MiSTer PCXT

[breiztiger]

yes but the same

[wark91]

ok so better stick to the build where the keyboard works until it is fixed again on your side

[kitune-san]

I have created a preliminary testing branch before the next beta, this time I add the proposal of @kitune-san... you can add that you comment, I don't know exactly how was the change or if I have to add it from a previous pull request that I rejected you, please feel free to make pull request to this branch, I accept it and generate a release for users to test it:

https://github.com/spark2k06/PCXT_MiSTe ... e_beta_1_2

Edit: I think I've made the two missing changes, I'll give it a try.

Thank you very much.
I will send you a pull request regarding the keyboard (interrupt) after my confirmation.

[spark2k06]

Thank you very much.
I will send you a pull request regarding the keyboard (interrupt) after my confirmation.

Great. I don't know when I will be able to test it and accept the pull request, so if you don't mind, please also provide the corresponding RBF file so that users can test it as soon as possible.

Thank you.

[kitune-san]

OK.

PCXT.zip

(1.03 MiB) Downloaded 150 times

I am currently experiencing a strange SDRAM error when I run the pcxt_uart_hdd script in my environment.
Until resolved, the pull request will be put off. Sorry.

[kitune-san]

Apparently it happens when the room(device) temperature is low.

[Mills]

Apparently it happens when the room(device) temperature is low.

I'm having the same issue with this last rbf. It happened to me with some of the older versions, and I don't really know what causes it in my case, it looks like random.

On the other hand, keyboard seems to work better, it did not disconnect or stop working.

[breiztiger]

@kitune-san with you last version keyboard seem to work

but sometime when load core i have continues beep

[MicroCoreLabs]

Apparently it happens when the room(device) temperature is low.

Temperature dependence is an indicator of a timing error in the FPGA. I suggest examining every instance where signals cross clock domains and also make an effort to completely constrain the design so you can achieve consistent results between synthesis passes and across devices.

Sometimes it makes sense to eliminate clock domains to simplify the design. Rather than clocking some modules at 4.77 Mhz, it might be simpler to clock them at a more common/faster clock domain like 100 Mhz and just sample the 4.7Mhz as if it was just a signal. This is what I do with the MCL86. Having a single or few clock domains greatly simplifies the design.

One other thing to check is the supported clock speed of the PLL(s)... Sometimes there is a minimum that they support, so perhaps 14 Mhz is too slow.

[spark2k06]

@kitune-san with you last version keyboard seem to work

but sometime when load core i have continues beep

Mystery solved, you were using Sergey's default BIOS, instead of Juko ST. That BIOS was already known to have the keyboard problem, which kitune-san has now solved.

The continuous beeps in Sergey's BIOS are due to RAM problems.

[kitune-san]

Apparently it happens when the room(device) temperature is low.

Temperature dependence is an indicator of a timing error in the FPGA. I suggest examining every instance where signals cross clock domains and also make an effort to completely constrain the design so you can achieve consistent results between synthesis passes and across devices.

Sometimes it makes sense to eliminate clock domains to simplify the design. Rather than clocking some modules at 4.77 Mhz, it might be simpler to clock them at a more common/faster clock domain like 100 Mhz and just sample the 4.7Mhz as if it was just a signal. This is what I do with the MCL86. Having a single or few clock domains greatly simplifies the design.

One other thing to check is the supported clock speed of the PLL(s)... Sometimes there is a minimum that they support, so perhaps 14 Mhz is too slow.

Honestly, I am surprised at how many clocks this core has.
When I saw your design, I certainly considered changing the design to use 100 MHz. However, I gave up because I did not have enough time to work on it and I could not change the design of other modules.
Then we have to constrain the timing. It will be difficult. But we have to do it...

[breiztiger]

RAM failure…

But when it’s arrive just reload the core solve thé beep problem

Is there a way to have real 5160 bios with xt ide ?

[spark2k06]

RAM failure…

But when it’s arrive just reload the core solve thé beep problem

Is there a way to have real 5160 bios with xt ide ?

Yes, it will be possible when it is stable and faultless 101 in this BIOS, but XTIDE will have to be placed in the E800 segment, in F000 it is not possible... therefore, with this BIOS we will have 16Kb less UMB memory. Juko ST is also a very good alternative at the moment, and XTIDE can be placed in F000 without any problem.

[MicroCoreLabs]

Apparently it happens when the room(device) temperature is low.

Temperature dependence is an indicator of a timing error in the FPGA. I suggest examining every instance where signals cross clock domains and also make an effort to completely constrain the design so you can achieve consistent results between synthesis passes and across devices.

Sometimes it makes sense to eliminate clock domains to simplify the design. Rather than clocking some modules at 4.77 Mhz, it might be simpler to clock them at a more common/faster clock domain like 100 Mhz and just sample the 4.7Mhz as if it was just a signal. This is what I do with the MCL86. Having a single or few clock domains greatly simplifies the design.

One other thing to check is the supported clock speed of the PLL(s)... Sometimes there is a minimum that they support, so perhaps 14 Mhz is too slow.

Honestly, I am surprised at how many clocks this core has.
When I saw your design, I certainly considered changing the design to use 100 MHz. However, I gave up because I did not have enough time to work on it and I could not change the design of other modules.
Then we have to constrain the timing. It will be difficult. But we have to do it...

The MCL86 BIU, for example, samples the 4.77 Mhz clock at 100Mhz as a regular signal. It is not treated as a clock. So a problem arises because you are routing 4.77 Mhz over the clock distribution network. So how does this signal get off of this network to be used as the input to a register?

There has been tremendous and surprisingly fast progress so far integrating this design; however it may be time to go back and re-examine some of the cores to simplify them into fewer clock domains, consider and minimize the number of clock domain crossings, examine what signals actually need to be clocks, pipeline signals with multiple loads, and fully constrain the design.

[kathleen]

Thanks a thousand of times for this amazing progress on this core.
I've already tried some applications that I'm using on my real Tandy and so far they all run like a charm

* Personal Inventory
* Logo IBM
* Eureka the solver
* Turbo Pascal
* GWbasic
* Visicalc
* PC-Calc

Will test more apps for sure. But I'm already so happy to see all of those working, just had some keyboard issues after a reboot (CTRL+ALT+ALT) where the only way to recover the keyboard was to restart the core.

Looking forward to see the 720*348 Herc. mode.

Again, thank you !

[NML32]

Could someone share their Config.sys and Autoexec.bat settings they are using for the Lo-Tech 2Mb EMS?

I downloaded the drivers from https://www.lo-tech.co.uk/wiki/LTEMM.EXE

[suww37]

I have put the configuration you show in MiSTer.ini and it works perfectly, both HDMI and VGA, look at the resolutions that the video module requires... maybe they are not compatible with your LCD monitor?

photo_2022-06-26_17-40-52.jpg

And once the splash screen is finished, in the BIOS:

photo_2022-06-26_17-45-08.jpg

Early on, your PCXT cores (PCXT_20220507.rbf, PCXT_20220510.rbf, PCXT_20220512.rbf) displayed good screen output through the 15pin vga port of the io board. The recently updated core (PCXT_20220523.rbf ~ PCXT_20220626.rbf) is not able to output screen . If I connect the hdmi port of the d-nano board to my lcd monitor, it outputs normally. Therefore, I think that there is no problem with the supported resolution of my lcd monitor. The monitor on the left in the picture I uploaded is a crt monitor that only pushes 15khz. It is connected to the d-nano board. If I set direct_video=0, the screen doesn't support 31khz, so the screen shakes and looks weird. I think that the LCD monitor connected to the io board connected at the same time should come out normally. Your early cores ( PCXT_20220507.rbf, PCXT_20220510.rbf, PCXT_20220512.rbf) were displayed normally. I can't understand why your latest cores don't support normal screen output in io board.

Did you try vsync_adjust=0 in mister.ini? It usually makes everything work (at least on lcd), at the cost of (maybe) choppy scrolls in any smooth scrolling game/program (very few for CGA). But for the moment It can be used to test. My lcd, connected using vga or dvi adapters, did not display the MDA graphics if vsync_adjust is not 0.

Also this is still a beta, bugs will be common for the moment.

Thank you. I solved it your way (mister.ini: vsync_adjust=0).

[flynnsbit]

Could someone share their Config.sys and Autoexec.bat settings they are using for the Lo-Tech 2Mb EMS?

I downloaded the drivers from https://www.lo-tech.co.uk/wiki/LTEMM.EXE

https://github.com/spark2k06/PCXT_MiSTe ... CONFIG.SYS

or I can send you my image I am working on if you ping me on discord. It's not perfect but I ran through all the emstests successfully. 4dos shell loads into ems which is nice. I'm still getting some small crashes depending on the app. I've been slowly collecting all these tools that people have changed over the years. There is a version of LTEMM that silences the memory ticks, there is a version of USE!UMBS.SYS that allows for multiple memory ranges if we ever need that. etc...

[kitune-san]

The MCL86 BIU, for example, samples the 4.77 Mhz clock at 100Mhz as a regular signal. It is not treated as a clock. So a problem arises because you are routing 4.77 Mhz over the clock distribution network. So how does this signal get off of this network to be used as the input to a register?

At the very least, the modules 8288, 8237, and ready signal, which control the bus, need to be redesign.
Other modules need to be considered carefully, but may only require a change in clock (and edges).

I'll think about it some more. Thank you!

[kitune-san]

Deleted.

[suww37]

I have very little computer knowledge so it's so hard to make hdd.img. If there is a pcxt games pack hdd.img file such as "MiSTer AO486 Top 300 DOS Games Pack" or a hdd.img file made by someone else, it would be appreciated if someone upload the file.

https://archive.org/details/top-300-final

[flynnsbit]

I have very little computer knowledge so it's so hard to make hdd.img. If there is a pcxt games pack hdd.img file such as "MiSTer AO486 Top 300 DOS Games Pack" or a hdd.img file made by someone else, it would be appreciated if someone upload the file.

https://archive.org/details/top-300-final

I don't think we are close to that yet and releasing something like that would be unfair for the developers who are actively working this publicly here. I feel like having access to something like that for those who can't dive into the os options and understand when something is a core bug and when something is "DOS being DOS" would cause too many false reports of stability.

I am putting effort into it though and trying to find the best combination of OS + Games + frontend (MyMenu).
I've built the following images:
IBM DOS 5
FreeDOS 1.3 (XT specific build from this video: https://www.youtube.com/watch?v=EOVLlMQs9f8)
MS DOS 6.22
Compaq DOS 3.31
(working on PC-DOS 2000)

There are issues with each above depending on what you are trying to accomplish. I would suggest downloading 86Box and then you can configure it very close to the core:
Machine type: 8080
Machine: [8088] Juko ST
Speed: whatever
Display: [ISA] CGA
Sound: [ISA] Adlib
HD Controller: [ISA] PC/XT XTIDE
FD Controller: Internal controller
Hard Disks: IDE mount a hdd.img
Floppy: mount any floppy img or install from Winworld

I've attached an example of my 86Box config for reference. Doing it this way, you can quickly install different os from floppy image and then just copy the hdd.img to your mister and test. This allows easy access to both the hdd.img and floppy drives to changing the images quickly.

Plug-and-play images will eventually come. It needs some love, and probably IDE support before the larger audience gets a hold of this. My opinion.

[Malor]

Without any XT experience, learning what's up using 86Box would be a great idea. That emulator is very solid, so if you think you see a bug in the Mister core, make sure to verify the behavior against 86Box. If it fails both places, it's probably operator error.

FPGA emulation in general tends to be less polished, and a core that's still being developed is likely to be a minefield of gotchas. Combining that with new-user problems may be frustrating. Don't be afraid to fall back to 86Box and learn how things work there; being on virtual hardware that's quite close to correct will be easier while you figure things out.

As an aside, it still kind of amazes me that the PC won the computer war. The XT was such a pile of crap. Everything about the design was terrible. But it had three main strengths; it was an open architecture, it was built to a very high quality standard, and it had an *outstanding* keyboard. Combine that with the ability to field 640K, and you could get a heck of a lot of work done a heck of a lot more easily than on the true 8-bits.

But, man, the 68000 machines were so much better on a technical basis. Motorola really should not have lost that war.

[jordi]

Without any XT experience, learning what's up using 86Box would be a great idea. That emulator is very solid, so if you think you see a bug in the Mister core, make sure to verify the behavior against 86Box. If it fails both places, it's probably operator error.

FPGA emulation in general tends to be less polished, and a core that's still being developed is likely to be a minefield of gotchas. Combining that with new-user problems may be frustrating. Don't be afraid to fall back to 86Box and learn how things work there; being on virtual hardware that's quite close to correct will be easier while you figure things out.

As an aside, it still kind of amazes me that the PC won the computer war. The XT was such a pile of crap. Everything about the design was terrible. But it had three main strengths; it was an open architecture, it was built to a very high quality standard, and it had an *outstanding* keyboard. Combine that with the ability to field 640K, and you could get a heck of a lot of work done a heck of a lot more easily than on the true 8-bits.

But, man, the 68000 machines were so much better on a technical basis. Motorola really should not have lost that war.

At the end it was the industry support, and IBM had a huge commercial power in companies.

Computer war had a lot of war fronts:
- First, microcomputers like the Spectrum, CPC, MSX and C64 won the houses war before PC became cheaper.
- Then Amiga and Atari ST kept the home market. I would say, the Spectrum kept in countries like Spain for sooo long.

Macintosh, also in this competition. I have one since the end of the 80s, and it was a nightmare of limitations as well. Also, no disk compatibility at all.

[suww37]

If I run "make_boot_with_jukost.py", is the "boot.rom" file formed? When running with python, "boot.rom" is not formed and the following error occurs.

Python 3.10.5 (tags/v3.10.5:f377153, Jun 6 2022, 16:14:13) [MSC v.1929 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license()" for more information.

==================== RESTART: D:\SW\make_boot_with_jukost.py ===================
Traceback (most recent call last):
File "D:\SW\make_boot_with_jukost.py", line 4, in <module>
import requests
ModuleNotFoundError: No module named 'requests'

[suww37]

I have very little computer knowledge so it's so hard to make hdd.img. If there is a pcxt games pack hdd.img file such as "MiSTer AO486 Top 300 DOS Games Pack" or a hdd.img file made by someone else, it would be appreciated if someone upload the file.

https://archive.org/details/top-300-final

I don't think we are close to that yet and releasing something like that would be unfair for the developers who are actively working this publicly here. I feel like having access to something like that for those who can't dive into the os options and understand when something is a core bug and when something is "DOS being DOS" would cause too many false reports of stability.

I am putting effort into it though and trying to find the best combination of OS + Games + frontend (MyMenu).
I've built the following images:
IBM DOS 5
FreeDOS 1.3 (XT specific build from this video: https://www.youtube.com/watch?v=EOVLlMQs9f8)
MS DOS 6.22
Compaq DOS 3.31
(working on PC-DOS 2000)

There are issues with each above depending on what you are trying to accomplish. I would suggest downloading 86Box and then you can configure it very close to the core:
Machine type: 8080
Machine: [8088] Juko ST
Speed: whatever
Display: [ISA] CGA
Sound: [ISA] Adlib
HD Controller: [ISA] PC/XT XTIDE
FD Controller: Internal controller
Hard Disks: IDE mount a hdd.img
Floppy: mount any floppy img or install from Winworld

I've attached an example of my 86Box config for reference. Doing it this way, you can quickly install different os from floppy image and then just copy the hdd.img to your mister and test. This allows easy access to both the hdd.img and floppy drives to changing the images quickly.

Plug-and-play images will eventually come. It needs some love, and probably IDE support before the larger audience gets a hold of this. My opinion.

I am not capable of doing the method you described. I have a hard time doing it. "junko ST" bios (boot.rom) cannot be formed, nor can hdd.img be formed. I don't know how to do it with 86box. I would be grateful if you upload the hdd.img and boot.rom(juko st bios) files that are just bootable.

[Mills]

You have to install a library called "requests" for python.

[suww37]

You have to install a library called "requests" for python.

I got "junko st bios" and booted into hdd.img. However, the boot continues to fail. Please help

[jca]

Your hdd.img is not correct. How did you make it?

[suww37]

Your hdd.img is not correct. How did you make it?

I found and downloaded the ms-dos img file that booted from google and renamed to hdd.img.