VOGONS


Reply 2020 of 2030, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++

>enough spare GPIOs

$1 RP2350B 48 IOs. Its actually more cost effective to slap two RP2350 chips with one 100% dedicated to ISA IO than use 74x logic IO extenders.

>require extra level shifters

You shouldnt need any level shifters to begin with. Picos are 5.5V tolerant just like the level shifters used currently, its placebo now.

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 2021 of 2030, by jmarsh

User metadata
Rank Oldbie
Rank
Oldbie
rasz_pl wrote on 2026-11-17, 23:20:

You shouldnt need any level shifters to begin with. Picos are 5.5V tolerant just like the level shifters used currently, its placebo now.

No it isn't. It's only "5v tolerant" if the chip is fully powered - meaning the bus has to be isolated until the startup sequence is complete = external circuitry is still required. This is no different from the 2040.

Reply 2022 of 2030, by rasz_pl

User metadata
Rank l33t++
Rank
l33t++
jmarsh wrote on 2026-11-18, 00:29:

No it isn't. It's only "5v tolerant" if the chip is fully powered - meaning the bus has to be isolated until the startup sequence is complete = external circuitry is still required. This is no different from the 2040.

14.9.1 (Absolute Maximum Ratings) shows
IOVDD=0 tolerance 3.63V
IOVDD=2.5V tolerance 4.2V
IOVDD=3.3V tolerance 5.5V
With linear reg dropout of 1.1V you are always above that margin. At the moment ISA 5V rail rises to 3.6V into linear reg you are already powering IOVDD with 2.6V giving you 1V margin for rise times.

Of course outputs still need buffers, but thats because of ISA bus loading.

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 2023 of 2030, by crusher

User metadata
Rank Member
Rank
Member
NeoG_ wrote on 2026-11-17, 07:16:
crusher wrote on 2026-11-17, 07:01:

An enhanced PicoGUS with MT32 (MUNT) functionality would just be a dream.
And if it can handle GM/.sf2 soundfonts (via FluidSynth?) as well this would be the ultimate killer card!!!

You can have that now if you add a wavetable-pi to the card's wavetable header

I know. But not having to add a daughterboard would be absolutely brilliant.

Reply 2024 of 2030, by Damiano.78

User metadata
Rank Newbie
Rank
Newbie

Wow, congratulations on your skills! I'm glad that you too would like to have the possibility to switch IRQ and DMA without having to physically put your hand on a new PicoGUS... Gold!

I agree too...the top of the top would be to have all the various MT-32 emulations etc... on board on the PicoGUS Gold, without having to buy an additional daughterboard with another RP on board...

Reply 2025 of 2030, by theelf

User metadata
Rank Oldbie
Rank
Oldbie
theelf wrote on 2026-11-08, 17:53:

Hi, any ideas this can be solved? https://github.com/polpo/picogus/issues/149

it happen all time, and it is becoming a real problem to using my system properly

Hi guys, answer to myself

Like i said before I have this problem with the PicoGUS CD emulation and was very annoying me. I do pgusinit /cdload 51, dir d: ok, then pgusinit /cdload 52, dir d:... and it shows me the files of 51!! pgusinit says the 52 is loaded, no error nothing. If I wait some seconds and try again, then it works.

So first I think is the PicoGUS firmware, but no. I installed the toolkit in a modern PC, compile, but realize firmware loads the new image fine and says the disc changed.

The problem was same with MSCDEX and SHSUCDX, i tested mscdex from dos 6.22 and pcdos 7. But lucky SHSUCDX have source code, so looking there I find the reason and the fix. It only ask the CD driver "hey, the disc changed?" if it pass around 7 seconds (128 ticks) since the last time you touch the drive. If you change the image and go to the drive before that, it gives you the old directory from the cache.

What I modified in shsucdx.nsm:

when you do directory stuff (FindFirst, Open, ChDir, GetAttr, GetSpace) now it always ask the driver for media change. It only drops the cache if the driver says "changed". File reads, FindNext and the MSCDEX calls (games reading raw sectors etc) still use the 7 seconds, so nothing goes slower.
New option /T:n for the timeout in seconds (0-99). Default still around 7 seconds. /T:0 = check every time. You can change it also when SHSUCDX is already loaded, butt at least for now, work fine with default in my 286 now

In the zip you have full source and two builds: SHCDX86.COM (8086/286) and SHSUCDX.COM (386). Compiled with NASM. All credit to Jason Hood for SHSUCDX (https://github.com/adoxa/shsucd), I only touch a little thing.

Also one idea for pgusinit, after a good /cdload or /cdloadname, pgusinit can send an IOCTL "media changed" (INT 2Fh AX=1510h) to the CD drives. SHSUCDX drops the cache immediately when it receive this, so people with normal SHSUCDX no need any patch. With MSCDEX I dont know if it works the same, need to test.

PicoGUS issue: https://github.com/polpo/picogus/issues/149

greetings!

The attachment shsucd-master.zip is no longer available

Reply 2026 of 2030, by NeoG_

User metadata
Rank Oldbie
Rank
Oldbie

I wonder if the behaviour is different in DOS7? I am able to see updated directory contents immediately after changing disc images with MSCDEX so disc change messaging must be working in some capacity

I wonder if there is a TSR or overlay you like to run which could be interfering with IOCTL?

98/DOS Rig: BabyAT AladdinV, K6-2+/550, V3 2000, 128MB PC100, 20GB HDD, 128GB SD2IDE, SB Live!, SB16-SCSI, PicoGUS, WP32 McCake, iNFRA CD, ZIP100
XP Rig: Lian Li PC-10 ATX, Gigabyte X38-DQ6, Core2Duo E6850, ATi HD5870, 2GB DDR2, 2TB HDD, X-Fi XtremeGamer

Reply 2027 of 2030, by theelf

User metadata
Rank Oldbie
Rank
Oldbie
NeoG_ wrote on 2026-11-18, 11:59:

I wonder if the behaviour is different in DOS7? I am able to see updated directory contents immediately after changing disc images with MSCDEX so disc change messaging must be working in some capacity

I wonder if there is a TSR or overlay you like to run which could be interfering with IOCTL?

Hi, about TSR, no, i dont have any, and one of first test was removeee all drivers, like mouse, etc tried different himem.sys neither work

Thats why i decide to spend this weekend on this, because it was VERY nasty problem to me, almost every time i need to change CD, need to do a reset! even lend a laptop with win10 from my brother, installed pi pico sdk etc and compiled firmware thinking the bug can be there

for me is solved, hope if someone have similar problem can help

Reply 2028 of 2030, by theelf

User metadata
Rank Oldbie
Rank
Oldbie
polpo wrote:

...

WOW!!! I want to say THANK YOU for the latest beta firmware with Creative ADPCM support!!!! Thanks a lot!

The only reason I kept my SB16 was this: I have a VOC player in my shell and I love playing my music with no CPU use... jeje

Thanks! Great work!

Reply 2030 of 2030, by polpo

User metadata
Rank Member
Rank
Member

PicoGUS firmware v4.2.0 is now released! https://github.com/polpo/picogus/releases/tag/v4.2.0

Fixes/improvements

GUS mode

GUS emulation has received a ton of fixes under the hood, solving a lot of long-outstanding issues and annoyances. Huge thanks to nukeykt's LPC-GUS gf1.v, reverse engineered from a die shot of a GF1 as a reference.

  • Hocus Pocus now works with the Ultramid version bundled with it. It polls the DRAM peek command at the same time as it DMAs new sound effect samples, which could hang the rp2040-psram library.
  • Fixes sound stuttering in Hocus Pocus and probably other games that stream audio: when the DMA interval is 1, DRQ is now triggered immediately instead of via alarm which saves IRQ round trips.
  • The line-out disable bit now mutes the output. This fixes hanging notes still playing when paused in CapaMOD.
  • The GF1's voice IRQ latch and its timing is emulated. This fixes music occasionally stopping in the demo Dope (and perhaps others).
  • Volume ramps are now applied once per GF1 frame, which is more accurate and more efficient. A stopped ramp now retriggers IRQs and a finished non-looping ramp stops one increment short of its end value.
  • Pan attenuation now uses the GF1's ROM values from gf1.v. DOSBox's volume implementation is also fixed and matches GF1.
  • Voices playing back faster than the PSRAM sample cache simply bypass it. Fixes crackles/stutters in very fast playing samples.
  • GF1 timer accuracy: writes to 2X9 are routed to the timers, stopping a timer immediately cancels its IRQ, and the 5us fudge on timer periods is removed.
  • DMA fixes: the DMA rate divider is now in the correct order, DMA address wraps at 1MB, the address advances past the terminal count byte, and unsupported GUS->host DMA requests are ended early instead of getting the ISA DMA state machine stuck.
  • Only active voices set the wave/ramp IRQ status bits.
  • Ramp registers can now be read back, and register writes during reset are dropped.
  • Register 0x8E reads back the active channel count as written, even if we clamp to 14 channels internally.
  • The IRQ line is deasserted when mix control disables the IRQ latches.
  • GUS resets no longer clear the DMA interval override set by pgusinit.
  • Fixes in inter-core synchronization.
  • GUS RAM is cleared at power on. This works around a bug in Star Control II where it plays garbage after the first sample in GUS RAM.

MPU-401

  • UART mode and reset now behave like a real MPU-401. Bytes go straight to the MIDI output, DRR reads busy for about 100us after each data write, and a UART mode reset delays about 33ms while the output drains. This fixes Frederik Pohl's Gateway and its buggy driver with no need for workarounds.
  • Sequencer ticks keep counting while a data request is outstanding. This fixes the too-slow tempo in Gateway and probably other Legend Entertainment titles.
  • When the MIDI buffer is full, intelligent mode track output is deferred instead of blocking the ISA bus handler.
  • New pgusinit /mpuwait 1|0 option. Some games like A-10 Tank Killer give up after only 256 busy status reads. Now that PicoGUS acts even more like a real MPU-401, those titles time out on fast PCs just like with a real card. When mpuwait is enabled, each busy status read raises IOCHRDY an extra 2us to give the MIDI buffer extra time to drain.

SB mode

  • Fixes SB DMA transfers getting out of phase, which caused flipped stereo on SBPro or garbage audio on SB16.
  • When in SB Pro mode (pgusinit /sbmode 4), PicoGUS now properly emulates the SB Pro's flipped stereo. This means that titles that take the flip into account now have proper stereo. A pgusinit /sbswapstereo option can "un-flip" SB Pro stereo for titles that don't take the flip into account!
  • Blinkenlights on SB DSP activity, once per DMA block, and blinkenlights on OPL music is back.
  • Brings in improvements to dbopl3 from HitoriSensei's PicoMemBlaster fork of PicoMEM.

CD-ROM

  • Blinkenlights on every sector transfer.

USB

  • Basic (hacky) support for the Logitech G920 wheel as a joystick. A full overhaul of joystick support is coming...

ISA analyzer

  • Captured ISA bus transactions are now timestamped.

General/pgusinit

  • Overhauled firmware flashing to be more resilient.
  • Much better error messaging if the PicoGUS settings don't agree with your ULTRASND or BLASTER environment variable.
  • Changing settings can be combined with /mode changes in pgusinit. They'll be applied after the card changes modes.

creator of PicoGUS and PicoIDE