each sector on the disk
c. What the driver calls them
d. What the operating system calls them
For example, if the OS/driver combination uses a transparent sparing
method, there is a translation between (c) and (d) that maps out bad
sectors, so the disk appears to be a continuous span of unflawed
sectors. Another example is the insistence of CP/M on 128 byte
sectors--there's a blocker/deblocker that maps 128 byte sectors to
256 byte or larger sectors.
For (b) to (c), the driver generally performs a translation from some
sort of linear consecutive addressing to whatever the floppy sector
addresses are. Usually, this is a form of sector "skewing" , but it
needn't be--I've got at least one OS sample that places sector 0 on
track 20 on a 40 track disk (SSDD, so tracks=cylinders in this case).
The sequence then proceeds from track 20 outward to track 0, then
inward from track 21 to track 39. Similarly, on double-sided
diskettes, addressing can proceed from sector, then side, then
cylinder or sector, cylinder then side. Some formats go from
physical cylinder 0 to cylinder n on one side, then from cylinder n
in reverse down to cylinder 0 on the other side. Others start at
cylinder 0 again. Some formats start on the bottom surface; others
on the top.
For (a) to (b), just about anything in the address fields will
generally work, unless a controller insists on a particular
convention. A skew may be built in, so that driver software doesn't
have to deal with it. Controllers that can do multisector will
benefit from sector addresses that are consecutive by 1, but there's
no inherent demand that it will be so.
All of which makes talking about cylinder, track and sector on
floppies extremely confusing. I try to disambiguate a bit by using
the term "physical" to refer to the absolute location of sides and
tracks, but use the header ID value to refer to sector numbering.
Thus, if the first sector on a track is addressed as 129, I'll talk
about sector 129, not sector 0.
Cheers,
Chuck
5:19 p.m.
Out of interest, which is the correct terminology when defining a single point
on a disk's surface - is it better to talk in terms of cylinders, or tracks?
I always talk in terms of cylinders, even for a floppy. Well, 'always'
apart from calling the home signal on the inteface connector 'Track00'
and the write-current-reduction signal on a 8" drive 'Track Greater than
43'. Those names are so well-known (and given in just about every drive
service manual) that to call them 'Cyl00' and 'CG43' would just cause
confusion.
People often seem to talk about floppy drives in terms of tracks, heads and
sectors (e.g. for an 80 cylinder floppy with two sides they'll still talk in
terms of tracks 1-80 in conjunction with a side number, even though the media
has 160 tracks in total)
One of the most confusion things relating to this was a manual that
claimed an IBM PC 360K disk had 80 tracks. It does (in 40 cylinders), but
it's still confusing.
Talking in terms of cylinders seems to be generally the norm with hard disks
though, and perhaps seems a bit more sensible - but remember that hard disk
manufacturers are the people who introduced decimal megabytes into common use :-)
Personally I prefer to talk in terms of 'surface number' rather than 'head
number', as a given surface could conceivably have more than one head in order
Well, if I had a drive with, say, 2 surfaces, 2 heads per surface, and
100 positions those heads could be moved to, I'd probably call it 100
cylinders, 4 heads, since logically it's the same as 4 surfaces with one
head per surface. Calling it 200 cylinders, 2 surfaces, while correct,
would seem to be as confusing as clalling it 400 tracks
-tony
6:13 p.m.
Tony Duell wrote:
Hmm, that's interesting. I suppose it depends on whether those multiple heads
are referenced independently, or whether they're only there to improve latency
(and hence from an interfacing point of view you still only reference two
surfaces, but the electronics are smart enough to 'use' the head which happens
to be nearest to the data that you want)
Personally I
prefer to talk in terms of 'surface number' rather than 'head
number', as a given surface could conceivably have more than one head in order
Well, if I had a drive with, say, 2 surfaces, 2 heads per surface, and
100 positions those heads could be moved to, I'd probably call it 100
cylinders, 4 heads, since logically it's the same as 4 surfaces with one
head per surface. Calling it 200 cylinders, 2 surfaces, while correct,
would seem to be as confusing as clalling it 400 tracks 
3:46 p.m.
Tony Duell wrote:
Hmm, that's interesting. I suppose it depends on whether those multiple heads
are referenced independently, or whether they're only there to improve latency
(and hence from an interfacing point of view you still only reference two
surfaces, but the electronics are smart enough to 'use' the head which happens
to be nearest to the data that you want)
I had a ST412, didn't remember the exact model close enough, not the
full height type, that had one platter, 4 heads. Two heads per
side. Fe oxide (orange platter).
Cheers, Wizard
Personally I prefer to talk in terms of 'surface number' rather than 'head
number', as a given surface could conceivably have more than one head in order
Well, if I had a drive with, say, 2 surfaces, 2 heads per surface, and
100 positions those heads could be moved to, I'd probably call it 100
cylinders, 4 heads, since logically it's the same as 4 surfaces with one
head per surface. Calling it 200 cylinders, 2 surfaces, while correct,
would seem to be as confusing as clalling it 400 tracks 
7:11 p.m.
On Fri, 22 Dec 2006, Tony Duell wrote:
Well, if I had a drive with, say, 2 surfaces, 2 heads
per surface, and
100 positions those heads could be moved to, I'd probably call it 100
cylinders, 4 heads, since logically it's the same as 4 surfaces with one
head per surface. Calling it 200 cylinders, 2 surfaces, while correct,
would seem to be as confusing as clalling it 400 tracks
I have a Seagate drive (225?) with a clear plastic cover (trade show demo
drive). It appears to have 17 sectors per track, 610? cylinders, 4 heads,
but only one platter! The computer treats it as if it had 2 platters,
with one head on each side of each platter.
On each side of the single platter, there are two heads!
One head "homes" (cylinder 0) near the outer edge, and the other homes
halfway to the middle.
--
Grumpy Ol' Fred cisin at xenosoft.com
23 Dec
23 Dec
6:33 a.m.
Fred Cisin wrote:
On Fri, 22 Dec 2006, Tony Duell wrote:
You know, I never realised that it was done on such 'consumer grade' drives. I
suppose the question is whether any such drive with multiple heads per platter
allows the user to control them independently, or whether the number of data
surfaces (along with cylinders and sectors) is what's exposed to the user.
Doubtless there are some oddball systems that might drive the heads
individually (such as the video replay system that Tony has), but I suspect
that for typical computer storage systems geometry is expressed in terms of
surfaces even for a drive that has multiple heads per surface.
Of course, I can just make an executive decision and say I'm not supporting
any device in my software which can't be fitted into the cylinders / surfaces
/ sectors paradigm (I've already decided that I don't want to go near anything
that has a varying block size [1] - although I can't see a problem with
variable sectors per cylinder)
[1] I can't think of anything that does vary the block size on a device
according to position; nor does it really make sense to do so from an OS point
of view - but doubtless there's some systems out there that do :)
cheers
Jules
Well, if I had a drive with, say, 2 surfaces, 2
heads per surface, and
100 positions those heads could be moved to, I'd probably call it 100
cylinders, 4 heads, since logically it's the same as 4 surfaces with one
head per surface. Calling it 200 cylinders, 2 surfaces, while correct,
would seem to be as confusing as clalling it 400 tracks
I have a Seagate drive (225?) with a clear plastic cover (trade show demo
drive). It appears to have 17 sectors per track, 610? cylinders, 4 heads,
but only one platter! The computer treats it as if it had 2 platters,
with one head on each side of each platter.
On each side of the single platter, there are two heads!
One head "homes" (cylinder 0) near the outer edge, and the other homes
halfway to the middle.
25 Dec
25 Dec
1:33 a.m.
Well, if I had a drive with, say, 2 surfaces, 2 heads
per surface,
and 100 positions those heads could be moved to, I'd probably call it
100 cylinders, 4 heads, since logically it's the same as 4 surfaces
with one head per surface.
Only if the (physical) tracks accessible by the two heads on a given
surface are disjoint. If the heads range over the same physical tracks
and exist either to just give either lessened rotational latency or to
give simultaneous access to distinct physical tracks, then it is
logically very much different from a 4-surface 1-head/surface drive,
and there is then no accurate way to describe it except as 2 surfaces
with 2 heads each.
/~\ The ASCII der Mouse
\ / Ribbon Campaign
X Against HTML mouse at rodents.montreal.qc.ca
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
6393
days inactive
6396
days old
10 comments
7 participants
participants (7)
-
ard＠p850ug1.demon.co.uk -
cclist＠sydex.com -
cisin＠xenosoft.com -
dave06a＠dunfield.com -
jpero＠sympatico.ca -
julesrichardsonuk＠yahoo.co.uk -
mouse＠Rodents.Montreal.QC.CA