|
GTIClub/K056800/RF5C400 Sound Fix
#303215 - 01/25/13 09:26 AM
|
|
|
I've been playing around with the GTIClub driver and while it's almost there (before I looked at it), there are some problems.
I'm not going to pretend I have a clue what I'm doing with the graphics but I have stuck a hack in there to remove the floating FBI warning in the sky. Just reduced the view port of the roz layer 1 to 32px and started the tile mapper for roz 1 looking over an extra 32px. This probably breaks other games however.
The sound is where I think I can actually do something submittable.
First of all the crystal value of the PCM chip was wrong. It should be 16.9mhz, not 16. This fixes the pitch (for all GTIClub based games it seems) but the 056800 was sequencing the music too slowly. It sounded wrong compared to the PCB anyway but you can tell as the announcer's "GTI CLUB!" exclamation was arriving too early. I changed the sound rate in the 056800's driver which works great for GTI Club but this made Thrill Drive's music too fast. Can't be sure about other games as I have no reference.
So any ideas from here?
Edited by RetroRepair (01/27/13 03:34 PM)
|
http://www.youtube.com/retrorepair
|
|
| Embryoice |
|
I am the Master Warrior!
|
|
|
|
|
|
Reged: 12/30/12
|
|
Posts: 74
|
|
Loc: Los Angeles, CA
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303234 - 01/25/13 06:59 PM
|
|
|
Also the title screen not shown behind the 3d graphics due to the layer priority issues that the model 3 board does.
|
|
|
|
Re: GTIClub
[Re: Embryoice]
#303235 - 01/25/13 07:05 PM
|
|
|
> Also the title screen not shown behind the 3d graphics due to the layer priority
> issues that the model 3 board does.
Yes I believe this is also the real reason for the FBI warning in the sky. Is this a known issue with the PPC core then?
|
http://www.youtube.com/retrorepair
|
|
| Embryoice |
|
I am the Master Warrior!
|
|
|
|
|
|
Reged: 12/30/12
|
|
Posts: 74
|
|
Loc: Los Angeles, CA
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303240 - 01/25/13 08:08 PM
|
|
|
> > Also the title screen not shown behind the 3d graphics due to the layer priority
> > issues that the model 3 board does.
>
> Yes I believe this is also the real reason for the FBI warning in the sky. Is this a
> known issue with the PPC core then?
Maybe, because the GTI Club hardware is made from PPC. Not to mention that the sun in the sky during gameplay was kinda cut in half for some reason and we show the Black transparency in the result screen. This issue also happens in Operation Thunder Hurricane (despite the lightgun emulation is not implemented yet)
BTW, are you talking about emulation or the real hardware itself? Cause when I hear the music on MAME, it kinda sounded like a rock/action-esque music that didn't play a final note as opposed to a high pitched European song that can finish the song with 2 notes.
Edited by Embryoice (01/25/13 08:23 PM)
|
|
|
|
Re: GTIClub
[Re: Embryoice]
#303247 - 01/25/13 10:16 PM
|
|
|
The music in MAME currently is wrong. Here's a comparison video, the first part is MAME now and the second is with my "fixes" (which currently breaks other games):
http://youtu.be/Bg8xulSU4p0
This has been verified against the PCB.
BTW The music now also sounds right in Operation Thunder Hurricane. The tile map hack causes problems here though.
Edited by RetroRepair (01/25/13 10:36 PM)
|
http://www.youtube.com/retrorepair
|
|
| Embryoice |
|
I am the Master Warrior!
|
|
|
|
|
|
Reged: 12/30/12
|
|
Posts: 74
|
|
Loc: Los Angeles, CA
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303251 - 01/25/13 10:55 PM
|
|
|
The credits screen are transparent on MAME due to layer priority, but in PCB, it shows up on a black background.
|
|
|
|
Re: GTIClub
[Re: Embryoice]
#303258 - 01/25/13 11:58 PM
|
|
|
I'm starting to think it's more like it doesn't clear the screen when it's meant to. I'm not sure though and it's not easy when there's no docs or info available for the tile generator.
|
http://www.youtube.com/retrorepair
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303290 - 01/26/13 01:41 PM
|
|
|
Responding to what RetroRepair said:
Quote:
The music in MAME currently is wrong. Here's a comparison video, the first part is MAME now and the second is with my "fixes" (which currently breaks other games):
http://youtu.be/Bg8xulSU4p0
This has been verified against the PCB.
BTW The music now also sounds right in Operation Thunder Hurricane. The tile map hack causes problems here though.
Nice... now what are those "fixes"? 
Also, I think Ville is working on that hardware.
|
What? There's nothing here to say.
|
|
| Embryoice |
|
I am the Master Warrior!
|
|
|
|
|
|
Reged: 12/30/12
|
|
Posts: 74
|
|
Loc: Los Angeles, CA
|
|
|
|
Send PM
|
|
|
|
> Responding to what RetroRepair said:
> The music in MAME currently is wrong. Here's a comparison video, the first part is
> MAME now and the second is with my "fixes" (which currently breaks other games):
>
> http://youtu.be/Bg8xulSU4p0
>
> This has been verified against the PCB.
>
> BTW The music now also sounds right in Operation Thunder Hurricane. The tile map hack
> causes problems here though. Nice... now what are those "fixes"?
> Also, I think Ville is working on that hardware.
The fixes on this portion are music and the roz layer. But Ville worked last time before 0.147 are Gourad Shading, Fog and even pretty cool stuff. The only thing left to accomplish with PPC emulation was 2D priority that doesn't clear the whole 3D stuff that supposed to fill the entire screen.
|
|
|
| R. Belmont |
|
Cuckoo for IGAvania
|
|
|
|
|
|
Reged: 09/21/03
|
|
Posts: 9713
|
|
Loc: ECV-197 The Orville
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303321 - 01/27/13 03:45 AM
|
|
|
> Is this a known issue with the PPC core then?
The PPC core in current versions gives a full pass on PPCTorture's hundreds of thousands of opcode tests (the same validation is used by Supermodel, Dolphin, SheepShaver, and QEMU). So it's strongly unlikely CPU core bugs are the cause of any issues with the graphics; Konami graphics hardware frequently has very unusual functions and they're quite difficult to reverse-engineer.
|
|
|
| Embryoice |
|
I am the Master Warrior!
|
|
|
|
|
|
Reged: 12/30/12
|
|
Posts: 74
|
|
Loc: Los Angeles, CA
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: R. Belmont]
#303331 - 01/27/13 08:07 AM
|
|
|
> > Is this a known issue with the PPC core then?
>
> The PPC core in current versions gives a full pass on PPCTorture's hundreds of
> thousands of opcode tests (the same validation is used by Supermodel, Dolphin,
> SheepShaver, and QEMU). So it's strongly unlikely CPU core bugs are the cause of any
> issues with the graphics; Konami graphics hardware frequently has very unusual
> functions and they're quite difficult to reverse-engineer.
Is there any kind of Protection that we're going to fiddle around with it by decapping or it needs redumping by Guru? BTW, I'm just a beginner on engineering.
|
|
|
|
Re: GTIClub
[Re: R. Belmont]
#303333 - 01/27/13 11:44 AM
|
|
|
I do remember now actually Bart said the core was good afterall as I recall he suspected it at one point.
I'm not going to be submitting anything to do with the gfx issues anyway, I just patched the memory so I can play it on the cab without it niggling me.
I think I've made a bit of progress with the sound though.
First of all I added the correct clock for the PCM chip which fixed the pitch but obviously the sequencer IC was still too slow. Looking at Thrill Drive which uses the same setup for sound, the irq was clearly being handled differently so I decided to port this over to the gticlub driver which helped but was still a bit out. Then I noticed Thrill Drive is a bit out too so the timer in k056800.c is possibly wrong.
I adjusted it a bit so it at least sounds right for now. The "slam" sample at the end of the thrill drive attract demo now triggers at the same time the title screen appears and the "gti club" announcer voice is in the right spot too.
Having checked out a few videos on youtube however on thrill drive it seems the k056800 envelopes for at least decay and release are also incorrect (by a long shot actually) so I suspect the attack envelope is also wrong. I've adjusted them a bit (to sensible levels anyway).
At this point I need proper recordings from PCBs though as I no longer have my GTI Club board so if anyone has a link to some good recordings it'd be a great help. Any of the konami boards which have the 056800 chip on the CPU board.
I think I can get these games sounding much better with the right source material 
Edited by RetroRepair (01/27/13 11:48 AM)
|
http://www.youtube.com/retrorepair
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303340 - 01/27/13 01:45 PM
|
|
|
Just realized hornet hardware uses the same setup also. Using Gradius 4 as an example I changed the irq timer to 5 micro seconds which syncs it up with the original recording posted here: http://mametesters.org/view.php?id=3275 (in the current build the announcer says "gradius 4" at the wrong point).
This is the same timing thrill drive and gti club use so I think I've found the sweet spot.
Still not 100% sure on the envelopes though. Now I listen to it side by side with the original recording it's obvious the decay and release were turned right up as the reverb effect isn't working so it sounded closer to the real thing (though caused other more intrusive problems, gticlub skid sound looping for example). It looks like reverb effects aren't hooked up yet and that's a bit beyond my scope.
It does sound better IMHO, though maybe a little empty without reverb.
I'll post a diff in a bit, it'd be good if someone can look at it and make sure it's ok.
Edited by RetroRepair (01/27/13 02:11 PM)
|
http://www.youtube.com/retrorepair
|
|
|
|
Ok I've attached the diff. So I've:
Fixed the sample rate in gticlub.c
Updated other drivers that use the k056800.c
Changed the timer in k056800.c
Altered the envelopes in rf5c400.c to avoid overrun samples
I've also labeled the inputs for GTI Club and added PORT_TOGGLE to the view switch so you can now select manual transmission if required.
Aside from the 3D issue then I think all GTI Club needs now is DIP 4 looked at (should disable wheel check instead fails to go into demo mode) and that reverb added to the k056800 driver.
If someone can check the diff for me to make sure it can be submitted I'd appreciate it.
Edited by RetroRepair (01/27/13 03:12 PM)
|
http://www.youtube.com/retrorepair
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303388 - 01/28/13 09:36 AM
|
|
|
> Ok I've attached the diff. So I've:
>
> Fixed the sample rate in gticlub.c
> Updated other drivers that use the k056800.c
> Changed the timer in k056800.c
> Altered the envelopes in rf5c400.c to avoid overrun samples
>
> I've also labeled the inputs for GTI Club and added PORT_TOGGLE to the view switch so
> you can now select manual transmission if required.
>
> Aside from the 3D issue then I think all GTI Club needs now is DIP 4 looked at
> (should disable wheel check instead fails to go into demo mode) and that reverb added
> to the k056800 driver.
>
> If someone can check the diff for me to make sure it can be submitted I'd appreciate
> it.
I don't understand the PORT_TOGGLE for the AT/MT switch though.
On the real hardware is IS in fact a simple button (like the VIEW/START) button.
You start off in AT mode, hit it once, and continue in MT mode.
No need for a toggle.
|
|
|
|
Re: GTIClub
[Re: SailorSat]
#303389 - 01/28/13 09:50 AM
|
|
|
Ah ok, I seemed to recall it was literally a toggle switch on the cabinet, thanks for that. This is why I didn't want to just go ahead an submit it. Everything else look ok to you?
|
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303393 - 01/28/13 11:56 AM
|
|
|
|
|
| R. Belmont |
|
Cuckoo for IGAvania
|
|
|
|
|
|
Reged: 09/21/03
|
|
Posts: 9713
|
|
Loc: ECV-197 The Orville
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303424 - 01/28/13 11:48 PM
|
|
|
I'll save you the trouble and reject this patch right here ;-) Don't worry, it's pretty easy to fix. And don't forget recordings from your PCBs showing that this actually improves things.
(Aside: the 056800, which Konami calls MIRAC, is a "mailslot" chip; it allows the sound 68000 to communicate with the main CPU, and was first used on System GX. Music sequencing is done entirely in software on the 68000).
attotime timer_period = attotime::from_hz(46000) * 128; // not sure what this is in usec but sounds correct
Digital logic can't derive clocks from thin air. All timing values are always a function of the input clock divided by an integer (usually a power of 2, but on newer chips like this odd dividers are possible). Something like attotime::from_hz(device->clock() / 3); would yield a very similar rate and have the benefit of not being a number that was obviously pulled out of someone's butt.
- machine.scheduler().timer_set(attotime::from_usec(1), timer_expired_delegate(FUNC(zr107_state::irq_off),state), line);
+ state->m_sound_irq_timer->adjust(attotime::from_usec(5), line);
Changing the "from_use(1)" to "from_usec(5)" would yield the same result without the unnecessary copypasta of that poor other driver's code.
|
|
|
| R. Belmont |
|
Cuckoo for IGAvania
|
|
|
|
|
|
Reged: 09/21/03
|
|
Posts: 9713
|
|
Loc: ECV-197 The Orville
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: R. Belmont]
#303426 - 01/29/13 12:46 AM
|
|
|
I just noticed something important about the envelope rates: they're based on the MAME output sample rate, not the chip's output sample rate. That's gibberish, and it means I have faster envelopes than stock (I run MAME at 48000 Hz output) while people running 22050 for performance have 1/2 speed envelopes. What output rate were you running at when you made those adjustments?
|
|
|
|
Re: GTIClub
[Re: R. Belmont]
#303440 - 01/29/13 11:18 AM
|
|
|
Thanks for all the input, it's helped me out quite a bit I've learned a lot about mame the last few days, hopefully I can make some useful contributions in the future.
I didn't much like changing the clock on the MIRAC like that and felt it was probably wrong which is why I posted it here first 
I don't have gticlub anymore so I can't provide good recordings. There is a PCB recording posted above from gradius 4 but it's not the best example (though for timing purposes it's a great example). You can hear the difference though, reverb vs lots of release and decay which exacerbates a few of the audio issues the driver still has. Not to mention is just sounds bad and is pretty much a hack. It's most noticeable on thrill drive on the car select screen and the coin up sfx. It would be good to get a line recording of thrill drive for comparison. I may source a board to do just that.
I didn't notice how the envelopes were timed though. I'm running mame at the default 48khz.
Edited by RetroRepair (01/29/13 11:20 AM)
|
|
|
| R. Belmont |
|
Cuckoo for IGAvania
|
|
|
|
|
|
Reged: 09/21/03
|
|
Posts: 9713
|
|
Loc: ECV-197 The Orville
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303446 - 01/29/13 05:29 PM
|
|
|
> I didn't notice how the envelopes were timed though. I'm running mame at the default
> 48khz.
Could you run mame at 44100 and re-adjust the envelopes for your resubmit? That way when I change them to be correctly timed off the chip it won't screw up your adjustments
Thanks!
|
|
|
|
Re: GTIClub
[Re: R. Belmont]
#303495 - 01/30/13 01:25 PM
|
|
|
I ran it at 44100 but it didn't sound any different, even comparing recordings with both frequencies.
I'm having a bit of trouble with this though:
Code:
attotime timer_period = attotime::from_hz(device->clock() /3);
This should work fine but it's hanging before emulation starts. Any idea why?
|
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303503 - 01/30/13 02:58 PM
|
|
|
> I ran it at 44100 but it didn't sound any different, even comparing recordings with
> both frequencies.
>
> I'm having a bit of trouble with this though:
>
> attotime timer_period = attotime::from_hz(device->clock() /3);
>
> This should work fine but it's hanging before emulation starts. Any idea why?
1. The current K056800 implementation doesn't take a clock parameter so it defaults to 0. This is easily fixed with:
Code:
#define MCFG_K056800_ADD(_tag, _interface, _clock) \
MCFG_DEVICE_ADD(_tag, K056800, _clock) \
Unfortunately the K056800 clock isn't documented. It could be 16MHz, 16.934MHz or something else entirely (though unlikely).
2. For a 16.934MHz clock, attotime::from_hz(device->clock() /3) produces a 5.3MHz timer, which is far too high Ideally we want something like:
Code:
attotime::from_hz(device->clock()) * 16384 * 3
This would give a frequency of 344.5Hz (period 2.9ms - NOT usec!). This of course is MAME's current timer value so some experimentation is in order.
Ideally we need somebody with a boardset and frequency counter to measure the actual values.
|
|
|
|
|
Ah this all makes a lot more sense to me now
I am curious though where this number comes from: 16384
I've added that in anyway so I'll experiment a bit.
|
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303509 - 01/30/13 05:20 PM
|
|
|
> Ah this all makes a lot more sense to me now
>
> I am curious though where this number comes from: 16384
>
> I've added that in anyway so I'll experiment a bit.
As RB said, clock dividers are typically powers of 2 and in this case we need a particularly large one to divide down to the required frequency range.
|
|
|
|
|
Ok well it looks like 14700 (44100/3) is correct which means there's a pretty good chance the 056800 runs at 16mhz as the 68k does.
I've updated all the drivers concerned to use this timing. If I've missed any let me know.
Diff is attached, let me know if it's ok
|
|
|
| R. Belmont |
|
Cuckoo for IGAvania
|
|
|
|
|
|
Reged: 09/21/03
|
|
Posts: 9713
|
|
Loc: ECV-197 The Orville
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303727 - 02/03/13 05:50 PM
|
|
|
Applied along with my fix so the MAME output rate doesn't change envelope speeds.
And then Ville woke up and fixed the title screen and the misplaced FBI logo.
|
|
|
|
Re: GTIClub
[Re: R. Belmont]
#303732 - 02/03/13 06:24 PM
|
|
|
Excellent, some real nice progress made on this driver now then
I compiled both of Ville's commits and it looks great now, though the FBI logo is still in the sky I think it's either the tile is rendered too narrow or the read address is slightly off or something. shifting it over by 32px fixes it (not the split sun though, that's a different layer).
Edited by RetroRepair (02/03/13 06:24 PM)
|
|
|
| Embryoice |
|
I am the Master Warrior!
|
|
|
|
|
|
Reged: 12/30/12
|
|
Posts: 74
|
|
Loc: Los Angeles, CA
|
|
|
|
Send PM
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303735 - 02/03/13 07:42 PM
|
|
|
Can you upload the video with his fix.
|
|
|
|
Re: GTIClub
[Re: RetroRepair]
#303736 - 02/03/13 08:13 PM
|
|
|
> Excellent, some real nice progress made on this driver now then
>
> I compiled both of Ville's commits and it looks great now, though the FBI logo is
> still in the sky I think it's either the tile is rendered too narrow or the read
> address is slightly off or something. shifting it over by 32px fixes it (not the
> split sun though, that's a different layer).
I've mostly figured it out now.
ROZ layer 0 needs to be 2048x1024. This makes the FBI logo go in the right place (X=1024, Y=384) and makes the sun correct.
ROZ layer 1 needs to be 512x1024.
Hopefully with this info I can identify the ROZ size register somewhere and get rid of a lot of hacks...
|
|
|
|
Re: GTIClub
[Re: Ville Linde]
#303791 - 02/04/13 03:55 PM
|
|
|
That's great, so this driver should look pretty much perfect now, unless there are other effects I'm not aware of
I'm talking with someone ATM who is going to do a line recording from their GTI Club PCB so I can hopefully nail down some more of the sound specifics. I think the main thing it's missing though is the effects such as reverb.
|
|
|
|
Re: GTIClub
[Re: RetroRepair]
#304867 - 02/25/13 12:08 AM
|
|
|
Just got a GTI Club twin, I will be doing some recordings of the audio for archival purposes.
Is there anything else I should do while I have this here? Anything useful for the driver? The cabinet itself is in pretty poor shape so I'll be parting it out very soon.
I did notice that it seems to be notably easier than MAME with default settings. I'm terrible on either, but I find I get 1st pretty much every time on the real thing and only once in a blue moon on MAME (I have MAME set up in my sega driving cab, the controls don't feel any different except for force feedback). It's like the CPU cars are slower on the PCB, or the player car is faster?
I've also noticed there's elements to the music that are in the PCB that are totally missing in MAME. I guess it could be a different rom revision, but I'll be dumping the roms anyway in case.
Edited by RetroRepair (02/25/13 12:09 AM)
|
|
|
|
Re: GTIClub
[Re: RetroRepair]
#304946 - 02/26/13 07:48 PM
|
|
|
> Just got a GTI Club twin, I will be doing some recordings of the audio for archival
> purposes.
>
> Is there anything else I should do while I have this here? Anything useful for the
> driver? The cabinet itself is in pretty poor shape so I'll be parting it out very
> soon.
>
> I did notice that it seems to be notably easier than MAME with default settings. I'm
> terrible on either, but I find I get 1st pretty much every time on the real thing and
> only once in a blue moon on MAME (I have MAME set up in my sega driving cab, the
> controls don't feel any different except for force feedback). It's like the CPU cars
> are slower on the PCB, or the player car is faster?
>
> I've also noticed there's elements to the music that are in the PCB that are totally
> missing in MAME. I guess it could be a different rom revision, but I'll be dumping
> the roms anyway in case.
It's possible that the network IRQ can affect the car speed. It's mostly guesswork so I wouldn't be surprised if it was wrong.
|
|
|
| Embryoice |
|
I am the Master Warrior!
|
|
|
|
|
|
Reged: 12/30/12
|
|
Posts: 74
|
|
Loc: Los Angeles, CA
|
|
|
|
Send PM
|
|
|
|
What I noticed on the replay is that the cars and the camera angle doesn't move it's like turning the race into a slideshow (except for CPU Cars)
|
|
|
|
Re: GTIClub
[Re: Embryoice]
#305003 - 02/27/13 03:56 PM
|
|
|
Quote:
What I noticed on the replay is that the cars and the camera angle doesn't move it's like turning the race into a slideshow (except for CPU Cars)
I also noticed that, too. Probably some bug in the driver.
|
|
|
|
|
are any these fixes in beta binaries anywhere? or do I need to learn to compile myself?
my lad loves GTi on PS3 and been watching all gti on youtube he loves the long french one for some reason.
anyhow any way I can play gti club 1, 2 (italaino) and xtrail
|
|
|
|
Re: GTIClub/K056800/RF5C400 Sound Fix
[Re: RetroRepair]
#317183 - 11/09/13 04:55 PM
|
|
|
I've been fixing sound issues across the mid-late 90's Konami stuff and made some discoveries relating to GTI-Club and co.
Firstly, everything here is wrong
The 056800 does not generate any timer interrupts.
For games with an RFC5400, the timer interrupt is generated by dedicated logic. The frequency seems to be fixed but differs between hardware platforms.
We need frequency measurements from GTI-Club, a Hornet-based game and maybe something on NWK-TR too if we're to do this properly.
|
|
|
Previous 
Index 
Next 
Threaded 
GTIClub/K056800/RF5C400 Sound Fix
Reply
[Re: 
