14 replies to this topic

[WhenLit]

This concerns (at least) Eight Ball Deluxe and doesn't appear with previous versions of VPinMAME.
In this pinball you have voice and electronic sounds.
During voice talking, the electronic sounds are not played.

[Gaston]

That's correct. It has never been any different on other versions though.

[WhenLit]

That's correct. It has never been any different on other versions though.

Do you plan to restore the good behavior in VPinMAME 2.2?

[Gaston]

It has never been any different on other versions

[MDG79]

Mentioning the voice in the earlier posts. It was generated by the TMS5200 as we all know.

I was wondering about the progress on getting the emulation of this speechchip right, cause I haven't heard about it in quite some time.

The topic called "speech problems" from the old Vpforums.com ceased to exist, so I chose to post it here.

[Lord Nightmare]

Mentioning the voice in the earlier posts. It was generated by the TMS5200 as we all know.

I was wondering about the progress on getting the emulation of this speechchip right, cause I haven't heard about it in quite some time.

The topic called "speech problems" from the old Vpforums.com ceased to exist, so I chose to post it here.

Yeah, I loved that old thread. Well, in regards to TMS5200, not 2 hours ago i finished typing in the lpc tables from dumps i made back in april 2008 of the PROMOUT pin of the chip. It will be included (along with the extensive tms52xx rewrite from the version released today (0.137u3)) in the next version of mame, 0.137u4, probably to be released in about 2 weeks.

The sound quality improvement is astounding. Sark in Discs of Tron actually sounds like a human being, and jack rabbit and money money both sound much more correct. (Those are all the MAME games which use the TMS5200)

If you want to play with the new tms5200 sound NOW (or to import the tms5200 table into qboxpro and encode your own speech or stuff like that), the new tms5110r.c (yes the tms52xx and tms51xx chips now share one file for their lookup tables) can be found at: http://pastebin.com/GrFTXbdC
Though it will expire a month from now, by then it should be in baseline mame and there shouldn't be any issues finding it.

Happy Pinballing, and let me know how things work out with this and the bally tables which use it! Post before and after samples!

LN

Edited by Lord Nightmare, 22 April 2010 - 04:03 AM.

[MDG79]

Good job and thanks for the effort.

I've already tried MAMEXT 0.137u3 and 0.137u4. It indeed sounds far more better than before.
unfortunately the speech in Jack Rabbit and Money Money isn't working in 0.137u4 (and still not in 0.138)
but it is in the other games that use the chip.

As for the link to pastebin.com, since I'm no programmer I don't know how to use the data posted.
I really love to test it with the Squawk and Talk and Zaccaria tables
I've seen from almost all of them movies on Youtube (like these* ),
so I know what to expect so I can hear both the real thing and the improved emulated part and hear the improvement.

*
*
*

I'm looking forward for VPinMAME 2.3 to be released. When can it be expected?

Edited by MDG79, 17 May 2010 - 09:22 PM.

[Lord Nightmare]

Good job and thanks for the effort.

I've already tried MAMEXT 0.137u3 and 0.137u4. It indeed sounds far more better than before.
unfortunately the speech in Jack Rabbit and Money Money isn't working in 0.137u4 (and still not in 0.138)
but it is in the other games that use the chip.

It actually does work in those games; those games expose what is either a bug in MAME's 6821 PIA implementation involving the CA2 and CB1 lines, which are used to show the /int and /ready state of the 5200 to the main cpu, OR on the 52xx /READY timing when certain commands are executed; it seems unless I make the /READY line, after /RS is activated by the PIA, go inactive for more than twice as long as the real chip makes it go inactive during speech data writes, speech won't work at first in those games; however after playing for a while (sometimes two or 3 games) speech usually starts working again. According to tests run on a real pcb, it is a mame bug; the pcb speaks immediately after bootup.

Fixing this is on my long list of things to do eventually.
Note that the hookup in money money/jack rabbit is almost (but not quite) identical to the later zaccaria pinball 1B1170 and 1B11136 boards. (the 1B11178 board is slightly different and might not be affected) However, this means if the pinball sound code is similar to the arcade code, that boards in pinmame which use the pinball boards will probably also fail to speak properly with the new tms5220.c code. I'm looking into it.


LN

Edited by Lord Nightmare, 18 May 2010 - 08:32 AM.

[MDG79]

I've tried both Money Money and Jack Rabbit again, speech in monymony starts earlier than in jack rabbit (at the 4th game)
but monymony sounds exact as here: and Jack rabbit now sounds also more human.

The new LPC tables will do very good for pinball games as Centaur, Elektra and Vector and the other ones.
I like the voices heard on the Squawk and Talk games more than those on the Zaccaria Machines.

Unbelievable how you did it.
That must require some very high education to be able to understand such hardware en reproduce it in a software environment.
Hats off for that.

[destruk]

Yes, good job on that. I don't have much of a clue how to port it to our system for PinMAME, but I'm sure others will be able to get it going with time.

[Lord Nightmare]

It actually does work in those games; those games expose what is either a bug in MAME's 6821 PIA implementation involving the CA2 and CB1 lines, which are used to show the /int and /ready state of the 5200 to the main cpu, OR on the 52xx /READY timing when certain commands are executed; it seems unless I make the /READY line, after /RS is activated by the PIA, go inactive for more than twice as long as the real chip makes it go inactive during speech data writes, speech won't work at first in those games; however after playing for a while (sometimes two or 3 games) speech usually starts working again. According to tests run on a real pcb, it is a mame bug; the pcb speaks immediately after bootup.

hap fixed this, will be fixed for 0.138u1, was indeed related to the PIA missing /READY state changes, and required a new callback.

LN

[lettuce]

Hopefully some work around can be implmented for Data East Tables

[Lord Nightmare]

Unbelievable how you did it.
That must require some very high education to be able to understand such hardware en reproduce it in a software environment.
Hats off for that.

I don't find it incredibly hard, it just requires studying the patent schematic of the chip, reading the datasheets, and understanding what the hardware is actually trying to do. The chip provides a convenient PROMOUT pin (which shows the current bit of the internal LPC data rom being read), and a ROMCLK clock which PROMOUT is valid on the rising edge of. This makes logging the LPC data as it is read very easy. The other really important thing is MS excel 2007, which can handle staggeringly large data sets correctly and was invaluable for decoding the data.
Also, borrowing time on a university logic analyzer helps a lot.

LN

[MDG79]

If you want to play with the new tms5200 sound NOW (or to import the tms5200 table into qboxpro and encode your own speech or stuff like that), the new tms5110r.c (yes the tms52xx and tms51xx chips now share one file for their lookup tables) can be found at: http://pastebin.com/GrFTXbdC
Though it will expire a month from now, by then it should be in baseline mame and there shouldn't be any issues finding it.

Is there a chance you can compile a temporary addition to VpinMAME to play with the new TMS5200 sound until release of the next PinMAME?
Not just for me, but for all members of this forum and users of this emulator. I just have no clue how to do it. It's beyond my education.

It's been a while since the last post on this thread by the way.

Edited by MDG79, 19 November 2010 - 03:42 PM.

[sleepy]

I tried substituting the lookup tables in PinMame source and quickly realized that the use of the tms5110r.c file, renamed to tms5220r.c, requires a rewrite of the PinMame source, probably an update of the Mame core to v.139.
To date, VPinMame/PinMame is still using something like Mame .75 for the core.

Of course I haven't tried copy/pasting the individual tables under the former var names in the older version though...

Edited by sleepy, 12 December 2010 - 10:09 AM.