Sprinter/mame
2
0
Fork 0
mirror of https://github.com/holub/mame synced 2025-04-23 00:39:36 +03:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

hp9845: TACO doc updated

Browse Source
This commit is contained in:
fulivi 2016-02-17 10:10:01 +01:00
parent 02ec85119d
commit 63b58daec9
1 changed files with 100 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

113
src/devices/machine/hp_taco.cpp
View File

@ -8,17 +8,48 @@
*********************************************************************/
// This device has been reverse engineered entirely through documents & study of HP software.
// I had no access to the real device to experiment.
// Available documentation on the internal working of TACO chip is close to nothing. The best
// I could find is [1] (see below) where all that's described is a (too) brief summary of registers and little else.
// In other words, no description of the commands that can be issued to TACO chips.
// So, my main source of information was the careful study of HP software, especially the 9845 system test ROM (09845-66520).
// The second half of this ROM holds a comprehensive set of tape drive tests.
// The main shortcomings of my approach are:
// * I could indentify only those TACO commands that are actually used by the software. I managed
// to identify 17 out of 32 possible commands. The purpose of the rest of commands is anyone's guess.
// * I could only guess the behavior of TACO chips in corner cases (especially behavior in various error/abnormal
// conditions)
//
// Documentation I used:
// [1] HP, manual 64940-90905, may 80 rev. - Model 64940A tape control & drive service manual
// [2] US patent 4,075,679 describing HP9825 system (this system had a discrete implementation of tape controller)
// [2] US patent 4,075,679 describing HP9825 system (this system had a discrete implementation of tape controller). The
// firmware listing was quite useful in identifying sequences of commands (for example how to find a specific sector etc.).
// [3] http://www.hp9845.net site
// [4] April 1978 issue of HP Journal. There is a one-page summary of TACO chip on page 20.
// This is an overview of the TACO/CPU interface.
//
// Reg. | R/W | Content
// =====================
// R4 | R/W | Data register: words read/written to/from tape pass through this register
// R5 | R/W | Command and status register (see below)
// R6 | R/W | Tachometer register. Writing it sets a pulse counter that counts up on either tachometer pulses or IRGs, depending
// | | on command. When the counter rolls over from 0xffff to 0 it typically ends the command. It's not clear to me
// | | what value could be read from this register, if it's just the same value that was written last time or the internal
// | | counter or something else entirely.
// R7 | R | Checksum register. Reading it clears it.
// R7 | W | Timing register. It controls somehow the encoding and decoding of bits. For now I completely ignore it because its
// | | content it's totally unknown to me. It seems safe to do so, anyway. I can see that it's always set to 0x661d before
// | | writing to tape and to 0x0635 before reading.
//
// Format of TACO command/status register (R5)
// Bit R/W Content
// ===============
// 15 RW Tape direction (1 = forward)
// 14..10 RW Command
// 9 RW ? Drive ON according to [1], doesn't match usage of firmware
// 8 RW ? Size of gaps according to [1]
// 14..10 RW Command (see the "enum" below)
// 9 RW ? Drive ON according to [1], the actual use seems to be selection of gap length
// 8 RW ? Size of gaps according to [1], N/U in my opinion
// 7 RW Speed of tape (1 = 90 ips, 0 = 22 ips)
// 6 RW Option bit for various commands
// 5 R Current track (1 = B)
@ -28,6 +59,71 @@
// 1 R Cartridge out (1)
// 0 R Hole detected (1)
// Here's a summary of the on-tape format of HP9845 systems.
// * A tape has two independent tracks (A & B).
// * Each track holds 426 sectors.
// * Each sector has an header and 256 bytes of payload (see below)
// * Sectors are separated by gaps of uniform magnetization called IRGs (Inter-Record Gaps)
// * The basic unit of data I/O are 16-bit words
// * Bits are encoded by different distances between magnetic flux reversals
// * The structure of tracks is:
// - Begin of tape holes
// - The deadzone: 350x 0xffff words
// - 1" of IRG
// - Sector #0 (track A) or #426 (track B)
// - 1" of IRG (2.5" on track A)
// - Sector #1 (track A) or #427 (track B)
// - 1" of IRG
// - Sector #2 (track A) or #428 (track B)
// - ...and so on up to sector #425/#851
// - 6" of final gap
// - Non-recorded tape
// - End of tape holes
// * Sector #0 is not used
// * Sectors #1 and #2 hold the first copy of tape directory
// * Sectors #3 and #4 hold the second/backup copy of tape directory
// * User data are stored starting from sector #5
// * There is no "fragmentation" map (like file allocation table in FAT filesystem): a file
// spanning more than 1 sector always occupy a single block of contiguous sectors.
//
// A sector is structured like this:
// Word 0: Invisible preamble word (always 0). Preamble comes from 9825, don't know if it's
// actually there in TACO encoding. I assumed it is.
// Word 1: Format/sector in use and other unidentified bits.
// Word 2: Sector number
// Word 3: Sector length and other bits
// Word 4: Checksum (sum of words 1..3)
// Words 5..132: Payload
// Word 133: Checksum (sum of words 5..132)
//
// Physical encoding of words is borrowed from 9825 as I wasn't able
// to gather any info on the actual encoding of TACO chips.
// This is how 9825 encodes words on tape:
// - the unit of encoding are 16-bit words
// - each word is encoded from MSB to LSB
// - each word has an extra invisible "1" encoded at the end
// - tape is read/written at slow speed only (21.98 ips)
// - a 0 is encoded with a distance between flux reversals of 1/35200 s
// (giving a maximum density of about 1600 reversals per inch)
// - a 1 is encoded with a distance that's 1.75 times that of a 0
//
// This driver is based on the following model of the actual TACO/tape system:
// * Tape immediately reaches working speed (no spin-up time)
// * Inversion of tape direction and change of speed are immediate as well
// * Time & distance to stop the tape are modeled, though. Firmware is upset by
// a tape with null braking time/distance.
// * Speed of tape is exceptionally accurate. Real tape was controlled by a closed loop
// with something like 1% accuracy on speed.
// * Storage is modeled by one "map" data structure per track. Each map maps the tape position
// to the 16-bit word stored at that position. Gaps are modeled by lack of data in the map.
// There is no model of the physical encoding of bits (except to compute how long each word
// is on tape).
// * Read threshold is ignored. Real tapes could be read with either a low or high threshold.
// * "Flag" bit is used as a busy/ready signal in real TACO. Here I assumed the device is
// always ready, so Flag is always active.
// * I tried to fill the (many) gaps on chip behavior with "sensible" solutions. I could only
// validate my solutions by running the original firmware in MAME, though (no real hw at hand).
//
#include "emu.h"
#include "hp_taco.h"
@ -759,15 +855,6 @@ hp_taco_device::tape_pos_t hp_taco_device::word_length(tape_word_t w)
zeros = 16 - ones;
// Physical encoding of words is borrowed from 9825 as I wasn't able
// to gather any info on the actual encoding of TACO chips.
// This should be enough for emulation.
// Anyway, this is how 9825 encodes words on tape:
// - the unit of encoding are 16-bit words
// - each word is encoded from MSB to LSB
// - each word has an extra "1" encoded at the end
// - a 0 is encoded with a distance between flux reversals of 1/35200 s
// - a 1 is encoded with a distance that's 1.75 times that of a 0
return zeros * ZERO_BIT_LEN + (ones + 1) * ONE_BIT_LEN;
}