does sound using something sound blaster compatible in dma mode work?

rasz_pl wrote on 2023-03-02, 17:17:

does sound using something sound blaster compatible in dma mode work?

This made me think. I have replaced the Aztech MM (ISA) I had in there with a SB Vibra 16 (ISA) and installed their official drivers. Since then I also get repeated errors when trying to read from CD. The errors are something like "Wrong CD format""Not an ISO" or similar. I tried loading the vide-cdd driver before the SB-Drivers which helps in some occasions. But only unloading everything SB16 and then even removing the card fixes Floppy and CD read errors consistently for me. I then configured the SB16 to use the same IRQ/DMA (I5 D1) settings the Aztech used before which helped a little but not always. Before that it was I10 and D7 which basically made the system unusable...see above. Also for example Beneath a Steel Sky runs fine (if I get the CD-Rom to load before the SB - or just a coincidence) as long as I do not try to use speech. When I have Speech selected and "talk" to somebody in the game it fails with

1Fatal Sustem Error
2cs:ein = 2b:8975
3Sorry!

I didn't find out what this hex code means but I suspect/hope it has to do with the SB interfering with something.

Funny thing I have the exact same SB Model (4170) in my Baby AT SSocket 7 k6-2+ PC where it does not create any problems and games run fine (even BaSS) but here I think Win98 is smart enough to sort out the IRQs/DMAs. Since the sound and even midi sound so much better with this SB I replaced the Aztech in my 486 build, too. Dang.
The card is a PnP card with soft config btw and does not feature a High DMA setting (not needed according to Creative) even though they configure one on installation.

I will try and configure the SB with the sbbase driver but I don't think this will fix the problems? We'll see.

Sound card is not the source of your problems. Cdrom is irrelevant/symptom of another problem.
"talk" in Beneath a Steel Sky is DMA sound, and that crashing makes me think you lost ISA DMA controller somehow.
looking at https://theretroweb.com/motherboards/s/fic-486-pio2 VT82C496G chip has build in 82C206, and thats the DMA controller. That is a shitload of dense PQFP chips that like to lose connections with time.

https://theretroweb.com/chip/documentation/vt … 56711526343.pdf page 28 has pins responsible for ISA bus. You will need to get a multimeter and cross reference those pins with ISA slot pins checking if every single one has solid connection with good conductivity and if all chipset pins are soldered well and dont move (gently poking with needle works)

Alright will have a look into that. Just for reference: Background sound FX and Music works in BaSS. Also I initially had an ISA GPU on the board but it always helps to check again. Thanks. Will report back.

To ask a stoopid question ... the PIN number I suspect is the pin on the VT82C496G? But how do I find the corresponding pin on the 82C406? At least the only charts I found when looking up that chip are the those you've already sent me. What I did by now: I checked every solder joint of the chipset and the 406 by hand and also used a magnifying glass I use for checking my solder joints and every pin is connected. None is corroded. Then I went ahead with my multimeter and checked the connection between the 406 and the ISA ports. At least for every ISA-pin I get a beep for 1 to multiple pins on the 406. None of the "multiple" pins appear to be shorted at least.

406 is keyboard controller rtc and some other devices. isa logic is in 82C496

>Background sound FX and Music works in BaSS

you mean the door in the first scene? thats adlib sound effect, doesnt count https://www.dosdays.co.uk/topics/Games/game_bass.php not so sure anymore about the door being adlib sound effect

Alright. I did the debugging and at least learned something new. But unfortunately (or fortunately I don't know) it does not seem to be the ISA DMA.

Here is the "protocol" of my debugging session:

. I leave it here in case somebody else need's a quick way of testing.
I tested every pin from the data sheet back and forth. I found a short between A11 and D17 and fixed it but unfortunately it didn't change anything 🙁 ... An easier way other than starting BaSS is just using CTCU to Test my config and the whole system freezes when it tries to start the DMA.

I will search the internet a bit more and hope I find some inspiration *sigh*

I tried the a different driver. Namely the Vibra 16 Drivers from Vogons. Instead of the ones I found on the internet. BaSS is still crashing but this driver's diagnose tool finds the DMAs and even lets me test 8bit, 16bit and FM audio. Unsure here but should that mean that DMAs work and it might be something else entirely? Just hoping 😉
I also tried Larry 6 and here speech works as well. hmm

Can you post a picture of the motherboard?
I'd like to check the CPU jumpers, perhaps one is wrong or missing.

marbury wrote on 2023-03-03, 22:22:

I found a short between A11 and D17 and fixed it but unfortunately it didn't change anything 🙁

A11 AEN | D17 /MASTER ? That would definitely explain crashed during attempted dma. How was that shorted? maybe short caused permanent damage to the chipset.

rasz_pl wrote on 2023-03-04, 08:53:

A11 AEN | D17 /MASTER ? That would definitely explain crashed during attempted dma. How was that shorted? maybe short caused permanent damage to the chipset.

It was some trace of solder. Never soldered there but seen solder trace weird paths on really old boards. But TBH I am not so sure anymore that it’s broken. I am using a different driver now and that detects high and low DMA‘s fine and I can run 8 bit,16bit and FM sound tests fine. I repeatedly ran the Diagnose program to make sure it’s not just a fluke. I also tried some other games by now and they all produce speech and sound. Fixing the short was good. But I shouldn’t just have focused on just BaSS I think. But doing the debugging session was a nice exercise nonetheless 😉
Regarding BaSS I will Fokus on autoexec.bat and config.sys things now and see what's going on there.

Disruptor wrote on 2023-03-04, 07:05:

Can you post a picture of the motherboard?
I'd like to check the CPU jumpers, perhaps one is wrong or missing.

Of course. Here it is

I am currently running a 5x86 P75 overclocked to p75+@150MHz. But as I have stated earlier it does not matter if I have a 486dx4100 in there or run the 5x86 correctly clocked. What makes a difference though is JC1 and JC5. In the Screenshot JC1 is currently open which would make for the correct WB config the 5x86 (or my &ew dx4) needs imho. But only WT works correctly with that board. But maybe you can spot something off. especially in range of the JC1-5 Jumpers. RAM seems to be fine as well. It was the first thing I suspected and a multi-pass memtest for a whole day without errors. Funny thing is that everything started (maybe a coincidence) when I added the SB16 Vibra. Before that I had a Aztech MM Pro which never made any problems. Unfortunately I have already sold it ... dang. The SB16 is currently running in H240, I7, DMAs 0,0 which might create problems with some games anyway. Would really like to have it in DMA 3 or 5 but the Diagnose tool doesn't let me set it. Maybe I should throw this out and get an aztech again. Just wasted too much time already 🙁

marbury wrote on 2023-03-04, 18:59:

It was some trace of solder. Never soldered there but seen solder trace weird paths on really old boards.

Could have been tin whiskers, grow on old solder on its own. If thats it then there might be more hiding somewhere inaccessible to naked eye.

marbury wrote on 2023-03-04, 19:05:

I am currently running a 5x86 P75 overclocked to p75+@150MHz. But as I have stated earlier it does not matter if I have a 486dx4100 in there or run the 5x86 correctly clocked. What makes a difference though is JC1 and JC5. In the Screenshot JC1 is currently open which would make for the correct WB config the 5x86 (or my &ew dx4) needs imho. But only WT works correctly with that board. But maybe you can spot something off. especially in range of the JC1-5 Jumpers.

According to the manual, you are supposed to set JC5 only for a Pentium overdrive processor. Both the &EW DX4 and the 5x86 are supposed to be jumpered according to "P24D" settings (except for JC2 which should be set to 2-3 for 33*4, but is correctly set at 1-2 for 50*3). Your photo shows JC5 installed, which is surprising. While JC1 is used to enable/disable WB cache, the magic that makes it really work is RNA/RNC/RNI. And that's where thinks get strange. P24D, AMD Enhanced DX4, 5x86 and M1SC are supposed to be compatible in this regard, but the FIC manual suggest to left-align RNI for the P24D, but right-align it for AMD Enhanced DX4/M1SC. You might try right-aligned RNI if left-aligned RNI makes trouble with L1WB.

rasz_pl wrote on 2023-03-04, 19:21:

If thats it then there might be more hiding somewhere inaccessible to naked eye.

Pretty sure that's that. Know it from old PCBs. I checked all SMT's with some magnifying glass and were needed scraped the whiskers off with a thin needle. I mean you never know but at least the ones I could see are gone now. So far no problems here.

mkarcher wrote on 2023-03-04, 19:45:

According to the manual, you are supposed to set JC5 only for a Pentium overdrive processor. Both the &EW DX4 and the 5x86 are supposed to be jumpered according to "P24D" settings (except for JC2 which should be set to 2-3 for 33*4, but is correctly set at 1-2 for 50*3). Your photo shows JC5 installed, which is surprising. While JC1 is used to enable/disable WB cache, the magic that makes it really work is RNA/RNC/RNI. And that's where thinks get strange. P24D, AMD Enhanced DX4, 5x86 and M1SC are supposed to be compatible in this regard, but the FIC manual suggest to left-align RNI for the P24D, but right-align it for AMD Enhanced DX4/M1SC. You might try right-aligned RNI if left-aligned RNI makes trouble with L1WB.

Yeah the manual is a bit lacking here. But even with triple checking I was never able to switch on WB with my 3x DX4 &ew and left it off (maybe the DX4 specifically has a borked L1 cache...dunno). Regarding the 5x86 there are like 3 working configs to have it being recognised correctly. But I never tried re-setting the RNI because I couldn't find any good explanation for what this really does. Or no, I set it to right-aligned once but didn't notice anything different so I set it back. I have it in P24D WB setting with the RNI right-aligned now. So far no crashes. With RNI Left-aligned and jc 2,3,4 set to 1-2 the following would happen: JC1 open: no DMA config/crash on DMA discovery - and intermittent crashes for seemingly no reason. But in hindsight it was always something with audio. JC5 open: insta-crash on CD or floppy access. Before upgrading the BIOS either JC1 or JC5 open and the DX4 installed would cause the PC to stop/hang at BIOS-to-CPU handover. Could just use my MMX or k6-2+ to play everything but nerd-pride dictates to get this little sucker running as smooth as possible.

And as an anecdote: I have re-ordered an Aztech MM PRO ABI which was in there before the SB16 and just worked with the common settings (Midi: 330, Audio: 220, I5, L1, H5). The Vibra is not willing to set to anything other than M:300, I7, L/H:0 which makes basically all Lucas arts games mute in that case. Should never have sold the Aztech. But since this FIC build resided in the basement for a couple of month and my other builds work so well with the Vibra I decided I couldn't go wrong and I never could afford a real SB as a 90s teen so there you have it. Shouldn't be so quick to sell things. Oh well.

Alright, this is finally fixed.

I debugged sky.exe and first checked what is at 2b:8975 (the address from this post above Re: 486 crashes when accessing a file on a floppy disk). It didn't yield anything. Except for some gibberish random bytes of ascii
Then I cleaned out my config.sys for good and removed every line one by one to no avail. Removing EMM386 would instantly crash the game while removing himem.sys did nothing as well except not loading any EMS and XMS. But the error stayed the same. So I changed everything back to what it was before.

In the meantime the re-ordered Aztech MM Pro16 ABI (2316-S) arrived and after replacing the Vibra for the Az and installing the drivers nothing changed except I got a different error message when the system tries to play speech

12b:ac91

I ran that through the debugger as well and it reported (paraphrased) "deinitialising extended memory" which got me thinking in terms of me loading EMS/XMS but the game bringing its own memory management. And because I'd already tried with disabling himem/emm386 which did nothing so I figured BaSS would just not care. So I commented out himem.sys and emm386 again, reboot and voila!
IT FREAKING WORKS ... Music, FX, Speech \o/. Now I can even start Beneath a Steel Sky from the cd without even installing or configuring and still have all sound.

OMG I am so happy I could finally solve this stupid mess. So thanks guys for the hints and the learnings 😀

I almost forgot about floppy, does it work now? 😀

yeah it does. I think it was JC1 or JC5. Can't really tell anymore because trying to fix the other error I changed config so much that at some point it worked when I needed it 😁