QIC-xx tape standards

[Flavio Spada]

QIC standards have a number (grows with time...). Standards exist for
recording forma, cartridge size, adapter to formatter interfacing, etc...

A QIC setup consists of:

An adapter that interfaces a system bus to a device bus
A formatter that interfaces a device bus to the tape unit

The device bus physically is a 50 pin cable to which up to four
formatters/tape units can be attached.

The system bus can be something like the VM, or the AT, or the SCSI.

The recording standards are fairly simple. Essentially all the recent ones
specify a fixed track size, and a fixed horizontal recording density
(tipially 8k bits per inch). What vaies is the number of tracks across
the width of the tape.

The very first QIC unit (the DEI 20-40MB ones in the ONYX C8002) had four
track, which could be read and written independently; current ones
have from 9 to 15, arranged in a serpentine. When you load the
cartridge the head is moved up and down to locate the first tracks.
15 seems to be the current maximum # of tracks across the width of the tape,
and further improvements will have to come from the horizontal recording
density.

The two most recent standards are for 125 MB capacity and 150 MB capacity;
apparently the 125MB standard is a fluke, because the same type of
unit can be modified to the 150MB one, and of course one would prefer the
latter.

The QIC recording format is a stream format; no preformatting of
tapes is necessary, data are simply appended to the tape. If the write
fails it is repeated by the formatter automatically until it succeeds,
i.e. after having passed over a bad section of the tape.

This is very different from the 3M/HP cartridges that have the same
physical size; these are formatted into discrete records, and a during
preformatting a bad block table is written, etc..., just like a disk.
This preformatting enables higher recording densities and random access
(you can put a unix filesystem on such a cartridge, and it will behave
just like one on disk -- but for speed...). It also has a couple
of drawbacks, that the format pattern is copyright 3M, so you have to
buy their preformatted tapes, and that since the blocks are small (one
kilobyte) the tape unit is abysmally slow.

The other standards of interest are the QIC standards for interfacing
the adapter to the formatter and those for command sequences.

They are the QIC-02 and QIC-36 standards. Older units tend to have
the QIC-02 interface and newer ones the QIC-36. QIC-36 is somewhat
more sophisticated. The command sets are defined like SCSI, a core
set of commands, and several optional ones. The core set is truly
minimal. Virtually no drive, even the most sophisticated, implement
some useful commands, such as backspace record or file, and append
to the end of the tape (except in restricted even if useful cases --
too long to discuss).

Tape interchangeability is very good -- QIC-24 (9 track) units can
read QIC-11 (6 tracks?) ones, and the newer 125/150MB ones can read
QIC-24 tapes. Writing with earlier recording formats is usually
either not supported or not very reliable.

Some obstacle to interchangeability is that adapters occasionally
differ as to byte ordering or byte signedness (to read a plexus
cartridge on a ncr you must subtract replace each byte by 255
minus its previous value...) but usually writing a simple
filter is sufficient.

Many adapters differ as to the maximum block size supported; you
can safely bet that every adapter will support blocks to 32767
bytes in length, but after that you take your chances (the plexus
P55 adapter and its driver will crash the system if you try reading
or writing long blocks).

As to speed, all recording standards so far specify 90 inches per second
that multiplied by the horizontal density gives a maximum thruput
of 5 MB per minute. This is rarely attained, in practice, because it
requires the adapter to supply data at that rate, and most discs and
filesystems and backup utilities have difficulty with that. This emans
that the tape will occasionally have to stop recoding or reading, and
wait for more data to arrive. The cheap mechaniscs of QIC units imply
that stopping will take a long time, and this would waste a lot of
tape at each stop, so stopping actuallyis followed by a backspace.
In practice the speed of a QIC unit is dominated by the time it
takes to stop the tape after each block and that to backspace. Writing
long blocks is advisable, because it minimizes the occurence of
these actions, and selecting a unit that is quicker at this helps
as well, as all other paramters are standardized.

I Hope that this long and necessarily skimpy introduction is useful;
further details may be found in several back issues of magazines
such as EDN, Mini-Micro Systems, etc..., that often (usually every year)
publish a survey of current QIC technology and products.

	Piercarlo Grandi


Until 88/JUL/05:
	...!mcvax!i2unix!irst!sabi!piercarl
	sabi!piercarl at irst.it

After 88/JUL/05:
	...!mcvax!ukc!aber-cs!piercarl
	piercarl at cs.aber.ac.uk