> > Haave you considered using magnetic tape to build your own drum memory?
> > Would this even work? I don't know, I've only considered it.
> kinda doubtful
Well, as long as you think along a 'real' drum with flying heads
etc. you're right, but there is a simple solution - take a card
board tube and warp regular cassete tape around, and let a head
read/write this tape ... if you take an old style washing powder
drum (ca. 30 cm diameter / 1 ft) you get a lenght of about 1m
(3 ft 4") - what was the usual speed of cassetes ? 4 cm/s ? I'm
not shure (Tony ?), as a rough estimation you get a track size
of some 3 kByte and an average access time of 12s not to bad ...
almost like a floppy (well, except the access time) and of course
we can add more tracks ... and with seperate heads for each track
the over all performance would be comperable to early floppy drives.
The only real problem might be the head(s) - or better the acuracy
of the tape, since the head is fixed. You may need a real good
way to adjust the tape. Otherwise you need some head mechanics
so the head may follow your tape ...
Now if you invest this much effort, you may also glue the tape
in a spiral along the drum - like a good old vinyl disk (you know,
the courious black stuff they had before CDs) or well, on a CD.
All you need is to add a mechanic to lift the head, and you get
one big tape with direct access and fast loading, since you don't
need to switch tracks within a read operation.
For the head movement a simple thread rod and a stepping motor
could work fine.
With our washing powder drum we could get around the same storage
than on a 8" disk (~500K) with still acceptable access time (no
idea hof fast the head positioning will be, but at least it must
be able to move the head within one rotation as fas as the tape
is wide (plus a bit for the 'track gap') - I just assume it would
be possible to move the head 10 times faster when lifted. putting
this together (140 Rounds, 25s per round, 2,5s per track lifted
movement) we get a maximum, seek time of <400s, and an average
seek of <200s or 3 minutes ... not so bad for a old card board
drum :)) A real killer would be a TOC at the start of your drum.
But with a little inteligent programming the average timing may
be quite good (for example do seek operations already when you
are still computing the last data chunk, etc.)
Also you may turn the head at the end ... like car cassete systems,
and have a second 'side' backwards - now we have a whooping Megabyte.
And we need a programm controled direction change for the drum
(plus brake and spin up wait).
Maybe place it in the middle - or just forget about it at all.
Oh ja, for your tape thing - putting a directory at the begining of
the tape is quite a pain - maybe you should take a look at old
IBM manuals for tapes. They had fast tapes, with automatic loading,
and _realy_ fast for/backwinding, Reading in both directions, etc. pp.
but still their tape data structure (VOL1/2/3, HDR1/2/3, etc.) didn't
include any central directory structures.
Last but not least, there is still the endless tape! This device
is way more like what you think about ... 5 min endless cassettes
should be still available for some answering machines - you get
a storage of some 35 kByte which should be fair enough.
Gruss
H.
--
VCF Europa 3.0 am 27./28. April 2002 in Muenchen
http://www.vcfe.org/
Received on Wed May 16 2001 - 14:15:57 BST