[tech] Neogeo the mystery ?!

All around the internet you find more or less easily technical stuff about your preferred console. This place it's heaven for guys (me first) who want absorb a lot of obscures technicals specifications like devkit, development tips, chips datasheet, diagram...

However, there is a minority of console who are not well documented, even on common point like graphic capabilities, chips function, memory, datasheet. It's the case of the Neogeo. How a very famous console like Neoego have only so few documentation ? My best friends, Google, Yahoo, Altavista and Co. refuse to give me any information about old crap dusty Neogeo chips :banghead: !

This is why I ask here, question is very "simple" : what is it ?

On Neogeo AES (board rev2)

- LSPC (sprite processor ??) http://wellby.dyndns.org/dnl/img/mb_lspc.jpg
- NeoB1 (unknow) http://wellby.dyndns.org/dnl/img/mb_neob1.jpg
- neo-go (unknow) http://wellby.dyndns.org/dnl/img/mb_neo-go.jpg
- neo-d0 (unknow) http://wellby.dyndns.org/dnl/img/mb_neo-do.jpg
- neo-c1 (unknow) http://wellby.dyndns.org/dnl/img/mb_neo-c1.jpg

On cart (MVS, but AES contains same chips)

- PCM (sound ? no... Bankswitch, probably...) http://wellby.dyndns.org/dnl/img/car_pcm.jpg
- neo-273 (there too... unknow...) http://wellby.dyndns.org/dnl/img/car_neo-273.jpg

Humm... Others questions... yes, about YM2610, PSG sound is a part of this sound processor or it's built in another chip (like Genesis VDP) or stand alone chip.
Thank ! I am thirsty of knowledge :drink:

 

There's a good change that somebody here might be able to answer your questions, but if not check www.neo-geo.com

There's a huge community there with lots of people who know much about the system and its hardware.

 

Oh oh, you really believe that I do not already know this site ? Nothing about Neogeo chips in these forums. And when you ask, it answer you that you are a lazy boy, google is your friend and many other pleasant useless things.

If I ask here it's because on neo-geo.com they don't know (or don't want... or don't have time... choose one !). Ok, these informations are boring technicals specifications, but I just curious and my curiosity is not filled by google, neo-geo and many others sites.

 

You won't be able to find anything online about the chips; I've tried everything. It's been a year or more since I last looked at the boards but I can tell you that the NEO-273 replaces 4x '273 on early boards off the top of my head, that's a very obvious one. I have some notes I made at home I can post later.

Edit: NEO-ZMC is "standard" bankswitching used until protected games.
IIRC NEO-PCM is a bunch of random stuff for decoding/latching
PRO-CT0 is some sort of character MUX

 

Calpis --> Oh ! Thank for these informations. I have no idea how you obtained them but it is a beginning for me.

Eh eh, me too :icon_bigg

 

It is clear that NEO-ZMC is character bankswitching because it is used on games with I think over 32M character graphics, not on early games, I have also seen some bootleg carts with discrete registers to simulate it which surely means bankswitching. (I sadly don't own one of these bootleg carts to learn from)

PRO-CT0 is only on early Neo AES carts, this is to reduce the number of pins on the cartridge connector and provide a protection mechanism so that people can't easily adapt MVS -> AES. Unfortunately (of course) there is no discrete chip equivalent. The Phantom-1 works by using a NEO-CMC chip which contains the PRO-CT0 logic as well as protection stuff. The new Neo SNK converter AFAIK uses a CPLD to simulate all this.

NEO-PCM I'm sure is simply registers/latches because early games don't have this chip; instead they have the equivalent discrete circuit. The same thing applies to NEO-273. These chips are used by all games, not for protection but for just operation. Again, later huge chips which contain protection also contain this logic inside.

 

NEO-ZMC2 is PRO-CT0 combined with NEO-ZMC.

 

After a bit of multimeter action but mainly from the top of my head...

The NEO-PCM is a collection of approximately 8 x 74x364 Octal D Type latches in a single package used to buffer PCM data from the V Roms to the YM2610.

The NEO-273 (as Calpis already mentioned) is a set of 4 x 74x273 D Type FF's arranged in 2 pairs. Its function appears to be to buffer data from both the C Roms AND the S Roms when needed.

The NEO-ZMC At a wild guess is a (Z)80 (M)emory (C)ontroller. It is used to page in and out the 64k banks of the M Rom used by the Z80 sound controller.

MrS

 

Looks like their server is down at the moment but you may want to talk to the guy who runs this site.

www.snk-capcom.com

He's owned and tweaked a few prototype [ Ghost Lop ] and homebrew MVS games and knows the folks who travel in those circles pretty well, so he may have that info or be able to point you in the right direction.

I havent talked to him in while but I bellieve his name is Venn Trieu. ;-)

 

Oh thanks for all these usefull information. My main purpose is to know how work the audio pass through we found on somes MVS to AES converter. Where is taking the audio signal ? It's strereo analog out but on the cart !

according to this topic the pinout don't show any audio through the cart port.

I Wondering where is the audio carrier in cart port and why use this process to get audio out (Neogeo AES has builtin plug for audio out).

 

My pinout is not 100% complete, I've also been wondering which pins are audio but I don't want to open my AES to find out the easy way. You'll have to find someone with a Phantom-1 to trace which pins are the audio out, I've tried but haven't found anyone willing to do so.. it can't be too difficult to find though since you can just touch a wire to an amplifier to each NC pin, I just don't want to ruin a cable

BTW, I since bought a Super MVS Converter II and it doesn't have a CPLD, it has a custom ASIC from Neo Team for the CHA board, and on the PROG board there is only a 16V8 and 2x74245s.

Mr Sporty was right about the ZMC, it bankswitches the M ROM not C ROM, almost all other bankswitching is internal to the console! The Neo appears to be able to address 1Gbit of C ROM also which I don't think is well known, but this is probably where "Giga power" comes from!

Also now I know for sure that:
-273 simply latches the addresses for C ROM and S ROM on either edge of a clock, just the same as the original discrete circuit, but it has a few more address bits (128M I think)
-PCM does a little more than latch addresses for V1 and V2 ADPCM buses, it latches and arbitrates both ADPCM ports using one ROM.
-ZMC consists of VERY simple NES-like bankswitching

Approx logic size:
-273 + 7474 to address 512M: 273,273,273,273,174,04
-PCM: 374,374,374,374,374,374,374,374,161,14,14,245,373
-ZMC: 273, 245, 157, 04, 32

I got these figures from a bootleg with discrete workarounds. I haven't pulled it apart yet to learn the logic but I will sometime...

 

Thanks Calpis ;-)

You can read this old topic about audio out through cart port. But it does'nt give any definitive answer on this feature.

OK, it's a good forward for me on knowledge about NG hardware. Now, I come back on another question :

about YM2610, PSG sound is a part of this sound processor or it's built in another chip (like Genesis VDP) or stand alone chip ?

 

Yes YM2610 has 3 separate soundoutputs... one for analog PSG and two digital for FM/PCM.

 

Hmm, I'm not so sure they are correct but now I'm really glad I didn't open my AES... I hope it's not true that this method can't be used on all consoles because I think it's nicer than using the headphone jack.

Because there are two pins that loop back on the PCB, I assume that's the audio channels, but I don't know which is in and out or which is which channel. These pins are A23/A24 and B23/B24.

 

I might be stating the obvious, but there might be some information of interest in the MAME sources. At least, it should give you an idea what the system does and supports internally, that maybe you could trace the function back to the chips.

 

You'd think so, since it's aim is to "document hardware", but MAME source is pretty useless for determining practically anything about the hardware other than what processors/video/sound controllers it consists of and a fuzzy memory map. MAME drivers themselves are almost always blackboxed and verified against real hardware since it's far more practical than deconstructing and reverse engineering every single PCB, but this also means there's no documentation other than high level approximations of how a tile gets rendered, and honestly it doesn't really matter to just play or develop a game, it does matter however if you want to accurately interface to/clone/fully understand a game.

 

We do try. There are a lot of deficiencies, I for one would like to see the 68000 emulation use the microcode. However some would argue that this documents the hardware less, even though it's working more like the real thing. I'm not sure if we have the actual microcode right now, the one in the patent is supposedly not quite the same as the released version.

What each chip does on a neogeo, I don't know. I doubt anyone outside of SNK knows & I don't know how we could find out.

MAME has fallen behind in alot of ways, compared to Raine and NeoRageX it is more accurate. But MESS ( which is the console equivalent and shares alot of code with MAME ) is less accurate than many dedicated emulators & architecturally it's hard to achieve some things.

However MAME is still active, so it can be improved. We let you have the source too, so you can try and improve it too.

 

Perhaps the next step should instead be the inclusion of VHDL/Verilog source? I mean, if you're going to go for microcode, why not shoot for clock edges?

Here's my process: first step -- obtain/make a schematic of the console and see which identifiable parts the unknown chips interface with. Second step -- identify which high level logic gets applied in which chip and verify/bring it low level with a logic analyzer (or find fully operational bootlegs with discrete circuits). Third step -- work out the kinks (easier said than done I know). Yeah this is a crazy time consuming, destructive/expensive and admittedly stupid process but it works.

I think more people would like to but find MAME difficult or limiting as opposed to stand alone emulators.