My scope didn’t save the test capture .. Next time.
But I could see the clear ones and zeros.
Dirk cleaned up a OCR version of the program, which I cleaned up some more, and found some errors.
So that should be okay now.
A dentro (Combination of demo and intro.) from 1995
It using only background colors behind existing text to display the dentro text. There is (somewhere) a version with animations.
New work:
Booting from floppy, showing a flash screen in mode 13h.
Starting a trackloader, which loads a raw adlib file and plays it.
All sectors written to floppy using my new sector writer.
WOOT .. music at boot time!
Sector writer
use16 org 0100h ; 0 0 1 ; bootblock with flashcode mov cl,1 ; start mov al,1 ; # sectors mov dh,0 ; head mov bx,bootblock call wrtsector ; flash image mov cl,8 mov al,4 mov dh,0 mov bx,gfx call wrtsector ; Music loader mov cl,7 ; start mov al,1 ; # sectors mov dh,0 ; head mov bx,nextpart call wrtsector ; Music raw mov cl,13 ; start mov al,4 ; # sectors mov dh,0 ; head mov bx,musicraw call wrtsector jmp do_exit printerror: push cs ; make ds same as cs pop ds MOV DX,TxtErr1 ; error MOV AH,09h INT 21h mov ax,4c00h int 21h do_exit: mov ax,4c00h int 21h wrtsector: ; On entry: AH 03h ; AL Number of sectors to write ; CH Cylinder number (10-bit value; upper 2 bits ; in CL) ; CL Starting sector number ; DH Head number ; DL Drive number ; ES:BX Address of memory buffer ; ; Returns: AH Status of operation (See Service 01h) ; AL Number of sectors written ; CF Set if error, else cleared cld mov ah, 3h ; int13h function 2 mov ch, 0 ; from cylinder number 0 mov dl,0 push cs pop es int 13h jc printerror ret TxtErr1: DB "Error!",7,10,13,"$" bootblock: file 'flash_b.bin' gfx: file 'flashgfx.raw' nextpart: file 'nextpart.bin' musicraw: file 'LVLINTRO.RAW'
I wanted to know how a floppy differs from a floppy.
So i wrote below code to fill each sector on a floppy disk or image with information at the start of each sector.
(Head, Cylinder and Sector)
empty.bin was made using
dd if=/dev/zero of=empty.bin count=1 bs=512
use16 org 0100h mov ch,0 ; cyl mov cl,1 ; sector mov dh,0 ; head nextsector: mov bx,empty mov [empty],dh mov [empty+1],ch mov [empty+2],cl push cx push ax push dx call printer call wrtsector pop dx pop ax pop ax inc cl cmp cl,19 jnz nextsector mov cl,1 inc ch cmp ch,79 jnz nextsector ; other side mov dh,1 mov ch,0 mov cl,1 nextsector1: mov bx,empty mov [empty],dh mov [empty+1],ch mov [empty+2],cl push cx push ax push dx call wrtsector pop dx pop ax pop ax inc cl cmp cl,19 jnz nextsector1 mov cl,1 inc ch cmp ch,79 jnz nextsector1 jmp do_exit printerror: push cs ; make ds same as cs pop ds MOV AX,3 ; default text mode 3 INT 10h MOV DX,TxtErr1 ; error MOV AH,09h INT 21h mov ax,4c00h int 21h do_exit: mov ax,3 int 10h mov ax,4c00h int 21h printer: push cx push dx mov dl,dh mov ah, 02h add dl, 30h int 21h mov dl,ch add dl, 30h int 21h mov dl,cl add dl, 30h int 21h mov dx,13 mov ah,2 int 21h mov dx,10 mov ah,2 int 21h pop dx pop cx ret wrtsector: ; On entry: AH 03h ; AL Number of sectors to write ; CH Cylinder number (10-bit value; upper 2 bits ; in CL) ; CL Starting sector number ; DH Head number ; DL Drive number ; ES:BX Address of memory buffer ; ; Returns: AH Status of operation (See Service 01h) ; AL Number of sectors written ; CF Set if error, else cleared ; cld mov ah, 3h ; int13h function 2 mov al,1 mov dl,0 push cs pop es int 13h jc printerror ret TxtErr1: DB "Error!",7,10,13,"$" empty: file 'empty.bin'
Viewing the floppy image with ghex
Offset 7000 = Head 1, Cylinder 1 and sector 3
When doing
times 512 – ($-$$) db 0
to fill binaries to 512 bytes, you could cat the sectors to a disk/file with this knowledge.
Writing tools and effects for my new boot demo.
Below the output of the sin/cos generator ( see used in video below )
(It also shows a visual plot of the function)
Source code python script
# importing the required module
import matplotlib.pyplot as plt
import numpy as np
import math
# Change these
numberofdatapoints = 360
maxamp = 180
howmuchfromwave = 0.5
numberofharmonies = 1
# Number of harmonies are sin/cos additions in the calculation line below
# not here
step = 360/numberofdatapoints*howmuchfromwave
offset = maxamp
maxamp = maxamp / numberofharmonies
offset = 0
x = [ ]
for xv in range(numberofdatapoints):
xvstep=xv * step
# Calculation line
# datapoint=np.sin(math.radians(xvstep))
# Double harmony example
datapoint=np.sin(math.radians(xvstep)) + (np.sin(math.radians(xvstep*3))/2)
datapoint=datapoint * maxamp
datapoint=datapoint + offset
x.append(int(datapoint))
print(" db ", end="")
print(*x, sep = ",")
# plotting the points
plt.plot(x)
# naming the x axis
plt.xlabel('x - axis')
# naming the y axis
plt.ylabel('y - axis')
# giving a title to my graph
plt.title('Example')
# function to show the plot
plt.show()
Minimalistic very fast boot loader flash screen effect
Graffiti bouncher test (probably ends up bounching a 320×400 image)
This one uses the generated sintab (Using the python script above)
Test code for a text scroller
Code optimalisation/tricks
clear a (double) register?
xor ax,ax
is faster than
mov ax,0h
Want to make ds pointer same as cs?
Instead of
mov ax,cs
mov ds,ax
use
push cs
pop ds
self modifying code
mostly we just move data around, but you also can change the runtime code (instructions)
Speedcode
Populair on the C64 where resources are even more limited, you could use speedcode.
Most of the speedcode you generate, due to its repeating lines.
When looking at clock cycles you can save some extra cycles, by using a little more memory but saving on “expensive” loops.
Simple example
Left a funtion with a loop, right is the same but all instuctions sequencial
Left 15 bytes but 284 cycles
Right 39 bytes but only 102 cycles!
4 4 ; below part 9 times 9 3 4 16 or 4 = 284 cycles Speedcode 4 2 ; xor is faster 9 3 ; even 2 when you can use BX pair! 9 3 9 3 9 3 9 3 9 3 9 3 9 3 9 = 111 cycles (or 102 BX pair)
Moving memory blocks (No DMA)
;DS:(E)SI to address ES:(E)DI
push cs ; example to set es to code segment
pop es
mov si,1000 ; offset pointer source
xor di,di ; destination offset
mov cx,320*100 ; number of transfers (See below words)
mov ax,0a000h ; Destination
mov es,ax ; destination segment
cld ; Clear direction flag set SI and DI to auto-increment
rep movsw ; repeat mov words! until number of transfers is 0
;
Short binary > bcd > dec (ascii) convert for numbers (0-99)
mov ax,01ch ; = 28
mov bx,0ah ; = 10
div bl ; divide BL by AX
; AX = 0802 ; Remainder/Divider
xchg ah,al ; change around (dont use if you want to keep little endian)
add ax,3030h ; offset to ascii 30=0 31=1
; ax ends up with 3238 .. 28
(NOTE, Dosbox can’t cope with the register speed, use real HW or PCem)
Generating a RAW image and Palette, a in a new way
This bash script to convert BMP to Raw and a compiled colorpalette.
(Note: this converts to 8 bit depth, the assembly code in the final assemby program converts to 6 for VGA mode 13h
So this time, i won´t have to use the standard VGA palette as mentioned in previous posts.
(Gimp colors > indexed (255 colors) ; save as BMP, exclude colorspace information)
I’m using identify to extract the colorpalette, which i’m converting to DB entries for the fasm compiler
#!/bin/bash
if [ $# -lt 1 ] ; then
echo "$0 filename"
exit 0
fi
size=$(stat $1 | grep Size | awk '{ print $2 }')
skipsize=$(( $size - 64000))
dd if=$1 of=$1.raw skip=$skipsize bs=1
identify -verbose $1 | awk '/Colormap:/,/Rendering/' | grep -v Colormap | grep -v Rendering | awk '{ print $2 } ' | tr -d '()' | while read ; do echo "db $REPLY" ;done > data.asm
fasm data.asm
Code
use16
org 0x100
; variables
CRTC_INDEX = 0x03D4
CRTC_DATA = 0x03D5
INPUT_STATUS = 0x03DA
HRETRACE = 0x01
VRETRACE = 0x03 ; bit 3 =8 ?
MAXIMUM_SCAN_LINE = 0x09
LINE_OFFSET = 0x13
; bar
upperbar = 1
lowerbar = 399
jmp start
; memory locations for data
updown dw 1
direction DB 0
filename DB "firework.raw",0
oldline db 0
start:
; set mode 320x200 256 colors palette
mov ah,0x0
mov al,13h
int 10h
; clear screen routine, not really needed
clearscreen:
push ax
mov ax, 0a000h
mov es, ax
pop ax
xor di, di
inc ax
mov cx, 64000 ; 320x200
rep stosb
; set colors
; call file loader
call Loadfile
call setpalette
; Move loaded file to Screen memory
mov ax,0a000h
mov es,ax
mov ax,6000h
mov ds,ax
mov si,0
mov di,0
mov cx,320*200/2
rep movsw
push cs
pop ds
; store org effect2 values
mov dx, CRTC_INDEX
mov al,LINE_OFFSET
out dx,al
mov dx, CRTC_DATA
in al,dx
mov [oldline],al
; after displaying the image or displaying an error, wait for keypress to exit
waitforkeyloop:
call effect ; Calling the effect
MOV AH,1
INT 16h
JZ waitforkeyloop
XOR AH,AH
INT 16h
Exit:
MOV AX,3 ; default text mode 3
INT 10h
MOV AX,4C00h ; exit to dos (terminate process)
INT 21h
; loop ends here
; Loads raw 64000 bytes image to screen memory
Loadfile:
push ds
MOV DX,filename
MOV AX,3D00h ; open filehandle
INT 21h
JC Err1
MOV BX,AX ; filehandle
MOV CX,64000
mov dx,06000h ; destination 0000:a000h - Screen memory
mov ds,dx
MOV DX,0
MOV AH,3Fh ; read from file
INT 21h
JC Err1
MOV AH,3Eh ; close filehandle
INT 21h
pop ds
RET
; print error
Err1:
push cs ; make ds same as cs
pop ds
MOV AX,3 ; default text mode 3
INT 10h
MOV DX,TxtErr1 ; error
MOV AH,09h
INT 21h
RET
effect:
cli ; stop interrupts
call waitvretrace ; wait for vertical retrace
; gets start scanline and direction
mov ax,[updown]
mov cl,[direction]
cmp cl,0 ; 0 move down
jz addcounter
dec ax
dec ax
cmp ax,upperbar ; reached upper bar ?
jnz gohere ; jnz
mov cl,0
mov [direction],cl
jmp gohere
addcounter:
inc ax
inc ax
cmp ax,lowerbar ; reached bottom bar?
jnz gohere ;jnz
mov cl,1 ; change direction
mov [direction],cl
gohere:
mov [updown],ax ; store new location
; al = scanline, call wait for scanline
call longwaithretrace
; other effect
mov dx, CRTC_INDEX
mov al, LINE_OFFSET
out dx,al
mov dx, CRTC_DATA
mov al, 0
out dx,al
; wait scanlines (height of bar)
mov ax,400
mov cx,[updown]
sub ax,cx
call longwaithretrace
; restore effect2
mov dx, CRTC_INDEX
mov al, LINE_OFFSET
out dx, al
mov dx, CRTC_DATA
mov al, [oldline]
out dx,al
sti ; start interrupts again
ret
; routine that fixes 8 to 6 bits and sets palette
setpalette:
; 8 bits to 6
mov si,coltab
mov cx,256*3
rest:
mov al,[si]
shr al,2
mov [si],al
dec cx
inc si
cmp cx,0
jnz rest
; now set colors
mov dx,3c8h
xor al,al
out dx,al
inc dx
mov si,coltab
mov cx,256*3
rep outsb
ret
; this waits for vertical retrace
waitvretrace:
mov dx,INPUT_STATUS
waitv1:
in al,dx
test al,8
jnz waitv1
waitv2:
in al,dx
test al,8
jz waitv2
ret
; routine that waits for horizontal retrace
waithretrace:
mov cl,al
mov dx,INPUT_STATUS
waith1:
in al,dx
test al,1
jnz waith1
waith2:
in al,dx
test al,1
jz waith2
dec cl
cmp cl,0
jnz waith1
ret
longwaithretrace:
mov cx,ax
mov dx,INPUT_STATUS
lwaith1:
in al,dx
test al,1
jnz lwaith1
lwaith2:
in al,dx
test al,1
jz lwaith2
dec cx
cmp cx,0
jnz lwaith1
ret
TxtErr1 DB "firework.raw not found!",7,10,13,"$"
coltab:
include 'data.asm'
Plays RAW adlib songs in 100 lines of code … kindda
Using information from here:
https://moddingwiki.shikadi.net/wiki/RAW_Format_(Adlib)
And using fasm to compile I can play captured raw songs.
But something is still off ?!?
It sounds a little different, and I need to implement a better timer routine. (Below my version and opencubicplayer)
This is a test for my bootloader, playing music from my bootblock!
CODE
use16 org 0x100 ctrlreg=0388h datareg=0389h mainloop: ; set speed (Byte 8,9 from the raw file) mov ax,[tune+8] mov [clockspeed],ax ; call player call rawreg ; wait 0.5 sec for exit mov bx,6 call waitmore jmp exit rawreg: mov bx,tune+0ah ; start of song at offset ah ; order registerdata, register! ; Are there more control codes? ??? lraw: mov cx,[bx] ; reg = 2 - check data cmp ch,2 je checkreg2 ; reg = 0 - cyclewait cmp ch,0 je cyclewait ; data = FFFF - end song - end play routine cmp cx,0ffffh jne skipr ret cyclewait: ; waits cl times waitroutine cylloop: call waitlong dec cl jnz cylloop inc bx inc bx jmp lraw checkreg2: ; check low opl cmp cl,1 jne checkh mov ch,0 mov [highlow],ch jmp incandret checkh: ; check high opl cmp cl,2 jne check00 mov ch,1 mov [highlow],ch jmp incandret check00: ; set new speed cmp cl,0 jne incandret inc bx inc bx mov ax,[bx] mov [clockspeed],ax incandret: ; next double byte in the song please inc bx inc bx jmp lraw skipr: ; sends data to the control and data registers mov dx,ctrlreg mov al,[highlow] cmp al,0 je regokay inc dx inc dx regokay: mov al,ch out dx,al ; call waitshort ; not needed for newer adlib cards mov dx,datareg mov al,[highlow] cmp al,0 je regokay2 inc dx inc dx regokay2: mov al,cl out dx,al call waitlong inc bx inc bx jmp lraw waitshort: push ax push cx push dx mov cx, 0 ;HIGH WORD. mov dx, 010h ;LOW WORD. mov ah, 86h ;WAIT. int 15h pop dx pop cx pop ax ret waitlong: push bx push ax push cx push dx mov cx, 0 ;HIGH WORD. mov dx, [clockspeed] shr dx,1 mov ah, 86h ;WAIT. int 15h pop dx pop cx pop ax pop bx ret waitmore: ; in bx == 12h is 1 sec ; destroys ax,bx,cx,dx push ax push bx push cx push dx mov ax,0h int 1ah add dx, bx mov bx,dx waitloop: mov ax,0h int 1ah cmp bx,dx jnz waitloop pop dx pop cx pop bx pop ax ret exit: mov dx,0388h mov al,0b0h out dx,al inc dx xor al,al out dx,al mov ax,04c00h int 21h clockspeed: dw 0 highlow: db 0 tune: file 'RAWSONG.RAW'
Loading a 320×200 image from 14 cilinders. There is no msdos on the floppy!
9 sectors, 14 cilinders, 1 head * sector size (512 bytes) = 64512 bytes
Mode 13h (320×200 265 colors)
Cuting the raw part from a BMP (see previous post)
root@battlestation:/mnt/# ls -la MAD.bmp -rw-rw-r-- 1 fash fash 65078 Sep 14 15:57 MAD.bmp I need 64000 bytes (320x200) 65078-64000 = 1078 root@battlestation:/mnt/# dd if=MAD.bmp of=mad.raw skip=1078 bs=1 64000+0 records in 64000+0 records out 64000 bytes (64 kB, 62 KiB) copied, 0.441618 s, 145 kB/s
I use debug to write to the sectors
debug mad.raw -w100 0 9 7f (write from address 100 drive=0 startsector=9 (cylinder 1) and 7f sectors long
Wrote a little sector viewer to debug/view data written.
I will post the code after some code cleaning and adding some comments
This is a work in progress, below are my Lab notes.
I want to rewrite pieces we made for a demo, loading images and effects from a floppydisk bootloader.
Without looking at old code (which was written using Masm), I wanted to learn the steps using Fasm.
I started with a boot sector program, It should do the following.
I got it working, half that is.
In the past I used real disks, and now a virtual disk, maybe thats the difference?
First Code
use16 org 0x7c00 mov ah,0x0 mov al,0x13 int 10h mov ax, 0a000h mov es, ax xor di, di mov ax, 50 mov cx, 64000 rep stosb loophere: jmp loophere times 510 - ($-$$) db 0 dw 0xaa55
Compiling and starting:
fasm mybootblock.asm qemu-system-x86_64 --drive format=raw,file=mybootblock.bin
This works, it sets the graphical mode and clears the screen.
Second Code
Skipping the int 25h version
## Track read part
xor ax, ax ; DS = 0
mov ds, ax
cld
mov ah, 2h ; int13h function 2 track read
mov al, 2 ; number of tracks ( should be 2 for reading only palette)
mov ch, 0 ; from cylinder number 0
mov cl, 2 ; the sector number 2 - second sector (starts from 1, not 0)
mov dh, 0 ; head number 0
xor bx, bx ; BX = 0
mov es, bx ; ES = 0
mov bx, 7e00h ; Offset from above
int 13h
call setpal
## End part with setpalette routine
## appending palette.colors
## and a raw image
setpal:
mov dx,3c8h
xor al,al
mov di, ax
out dx,al
inc dx
mov cx,256*3
mov si,07e00h
rep outsb
ret
times 510 - ($-$$) db 0
dw 0xaa55
include 'palette.colors'
times 2048 - ($-$$) db 0
file 'image.raw'
Seems there is still a header on the RAW file, lets look at how I made this.
NOTE! .. Below converts an image with a STANDARD VGA palette, not a custom one as used above
Looking with ghex at the file I saw that there was a header 0x415 bytes large.
(Probably still palette colors in there)
dd if=shoes.bmp of=cutshoe.bmp bs=1 skip=1078 (0x415h + 3?)
worked for me
Loading the extra tracks didn’t work for me?!?!
But how could I define tracks/sectors and heads on a virtual floppy?
I tried to write sectors using debug.com
start dosbox imgmount a: /tmp/floppy.img -t floppy debug.com bootsector.bin -r bx 01 -r cx 512 ; set bx:cx for size -w 100 0 0 1 ; write from address 100, drive 0 (a), sector 0, number of sectors ; testing -l 100 0 0 1 ;load sector to addr 100 drive 0 sector 0 number of sectors This used to work with real disks on a real machine, not in dosbox ?!?! my way to create a disk in linux dd if=bootblock.bin of=disk1.img bs=512 count=1 seek=0 dd if=palette.col of=disk1.img bs=512 count=1 seek=1 # or 2? dd if=shoes.raw of=disk1.img bs=512 count=10000 seek=17
It looks like I can’t read futher than 18 sectors on a virtual floppy.
What next? Head=1? Cylinder=1?
Below the info from a floppy image before altering.
DOS/MBR boot sector, code offset 0x3c+2, OEM-ID “MSDOS5.0”, root entries 224, sectors 2880 (volumes <=32 MB), sectors/FAT 9, sectors/track 18, serial number 0x1c2a0d0f, unlabeled, FAT (12 bit), followed by FAT
Appending the RAW to a executable gave me problems to.
(Without making a bootdisk)
Above and below weird data. Appending the data to the executable needs some work also.
At least for today let me display this image correctly 🙂
use16 org 0x100 ; set mode 320x200 256 colors palette mov ah,0x0 mov al,13h int 10h ; clear screen routine, not really needed clearscreen: push ax mov ax, 0a000h mov es, ax pop ax xor di, di inc ax mov cx, 64000 ; 320x200 rep stosb ; call file loader call Loadfile ; after displaying the image or displaying an error, wait for keypress to exit waitforkeyloop: MOV AH,1 INT 16h JZ waitforkeyloop XOR AH,AH INT 16h Exit: MOV AX,3 ; default text mode 3 INT 10h MOV AX,4C00h ; exit to dos (terminate process) INT 21h Loadfile: MOV DX,filename MOV AX,3D00h ; open filehandle INT 21h JC Err1 MOV BX,AX ; filehandle MOV CX,0FFFFh ; size mov dx,0a000h ; destination 0000:a000h - Screen memory mov ds,dx MOV DX,0 MOV AH,3Fh ; read from file INT 21h JC Err1 MOV AH,3Eh ; close filehandle INT 21h RET ; print error Err1: push cs ; make ds same as cs pop ds MOV DX,TxtErr1 ; error MOV AH,09h INT 21h RET filename DB "shoes.bmp",0 TxtErr1 DB "shoes.bmp not found!",7,10,13,"$"
Tomorrow .. back to the track loader
Quiet days, I working on some art.
But here are the last ‘prutsen’
My current Wifi setup
I’ve got a Wifi outside of my network for guest and emergency. ( 2 SSIDs)
Then a main Wifi router in my livingroom, one in my workshop/studio and one in the Attic (Electronics Lab)
So three main Wifi AccessPoints. These all have the same SSID’s but on different frequencies. That way i’ve got roaming in and outside my house.
Also some virtual accesspoints are configured.
I’ve got a main, folkband, IOT, guest-inside all on 2.4Ghz and 5Ghz.
I watched a lot of YT presentations about Mikrotik Wifi.
So I ended up with DFS safe channels 20Mhz for 2.4 and 20/40Mhz Ce for 5Ghz. (subchannels for each after some frequency scanning)
(2.4 does a failback to 20Mhz whenever there is even one client detected which connects only on this band. Such as some old IOT stuff)
2.4 in only 1,6 and 11 no overlap, each on another device.
300Mbps is sufficient for my wifi 🙂
I’ve got accesslists in place and i’m going to read into kicking a client when the signal strenght is below -50dB
80386 (DX) Computer
Besides my 8088 and 8086 machines I needed a machine which could run our old demo’s. So I bought a new toy.
It has 8Mb Ram and runs at 40Mhz.
I’ve noticed that many of my VGA register manipulation code, can’t be run on a modern VGA monitor, I need to use a CRT for that .. Another thing to buy
Lilygo T-Display S3 Streaming
Not my code: https://github.com/Steve5451/esp32-stream-desktop
A very cool project!
Needed to fix arduino code, due to the TFT_eSPI library issues.
And I’ve got a S3 with another resolution, but that was an easy fix.
Then needed to reinstall nodejs with another version.
Had to modify the code because the tcp server would not start.
Weird errors logging, but in the end fixed … very cool
I probably end up designing a 3D printed case that looks like a monitor or tv.
Below you can find the files in a zip for writing assemby machine code in dosbox.
I’ve got mine extracted in ~/projects/dos
To automount this i’ve edited
~/.dosbox/dosbox-*.conf ;-------- bottom part [autoexec] # Lines in this section will be run at startup. # You can put your MOUNT lines here. mount c: /home/myusername/projects/dos c:
The files
Template
; everything in 64k CS, DS, and SS are the same
.model small
; start pointer
.stack 100h
.code
start:
; set mode 13 (320x200 * 265 colors palette)
mov ah,0
mov al,13h
int 10h
; set pixel in the middle color 2 (cyan)
mov ah,0ch
mov al,2
mov cx,160
mov dx,100
int 10h
; wait key input
mov ah,0
int 16h
; set mode back to text
mov ah,0
mov al,3
int 10h
; exit to dos
mov ax,4c00h
int 21h
end start
Usage
start dosbox
a gfx (edit gfx.asm example)
esc, q, e
gfx (run program)
In previous post :
I mentioned a 2 ROM setup because the 8086 is 16bits instead of 8.
So I was wondering that maybe a recompile was needed, or the data being split over two roms (odd/even)
The guy from GLABios was so kind to build me two interleaved roms.
So while working on a padded bench, I tested the ROMs.
Back to the roms, it didn’t work!
But I missed a detail in the technical manual (the bold text)
In Turbo XT /2 and Turbo XT /3, there are two 28-pin sockets for ROM,
both of them are occupied by 2764 which stored the BIOS. The contents
of the two 2764 are identical. One of them contribute the ODD Byte to the system and the other EVEN Byte. Together they support 16 Bit BIOS
access.
I don’t know why this is how it works, but when I flashed two the same 28C64’s it worked!
(I also tought that is was strange that both original roms had the same markings.
It workes!
Only remarks/observations:
UPDATE
GLABios was not updated for displaying 8086 yet.
Error 1701 was the (old spinning) harddisk not being connected.
