VOGONS


Reply 460 of 481, by Acorns

User metadata
Rank Newbie
Rank
Newbie

Catch Catch playing on a 16Mhz 286 Packard Bell Victory, fullscreen at 30FPS. With the Picomem V1 acting as the hard drive and sound card (and ethernet card, though it didn't want to auth this boot for some reason). On the custom firmware as patched a couple posts above.

https://www.youtube.com/watch?v=fEJPLT_rSRY

Reply 461 of 481, by Acorns

User metadata
Rank Newbie
Rank
Newbie
Acorns wrote on 2026-10-31, 00:24:

Catch Catch playing on a 16Mhz 286 Packard Bell Victory, fullscreen at 30FPS. With the Picomem V1 acting as the hard drive and sound card (and ethernet card, though it didn't want to auth this boot for some reason). On the custom firmware as patched a couple posts above.

https://www.youtube.com/watch?v=fEJPLT_rSRY

Apologies for the spam of posts, I cannot edit the existing one. Wasn't happy with the CRT's flashing in the phone camera, so reshot on an LCD.

https://www.youtube.com/watch?v=B2gtyRtmwMw

Reply 462 of 481, by Vivid Warthog

User metadata
Rank Newbie
Rank
Newbie
FreddyV wrote on 2026-10-23, 10:09:
Hi, Thanks 😀 Are you ok if I check and integrate some of this in next firmware ? I started also a Web interface on my side.... T […]
Show full quote
Vivid Warthog wrote on 2026-10-11, 11:20:
Hi! Just wanted to share that I've made some work on the PicoMEM (2.0) FW fork that focuses heavily on Sound Blaster compatibili […]
Show full quote

Hi!
Just wanted to share that I've made some work on the PicoMEM (2.0) FW fork that focuses heavily on Sound Blaster compatibility and audio quality, plus some QoL improvements (like full setup via config.txt, so no PMINIT call is needed or firmware update from SD and access to SD card with HTTP over WiFi - so no more need to unscrew the pc tower to update the card 😁)

There are also some bugs fixed, but all mentioned in CHANGES.md file.

So if you are using PicoMEM2 for SB/FM audio, you can check it out:

https://github.com/HitoriSensei/ISA-PicoMEM

Hi,
Thanks 😀
Are you ok if I check and integrate some of this in next firmware ?
I started also a Web interface on my side....
The public picomem source is not the latest one, so will be lot of work to manually include this.
I also have to ensure it does not use too many ram

Hello FreddyV!
Absolutely, go ahead, grab and integrate whatever you find useful 😁
I understand that some things might be conflicting, as I've based the changes on the older code. Seems we fixed some of the same bugs (IRQ7 for example), probably in a completely different way.

Reply 463 of 481, by RetroBard

User metadata
Rank Newbie
Rank
Newbie

I finally got the PicoMem working properly on the loaner EuroPC. I removed the PREBOOT from the config, so I was able to access the PicoMem settings and change them.

The downside to all this is that I may have to try to find another EuroPC to install the PicoMem into. I believe, that since I no longer can get an image out of mine, it’s only good for spare parts.

Reply 464 of 481, by RetroBard

User metadata
Rank Newbie
Rank
Newbie

IMG_1325.jpg?ex=6ae8a9b3&is=6ae75833&hm=96bc6d0e2657be8d18af064abef63fef453cb10d02240d65f932fdd34dcea6d2&

IMG_1326.mov?ex=6ae8a9b3&is=6ae75833&hm=7c0aa648cf6e5bc7b655e2a511329de2b8629092fea5bd72dfd00fe057c4bf29&

A couple of images of the PicoMem 2 (and 1) working on a loaner EuroPC. I’m excited to finally see it working. The only downside is that my own EuroPC seems to have deteriorated to good-for-spares condition. With time and patience, I might get it working again, but at the moment, I really don’t have the time and energy to do that.

Reply 465 of 481, by RetroBard

User metadata
Rank Newbie
Rank
Newbie

IMG_1331.jpg?ex=6ae8cb2d&is=6ae779ad&hm=e005d588fccffd9c0109ee9425cf96e0640cfbc91ba81351a89a424fc0213240&

IMG_1330.mov?ex=6ae8cb2d&is=6ae779ad&hm=d4ac50a6ee9cbc6c9cdf6a8d8e111093aa577d942a6023116ccb1badfc193d1c&

Well, talk about a Hail Mary working out… The above are from my EuroPC with Indy 500 running from the PicoMem2.

So, I did some more probing on the mainboard, and found out that one address line from the expansion slot wasn’t making contact to one logic chip. I figured I had nothing to lose if I simply soldered a long bodge wire across the board to connect the two points, as the system wasn’t even posting. After adding the bodge, it still wasn’t posting but tried and true knocking against the tabletop sorted out that at least temporarily, and lo and behold, PicoMem was identified and the system was booting into the hard disk image I had on the SD card. You cant imagine how good it felt to hear the Planet X3 intro out of the PicoMem’s audio.

This has been a long journey, and it still isn’t quite over as I still have the issue of the system not posting without a good knock, so I need to get that sorted out still, but I accept my victories as they come.

Reply 466 of 481, by RetroBard

User metadata
Rank Newbie
Rank
Newbie

Some more progress. I tried the probably the most obvious treatment to the non booting issue and pressed all the socketed chips, which seemed to fix it. Now the EuroPC seems to boot consistently.

I suppose I can now concentrate on configuring the PicoMem…

Reply 467 of 481, by NeilKnows

User metadata
Rank Newbie
Rank
Newbie
RetroBard wrote on 2026-11-01, 20:09:

Some more progress. I tried the probably the most obvious treatment to the non booting issue and pressed all the socketed chips, which seemed to fix it. Now the EuroPC seems to boot consistently.
I suppose I can now concentrate on configuring the PicoMem…

Glad you got it working

Reply 468 of 481, by WSkinny

User metadata
Rank Newbie
Rank
Newbie
FreddyV wrote on 2024-04-01, 13:58:
Hi, […]
Show full quote
PeacockXT wrote on 2024-03-31, 19:56:
STrRedWolf wrote on 2024-03-30, 13:17:

Well, if you multiplex the data and part of the address into 16 pins... could multiplex some of the control pins as well... Adafruit PSRAM only takes 4 pins plus power and ground... so relatively easy to do with the right hardware.

I have looked at the pcb pictures and believe that 10 GPIO's handle the address Input and 8 of them handle Data I/O too. Pico GPIO has to toggle between address select, Data select, Data direction, address multipexing , and all are time critical.
Excellent programming job.

Hi,

Close, it is is multiplexed on 12 Pin : A0-A7 / A8-A15 / D0-D7 then MR-MW-IOR-IOW /A16-A20
Plus 2 Pin to control the multiplexing > 14 Pin in total

Hi there FreddyV, hope you're not from Elmstreet 😀

Anway, I was looking to get the schematics of the picomem board, don't really need neither the lastest version but the earliest would be best avoided as well, but let me explain.

I love your project, yes, it does have a number of features that would help me with my "new" DX2/80, but it also has features I do not need at all like the memory related features. FDD/HDD, Audio and mouse support though are indeed features I do want.

Now, I have a few Pico's myself as well as a number of EncXxxx ethernet module, Cyclone 2 dev boards, ESP32, ESP8266 and a based Cyclone IV as well as Cyclone IV dev board with VGA output, PS/2 connector and a couple other things not relevant for my goals which involve a using about 2 of the Cyclone 2's or maybe the basic Cyclone IV (this would be the ISA interface management), one of the Pico's (mainly to have at least some work already done), the I2S module for audio output and a MilkyV-DUE 256Mb and an expansion port to allow maybe later on use the Cyclone IV dev board with VGA output to be a graphics card.

My objective is to use these components on a 16Bit card to implement the following support on ISA x86 systems:
- USB Mouse to PS/2 or BUS Mouse
- FDD Emulation
- HDD/CDROM emulation
- SB32 emulation (using the MilkyV in part due to the available RAM) ensuring full Yamaha OPL3 support, 8/16Bit Digital Audio Support and importantly, EMU8000 support.
- 1 Serial Port
- MIDI IN/OUT ports
- Ethernet Support 10/100Mbps (NE2000 emulation or maybe a 3COM or early Intel model)
- WiFi support via ESP32 (this and the ethernet are to be supported via the MilkyV since I can both connect the ENC adapter and the ESP32 to it, in particular the ESP32 has firmware that enables it to work as a Client or AP under linux making this a suitable solution)
- Expansion header allowing the board to be directly connected to the Cyclone IV dev card and have it implement MDA, Hercules, CGA, EGA, VGA, SVGA, XGA (maybe ECS/AGA, yes I now not PC modes) and if it fits on the available LEs some sort of supported 3D acceleration, if not that maybe Video decoding
- Front panel with a 2.8" Touch screen and additional SD card reader, the front panel card reader will be primarily for FDD and CDROM images, HDD images should if possible be implemented directly on the MilkyV

Another important detail I am interested in implementing is DMA support for the HDD and CDROM emulation at least for HDD if nothing else, this is in part why the FPGA is the chosen component to link and manage the ISA bus connectivity and settings. This will also enable software configuration of IRQs, DMAs, Addresses etc.

Now, I am quite aware that I will need help with a number of things but at a minimum I am certain to be capable of implementing the circuit design and have an initial working pcb if not immediately capable of having all desired features working at least their hardware implementation will be ready for the software to enable all features.
The secondary graphical card support feature in particular is certain to be the last thing to be implemented.

If you look at the feature set you will easily notice that this is basically a single (well dual) add-in board solution to get any x86 motherboard with only keyboard CPU and RAM to become a fully usable system.

One thing I will promise you is not to add to any support burden you may have from your own PicoMEM project and will for sure mention your help and base design sources where I'll be deriving part of the hardware and softare design from.
This is a project I will definitely make fully opensource and have no intent on making any profit from other than my own experience and the pleasure of getting some Retro hardware and related geeks happy 😉

The most important detail of all in my goals is that the whole shebang must run smoothly under OS/2 Warp 3 and newer 😀 Also, NT 3.51 though not as important as far as I am concerned.

If you can share the base circuit schematic in an EDA supported format I can work with and expand uppon I would be very very happy for your help, past that I won't be asking for anything else.

Best regards,
WSkinny

Reply 469 of 481, by douglar

User metadata
Rank l33t
Rank
l33t
WSkinny wrote on 2026-11-09, 22:43:

Another important detail I am interested in implementing is DMA support for the HDD and CDROM emulation at least for HDD if nothing else, this is in part why the FPGA is the chosen component to link and manage the ISA bus connectivity and settings. This will also enable software configuration of IRQs, DMAs, Addresses etc.

Edit -- Added some clarity--

8 bit DMA is useful for floppy drives and sound cards where there is a requirement for a minimum guaranteed data delivery rate. You don't have to worry about interrupts derailing the system. That's not the case for IDE hard drives or CD-ROMs. They don't have a minimum requirement. Doing a DMA implementation for IDE over 8 bit ISA is more complicated, would have more compatibility issues, and even in the most competitive scenario, like old processors that lacked the INSW instruction, the performance would still struggle to match the speed of FreddyV's memory mapped IO driver.

DMA became useful for hard drives when multitasking operating systems arrived, but only after Pentium chipsets allowed the processor to function while a DMA transfer was going on in the back ground. On almost everything older than a Pentium system, the CPU is essentially shut off while ISA DMA transactions are occurring, so there's no multitasking reward for DMA on ISA cards.

DMA for CD-ROM drives became useful to avoid buffer under runs when burning a CD ROM, but this isn't a CD-R, so that doesn't figure in either.

Reply 470 of 481, by WSkinny

User metadata
Rank Newbie
Rank
Newbie
douglar wrote on 2026-11-10, 12:26:
Edit -- Added some clarity-- […]
Show full quote
WSkinny wrote on 2026-11-09, 22:43:

Another important detail I am interested in implementing is DMA support for the HDD and CDROM emulation at least for HDD if nothing else, this is in part why the FPGA is the chosen component to link and manage the ISA bus connectivity and settings. This will also enable software configuration of IRQs, DMAs, Addresses etc.

Edit -- Added some clarity--

8 bit DMA is useful for floppy drives and sound cards where there is a requirement for a minimum guaranteed data delivery rate. You don't have to worry about interrupts derailing the system. That's not the case for IDE hard drives or CD-ROMs. They don't have a minimum requirement. Doing a DMA implementation for IDE over 8 bit ISA is more complicated, would have more compatibility issues, and even in the most competitive scenario, like old processors that lacked the INSW instruction, the performance would still struggle to match the speed of FreddyV's memory mapped IO driver.

DMA became useful for hard drives when multitasking operating systems arrived, but only after Pentium chipsets allowed the processor to function while a DMA transfer was going on in the back ground. On almost everything older than a Pentium system, the CPU is essentially shut off while ISA DMA transactions are occurring, so there's no multitasking reward for DMA on ISA cards.

DMA for CD-ROM drives became useful to avoid buffer under runs when burning a CD ROM, but this isn't a CD-R, so that doesn't figure in either.

That DMA detail is indeed interesting and important. That said, I have noted that I am implementing this on a 16Bit card 😀 which is why the FPGA frontend to the bus becomes handy. Now, DMA will be likely one of the last items on the implementation list and the only reason I am looking at that as a requirement is because I want this baby to work under OS/2 Warp 3 maybe Warp 4.x .

Would love as such the base schematic of the Picomem to give me a base to work upon and expand, yes that is me being a bit lazy but also relying on something that has worked very well for me a number of times, like for example when I made my first serial cable for my Nokia 5110 and Siemens phone, they had similar MAX232 based designs but there were some differences so I combined both designs into a single circuit needing only one single switch to select the device type from the Nokia 5110 to the Siemens (cant recall the model of that one sadly). This circuit worked beautifully for both phones so that was my first win in combining similar schematics and adapting them. In a way reminds me of Karnaugh maps hehehe (hope you know what I am referring to).

Best regards,
WSkinny, 8bits and one nibble! (possibly a nibble on the pie)

Reply 471 of 481, by douglar

User metadata
Rank l33t
Rank
l33t
WSkinny wrote on 2026-11-10, 22:35:

,DMA will be likely one of the last items on the implementation list and the only reason I am looking at that as a requirement is because I want this baby to work under OS/2 Warp 3 maybe Warp 4.x

Not sure why you think DMA is a requirement for OS/2. Could you give some additional insight into that?

Are you looking at implementing 3rd party DMA through the 8237 DMA controllers or are you looking at doing a full first party bus master design?

Reply 472 of 481, by Acorns

User metadata
Rank Newbie
Rank
Newbie

Tested more, and ended up needing the custom firmware on the Picomem V2 after all. On the 286 I had been able to work around it on the stock firmware by limiting to 8 sectors per request, and filling the SB buffer right before an interrupt mask. This cut I/O throughput by ~10%, but left the sound (mostly) working. On a 5150, the same strategy cut I/O throughput by ~50%. Fortunately the fix is the exact same set as last time:

* Increase the SB buffer sizes so the sound doesn't underrun during a standard operation (128 to 512, and 512 to 2048)
* Feed the buffer during an I/O operation

Plus the three bugfixes:
* Don't decode DMA cycles as register writes
* Don't leave the sound buffer filled with previous data
* Don't stop reading I/O data when the CPU is busy

And one performance lift:
* Batch read from the SD card while the CPU copies the previous data (instead of one sector at a time; ~10% throughput increase)

The only difference from the last branch is the name (since its not just targeting PM V1 anymore), and currently, a hint that lets an OS detect if these changes are present in the firmware (to turn off the 8 sector and masking workaround). Patch notes are still in the message on the previous page.

https://github.com/Elendilon/ISA-PicoMEM/tree/pm-pr
This remains entirely AI code.

@FreddyV would love to hear if you are interested or not in the work.

Demo of what this allows on a Picomem V2 inside a 5150:
https://www.youtube.com/watch?v=Aej64YzI_qw

Reply 473 of 481, by FreddyV

User metadata
Rank Oldbie
Rank
Oldbie
Acorns wrote on 2026-10-27, 23:10:
@FreddyV, […]
Show full quote

@FreddyV,

I have a patch for Picomem V1 on the branch at:

https://github.com/Elendilon/ISA-PicoMEM/tree/pm1x-pr

If you are interested. This is entirely AI work - I was getting the Picomem V1 to work (SB, Network, one drive bug and drive speed) with os8088. Picomem V2 didn't need anything, only V1 had a few issues to work through.

https://github.com/Elendilon/os8088/tree/elen … rmware/picomem1 - has the built firmware.

Description of the changes attached.

Hi,

Where can I find the source change ?

Also, why do you want to stop the audio mix during disk access ?
The Goal is to avoid sound sluttering

Reply 474 of 481, by drivera

User metadata
Rank Newbie
Rank
Newbie

Hi, all!

I've currently resurrected my Tandy 1000 SX (with the help of a second unit which provided some important parts for the resurrection!). My old Tandy had a 286 Express card, an XT-IDE HDD with a whopping 30MB (please hold the ooohs and aaahs! 😁), full 640MB RAM, and 2 serial ports. With the second unit I got a second 286 Express card, a second XT-IDE HDD, and an EMS board with some 300 (lost count) DRAM chips (I'm sure it totals to like 96KB ... but still 😀 ).

Now ... as a good retro-enthusiast, I obviously decided to modernize the whole setup. The HDDs were non-starters - neither of them spun up, both had severe stiction (apparently), and were losses. I'm sure their controller cards are fully operational, but I couldn't get the drives to cooperate. Oh well.

Enter @FreddyV's PicoMEM 2 ...

Got the HDD images built (2 x 128MB ... working just fine). Got the memory mapped out, etc. Got a few glitches and inconsistent bootups - i.e. it'd sometimes fail to POST ... power-cycle sometimes fixed it, sometimes the one-finger salute... but overall it works.

Until I plugged in the 286 board ... it is at that point that nobody was happy anymore.

I turned down the 286's caching range to 64KB so I'd be able to reconfigure the PicoMEM's settings. I had configured the PicoMEM for BIOS address D800 so I could assign C000-D7FF as a PM RAM extension (a single block of UMB worth 90KB), and enabled the EMS at port 268 and address E000. The PicoMEM is on DMA 1 and IRQ 7.

All was well and happy ... until the 286 board entered the picture. Currently the machine boots, but I'm unable to run the PCT tool, or load EXPRESS.SYS (even with the SETVER workaround to the "Must be on DOS 2.0 or newer" or whatnot error message).

I can't verify if the 286 is in play. I suspect it is, but I can't *VERIFY* it ... I've also observed some weird behvior in the PicoMEM's configuration screen: I can't turn off FastBOOT, and the Adlib is also stuck as "on".

Perhaps someone has a similar setup and has had more success?

FYI this is my config.sys:

[MENU]
MENUITEM=DEFAULT,Default Settings
MENUITEM=DEBUG,Debugging
MENUDEFAULT=DEFAULT,3

[DEFAULT]
INCLUDE=UMBEMS
INCLUDE=BASE

[DEBUG]
INCLUDE=BASE

[UMBEMS]
DOS=HIGH,UMB
DEVICE=C:\BOOT\STOPMAX.SYS
DEVICE=C:\BOOT\PMEMM.EXE /P:E000 /I:268 /N /Q
DEVICE=C:\BOOT\USE!UMBS.SYS C000-D7FF
DEVICE=C:\BOOT\DOSMAX.EXE /R+ /P- /Q-
SHELL=C:\BOOT\SHELLMAX.COM C:\COMMAND.COM C:\ /E:512 /P
DEVICEHIGH=C:\DOS\RAMDRIVE.SYS 1024 512 64 /A

[BASE]
DEVICEHIGH=C:\DOS\SETVER.EXE
DEVICEHIGH=C:\DOS\ANSI.SYS
FILES=50
BUFFERS=10
STACKS=9,256
LASTDRIVE=Z
NUMLOCK=ON
SWITCHES=/F

And this is my autoexec.bat:

@ECHO OFF
PROMPT $p$g
PATH C:\DOS;C:\XTGOLD;
SUBST T: E:\
SET TEMP=T:\
LH C:\DOS\DOSKEY.COM
LH C:\BOOT\FREESPT.COM CD
C:\BOOT\PMINIT.EXE /K

This is config.txt:

TANDY
BIOS D800
IRQ 7

The ironic thing is that the PicoMEM boot was much, MUCH less predictable/stable without the 286 board in place.

Anyhow, I very much appreciate any information y'all can offer.

Cheers!

Reply 475 of 481, by drivera

User metadata
Rank Newbie
Rank
Newbie

Aaaaaaand it turns out some files were corrupted which is why I couldn't do jack with the damn thing ....

Express.sys now loads quite happily, the PCT executable also does its thing, so ... yeah .... lots ado about nothing.

Sorry for the wasted bandwidth. 😀

Good news is that my Tandy is back from the dead, ~35 years after I last used it ...

Cheers!

Reply 476 of 481, by Acorns

User metadata
Rank Newbie
Rank
Newbie
FreddyV wrote on 2026-11-16, 15:42:
Hi, […]
Show full quote

Hi,

Where can I find the source change ?

Also, why do you want to stop the audio mix during disk access ?
The Goal is to avoid sound sluttering

The source change is on the linked branch:
https://github.com/Elendilon/ISA-PicoMEM/tree/pm-pr
1 commit ahead of your upstream repo. The name changed, as in my later post, because it stopped being only for picomem 1x, sorry!

This change does not *stop* the audio mix during disk access, but rather *keeps it going* (it stops right now during parts of the I/O operation, which causes stuttering).

* Increase the SB buffer sizes so the sound doesn't underrun during a standard operation (128 to 512, and 512 to 2048)
* Feed the buffer during an I/O operation

Plus the three bugfixes:
* Don't decode DMA cycles as register writes
* Don't leave the sound buffer filled with previous data
* Don't stop reading I/O data when the CPU is busy

And one performance lift:
* Batch read from the SD card while the CPU copies the previous data (instead of one sector at a time; ~10% throughput increase)

Reply 477 of 481, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++
Acorns wrote on 2026-11-18, 05:20:

* Don't decode DMA cycles as register writes

hey, isnt that what that bunch of gates are doing in hardware on the picogus?
https://github.com/polpo/picogus/blob/main/hw … S-schematic.pdf in the center under text "AEN IO MASKING"

https://github.com/raszpl/sigrok-disk FM/MFM/RLL decoder
https://github.com/raszpl/FIC-486-GAC-2-Cache-Module (AT&T Globalyst)
https://github.com/raszpl/386RC-16 ram board
https://github.com/raszpl/Zenith_ZBIOS Zenith Z-386 MFM-300 ZBIOS disassembly

Reply 478 of 481, by Acorns

User metadata
Rank Newbie
Rank
Newbie
rasz_pl wrote on Yesterday, 06:45:

hey, isnt that what that bunch of gates are doing in hardware on the picogus?

As I understand it, it is, yes! The PicoGus gates them with hardware, so the Pico itself never sees other device's DMA cycles. As with most things, the PicoMEM does it in software instead. It was already doing this, but there was a "leak". If A10 was used, it would result in the PicoMEM trying to copy data from an invalid address, which then meant static and possibly a system freeze. In this change, A10 is also masked off.

Reply 479 of 481, by FreddyV

User metadata
Rank Oldbie
Rank
Oldbie
rasz_pl wrote on Yesterday, 06:45:
Acorns wrote on 2026-11-18, 05:20:

* Don't decode DMA cycles as register writes

hey, isnt that what that bunch of gates are doing in hardware on the picogus?
https://github.com/polpo/picogus/blob/main/hw … S-schematic.pdf in the center under text "AEN IO MASKING"

Yes and it is present on PicoMEM 2
It simplify coding, but also prevent DMA in the other direction.