On 9.3.20 10:03, Jan Panteltje wrote:
> On a sunny day (Mon, 09 Mar 2020 06:59:02 GMT) it happened Jan Panteltje
>  wrote in :
>
>>> By the time the IBM PC appeared most CP/M-80 machines I was seeing
>>> wrote either 10 512-byte or 5 1024-byte sectors on a track on a 5.25"
>>> floppy -- making 400kB of data in all. The mostly used WD controller
>>> chips that allowed that. The floppy controller in the PC writes too
>>> much lead data on each track to achieve that, so the best it can do is
>>> 9 sectors of 512 bytes -- giving 360k (DOS originally wrote only 8
>>> sectors to keep the addressing simple(r), giving 320k in all).
>>
>> Not sure you should blame the IBM PC floppy controller,
>> as I used it in the CP/M clone I wrote, am talking about the 8272 chip.
>> I used the Kaypro II format with its 40 tracks x 10 sectors, 400kB.
>> Description of that system here:
>> http://panteltje.com/panteltje/z80/system14/system14.doc.txt
>>
>> Circuit diagram of the floppy controller board I designed:
>> http://panteltje.com/panteltje/z80/system14/diagrams/fdc-1.jpg
>>   IC1 is the 8272, circuit is dated 13-7-1984
>>
>> Wrote the drivers too of course.
>>
>>
>>> IBM really screwed up there!
>>
>> No, maybe it was Microsoft?
>
> PS, could have been an IBM BIOS limitation?


IBM *did* screw up: Already the first documentation on the 8086
family stated clearly that the first 32 (0x0 - 0x1f) exceptions
are reserved for hardware use. IBM succeeded in putting all
BIOS calls in there, making already the 80186/80188 unsuitable
for use with compatible software.

--

-TV

--- SoupGate-Win32 v1.05