From cca.ucsf.edu!ucsfcgl!ucbvax!ucsd!usc!samsung!uunet!uvm-gen!pegram Sat Jul 28 22:44:10 PDT 1990
Article 8 of comp.sys.atari.st.tech:
Path: wet!cca.ucsf.edu!ucsfcgl!ucbvax!ucsd!usc!samsung!uunet!uvm-gen!pegram
>From: pegram@uvm-gen.UUCP (Robert B. Pegram)
Newsgroups: comp.sys.atari.st.tech
Subject: How to upgrade your ST floppies to 1.44Meg 3.5" drives
Message-ID: <1572@uvm-gen.UUCP>
Date: 18 Jul 90 16:14:16 GMT
Sender: nobody@uvm-gen.UUCP
Organization: EMBA Computer Facility, Univ. of Vermont, Burlington.
Lines: 348


Tried to post this yesterday from GnuEmacs - no go.  Here it is,
Hartmut Semken(AKA hase)'s 1.44 meg Upgrade for Atari STs and Megas.
This is a compilation of several postings, edited by Bob Pegram, my
comments are indicated by  [..text..] <- RBPIII.  I have yet to do the
upgrade myself (lazy 8-), but it seems workable.  From the TT rumor 
about no HD floppy for now, it would seem that Atari is doing the same 
thing, only with a new (not yet fully available) floppy controller from 
Western Digital that *is rated* to handle the higher clock input that is
necessary.  Thus, in the future, this upgrade may become a more reliable 
and simple chip replacement.  Enjoy,
				Bob Pegram

originally From: hase@hase_1.UUCP (Hartmut Semken)
original Subject: Re: Floppy drive 1.44 meg woes!
original Date: 11 Feb 90 09:55:00 GMT
 References: <9002030029.AA23185@jade.berkeley.edu> <2103@ultb.isc.rit.edu>
 Organization: h.a.s.e., Berlin, West Germany


clf3678@ultb.isc.rit.edu (C.L. Freemesser) writes:

 Reply-To: hase@netmbx.UUCP (Hartmut Semken)
 Organization: netmbx Public Access Unix, Berlin, West Germany
 Lines: 28
 Posted: Sat Nov 25 17:57:46 1989


>>  I'm having a few problems hooking up my 1.44 meg teac. I was wondering
> From what I know, you can't use HD on the ST.  The controller chip
>can't handle it.  Just be glad it works in 720k mode.  :^)

Well, you can, if you hack the hardware.

Here we go again:

The trick is to double the clock frequency of the controller chip, so it
will talk twice as fast to the drive: the bit stream has "high-density"
format.
A clock of 16 MHz is well beyond specification (the WD1772 is rated to
run at up to 10 MHz), but it will work most of the time.

The hack consists of two parts: hardware and software and works with
( [3.5 inch] <- RBPIII ) high density drives only!

[The reason being that the drive must auto-sense the HD disk when it is 
 inserted and change its RPMs up accordingly.  5.25 (1.2Meg) drives do 
 *not* do this!  Neither do all HD 3.5" floppies (notably the ones on 
 true blue PS 2 s). ] <- RBPIII
 
Now: the hardware.

The clock pin of the WD1772 is pin 18. It is normally hooked to the 8
MHz clock.
Cut this pin and connect it to the output of a simple circuit. feed the
two inputs of this simple circuit with 8 MHz and 16 MHz clocks (16 MHz
can be gotten at SHIFTER pin 39).
The "simple circuit" mentioned above could be a simple, manual switch.
Now You are able to select two clock frequencies and therfore two
"modes" for the controller chip: double density and high density.

The simple circuit could also be a multiplexer (74LS157, I think)
controlled by the "I detected a HD-disk in me"-output of the floppy
drive (the TEAC FD 235 HFN provides this outout on pin 2 if a jumper is
set; the 235 HFN has a second jumper that enables internal detection of
HD disks: this must be set, too).
With the MUX, a disk inserted will select the appropriate mode of
operation automatically.

Do not worry about two drives in different modes: when copying from one
drive to the other, the clock rate will be switched once to twice a
second. This will not trouble the controller (remember the clock to be
in the range of Megahertz...).
If you want to connect one HD drive and one DD drive, you'll need two
more TTL gates (one 74LS00 will do) to make sure, the controller is in
DD mode for the DD only drive: invert the HD-detected-output (it is LOW
for DD before, HIGH for DD behind the inverter) and feed it into a NAND
gate together with drive select for the DD drive. Let a HIGH output of the
NAND gate select DD-clock (8 MHz). Now "not-HD-detected OR DD-drive-select"
will select DD mode, right?

With two HD drives, there is no problem at all :-) [except for the 
High Density floppy boot problem, discussed below.  You have to set 
the step rate from the boot sector.] <- RBPIII

Now for the software:

The controller uses its clock as base for all timings.
The pulse width and time between pulses for the step line are controller
controlled.                                                   ^^^^^^^^^
                                                       [WD 1772] <- RBPIII
To make it short: with the doubled clock the controller will try to make
the head step twice as fast: he will send a step pulse every 1.5
milliseconds if set to 3 milliseconds (which is TOS standard).
Just set the step rate to 6 milliseconds to make the controller step 3
milliseconds in HD mode (well, the 6 milliseconds steps in DD mode will
slow down floppy operation a bit, but not that much. Not stepping, [but]
reading and writing takes the time!).

How to set the step rate?
Well, there is a documented TOS variable for the step rate at 0x440
called seekrate.
The step rate in this variable will be used by the BIOS after a call of
the routine, pointed to by "hdv_init" (the hdv_init routine seems to
initialize the BIOS parameter blocks for the floppies). The pointer to
hdv_init can be found at 0x46A (this is a supported TOS variable, too).
Look at the MWC-Assembler source below.

[More information on the step rate from another posting - between the 
curly braces below] <- RBPIII

{
 In article <1346@uvm-gen.UUCP> pegram@uvm-gen.UUCP (pegram r) writes:
 >
 >Rats! I looked up the possible seek rates last night, 6 msec is as slow
 >as she goes. 

 Hmm, TOS will step at 6,12,2,3 msec (with rate codes 0,1,2,3).
 Setting the controller to 6 msec and doubling the clock works for almost
 all 3.5 inch HD drives.

 [I wanted to know if you could still use a (360 K) 5.25 drive, which 
 needs a true 6 msec step rate.  This was when no WD 1772 clock switching 
 (between 8 and 16 Mhz) was contemplated.  If your 1772 works this 
 fast, you can forget the switch, or use a properly buffered manual 
 switch for those rare occasions when it's necessary.  "Properly buffered"
 you ask?  See any of the usual designs that eliminate switch bounce -
 Radio Shack's pamphlets on digital electronics have the circuits needed.
 The 74LS00 mentioned above will work nicely as the debouncer and "glue".
 You still want that Multiplexer and you send in the debounced 16Mhz ("On" 
 or "TRUE") and 8Mhz (The inverted "FALSE" or "OFF") outputs from the manual 
 switch to select the clocks.  That is just a design off the top of my
 head, which needs to be improved.  Please do post a better one if you
 come up with one, I can think of a one chip (the LS00) solution if
 clock skew to the WD1772 is unimportant (as in 180 degrees out of
 phase 8-). The reason I mention it is because, I believe that the 
 problem mentioned at the beginning of the original posting was due to 
 either an incorrect jumper setting on the drive, or a noise problem 
 in the ST, having the clocks going all over the ST to the switch, 
 is not a good idea at all!  You don't have to worry about driving 
 more capacitance with such free standing wires, but you do have to 
 think about the effect of having some new high frequency antennas 
 inside your ST!] <- RBPIII

 >It would also be hard to boot from a floppy with this
 >mod (double - 16Mhz - clock for 1772 + formatter + "6"msec seek rate).
 >Do you have to have a seek rate setting boot sector on every boot
 >floppy and would that idea even work?

 Sure, that should work, since no steps schould be necessary to read the
 boot sector.
 I boot from harddisk...

 [Me too, but some utilities (DC formatter at least) can format disks 
  with boot sectors that can reset the step rate on boot up. Game disks
  could be similarly hacked to work with the new drive.] <- RBPIII

 If the HD-floppy does not work after modifying the ST, the step pulse
 width is a probable cause.
 In that case, use an edge-triggered monoflop to "resize" the pulse.
                                     ^^^^^^^^
 [hase means a monostable or "one-shot" available in TTL or with the "555"
  timer chip.] <- RBPIII  
}

The second thing is to get a formatter that will format 18 sectors per
track. HYPERFORMAT by Claus Brodt is such a thing, but a simple
formatter hacked from the example in the Mark Williams C Lexicon did the
same. It's included below.


For help try to write to hase@hase_1.mbx.sub.org, but it'll probably
bounce; our mail service is fading fast... 

[Check if hase's current address is the same, he does show up on c.s.a.s] 
<- RBPIII

I'll try to reply to anything, I finally get.

hase

------- cut for seek6.s -----------
/	module name seek6

	.shri

	.globl main_
main_:
	clr.l	-(sp)
	move.w	$0x20,-(sp)
	trap	$1
	addq.l	$6,a7
	move.l	d0,_save_ssp
	move.w	$0,0x440	/ 0 => 6 ms
	lea	0x46a,a0
	move.l	(a0),d0
	movea.l	d0,a0
	jsr	(a0)
	move.l	_save_ssp,-(sp)
	move.w	$0x20,-(sp)
	trap	$1
	addq.l	$6,a7
	.even
_save_ssp:
	.blkl	1
/ seek rate codes are 0,1,2,3 for 6,12,2,3 milliseconds

Sets seek rate to 6 Milliseconds for *all* drives (with *any* TOS
version; tested on Rainbow TOS (1.4), BLiTTER TOS (1.2) and TOS 1.0):

------------ cut for formath.c -------------

/* formath.c Formatter fuer High Density Disketten */
/* nur fuer angepasste Hardware! Floppycontroller und Laufwerk */
/* muessen High-Density tauglich sein! */
/* Hartmut Semken, Lupsteiner Weg 67 1000 Berlin 37 */
/* hase@hase_1.mbx.sub.org or hase@netmbx.mbx.sub.org */
/* 03-SEP-89 */
#include <stdio.h>
#include <osbind.h>
#include <time.h>

#define SLEEPTIME 1	/* 1: Zeitschleife, 0: Taste */
#define BLANK (0xE5E5)
#define MAGIC (0x87654321L)
#define BUFSIZE (20*1024) /* Platz fuer mehr als 18 Sektoren... */

#define DEVICE 0	/* 0 = Floppy A, 1 = Floppy B	*/
#define SIDES 2		/* je  				*/
#define SECTORS 18	/* nach				*/
#define TRACKS 80	/*Belieben			*/

#define TOTSEC (TRACKS * SIDES * SECTORS)

extern int errno;

main(){
	int track;
	int side;
	int status;
	short *bf;
	register char reply;
	short *middle;
	char buffer[512];
		
	printf("\033E\n");
	printf("Public Domain High Density Mini Formatter\n");
	printf("von H. Semken\nDer Autor garantiert in keiner Weise fr die Funktion\nDi
eses Programmes.\nBenutzung auf eigene Gefahr.\n");
	printf("\n\n\n");
	printf("\007\033pFormatiere Diskette in Drive %c\033q\n%d Seiten\n%d Sektoren p
ro Spur\n%d Spuren\n\n", (65+DEVICE), SIDES, SECTORS, TRACKS);
	printf("Wirklich formatieren? ");
	fflush(stdout);
	if ((reply = Crawcin()) != 'y' && reply != 'Y' && reply != 'j' && reply != 'J')
 {
		printf("Nein. Floppy nicht formatiert.\n");
		sleep(1);
		Pterm0();
	}
	printf("Ja.\n");
	printf("Diskette einlegen; Taste drcken...");
	fflush(stdout);
	Crawcin();
	printf("\n");
	bf = malloc(BUFSIZE);
	for (track = TRACKS-1; track >= 0; track--) {
		for (side = 0; side < SIDES; side++) {
			printf("Formatiere Spur %d, Seite %d", track, side);
			fflush(stdout);
			status = Flopfmt(bf, 0L, DEVICE, SECTORS, track, side, 1, MAGIC, BLANK);
			if (status) {
				middle = bf;
				printf("\t%d\n", status);
				while (*middle) {
					printf("\tDefekter Sektor %d\n", *middle++);
				}
			} else {
				printf("\tokay\r");
			}
		}
	}
	printf("\n\nAlle Spuren formatiert\n");
	printf("Initialisiere Directory\n");
	for (track = 0; track < (BUFSIZE>>1); bf[track++] = 0);
	for (track = 0; track < 2;track++) {
		for (side = 0; side < SIDES; side++){
		if (status = Flopwr(bf, 0L, DEVICE, 1, track, side, SECTORS)) {
			errno = -status;
			perror("Write Error");
		}
		}
	}
	Protobt(buffer, (long)Random(),3,0); /* Prototyp Bootsector fr 
	                                      * 80 * 2 * 9 Sektoren */
	/* Prototyp Bootsektor fr das neue Format anpassen */
	/* Bytes 19 und 20 enthalten die Sektoren pro Disk */

	/* unteres Byte von TOTSEC */
	buffer[19] = (char)(((TOTSEC>>8)<<8)^TOTSEC);
	/* oberes Byte von TOTSEC; es lebe das Intel int-Format */
	buffer[20] = (char)(TOTSEC>>8);
	buffer[24] = (char)SECTORS; /* Sektoren pro Spur */
	status = Flopwr(buffer, 0L, DEVICE, 1, 0, 0, 1);
	if (status) {
		errno = -status;
		perror("Write Error (Bootsector)");
	}
	status = Flopver(buffer, 0L, DEVICE, 1, 0, 0, 1);
	if (status) {
		errno = -status;
		perror("Verify Error (Bootsector)");
	}
	printf("Diskette in Laufwerk %c formatiert\n", (65+DEVICE));
	sleep(1);
	Pterm0();
}

sleep(seconds)
int seconds;
#if SLEEPTIME
{
	clock_t t;
	for(t = clock();clock() < (t + CLK_TCK*seconds););
}
#else
{
	printf("Taste druecken\n");
	fflush(stdout);
	Crawcin();
}
#endif
-- 
original idea and post by:

>Hartmut Semken, Lupsteiner Weg 67, 1000 Berlin 37 hase@hase_1.UUCP
>Dennis had stepped up into the top seat whet its founder had died of a
>lethal overdose of brick wall, taken while under the influence of a
>Ferrari and a bottle of tequila. (Douglas Adams; the long dark teatime...)

Additions by:

Bob Pegram   Internet: pegram@griffin.uvm.edu
	     UUCP: you figure it out!
	     Physical Mail: R. Pegram, R.R. #2, Box 2843, Green St.,
			    Vergennes, Vt. 05491, U.S.A.
	     Home Tel: (in N.A., add appropriate codes if elsewhere)
		 	(802) 877 0018