Re: System disk specific to the IIc+?

[Warren Ernst <wernst@gmail.com>]

On Feb 2, 3:38 pm, demp...@actrix.gen.nz (David Empson) wrote:
> <touchstone> wrote:
> > On 2008-01-31 08:15:58 -0700, schmidtd <schmi...@my-deja.com> said:
> > > And I now know why GSOS gives you the option to format 800k floppies
> > > with 1:2 or 1:4 interleave.
>
> > i have seen that but don't know what that means.
>
> When writing to a floppy (on an Apple II or Mac), the data has to be
> converted from its normal structure into an encoded form which is able
> to be stored reliably on the disk. This process takes some CPU time, and
> the writing process is also time critical, so the normal method is to do
> a two stage "encode, then write" sequence.
>
> The floppy drive is spinning all the time (once the motor is activated),
> so if the sectors on the floppy are adjacent and the computer wants to
> write two consecutive sectors, then after it has finished writing the
> first one and encoding the second one, the drive has already rotated
> part or all of the way through the next sector. This means it would have
> to wait an entire revolution of the drive before it can write the next
> adjacent sector.
>
> This would cause a major performance hit, so to improve performance, an
> interleaving technique is used. Sectors which are logically adjacent
> (numbered 0, 1, 2, etc.) are spread out with other sectors between them
> (on each track). The degree of separation is given in the interleave
> factor. 1:2 means that logically adjacent sectors are at least two
> physical sectors apart, and 1:4 means that logically adjacent sectors
> are at least four physical sectors apart.
>
> The reason for having different interleave factors is due to different
> overhead for various drives, and CPU performance.
>
> If the computer is fast enough and it has direct write access to the
> drive mechanism, then it is possible to write to a 3.5" disk with 1:2
> interleave at maximum speed, because the encoding can be completed in
> the time it takes one sector to pass under the head.
>
> If the computer is slower then the encoding process takes longer and
> more than one sector will have passed under the drive head. It is
> therefore necessary to use at least a 1:3 interleave to get best
> performance, and in some cases, 1:4 may be required.
>
> Some drives have even more overhead. A notable example is the UniDisk
> 3.5. This actually has a two-stage process for accessing the drive, with
> a microprocessor inside the drive acting as a go-between. For writing,
> the computer sends a sector of data to the drive's microprocessor, which
> then writes it to the disk.
>
> This is an alterating process: the UniDisk's microcontroller can either
> be transferring data from the host computer, or writing it to the drive.
> it also has to do the sector-level encoding. This means that the
> sequence of steps is to transfer the data to the drive, then encode it,
> then write it. The total time taken means that it can't even keep up
> with 1:3 interleave, so it is necessary to use 1:4 interleave to get
> best performance with a UniDisk 3.5.
>
> For consistency, the computer only offers 1:2 and 1:4 interleave, as
> this covers the fast ones (e.g. GS/OS and ROM 3 IIgs firmware, which can
> do 1:2), and the slowest ones (UniDisk 3.5). In principle, someone could
> create 1:3 interleave disks, which would be a little faster than 1:4 on
> many computers, but might be a lot worse on others.
>
> All of the above considers writing only. The same issues apply in the
> read direction, but there are some enhancements which can be made, which
> mean that for more advanced drivers it is possible to ignore the
> interleave and read any disk at full speed. This is done by reading an
> entire track into memory in a single pass, then sorting out the
> interleave order and decoding all the sectors in memory.
>
> For less capable drivers (ROM 1 IIgs firmware, or anything involving a
> UniDisk 3.5), the disk is read one sector at a time, and the same issues
> arise with interleave factors.
>
> > is this the same as "high speed interleave"?
>
> Yes.
>
> > any advantages or disadvatages?
>
> Speed.
>
> If you pick the wrong interleave factor (determined by the drives and
> computers in which the disk will be used) then you will get either a
> major or minor performance hit.
>
> If your computer and drive are fast enough to support 1:2 interleave,
> then writing to a 1:4 interleave disk will be about twice as slow,
> because the drive has to rotate twice as far to get to the next sector.
>
> If your computer and drive are only able to keep up with 1:4 interleave,
> then writing to a 1:2 interleave disk will be about ten times slower,
> because an entire revolution of the disk will be needed to write
> adjacent sectors.
>
> In both cases, reading will either be just as bad as writing, or be
> unaffected by the interleave, depending on the driver and hardware
> implementation (if it can read entire tracks in one pass).
>
> It is also worth noting that standard PC floppies are formatted with a
> 1:1 interleave, which they can achieve because the encoding step is done
> in hardware (inside the disk controller). If you write to a 1:1
> interleave floppy in an Apple ][ 3.5" drive, it will be about ten times
> slower than it would if you reformatted it as 1:2 (or 1:4, depending on
> your drive). The tradeoff is that the interleave will also reduce the
> performance for a PC (e.g. 1:2 will take twice as long to read or write)
> but on average you get best performance by increasing the interleave to
> suit the slowest computer.
>
> --
> David Empson
> demp...@actrix.gen.nz

David,

Great description! Thanks.

Now, do you happen to know, just for the sake of discussion, how
interleaving works with the SuperDrive and the SuperDrive Controller
Card?

-Warr