108 page document about CAVE 1st Gen. Hardware on FPGA

[djsheep]

Thankfully I own PCBs, so it saved me a lot of reading time Thought some of you would find it interesting perhaps...

https://t.co/UqgyFriFhH?amp=1

[chempop]

Very cool!

[ChurchOfSolipsism]

Man, I can't wait for some nerd to produce a batch of Cave consoles... just imagine you could play all of the Cave gen1 games in their original forms.

[donluca]

Man, I can't wait for some nerd to produce a batch of Cave consoles... just imagine you could play all of the Cave gen1 games in their original forms.

Now, *THAT* is something which would be really awesome and I'd be ready to pay a premium price for: a mini-CAVE cabinet with vertical screen with all the 1st generation CAVE games.

Enough with the other console Minis.

[Xyga]

Would be even more interesting for CV1000's which are fragile, insanely overpriced boards.

[spmbx]

cv1k would be more useful; cave68k is already emulated to such an extent that im sure you wouldnt notice the difference between an fpga implementation or a groovymame (mini) pc. I know i couldnt tell the difference between the original and groovymame side by side.

[Xyga]

IIRC even the 1st gen emulation isn't absolutely perfect, but you have to know a pcb very, very well to tell.
So emulation is a valid option there, I agree.

--

For cv1k I think we would need a special re-designed verion of MAME supporting wait-states and a dedicated driver, or low-level emulation on the fastest CPU money can buy lol.
(yet it wouldn't be 100% identical, but probably good-enough almost like 1st gen is)

Since neither is happening anytime soon, maybe even never, the FPGA alternative sounds way more logical and benefical then ?
Unless it can't actually do better than MAME ? (dunno, asking)

[spmbx]

Regardless, how cool that someone managed to do this as part of a masters study!

[zakk]

Since neither is happening anytime soon, maybe even never, the FPGA alternative sounds way more logical and benefical then ?
Unless it can't actually do better than MAME ? (dunno, asking)

As of right now it's likely beyond the capabilities of any reasonable priced consumer FPGA package. I know someone was looking at creating an sh-2 (or 3?) "hat" for the mister but that's probably never going to come to market.

[Necronom]

Man, I can't wait for some nerd to produce a batch of Cave consoles... just imagine you could play all of the Cave gen1 games in their original forms.


Now, *THAT* is something which would be really awesome and I'd be ready to pay a premium price for: a mini-CAVE cabinet with vertical screen with all the 1st generation CAVE games.

Enough with the other console Minis.

There is definitely a market for vertical oriented mini cabs or even handhelds!
Hell, Contra 4 or the arcade emulation of Mr.Do! work great on the DS due to the vertical orientation over two screens.

[avri62]

Would be even more interesting for CV1000's which are fragile, insanely overpriced boards.

Overpriced? I can see that, but only in a subjective sense (the free market decides the pricing based on supply and demand, so "overpriced" is not the best word to use here. Items that routinely, and quickly, sell for their asking prices cannot, by definition, be "overpriced"). Fragile? I absolutely don't agree there. They are very stable, long-lasting pieces of hardware in my experience. The small PCB makes board flex a non-issue, and the only components that ever really need replacing (every 20-30 years) are the condensers, which cost like five bucks (and maybe a few other small, cheap parts), and need to be replaced on any arcade hardware at some point. It is the most robust arcade hardware out there, really. And I think that's something that contributes to their current market value.

[spmbx]

Thats hugely oversimplifying things. Last cv1k board i got rid of when i was still actually using original pcbs was faulty because of the supposedly not too rare color problem:
http://blog.system11.org/?p=355
And its not like the other components are immune to failure; you’re pretty much fucked when any of the other components fail whereas on older pcbs you can replace almost anything and even have custom repro options for some customs.

Sorry for the offtopic.

[jandrogo]

It’s awesome! I am sure this project also opens the implementation for PGM board: DDP daioujou, Espgaluda and Ketsui.
I guess if Seibu SPI board (Raiden Fighters I&II, Jet and Vipher Phase) based in a 386DX cpu is far from the Mister capabilities

[awbacon]

Anything to bring CAVE games into the home at a smaller expense than original PCB's would be awesome to me

[heli]

Anything to bring CAVE games into the home at a smaller expense than original PCB's would be awesome to me

How much cost the FPGA they using ?

[6t8k]

Getting to do this as a part of a masters, how awesome is that.

Magnificent work.

[donluca]

CV1k games are completely out of the question due to the power needed in order to run those games.

PGM hardware might be stretching it, but could be doable.

Honestly, I'd just be happy with Dodonpachi & friends on a mini vertical cab (with proper microswitched controls, of course).

[MameHaze]

FPGA aren't magic, so a solution using an FPGA isn't always going to be better than emulation anyway - it's really just a different form of emulation.

The CV1k blitter is an FGPA in the first place (the games even upload some microcode for it - not used by the emulation) A lot of the limitations (timing etc.) come from the specific FPGA being used, possibly the RAM timings etc. so again unless you're using the exact same hardware for the blitter, it's not inherently going to be better.

[finisherr]

I JUST posted about this in the Shmups FB group. Thanks!

[Mantrox]

Im gonna read this, just cus, but hardware emulation is so good.
Good times ahead for fans of all tipes of obscure gaming titles.

Anything to bring CAVE games into the home at a smaller expense than original PCB's would be awesome to me

How much cost the FPGA they using ?

For development he used a 200$ board and the target hardware is the same board as the MiSTer FPGA, so about 130$.

[Bassa-Bassa]

cave68k is already emulated to such an extent that im sure you wouldnt notice the difference between an fpga implementation or a groovymame (mini) pc. I know i couldnt tell the difference between the original and groovymame side by side.

I thought that ESP RA DE has well known differences in the slowdown department between Mame and PCB? Or has that changed recently?

[finisherr]

Been doing a little research. With FPGA there are clones and functional replicas. Clones are extremely accurate to the original representation and operation of the board hardware where functional replicas might accurately translate the CPU and postulate the design of the other components, essentially leading to a replication that is good enough to pass for the real thing. What excites me about FPGA is that you actually have the ability to program the hardware to operate JUST LIKE the original PCB. With the right tools, access to a board, programming skills, and time, you can essentially create a clone of one of these arcade PCBs. The irony is that if you want to make an accurate clone, you typically need to have access to the PCB to reverse engineer the components.

[MameHaze]

As a developer I personally think it's a bit of a regressive step.

Sure, if you want to actually *bootleg* hardware, it's the way to go (and by no coincidence are a lot of bootlegs based off FPGA re-implementations)

But in terms of portability of what you're developing you're developing very specific solutions, that rely on hardware supplies of certain FPGAs being available if you want exact reproductions etc.

So yeah, if you're looking to produce a batch of bootleg boards / carts / hardware reimplementations and sell them, great, if you're trying to develop something for the future, maybe not so great.

It's like when people run pirate versions of PC based arcade games on their PC. In theory, cut out the emulator, same type of hardware as the original, so it's guaranteed to be 'perfect'. In reality, quite the opposite, sure, it can work, but there are many other factors involved, and those games require extensive hacking to run, and on some hardware configurations don't run at all.

With that said, I think FPGAs could benefit emulator, or computing in general, in a support role (and you do see this with FPGA based cards for bitcoin miners) but again, long term, I think it might be problematic.

[donluca]

Meh, I think I've expressed my views on FPGA and, just to reiterate, I believe that the optimal solution stands in between.

Modular boards where you can put some of the original ICs is, IMHO, the best solution available.

Say you're emulating a Mega Drive and got a spare 68000: pop it in on your FPGA board and it will be used and you'll have 100% accurate reproduction CPU-wise. If you don't have it, it will be emulated by the FPGA.
Same for Z80s, custom ASICs and so on.

Ideally, an FPGA board should work as a kind of "glue" between the original components and work at a logic level (how all those components are hooked up with each other) and add emulation wherever an original component is missing.

The whole point should be that an FPGA with all of the original components which made the target original hardware platform will behave 100% like the original board even at circuit level.
When one or more components are not available, the FPGA emulates those components striving to get as close to how the original component behaved as possible.

[spmbx]

cave68k is already emulated to such an extent that im sure you wouldnt notice the difference between an fpga implementation or a groovymame (mini) pc. I know i couldnt tell the difference between the original and groovymame side by side.

I thought that ESP RA DE has well known differences in the slowdown department between Mame and PCB? Or has that changed recently?

Im sure that if you study it frame by frame you will find a slight difference somewhere but i dare say that for 99,9% of the population it will be flawless.

This is (albeit non gameplay) a comparison thats already 4 years old:

https://youtu.be/EXF7wT7Hc60

[awbacon]

Thats an insane amount of work and super impressive. I for one don't hate FPGA, just due to the fact that at some point all the original hardware will be dead. I'd rather play SOME form of hardware implementation of the original vs straight emulation

[Bassa-Bassa]

cave68k is already emulated to such an extent that im sure you wouldnt notice the difference between an fpga implementation or a groovymame (mini) pc. I know i couldnt tell the difference between the original and groovymame side by side.

I thought that ESP RA DE has well known differences in the slowdown department between Mame and PCB? Or has that changed recently?

Im sure that if you study it frame by frame you will find a slight difference somewhere but i dare say that for 99,9% of the population it will be flawless.

This is (albeit non gameplay) a comparison thats already 4 years old:

https://youtu.be/EXF7wT7Hc60

That's Espgaluda, totally different hardware. And that video, if anything, tries to prove that you can get the same video frequency with Groovymame, which is entirely a different problem to the slowdowns issue, caused by unemulated features/hardware behaviour in particular situations (not seen in the attract modes, normally).

Too bad that we never got comparison videos for ESP RA DE, DDP and/or Guwange after all this time.

[spmbx]

You are right, remind me to never post from mobile :/ i sold my boards ages ago so i personally van’t do a comparison now either.

[Mantrox]

As a developer I personally think it's a bit of a regressive step.

Sure, if you want to actually *bootleg* hardware, it's the way to go (and by no coincidence are a lot of bootlegs based off FPGA re-implementations)

But in terms of portability of what you're developing you're developing very specific solutions, that rely on hardware supplies of certain FPGAs being available if you want exact reproductions etc.

So yeah, if you're looking to produce a batch of bootleg boards / carts / hardware reimplementations and sell them, great, if you're trying to develop something for the future, maybe not so great.

It's like when people run pirate versions of PC based arcade games on their PC. In theory, cut out the emulator, same type of hardware as the original, so it's guaranteed to be 'perfect'. In reality, quite the opposite, sure, it can work, but there are many other factors involved, and those games require extensive hacking to run, and on some hardware configurations don't run at all.

With that said, I think FPGAs could benefit emulator, or computing in general, in a support role (and you do see this with FPGA based cards for bitcoin miners) but again, long term, I think it might be problematic.

Why would it be problematic?
Can't we still have both?

Let the shmup enthusiasts have their FPGA board that allows them to play all the titles in a simplified, cheaper setup.
Let the arcade preservationists keep archiving and reverse engineering code.

Most of the work on this paper seems to have relied on the previous efforts of the MAME devs anyway.

Meh, I think I've expressed my views on FPGA and, just to reiterate, I believe that the optimal solution stands in between.

Modular boards where you can put some of the original ICs is, IMHO, the best solution available.

Say you're emulating a Mega Drive and got a spare 68000: pop it in on your FPGA board and it will be used and you'll have 100% accurate reproduction CPU-wise. If you don't have it, it will be emulated by the FPGA.
Same for Z80s, custom ASICs and so on.

Ideally, an FPGA board should work as a kind of "glue" between the original components and work at a logic level (how all those components are hooked up with each other) and add emulation wherever an original component is missing.

The whole point should be that an FPGA with all of the original components which made the target original hardware platform will behave 100% like the original board even at circuit level.
When one or more components are not available, the FPGA emulates those components striving to get as close to how the original component behaved as possible.

Has the deviation in behaviour between an FPGA modeled X68000 and a real IC been measured by anybody?
Would be cool to see what kind of order of magnitude we are talking about.

That hybrid could be the next step.

[donluca]

There are several inaccuracies for example in the 68000 emulator, one of the most important one being the lack of "wait states" emulation (you probably have already read about this) which results in incorrect speed (often compensated by lowering the CPU speed itself).

The fact that you're not able to discern the real thing from the emulation doesn't mean they're indistinguishable.
There are players who have put thousands of hours into some games and can tell you immediately.

AFAIK, only Pong and maybe another couple of very primitive games have been emulated with 100% accuracy (at circuit level).

EDIT: and, IIRC, it doesn't exist a CPU powerful enough to have them run at 100%. I remember reading an article some years ago where a water cooled overclocked i7 was able to reach 4% at most.