Indivision AGA Mk3 HDTV Audio

Hi All,

I've been experimenting with the new audio configuration in the latest release of the Indivision Tool and I found some things which I thought might be helpful to share. Apologies once more for the wall of text, and again none of this is a complaint. I think you will already know most of this but hopefully some of it might be useful info to others or to help for ongoing investigations.

At present the audio works great for me when configured at the 96kHz bitrate on the 4 HDTV's I've tested but at that bitrate only. The sound is not quite identical to the original analogue out but it is very good and very clean as you'd expect from digital – I don’t mind the filtering personally but I know that’s a subjective thing. I think I have noticed a little lag when I run the digital and analogue sound side-by-side but not so much that anything seems out of sync with what's happening on screen.

Sound seem to work fine for me in the Workbench, and in ProTracker, OctaMED and EaglePlayer, regardless of format (I think that’s because EaglePlayer processes all audio through the same pipeline before sending it on to Paula).

I've tried a couple of hundred or so games over the last few days (not exhaustively), and the audio seems to work great for a heck of a lot of them, the list below is the exceptions I found...

As a few other folks have noted, audio does not seem to work for me in any of the Turrican games after the initial Rainbow Arts/Factor 5 splash screen. I get no audio at all in B.C.Kid, , X-Out and Z-Out. Audio in R-Type and Denaris/Katakis seems fine.

Sound doesn’t seem to work for me in Rainbow Islands, Uridium 2, Fire and Ice, or Virocop. I think all were written by Andrew Braybrook/Graftgold so might all have the similar audio routines?

No audio in Xenon, Xenon 2, Cadaver, or Speedball 2 but Speedball, Gods, Magic Pockets, and Chaos Engine all seem fine and have full sound as far as I can tell.

No audio in Shadow of the Beast 1, Beast 3, Brian The Lion, or Puggsy. Audio seems fine in Beast 2, Agony, Awesome, Killing Game Show, Leander (maybe a missing sound effect in title music?), Lemmings, Lemmings 2, Obitus, Ork.

Audio seems to work for me in Rick Dangerous and Switchblade 2 but not in Rick Dangerous 2 or Switchblade.

From what I read in some of the other posts, I think you're already working something to improve compatability which I think will probrably fix most of the above. The next games I found are a bit different in that they have audio but it doesn't seem complete so I guess these might be a slightly different issue...

Dune has audio but only channels 1 and 2 seem to be output, channels 3 and 4 are missing – I played the Dune intro MOD in ProTracker to compare and it’s definitely channels 3 and 4 which are missing, so one from each stereo side.

No audio in Monkey Island 1 (TFMX like Turrican etc.?) but Monkey Island 2 and Indy Jones and The Fate of Atlantis have audio but are again missing notes/sounds – again it sounds to me like they're missing channels but I can’t be sure.

Same with Another World: About half of the audio is missing during the intro, no drums, most sfx missing – I don’t have the mod file for this one so I can’t be certain if it’s whole channels bit it is missing part of the audio for sure. Flashback seems fine, as do Future Wars, and Operation Stealth.

Frontier is another game with partial audio, you can hear it during the intro. Again this sounds like it’s missing a channel on each side.

New Zealand Story’s audio works on the title screen and intro but cuts out when starting a game after the walrus kidnaps the kiwis.

Rocket Ranger also seems to be missing some notes/channels.

The HDTV displays are now very occasionally (maybe after 1-2 hours or so) losing video sync and blinking out, which is something I've not seen since I swapped out my bad HDMI cable. I haven't changed my setup at all since the HDMI cable swap, I’ve only updated the Indivision firmware and Tool, so I think it might be something in the new firmware causing, it may even be the audio and my constantly messing with the settings to try things causing it – it only happens occasionally, once it starts the screen blinking out happens every few seconds for 5-6 times, then subsides but comes back again after a few minutes. Power cycling the Amiga also seems to stop it. I need to investigate this a little more as it hasn't happened since I unplugged the USB network adapter from my RapidRoad earlier today.

I've also noticed a little more pixel noise/errors on the Workbench since the latest update, not at all as bad as it was with the initial firmware but it seems to happen more frequently than it did in 1.4. I think it’s bad reads of the pixels on the Amiga side as they’re too big to be just moise on the output display. They seem more likely to happen if I launch something like DOpus4 which makes a clone of the Workbench screen, and they definitely occur more if I am copying something from the pcmcia slot (using a cf card and adapter).

At the moment I'm running the Workbench in DBLPAL progressive, 64 colours, and the Mk3 is set to scan that at 640x576 50Hz in SHiRes, and scale up to 1920x1080p50. Activating the CCKLine pull up seems to lessen the errors and gets rid of them most of the time, but not all of them, all of the time.

The slight screen zoom when in non-DVI mode, which moves both the Workbench border and the Indivision overlay partly off the screen still appears to be present. It’s not just a translation, it’s definitely being scaled slightly – I don’t know if the HDTVs are doing it due to there being less lines in the picture, or if it’s the Indivision scaling things differently in non-DVI mode. Also, the VSync ripple which happens every 10-12 seconds is still there, in both DVI and non-DVI mode – so something odd is going on with that. It might well be the TVs doing it and not the Indivision but the same thing does not happen when I plug in a Raspberry Pi or my PC’s graphics card into the same HDTV and set those to 50Hz modes. The PC graphics card might be interpolating the frames and smoothing things out but there's no way the Raspberry Pis are doing that, they're not powerful enough. If I try to activate VSync or Auto-Resolution (or both), I lose the picture completely and all TVs report an unsupported resolution.

Additionally, I've noticed that when I cycle through the audio/DVI modes on the Advanced options modal window in quick succession, the Indivision Tool sometimes seems to get itself into a state where it thinks it has already saved/synced changes I’ve made, when it has not. I have to quit out of it and redo them when this happens.

With all that said, I am really enjoying the great digital picture quality and the experimental audio too, so and thank you for all your hard work and efforts. I just hope some of the above is useful information.

Cheers!

Hiya, thanks for replying, sorry it is a Sunday, please feel free to ignore my nonsense until at least it's a work day, I've only just had time to try this stuff out. I'll try and keep this message a lot shorter!

The HDTVs are all in game mode, and also have as much of their additional extra-processing nonsense switched off as is possible. The audio lag isn't bad at all, and it was the analogue out on the Amiga I was comparing to, so I am not at all worried or surprised that there is a little lag since the audio is getting to the HDTVs via a more complicated route.

The audio in the Turrican games etc. is missing entirely, not just some channels or notes, but the Dune, Frontier, Another World and the non-TFMX LucasArts game audio seemed to maybe be a different issue. It wasn't like it was missing a note or sound effect here or there, it was like channels 3 and 4 were switched off completely, or had the volume turned down to zero but channels 1 and 2 were working fine. This could absolutely be the same thing as the Factor 5 games, it just seemed like different behaviour (I'd expect the audio replay code to send all channels using the same mechanism, so it'd either all work or all fail), so I thought I'd mention it. Fingers crossed it is the same thing though, as then one fix will solve all of them at once.

Whatever is causing the bad pixels, it has only previously seemed to be affected in a positive way when I switched on the CCKLine Pull-up, when I switch that off and try the PCLK in Pull-up mode instead, it doesn't seem to improve or clear the issue. The issue can manifest as just a single pixel but it is usually flickering on part of a line, like 8 or so bytes on a bitplane are being misread, or an entire line (which mostly happens on the DOpus clone of the Workbench screen if that is open. I've also seen part of some icons flicker and change colour, and this isn't line based it's like the palette entry for the particular colour has gone a bit crazy. The CCKLine pull-up definitely improves it whatever it is. Thinking about it, might it be palette and not pixel related? I never see it in a game or at low-res, only on the Workbench/DOpus screens which are set at DBLPAL no flicker at present.

The Lisa IC just says CBM on it, so I'm not sure what version it is but I don't think it's an HP one. I'll attach a picture of the mobo so you can take a look - it's an older picture, before the board was recapped but it is the same board. I guess PCLK is Pixel Clock, and CCLK CPU clock? If it is the CPU clock causing issues, might I need to perform the accelerator fixes to the bottom of the board as are required for some expansion cards to be stable? These have not been done as far as I can tell.

I tried rescanning the EDID modes but I'm still getting just the same three entries coming back as I did previously. Those are...
EDID 1920x1080@60

EDID 1280x720@60

EDID1280x1024@60

I do believe it must be the TVs not doing something correctly but I'm at a bit of a loss as to why, as they don't seem to do it for the other devices I've tried. When I instruct the RasPis or my PC to go into 50Hz mode they seem to do so and to sync up fully without screen tearing. Maybe the TVs are still doing something to compensate for those devices but they shouldn't be as all that stuff is manually disabled in the menus. The scaling in non-DVI mode I am totally at a loss to fix via the TVs menus, there's nothing at all additional that I can see which can be disabled, and the scaling is not present when in DVI mode. Are there still some additional display modes which are not yet listed by the Indivision tool, which might solve this? If I hard config a RaspPi to change it to a 50Hz mode, there are a lot more display modes which the TVs seem happy to switch into but which aren't on the available list in the Indivision Tool yet. Might the RaspPi be looking/using the CEA modes and the Individision looking/using the DMT modes? I don't know a lot about how that works, so please forgive me if I'm talking nonsense.

In case it helps with the cycling of the DVI/audio mode I am seeing, my current accellerator is an ACA1221lc, running (happily) at 40MHz. The cycle order is Pure DVI, no audio -> 48kHz ->44.1kHz -> 96kHz -> 88.2kHz, but I kept expecting the 96kHz to be last and accidentally skipping over it and then having to go around again, which is why I was clicking on it in quick succession... "user error" as Apple would say.

Hope some of this helps, please let me know if I can be of use with further investigations or information.

After some more experimentation, here are some pictures I managed to capture of the on-screen flickering in case it helps anyone else.

So on this occasion today, turning off the CCKLine pull up cleared the interference, which is the exact opposite of what normally happens. It didn’t seem to matter whether the PCLK was off, set to pull-up or pull-down... There is no pattern to this I can see other than switch on the Amiga, and when the flickering starts, toggle CCKLine mode from whatever it was to the opposite, and the problem goes away. Is it likely that my Amiga 1200 is just a steaming pile of 30 year old failure? I honestly don't know how you guys mange to design such reliable (and your stuff is very reliable) expansions for these temperamental $*%& machines! I have just one and it’s driving me nuts.

Hi,

In case it's useful for comparison purposes, I plugged a Pi4 in to capture the EDID block from one of the HDTVs, and got it to report what it thinks the HDTVs capabilities were, and what mode it thought it was in. I also put the EDID capture through parse-edid and edid-decode on my PC. I've attached the captured EDID data block and the parse logs from the Pi and PC so you can see the details.

The information from the Pi seems to indicate that I've got my manual 50Hz settings correct, as they're identical to those the Pi is saying it is outputting. In theory I think the VSync/Auto-resolution/both should work but for some reason they don't seem to. The EDID info also indicates that the HDTV can handle a wider set of audio bitrates than seems to work from the Indivision too - I'm not sure if the command is giving the audio bitrate per channel or for the audio stream as a whole but it seems to suggest 44.1kHz or 88.2kHz streams should work and they don't, only the 96kHz setting produces any actual sound from the HDTV. It also says something about supporting over and under scan and I'm wondering if this could be related to the scaling I'm seeing in non-DVI mode maybe? I don't seem to have a way of changing the setting to see if it produces the same effect I see on the Indivision in non-DVI mode. I can't quite work out why things are working differently for the Indivision compared to the Pi apart from that as the settings all looks the same.

Hopefully this is not completely useless info but apologies if it is. The EDID parse commands on my PC did give some proper front porch, sync, back porch numbers that I can enter into the Indivision and work nicely at 50Hz though, so quite a useful exercise!

No problem, thank you for looking. The problem was intermitent and has not happened for the last day or so - I will try to give better details to recreate it if it happens again. Maybe it was related to a setting leftover in flash from the previous firmware version or something.

Evening All,

I had time to investigate a few things further this week - I've spent the last few days trying to stabilise my A1200 with a new ACA1233n in place of my ACA1221lc, and have discovered a couple of things along the way which might be of use to others.

The first is related to the to the Mk3 audio....

When I was testing the games last week, I forgot to mention I was loading them from HDD via WHDLoad. The games I found to have partial audio with only 2 working channel, such as Dune, Another World, Frontier, Indy Jones 4 etc. all still behave the same when launched from the HDD, regardless of which accelerator is installed or if the MMU is active on the ACA1233n, as you'd expect. However, if they are launched from floppy/gotek, the audio works fully and properly. I know this is not a massive revelation but it is curious as I doubt the games will have had their audio routines re-written massively to run from HDD, so maybe they are just running from a different place in memory? I know this is nothing concrete but I thought maybe it might help with any investigations.

Next, the new accelerator card (or something I did when I fitted it) appears to have changed the behaviour of the previously mentioned on-screen multi-pixel and line corruption, and also the flickering palette colours – all have disappeared which is great but… now I have a glitching mouse pointer – and this is a really weird one because it’s not just flickering or corrupting graphically – it’s actually jumpy across the screen – always to the left, and if I am clicking an icon when it happens it drops the icon on the left of the screen too when the mouse moves, as if the OS thinks the mouse pointer is travelling 600+ pixels in a single step.

This new glitching only happens during the first few minutes after the Amiga is switched on from a cold boot, and it only seems to happen when I move the mouse. It will jump across the screen to the far left and then return to where it was, as I move the mouse pixel by pixel – it’s not on a timer, and doesn’t snap back if I keep the mouse still and wait. It’s as if the move causes the OS to read the mouse position and it gets zeroed by a bad read and then I move it again and it reads good and returns to where it was. Has anyone seen that behaviour before?

I ran the Amiga Test Kit's memory scan a few times (it takes a while with ~130MB of RAM), but it didn't find anything so far. I also totally flattened the HDD and reinstalled OS 3.1.4, ACATool, the MMULib etc. from scratch yesterday thinking it had to be software related somehow but it still happened again this morning for about the first 5 minutes after I switched the Amiga on.

The machine also locked up about 5-10 minutes into use this morning, and then failed to boot after that when power cycled. I turned it off, counted to 15, turned it back on, the ca-psu LED was green, the Indivision logo popped up on-screen, the Amiga power LED lit up, and the floppy emulator’s LED screen came on but no drive click, no HDD/FDD activity LED, just a black screen on both the digital and the analogue video out. I tried again, and the same thing occurred. On the third attempt the screen came up yellow (CPU error?), and I was like - woah, so I left it for 1-2 minutes... and then, on the fourth attempt it booted normally and has been fine all day since, including several power cycles, with no on-screen glitching, or software crashes, nothing bad at all, it has worked perfectly all day since - and I've been hammering it, re-installing a whole bunch of WHDLoad LHA archives, playing games and listening to music.

Something similar happened yesterday also, so I tried re-seating the accelerator card and the Indivision before I tried rebuilding the software. I will try again in the morning, once the Amiga has had a nice night’s rest and see if the pattern repeats. I will also try cleaning the edge connector again tomorrow, and will check if those pesky E123C and E125C are still in place underneath (and remove them if they are) – as I didn’t get chance to check that in the week, I was too keen to try out the new accelerator – which is awesome once I can get the darned Amiga to switch on!

Any ideas what else I might try, or if the E123C E1235C fix might help? I don’t think the problem is mechanical, like a loose connector or anything, I think it must be either thermal and/or electrical but I’m obviously not an expert like you folks are, and I haven’t seen enough wobbly Amigas to know if this is common. Any advice or ideas you might have would be very welcome, your educated guess will be far more likely to work than my attempts.

Thanks very much in advance!

Thanks for replying and for the advice - so far today the Amiga has been stable apart from some minor gfx glitches. I think I need more data to be at all confident about what's going on, so I'll keep things as they are for the next few days and just monitor, unless something horrible starts to happen again.

I've just realised that I did change additional things yesterday, after the boot failures happened. I tried removing MAPROM and VBRON from the acatool startup script, and re-ran the MMUlib installer which put MuFastRom and MuFastZero into the User-Startup - I'm wondering if I might have had both the ACATool and the MMULib both active previously, if that might cause the lack of stability? I understand conceptually how MMUs work but I'm not really sure how best to setup the software for one on an Amiga as it's not something I've had to do before. I know MMULib is not your software but do you know if any sort of idiot-proof guide exists to set it up along with an ACA1233n, sort of an "out of the box"?

As the Amiga has been stable all day, I've been trying out more games on floppy versus HDD to see if I could find anything new with the experimental audio feature - There's definitely improvements to audio when some games are run from floppy instead of via WHDLoad. I do think there's a significant difference, though I don't know what it is. Not all games have better audio when loaded form floppy but I've not yet found any that are worse.

In each case where the audio is improved but not perfect, there do not seem to be any individual missing notes, or parts of a sample. The audio runs continuously until a boundary event, such as a fade out between levels, or a title screen and in the games where the audio returns after a fail, it almost always does so (apart from in Megalomania) at another boundary such as the next level or the game over screen. It made me wonder if it was a specifc type of event being missed like a global volume change or similar.

It's a bit beyond my skills to trace exactly what is going on but if someone who's good with a debug monitor and/or an oscillascope takes a look to see what's different between the two versions, might there be a single fix possible for a lot of titles?

Anyhow, I hope some of this is useful to someone, and thanks again for all your help.

Cheers and have a great weekend!

I've managed to recreate the yellow screen with the help of the scummVM 030 1.5.0 from Aminet and the Discworld 2 game, which bombs out scummVM and the Amiga when I try to launch the game. The error is intermittent but repeatable. Is it possible it's just a fail in the MMULib? I haven't so far seen a RAM error when I run the memory scan in Amiga Test Kit but I am leaving it runnning for a while again to try and check.

No worries at all, I can imagine how many hours you guys must work each week maintaining and developing all of the products. I'm just grateful for the genuinely excellent support you guys provide.

I noticed something today which might (or might not) be useful. It is related to the partial audio I get via the digital output of the Indivision AGA Mk3 in some WHDLoad games. Depending on how I launch the games with partial audio, I can change whether I get full or partial sound.

Tiny Bobble is a good example (but there are others)...

When launched by booting from floppy at reset/power on, Tiny Bobble has full sound on all channels.

When launched from Workbench from the same floppy, again Tiny Bobble has full sound on all channels.

When launched from Workbench via WHDLoad install via WHDLoad slave, Tiny Bobble has sound on only two channels of audio.

...but when launched from Workbench directly from the WHDLoad install folder but without WHDLoad, Tiny Bobble has full sound on all channels.

So the only combination when Tiny Bobble is missing audio is when launched via WHDLoad, and I've found this to be the same for quite a few WHDLoad games as listed previously in this thread. Now at least I am fairly sure it's something WHDLoad is doing which is directly causing this and not a patched/hacked game file, or some library loaded by Workbench itself.

Other games which exhibit the same improved behavior include Indy Jones and the Fate of Atlantis, Dune, Frontier, Another World, Monkey Island 2... Basically anything that has partial audio but which you can launch directly from the WHDLoad data directory inside each install, rather than via its WHDLoad slave file, seems to have fully working audio when not launched using WHDLoad itself.

I haven't yet figured out exactly what WHDLoad is doing to confuse the Indi AGA Mk3 and make two of the audio channels go silent but I thought maybe if we users can figure that out, we might get a work around going at least for now, as there are a lot of WHDLoad games which use disk images instead of files and so can't be launched without WHDLoad. As a bonus we might also find something which makes it easier to figure out a future improvement for the Mk3 along the way.

Please may some other user/s check this when they have the time, and see if they get the same behavior? I've found a few things lately which only seem to happen on my Amiga, so I want to be sure it isn't another of those (or if it is, work out what's causing and fix my software!).

Cheers!