On 22 Sep 2012 at 19:33, Tony Duell wrote:
I wonder if this card does some kind of data
buffering. That could use
a microcontroller.
That was my suspicion. Performance of QIC drives on 4.77MHz 8088
machines can be somewhat anemic. As you know, starving a streamer is
not a great idea, so any buffering would help.
Indeed yes. The QIC-02 interfce, being a formatted interface, doesn;t
depend on the data rate, but if it can't transfer data fast enough to
prevent the data buffer in the tape drive from overflowing on a read, or
emptying o na write, the resultas are unpleasant (and not waht you might
intitiall expect). The drive starts 'washing machining' [1]. It reads a
block, the buffer fills, it reads the next block, etc. When the buffer
gets full, the tape doesn't stop instantly. It overruns. So the drive has
to stop, rewind, read that block again when the host has emptied the
buffer a bit. Perfomace goes throgu hthe floor. And wear on tape, ehads
and drive motor goes throguh the roof.
[1] Think of the motion of the agitator in an old-fashioned top-loader
washing machine.
The Archive sidewinder came in 2 versions, 30 ips and 90ips. the former
was used wit hte PERQ. If you use the latter, it washing-machines like
mad. I once covereted one from 90ips to 30 ips (a matter of fitting an
extra flip-flop IC on the controller board, changing a few jumpers, and
replacing lots of pasisbves in the data separator and read filter
sections, and then setting the whole thing up. I also used a Wangtek
90ips drive with my PERQ for a time. I found I could improve peformance
a lot, and get it to stop and rewind less frequently, by upgrading the
data buffer on the QIC02-QIC36 card from 2K to 8K. This was a matter of
changing the SRAM chip and moving a link IIRC.
-tony