9-track alignment (skew) tapes for R/W-head adjustment

________________________________ Von: Jon Elson <elson at pico-systems.com> An: cctalk at classiccmp.org Gesendet: 20:32 Sonntag, 7.April 2013 Betreff: 9-track alignment (skew) tapes for R/W-head adjustment 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