9-track alignment (skew) tapes for R/W-head adjustment - test-drb@ccmp.vtda.org

7 Apr 2013

Date: Sun, 7 Apr 2013 11:13:42 +0100
From: Colin Eby <colineby at isallthat.com>
To: "General Discussion: On-Topic and Off-Topic Posts"
	<cctalk at classiccmp.org>
Subject: Re: 9-track alignment (skew) tapes for R/W-head adjustment
Message-ID: <B5214C89-7571-4779-B1BC-9DB00FF5B119 at isallthat.com>
Content-Type: text/plain; charset=iso-8859-1

    Jon,

    First, lemme reassert my non-expert status on this.

    However, my understanding of PE -- phase encoding as a raw signal --
    is that a 0 is a low to high transition(Thomas or reversed high to
    low for IEEE) against a fixed clock time. Successive 0s have to be
    encoded taking the sign low to high, before going high to low again.
    The system is to ignore those signals and simply count the
    transition at the mid-point of the period. Setting the blocking
    aside, doesn't that mean you end up with two transitions in a period
    for every zero. And if you write zero to every channel you get a
    nice sine wave at the pre-amp, with the parity bit being the inverse
    signal (all ones).

    NZRI would of course be rather different. But for PE, this is my
    understand of the signal inside a block. I believe that's the signal
    form you were thinking of.

OH, you meant to write the tape in 1600 BPI (PE) mode!  yes, that would 
put 3200
transitions per inch on all data channels, but the parity channel would 
have its
transitions out of phase with the data channels.  PE mode will write two
transitions for every bit time when the same bit (1 or zero) is written,
but for alternating 1's and zeros, you only get one transition per bit time.
The polarity of the transition at the center of the bit time contains 
the data
bit, and additional transition needs to be added when the same data bit
follows.  See Page 4-11 of this doc for a picture :

<ftp://bitsavers.informatik.uni-stuttgart.de/pdf/digidata/0551711_1140-1640-1740_2-79.pdf>

Generally, skew is not worried about so much in PE mode, the drives have
FIFOs to resync the data.  But, of course after repairing the head 
mounting, it
could be WAY off, too far for the FIFO to correct.  They usually only have
9 bit times worth of skew correction.

Anyway, most older 800 BPI drives have circuitry built-in to assist in 
skew adjustment,
an analog summing circuit that adds the output of the 9 bit detectors 
together.
The stepped square wave is very easy to interpret and adjust on a scope.
I've never done skew adjustment on 1600 or 6250, I think it would be
harder than on 800 BPI.

I think the best bet is a digital or storage scope, looking at data channels
4 and 5 (the outermost ones on the tape) on two scope channels, and 
triggering
so the first transitions of a block are seen.  The preamble is 40 bytes 
of all
zeros, then one 1, followed by the data.

Jon