Next: Tape Backup for Small
This is the 4th part of this series and now we'll finally start
looking at how all of this is put together. First we'll examine the
QIC tape format, and in I'll explain other types in later
This section will discuss the QIC method - and it's probably a
good starting point as it's fairly easy to understand.
A slight digression here on design philosophy:
The 8" floppy disks and 3.5" disks also were designed with a
fail-safe 'attitude' while the 5.25" floppy format used a
The 8" disks have a write-enable tab - though a great many were
shipped with no write protect slot cut in the jacket. The 3.5"
diskette has a slideable tab. If either of these had a missing tab
the disk would become read-only, thus 'fail-safe'.
In the 5.25" design you had a write-protect tab. The disk had a
slot to enable writing on the disk - this was the default mode. If
you wished to protect the data you would place a tab over this, and
the first disks had a mechanical sensor that detected this tab.
[You may have noticed that many master diskettes were 'special' in
that there was no write-enable slot.]
If the tab accidentally fell off your master disk then you would
be in a fail-dangerously mode as you could write to the disk, which
you had assumed was non-writeable. More than a few people fell
victim to this.
A major manufacturer - 3M - also was bitten by designing to the
original implementations and making flexible tape style
write-protect tabs that didn't fall off like so many of the earlier
paper, or paper-metal combinations. This seemed like a step in the
However - some manufacturers started using optical sensors and
the 3M tabs were optically transparent to the infra-red of the
sensors and disks which were assumed to be write-protected,
Many things in the PC side of the world in the earlier days were
just 'thought of' and not really 'designed'. Those of us who
suffered through the mistakes then really appreciate what we have
to work with today.
The QIC [Quarter Inch Cartridge] is the data cartridge you have
seen which measures about 3" by 5", and with a metal bottom
This is a very rugged design and it's more complex mechanically
than you'd notice with a casual glance.
This style of backup media has grown from the early days of 20MB
per tape - which took between 30 minutes and one hour on the
machines I frist saw them running upon - to the current QIC style
cartridges which hold several Gigabytes.
The similarity to an audio cassette is that there is one reel
the tape feeds from and another reel for take-up. And at that point
- the similarity just about ends.
The aluminum bottom plate is used to give dimensional stability,
as the cartridge itself guides the tape, while in a cassette
machine there are tape guides in the transport but not in the
If you have a QIC cartridge you might examine it while I go
through most of the basics of the cartridge.
Holding the cartridge metal side down and facing the opening you
will see two openings in the plastic. On the right side you will
see a round object. Looking at this object from the the top of the
cartridge you will see an arrow and a slot. You can turn this by
using a coin in the slot, and the round tube will rotate to reveal
an opening on the edge of the cartridge. This mates with a sensor
in the tape transport and when it is open, the tape will not be
able to be written upon.
If someone wished to distribute master tapes that could never be
written upon you could remove this - though I have never seen it
done. This is a fail-safe type design.
In the center you can see an opening that is only goes down half
way. Behind this there is a rubber wheel - which is 'stepped'. The
top part of this being of a larger dimension than the bottom. The
top part of this wheel protrudes through the opening.
You'll also notice that you can see only part of the magnetic
tape and that it lies in a slot below the protruding part of the
This is another major area where data tapes diverge from the
audio format you are used to. The capstan [in a cassette the metal
shaft but in the QIC transport a belt drive shaft with a rubber
wheel] drives this rubber idler - but never touches the tape. In
the audio world the capstan pinches the tape between it and the
Now move this wheel and note that the tape reels move in a
direction opposite that of the wheel movement. Don't move the wheel
so that the right spool moves clockwise too far or you'll take the
tape right off the end of the spool. It is not fastened down. Move
it as far in the other direction as you wish. Don't worry about
getting it back to where it was, inserting it into the drive will
set it back correctly automatically.
The tape is being moved by a thin flexible ribbon. It is this
ribbon you see - about 1/8" of an inch wide when you look at the
back of the cartridge - and what you see going to the top of the
feed reel on the left. At first glance you may mistake this for the
tape in the cartridge as it is more visible.
This belt acts much like a conveyor belt as it supplies the
movement and supports the tape through its entire path through the
cartridge except for the portion crossing the head and for part way
up the take up spool just past the head.
In pro audio tape recorders - fully servo driven - I've seen
unsupported tape move at over 400"/second. I've also seen the
disaster when a reel hold-down fails - and you have a big pile of
brown spaghetti - or in the old days the air filled with brown
The reason for transport belt is guidance. The tape in current
machines moves well over 100" per second, or almost 6 miles per
hour [or about 9.5 kilometers/hour for the majority of the world].
Just imagine how hard it would be too keep this evenly spooled
without the belt. No home reel-reel audio tape recorder moves that
fast in home units in fast-forward or rewind. If you've ever had to
hand-wind something onto a spool you can appreciate this. [Brings
back memories of hand-winding 1200 feet of unexposed motion picture
film in a darkroom after dropping a loop while loading a
The other guidance in the cartridge is from two tape guides. The
moving white plastic wheels in the corner are for the belt. As far
as the guides go, you can see one between the write-protect roller
and the pinch-roller/idler, while the other is at the left side of
the tape cartridge. To see the left guide, just pull gently on the
plastic door to the left of the pinch-roller. This exposes the tape
and you can see the post to the left.
These two metal posts ensure that the tape crosses this area at
the EXACT height above the metal base-plate in very cartridge made.
All the guidance and and most of the tolerances are in the
cartridge, and the transport is fairly simple in comparison. This
measurement from the base-plate is similar to the measurement from
the top-plate in professional audio recorders. It is a stable
platform of reference.
The tape also has some small holes punched at either end of the
tape. If you look at the cartridge from the top between the
write-protect device and the pinch-roller, you will see a small
clear window and by moving the cartridge around in the light you
will see a small mirror at about a 45 degree angle.
There is a light and a sensor in the transport so that when a
hole in the tape passes this device, light goes through the hole
and is reflected back and signals the transport that it has reached
the end of the tape and to stop.
That covers the cartridge in more detail than you had probably
guessed, now let us examine the transport.
When you place the cartridge in the drive - even if was not
fully rewound - such as might in a power off situation and the tape
is removed before the power is applied - the tape will rewind until
it sees the end-of-tape holes and then will set itself for proper
When you place the tape in the drive - the resistance you feel -
at least in the older units - is from the ejection spring and at
the last push the plastic door you opened above is opened.
On the older units you then slide the tape door lock shut, which
then moves the tape head - including the erase head - up against
the tape. These mechanism are basically quite simple, which makes
them quite reliable. The head pushes the tape in by a very small
amount, and only the tension of the tape keeps it against the head.
There are no pressure pads such as you would find in an audio
So now that the tape is loaded we are ready to go.
You start your favorite backup program. The more sophisticated
programs read a label they have previously placed on the tape so
that it can warn you that you are writing over last night's backup
for example. They can also keep track of the count of times the
tape has been used so you don't exceed the maximum recommend use.
This is typically a conservative figure, but a new tape is far
cheaper than paying help to recreate data lost because of scrimping
on a $20 tape.
Once the tape starts moving, the erase head, which is in the
same housing as the read/write head, is sent current. The current
causes the FULL WIDTH of the tape to be erased.. At this time the
data flows from the computer to the transport electronics until the
tape sees the holes as the end of the tape.
At this time we stop the erase head. The mechanism holding the
record head is physically moved down, the tape then reverses
direction and the data is written on the next position on the
For each direction reversal the tape moves again so that if you
drew the data pattern on the tape on a piece of paper it would go
from one end to the other, move down and go back, and repeat, in a
You should hear the tape stop and restart only as many times as
there are tracks on the tape. Any more than this indicated you
aren't sending the data to the tape drive fast enough, and that the
tape has to wait for more data. This will increase the time it
takes to backup a system dramatically. Each time it stops, it then
has to go back an reposition itself at the correct position so that
the data stream appears to be continuous when you read it back.
This used to be called 'shoe-shining' but lately I've heard the
term 'back-hitching' used to describe this. Properly matching the
computer, hard drives, tape controller and tape device is required
to avoid this. A fully streaming tape that only stops at the end
usually surprises those who have only been exposed to backups in
the Microsoft environment for example.
The first QIC tape drives of which I used had only 4 data
tracks. So the tape would move across the heads four times - two in
The multiple tracks coupled with the full-track erase is why it
is impossible - for an ordinary user - to recover any tape which
has been partially written upon by mistake.
The full-track erase will erase the first part of tracks 1 and 3
and the tail end of tracks 2 and 4. So at least part of the data is
damaged on all of the tape. In addition, when you accidentally
start a write - typing 'tar cvf' when you meant 'tar tvf' or 'tar
xvf', the process of stopping the tape and rewinding puts a tape
marker at the end of the data, and the firmware of the tape will
prevent you from going past that area.
The only way you can write data on a 1/4" tape is starting from
the first of a tape - if the tape has data as in the disaster
situation above, or skipping past all the file marks to blank tape.
Multiple files can be written to tape using the no-rewind
If you accidentally write to a tape you are trying to restore ,
then your only real resort is a data recovery firm. ALWAYS
write-protect tape before you attempt to restore any critical
The first tapes had a capacity of about 20MB. To increase this
capacity to the point where we now have multi-gigabyte drive
capacity, and going from backing up at under 1MB minute to hundreds
of megabytes per minute we used evolutionary [as opposed to
As mechanical manufacturing precision became better, and
sophisticated designs became practical we were able to make the
tracks on the tape narrower. The first 4 tracks on quarter inch
tape were about the same width as the 9-track on 1/2 tape used in
the larger computers.
Once we could make the tracks narrower, we now could move the
tape head in a smaller vertical increments each time we changed
tape direction. I have not used any of the gigabyte tapes and have
not looked up the track density - but on the QIC formats I've
worked with the number of tracks was about 30 the last time I
Another step is to improve the electronics so we can switch the
data on and off faster. At this point we can move the tape faster
and increase the speed we write to tape, and thus reduce the amount
of time it takes to write it. Since we have not changed tape head
nor tape formulation each bit still occupies the same space, but we
just put it there faster.
Using the 5MB track capacity of the first 20 MB drives above - a
30 track tape gives you 150M-byte tape capacity by that alone.
Faster write means we back up in 1/2 the time we did originally. So
this is just a mechanical change and a data rate change.
The next step is to change the basic formulation of the tape. To
increase the density we have to have magnetic domains that are
smaller. This means we also have to change the tape head design.
That in turn means we have to redesign the electronics, as the
higher density tapes are higher coercivity and need more current to
be able to write the data.
If, in this example, we make the domains 1/2 the size as they
were above and make the tape head gaps 1/2 the size, we can now
double the data we have on the tape.
However - to make use of this we need to change one of two
things. Either we change the speed of the tape by making it move
one half the previous speed - or the smart thing is to redesign the
interface so we push the data to the tape twice as fast. Since the
data bits are smaller - by keeping the tape speed the same - we can
double the amount of data sent and written to the tape in the same
As Ted Nelson likes to say "Everything is intertwingled".
One by one the QIC style tapes are disappearing, and the only
major manufacturer I'm aware of in that market is Tandberg. Smaller
systems moved to DAT/DDS and larger systems to DLT. I'll look at
those next in this series.
As data storage increases the need to make timely removable
off-site storage becomes more critical. What was once an easy
overnight backup/verify, can now consume all the available spare
A friend of mine who worked at an engineering firm backed up
everything every night, until the 7PM start of backup didn't end
until 7AM when the employees started arriving. It will only get
I noted in a report from Sandia Labs [www.sandia.gov] that in an
article on sorting speed, they commented that at that time of the
report a typical data storage had about 250GB in use and predicted
that it to be at 6.5 terabytes by 2003. It's beginning to look like
those guesses might be conservative.