Difference between revisions of "765 FDC"

Latest revision as of 22:43, 18 November 2024

µPD765 - Floppy Disc Controller (used in CPC 664, CPC 6128, 6128 Plus and DDI-1 expansion).

The recommended ports used by Amstrad and compatible interfaces are:

Bit b10 of the address port is reset as the FDC is seen as an expansion, even if it is an internal chip. Bit b7 is reset to select the FDC. Bits b8 and b0 are used to select the specific mode of operation. All other bits should be set to 1 to avoid conflict. Source

The Vortex disc interface uses other ports. See its dedicated wiki page.

IC Models used in CPC

More than one manufacturer made 765 compatible ICs. These are the ones known to be used in the CPC by looking at pictures of CPC mainboards.

All should operate almost identically.

• NEC D765AC Source
• NEC D765AC-2 Source
• UMC UM8272A Source
• Zilog Z765APS Source
• Zilog Z0765A08PSC Source

The following data seperators are used:

• FDC9216 Source
• SED9420C Source

The CPC464, CPC472, 464 Plus and GX4000 are not equipped with a FDC chip.

All the floppy disk drive models used by Amstrad are referenced here: Amstrad FDD part

Accessing the FDC 765

The Main Status Register (Port &FB7E) signalizes when the FDC is ready to send/receive the next byte through the Data Register.

The Data Register (Port &FB7F) is used to write Commands and Parameters, to read/write data bytes, and to receive result bytes. These 3 operations are called Command-, Execution-, and Result-Phase.

Command Phase

A command consists of a command byte (eventually including the MT, MF, SK bits), and up to 8 parameter bytes.

Execution Phase

During this phase, the actual data is transferred (if any). Usually that are the data bytes for the read/written sector(s), except for the Format Track Command, in that case 4 bytes for each sector are transferred.

During data transfers between the FDC and the processor, the FDC must be serviced every 26µs (for MFM mode with CPC timings) or the FDC terminates the FDC command.

Result Phase

Returns up to 7 result bytes (depending on the command) that are containing status information. The Recalibrate and Seek Track commands do not return result bytes directly, instead the program must wait until the Main Status Register signalizes that the command has been completed, and then it must (!) send a Sense Interrupt State command to 'terminate' the Seek/Recalibrate command.

During the result phase, all the result bytes must be read. The FDC will not accept a new command until all the result bytes are read.

The 15 FDC Commands

Abbreviations used:

• N = 2^(n+7) bytes, with N between 0 and 7. For example, N=2 means 512 bytes. There is conflicting info about the meaning of N=8: CPC-Power thinks it means 32768 bytes Source. Same for Roudoudou Source. While Simon Owen thinks it means N=0 (128 bytes) Source. On the Intel 82078 datasheet, it is precised that all values up to 7 are allowable.
• MT = Multi-track (continue multi-sector function on other head)
• MF = MFM mode (1 = Double Density)
• SK = Skip deleted-data address mark (set if sectors with deleted DAM shall be skipped)
• HD = Head number select
• US = Unit select (drive select)

• HLT = Head Load Time: 2 to 254ms in 2ms increments
• HUT = Head Unload Time: 16 to 240ms in 16ms increments
• SRT = Step Rate Time: 1 to 16ms in 1ms increments (F = 1ms, E = 2ms, etc.)
• ND = Non-DMA mode

Notes:

• Specify: All timings will be doubled on CPC because the FDC runs at 4MHz instead of 8MHz for the datasheet
• Format Track: The processor must supply C, H, R, N to the FDC for each sector during execution phase
• Recalibrate: Walks up to 77 tracks, 80-track drives may need a second recalibrate if failed
• Seek / Recalibrate: All read/write commands will be disabled until successful sense interrupt

FDC Status Registers

The Main Status register can always be read through Port &FB7E. The other 4 Status Registers cannot be read directly, instead they are returned through the data register as result bytes in response to specific commands.

Main Status Register (Port &FB7E):

 b0..3  DB   FDD0..3 Busy (seek/recalib active, until succesful sense intstat)
 b4     CB   FDC Busy (still in command-, execution- or result-phase)
 b5     EXM  Execution Mode (still in execution-phase, non_DMA_only)
 b6     DIO  Data Input/Output (0=CPU->FDC, 1=FDC->CPU) (see b7)
 b7     RQM  Request For Master (1=ready for next byte) (see b6 for direction)

Status Register 0:

 b0,1   US   Unit Select (driveno during interrupt)
 b2     HD   Head Address (head during interrupt)
 b3     NR   Not Ready (drive not ready or non-existing 2nd head selected)
 b4     EC   Equipment Check (drive failure or recalibrate failed (retry))
 b5     SE   Seek End (Set if seek-command completed)
 b6,7   IC   Interrupt Code (0=OK, 1=aborted:readfail/OK if EN, 2=unknown cmd
             or senseint with no int occured, 3=aborted:disc removed etc.)

Status Register 1:

 b0     MA   Missing Address Mark (Sector_ID or DAM not found)
 b1     NW   Not Writeable (tried to write/format disc with wprot_tab=on)
 b2     ND   No Data (Sector_ID not found, CRC fail in ID_field)
 b3,6   0    Not used
 b4     OR   Over Run (CPU too slow in execution-phase (ca. 26us/Byte))
 b5     DE   Data Error (CRC-fail in ID- or Data-Field)
 b7     EN   End of Track (set past most read/write commands) (see IC)

Status Register 2:

 b0     MD   Missing Address Mark in Data Field (DAM not found)
 b1     BC   Bad Cylinder (read/programmed track-ID different and read-ID = FF)
 b2     SN   Scan Not Satisfied (no fitting sector found)
 b3     SH   Scan Equal Hit (equal)
 b4     WC   Wrong Cylinder (read/programmed track-ID different) (see b1)
 b5     DD   Data Error in Data Field (CRC-fail in data-field)
 b6     CM   Control Mark (read/scan command found sector with deleted DAM)
 b7     0    Not Used

Status Register 3:

 b0,1   US   Unit Select (pin 28,29 of FDC)
 b2     HD   Head Address (pin 27 of FDC)
 b3     TS   Two Side (0=yes, 1=no (!))
 b4     T0   Track 0 (on track 0 we are)
 b5     RY   Ready (drive ready signal)
 b6     WP   Write Protected (write protected)
 b7     FT   Fault (if supported: 1=Drive failure)

C, H, R, N values at result phase

If the processor terminates a read (or write) operation in the FDC, then the ID information in the Result phase is dependent upon the state of the MT bit and EOT byte:

• An empty cell means it is the same value as the one at the beginning of command execution
• LSB (Least Significant Bit): The least significant bit of H is complemented

Motor On/Off Flip-Flop

Writing 00h to Port &FA7E turns all disk drive motors off, writing 01h turns all motors on. It is not possible to turn on/off the motor of a specific drive separately.

An exception are the Vortex F1-S, F1-D, M1-S and M1-D drives. (How are they different?)

Another exception is the Gotek drives. They don't take the motor flip-flop into account and are always on. Source

Some FDC commands don't require the motor to be on. For example, the seek or recalibrate commands, that move the floppy drive head, work fine with the motor off. And seek/recalibrate also work with an empty drive (ie. without a floppy disk inserted). Source

The floppy disk rotates at a nominal speed of 300rpm, with some tolerance. This tolerance of the FDC has been measured by Roudoudou to be ±12% (it worked from 220 kbits/s to 283 kbits/s for a reference of 250 kbits/s). Source

The FDC speed is more impressive when you realize the Amstrad CPC’s tape loads at only 2 kbits/s in fast mode. Note: tape loading can be way faster than that if an MP3 player is used instead of a real physical tape, as it has been demonstrated here: https://youtu.be/MAIsOIwgJWA

Notes

Before accessing a disk you should issue a recalibrate command to the drive to move the head backwards until the track zero signal from the drive is sensed by the FDC. The FDC will also set its track counter for that drive to zero.

On an 80-track drive you may need to repeat that twice because some models of the FDC stop after 77 steps so if recalibrating from track 78 or above the controller might not reach track zero.

In order to format, read or write a sector on a specific track you must first seek that track using command 0Fh. That'll move the read/write head to the physical track number. If you don't do that then the FDC will attempt to read/write data from/to the current physical track, irrespective of the specified logical track ID.

The track, sector, and head IDs are logical IDs only. These logical IDs are defined when formatting the disk and aren't required to reflect the physical track, sector, or head numbers. However, when reading or writing a sector you must specify the same IDs that have been used during formatting.

Despite the name, a sector with a Deleted data Address Mark (DAM) is not deleted; the DAM-flag is just another ID bit. 'Deleted' sectors can be read/written just like normal data sectors and if that ID bit is specified correctly in the command.

Standard floppy disk formats

AMSDOS has been designed to complement CP/M, not to compete with it. They share the same structure and can read and write each other's files.

There are 3 standard floppy disk formats in AMSDOS. They are all single-sided and double density (MFM), with the following characteristics:

DATA format

It is the most commonly used format on Amstrad CPC as it offers the most available space among the standard formats.

SYSTEM format

The SYSTEM format is a VENDOR format with the addition of information to boot the CP/M on:

• Track 0: sectors &41 (boot sector), &42 (configuration sector), &48, &49
• Track 1: sectors &41, &46, &42, &47, &43, &48, &44, &49, &45

Type |CPM and you will find yourself under CP/M 2.2.

By the way, all the |CPM command does is load the 512 bytes of track 0 sector &41 into memory at address &100 and then execute it.

IBM format

The IBM format is only usable on early IBM PCs as it uses 8 sectors per track and the CP/M filesystem. This was the format used in MS-DOS 1.x. Source

With MS-DOS 2.0 (1983), PC floppies switched to 9 sectors per track and the FAT12 filesystem. Source

And MS-DOS 4.0 (1988) removed compatibility with MS-DOS 1.x formatted floppies. This is one of the reasons why MS-DOS 4.0 faced significant criticism.

Also, CP/M 2.2 supports the IBM format but CP/M Plus does not.

Catalog structure

In CP/M (and AMSDOS), files are stored on disks in a flat structure and can only be organized using USER numbers. CP/M does not have a hierarchical directory structure (with folders and subfolders) like newer operating systems.

The catalog structure supports up to 64 entries of 32 bytes. A file larger than 16KB will require multiple entries in the catalog.

CAT'art

The catalog structure can be hacked by using ascii control characters (codes 0 to 31) to create decorated catalogs (CAT'art).

CAT never loads and executes a boot sector. It just displays the chars of the files using "TXT OUTPUT" (#bb5a) which will execute control codes as well.

There were already tools from the 80ies for modifying a directory in a way that will display colourful chars placed at any location on the screen by adding dummy file entries with 0KB sizes. Or you hardcode it directly by yourself, which allows much more chars and effects.

The trick is to disable text output at the end of a filename to prevent printing file sizes/line feeds and enable it again at the beginning of the next filename; and you also have to use one hidden char for sorting them in the correct order.

See here for more details: All about CAT'arts creation (FR)

CP/M terminology

Record: The basic unit of data access in CP/M, 128 bytes in size. This was the standard size for transferring data between memory and disk.

Block: A group of records, 1 KB in size. Blocks were used to improve disk access efficiency by reading/writing multiple records at once.

Extent: A group of blocks, 16 KB in size. CP/M used extents to manage file allocation, with each extent corresponding to a portion of a file on disk.

Sector: The smallest physically addressable unit on a disk, 512 bytes in size on Amstrad CPC. CP/M accessed the disk at the sector level, though users typically worked with records.

File Headers

Cassette files are divided into 2KB blocks (then subdivided in 256-byte segments + CRC), each block being preceded by a 64-byte header.

AMSDOS files are usually stored with a 128-byte AMSDOS Header (based on cassette header) but can also be headerless, depending on the contents of the file. Unprotected ASCII files do not have header.

CP/M files do not have headers.

Custom floppy disk formats

Usually single sided 40-track 3" disk drives are used in CPCs. For practical purposes, 42 tracks could be used — the limit is specific to the drive and some support more tracks but 42 is a good maximum.

RSX programs exist on top of AMSDOS to get up to 208 KB of usable space per side, by using 42 tracks with 10 sectors per track. The problem with that is there is not much gap between sectors anymore, and so the diskette becomes very sensitive to rotation speed.

The FDC controller can be used to control 80-tracks and/or double sided drives, though AMSDOS doesn't support such formats. To use 80-track drives and/or dual-head drives, you need to use another CPC DOS instead of AMSDOS. The most popular one is ParaDOS. It can handle up to 796 KB instead of 178 KB for AMSDOS and it supports 22 different disk file formats, including the three standard AMSDOS ones.