Explanation of what a NOAC (NES on a chip) is

[Atariboy]

I had an argument with an individual that insists that a NOAC chip is a modern piece of electronics that's running a NES emulator. He has a NOAC based clone and believes that the fact that the plastic casing is essentially empty inside with a single chip unlike an original NES (Despite the fact that three decades have now lapsed since the hardware was first designed) and the fact that it isn't perfectly accurate are proof positive that it's emulation.

 

I tried to explain about the miniaturization of electronics that has made possible things like the 2600 on a chip on the Flashback 2, that a hardware reproduction can be inaccurate just as a software emulation can be, and that other plug and plays have utilized NES on a chip hardware with rewritten software like the various Intellivision plug and plays and the original Atari Flashback to no avail.

 

Can anyone post an explanation that I can link this individual that thinks he's an expert?

 

Thanks

Edited December 5, 2013 by Atariboy

[TheGameCollector]

Easy:

 

The reason you have inconsistencies in the way the games run is because the new version of the unit is using different chips in the place of the old ones. These chips are programmed to run the same way as the originals, but because they are not exactly the same or are missing certain parts altogether the results are different. Different companies use different replacement chips or program them differently so results vary. In example Yobo FC Twin versus Retro Duo versus Retron 2 all having different issues. After all, many of these parts are cheaper and are manufactured in China rather than Japan.

Edited December 5, 2013 by TheGameCollector

[godslabrat]

Well there was this thread a while back, where it was said that clone hardware is, in itself, a form of emulation. I kind of bowed out of the discussion because it kind of split hairs, but it sounds like it might be useful for your question.

Edited December 5, 2013 by godslabrat

[+Gemintronic]

Reverse engineering hardware isn't emulation. A NOAC is as much an NES as one revision of an NES is to another. Emulation means there is no attempt to clone the actual hardware - the focus is running software that translates game ROMs into something a foreign machine can handle.

[Osgeld]

I call it neither, its not just software emulating a system, nor is it really a clone, cause who in their right mind is going to design and produce chip level hardware to mimic a 20 something year old machine. I suspect most of it is FPGA technology which is a software description of the original hardware running on a generic gate array. It may be masked to do this, making it seem like some magical chip but its not

 

It puts those systems into a interesting purgatory of having none of the benefits of either and all the problems of both, which is why 99.99999% of X on a chip systems tend to suck.

Edited December 6, 2013 by Osgeld

[Atariboy]

All these NOAC's over the years definitely haven't been FPGA technology. To replicate the NES in that way is a fairly expensive proposition where as a NOAC chip costs hardly anything.

[Osgeld]

more expensive than designing an entire system on a chip, having it fabricated, tested, tweaked, having it fabricated, found a bug, change design, having it fabricated? I think you over estimate the capaicity of jim bob's knock off shop in china's willpower, financial status, and connections to accomplish a feat like that when even 10 years ago less than 500 bucks would get you a decent FPGA development system more than capable... and if you screw up its a simple erase and go.

 

sure the end product is a chip on board mask programed epoxy blob that cost next to nothing, but it did not get there by doing things the way it was done in 1989 either.

Edited December 6, 2013 by Osgeld

[Atariboy]

Sorry but it definitely isn't FPGA technology that has gone into things like the Retro Duo, the Atari Flashback, Generation Nex, etc. It's ASIC which replicates rather imperfectly the original hardware.

 

So in other words, it's a NES on a chip. All the major components are replicated on a single integrated circuit. No substituting alternate chips for the originals, no FPGA, and no emulation. They've probably made over a million of these things over the years so the production run quantity is certainly there to justify the expenditure and the cost is probably why problems persists instead of being improved upon.

 

But it most certainly can't be reprogrammed after the fact. It's expensive and no point to it in this application. FPGA's are valuable for small production runs or something that you want to be able to modify after the fact. But enough of these are made and the cost of a powerful enough FPGA are enough where it's no contest.

 

And I doubt that ever changes. I suspect that the Retron 5 shows the future for the classic gaming clone business. FPGA's place in the classic gaming world probably won't go past hobbyist and homebrewers.

Edited December 6, 2013 by Atariboy

[Osgeld]

microchip will sell you a pic with your program burned onto it permanently, just cause later one can not change the program does not mean its not a pic, but whatever, I am not here to argue with you

[onmode-ky]

I call it neither, its not just software emulating a system, nor is it really a clone, cause who in their right mind is going to design and produce chip level hardware to mimic a 20 something year old machine. I suspect most of it is FPGA technology which is a software description of the original hardware running on a generic gate array. It may be masked to do this, making it seem like some magical chip but its not

more expensive than designing an entire system on a chip, having it fabricated, tested, tweaked, having it fabricated, found a bug, change design, having it fabricated? I think you over estimate the capaicity of jim bob's knock off shop in china's willpower, financial status, and connections to accomplish a feat like that

As Atariboy has said already, NOACs are ASICs, not FPGAs. If you want specifics, the NOAC that was used in the first model of Atari Flashback is the Novatek NT6578 chip. Another known NOAC is the SinoWealth SH6578. Look them up if you want. I don't know why you think Chinese semiconductor factories are limited to "jim bob's knock off shop" rather than actual IC manufacturers.

 

As for "who in their right mind is going to design and produce chip level hardware to mimic a 20 something year old machine," I suppose Curt Vendel must really not have been in his right mind when he designed and produced the Michele ASIC in 2005, when the Atari VCS/2600 was nearly 30 years old, for the Atari Flashback 2. Atari produced 860,000 of them. Doesn't seem like a good fit for FPGA.

 

They've probably made over a million of these things over the years

Was that a typo? You seem to be shooting a bit low. As I noted over in the AtGames Intellivision plug-n-play topic, Techno Source produced about 4 million NOAC-based Intellivision plug-n-play systems between 2003 and 2005. The first Atari Flashback must have sold somewhat comparable numbers to the Flashback 2, considering it greenlit its successor project. Let's somewhat conservatively call the running total 4.5 million. And at this point, we haven't included Techno Source's other NOAC-based plug-n-play systems (card games, Coleco LED remakes), Radica's non-Sega plug-n-play systems (Tetris and Taito--I think these were NOACs), Majesco's NOAC-based 2004 plug-n-play lineup (Konami, Frogger standalone, casino), Dreamgear's and Senario's plug-n-play lines (Jungletac et al.), and, oh yes, the flood of pirate Famiclones over the past decade-plus. I think 5 million as an estimate of the global total would absolutely have to be on the low side. Definitely not numbers suited for FPGA.

 

Going back to the original question (sorry I'm late, Atariboy), I think recommending that your acquaintance do a Google search for "noac" and "6578" should be a sufficient start. If he still doesn't believe you, then . . . shake your head and sigh heavily.

 

onmode-ky

[Osgeld]

mk so I looked up the 6578 and what little I found on it, it seems to be a generic 6502 core micro, with ram and rom, intergrated video (undescribed) and a generic gate array (aka small fpga) mostly used in picture frames and small embedded touchscreen solutions for kiosks, not exactly what I would call a from the ground up recreation of the NES hardware.

 

NOAC that was used in the first model of Atari Flashback is the Novatek NT6578 chip

 

 

yea cause a 2600, 7800, NES and a Colby picture frame all have identical hardware originally, that should have been your fist tip off that it is not a from the ground up ASIC repoduction of the original hardware, its a half breed partically recreated in software and gate logic

Edited December 11, 2013 by Osgeld

[onmode-ky]

mk so I looked up the 6578 and what little I found on it, it seems to be a generic 6502 core micro, with ram and rom, intergrated video (undescribed) and a generic gate array (aka small fpga)

Can you post your source on the generic gate array part?

 

 

NOAC that was used in the first model of Atari Flashback is the Novatek NT6578 chip

yea cause a 2600, 7800, NES and a Colby picture frame all have identical hardware originally, that should have been your fist tip off that it is not a from the ground up ASIC repoduction of the original hardware, its a half breed partically recreated in software and gate logic

I was not saying that the Atari Flashback ran VCS/2600 binaries on a NOAC, which seems to be what you thought I meant, given your reference to the 2600 above. The first model Flashback ran NES ports of Atari games, because the games needed to utilize the capabilities of the NOAC. It could not run 2600 code.

 

And speaking of NOAC capabilities, I have attached the datasheet for the SinoWealth SH6578 chip to this post. If you can point out where in the specifications there is a generic gate array (which surely would be mentioned in the manufacturer datasheet, no?), rather than unmodifiable hardware components, I will concede this argument. In the meantime, I will quote the portions of the datasheet that specifically mimic the NES' video and audio systems:

 

Page 1

- "Resolution: 256 x 240 pixels"

- "64 sprites in one frame"

- "64 color palettes out of 53 display colors"

- "Audio: SH6578 supports 3 melody channels, one noise generator, and one PCM voice channel."

 

Page 9

- "Each sprite is composed of 8 x 8 or 8 x 16 pixels"

 

Page 10

- the diagram on this page shows that the 64-color palette (only 53 are usable for TV display, same as the NES PPU) is defined via 2 bits for luminance and 4 bits for color (2 ^ (2+4) = 64)

 

Page 20

- "The waveform of channel 1, 2 are square wave. Channel- 3's is a triangle wave, and channel 4 is a noise generator."

 

If you look at the datasheet itself, you can see specific details of how the SH6578 video and audio systems work, what pins are for what, what registers are for what, etc. I'm not an NES programmer, but I'm confident that all those details are the same as for the NES (minus whatever differences render NOACs not fully compatible).

 

If all the above characteristics were merely gate array logic, why would they be in a manufacturer-published datasheet as hard-defined, unmodifiable attributes? This chip is an ASIC specification, meant for high volume mass production, to be used to run end-user programs, with inputs including joysticks, that output to a television. Those programs happen to NES compatible, and thus, this is a NES-on-a-chip.

 

If the manufacturing volumes I described in my previous post and even the chip's datasheet are not enough to convince you that the NOAC is an ASIC, then please tell me what further proof you require. I'd ask our resident chip designer Curt to step in and discuss this with you, but he seems to be away from AtariAge right now.

 

onmode-ky

SH6578_Spec_V980826.pdf

[Osgeld]

well I never saw the datasheet, and that perticular chip model pops up in a ton of applications, and based on what I did find is what I said, not unlike a cypress PSOC with a 8051 intel core and a large clpd

 

so fine someone actually spent the time to design a already designed generic part from the ground up just to mimic a shitty pirate NES (unlike the more likely option that this specific make and brand has just been mask programmed)

 

grats you win

Edited December 12, 2013 by Osgeld

[JamesD]

Emulation implies running software that imitates the hardware.
While there have been a few videogames that have been recreated in a joystick with a small RISC cpu and some additional hardware, I don't believe I've heard of any NES clones that do that.
My NOAC unit certainly doesn't and it uses a custom ASIC, not an FPGA.

FPGA's were certainly used to develop the chips but they are way too expensive to include in these cheap consoles.

I think part of the reason some people are confused is because FPGAs are programmed in languages and and people see that like software.
The key difference vs software is VHDL and Verilog generate an actual description of the hardware that does not require running on a CPU.
It just skips the tedious steps of designing the hardware one gate at a time by hand. That is generated by software when the code is compiled.

Approximation might be a better word than emulation since the hardware isn't identical but I don't think dictionaries have caught up with this use of either word.
Things like analog filters on the sound aren't practical due to the added cost so designers just try to come close digitally.

[Robot2600]

No NES clones use emulation. Like you say, it's just a tiny NES.

 

Emulation would mean running some sort of OS and running an NES emulator on it like FCEU or something. Nothing does that. NES ports have often done that. The VC games run inside Nintendo's emulator, which is pretty accurate as far as I can tell.

 

Those pre-programmed multi-games (that look like n64 controllers) are running an emulator. It's probably some small custom linux OS with a simple frontend (list of all 64000 games) to load the roms and play them in whatever pirated emulator they use. They don't use carts though.

Edited December 12, 2013 by Robot2600

[MaximRecoil]

How expensive would it be for a company to make a real NES clone, that is, with a real 2A03 processor, 2C02 PPU, and everything? Or, using a real CPU and PPU and doing the rest in an FPGA, like Jrok did with his Mult-William's arcade board (an original 6809 CPU in that case)? Everyone loves Jrok's Multi-Williams board, as it executes game code exactly like the original hardware, unlike emulation or NOACs.

 

If I were rich I'd bring something like that to the market; hire someone like Jrok to create the hardware design; make it a toploader from the getgo, and maybe include RGB output (there are various approaches to doing that).

[CatPix]

Probably not that expensive. such clones were made as early as 1990. The most famous examples are the Micro Genius clones, also sold as the Steepler Dendy in Russia and as the Terminator II in Eastern Europe.

They aren't 100% faithful because they use clones of the CPU and PPU. But they use "real" hardware.

Also noteworthy is the fact that those systems run games at 60 Htz, but they drop it somehow to 50 Htz for PAL/SECAM compability. (make you wonder why they couldn't do that on SNES/Megadrive if a cheapo clone could do it without modifying the video chip)

Also, those names are still used today, but you will get a NOAC clone so one gotta really hunt for the old models (Dendy are the easiest to figure as the "real hardware clones" have a SECAM RF output, whereas NOAC versions have only PAL composite or PAL RF output).

 

[MaximRecoil]

Well that's interesting. I thought all of the bootleg systems used an NOAC. I wonder how good the overall quality was, specifically the quality of the card edge connector and the audio/video output. I know that the original U.S. front-loader NES has excellent quality A/V output (so good that I'm not sure that even RGB would be a noticeable improvement from a typical TV-viewing distance), even its RF output is very good, better than the RF output of an Atari 2600 or 7800 in my experience. The RF output of the official NES toploader sucks though (prominent jailbars), and there is of course no A/V output option.

Edited December 12, 2013 by MaximRecoil

[JamesD]

How expensive would it be for a company to make a real NES clone, that is, with a real 2A03 processor, 2C02 PPU, and everything? Or, using a real CPU and PPU and doing the rest in an FPGA, like Jrok did with his Mult-William's arcade board (an original 6809 CPU in that case)? Everyone loves Jrok's Multi-Williams board, as it executes game code exactly like the original hardware, unlike emulation or NOACs.

 

If I were rich I'd bring something like that to the market; hire someone like Jrok to create the hardware design; make it a toploader from the getgo, and maybe include RGB output (there are various approaches to doing that).

The separate parts would definitely cost more due to the need for a larger board, more drill holes, more parts for assembly, etc.. but it's not a huge price difference if you can use off the shelf parts.

Once you start talking about custom changes or having to produce your own ICs you might as well make a custom ASIC for the entire thing.

If you wanted to play games from an SD card instead of using carts then a custom ASIC for the entire NES would be a lot cheaper. The separate parts version would need a custom ASIC for different mappers anyway.

 

Probably not that expensive. such clones were made as early as 1990. The most famous examples are the Micro Genius clones, also sold as the Steepler Dendy in Russia and as the Terminator II in Eastern Europe.

They aren't 100% faithful because they use clones of the CPU and PPU. But they use "real" hardware.

...

An ASIC is real hardware. An FPGA is real hardware. If by "real" you mean separate parts rather than logic integrated into a single part then yes and they are only as accurate as the clone chips which is the same issue as the FPGA or ASIC.

 

[MaximRecoil]

The separate parts would definitely cost more due to the need for a larger board, more drill holes, more parts for assembly, etc.. but it's not a huge price difference if you can use off the shelf parts.

Once you start talking about custom changes or having to produce your own ICs you might as well make a custom ASIC for the entire thing.

If you wanted to play games from an SD card instead of using carts then a custom ASIC for the entire NES would be a lot cheaper. The separate parts version would need a custom ASIC for different mappers anyway.

 

An ASIC is real hardware. An FPGA is real hardware. If by "real" you mean separate parts rather than logic integrated into a single part then yes and they are only as accurate as the clone chips which is the same issue as the FPGA or ASIC.

So why can't anyone make a NES ASIC that is 100% accurate? Are there unknowns pertaining to the original hardware, or do the NOAC designers simply intentionally take shortcuts?

[+FujiSkunk]

So why can't anyone make a NES ASIC that is 100% accurate? Are there unknowns pertaining to the original hardware, or do the NOAC designers simply intentionally take shortcuts?

 

Money. The existing chips are cheap, and everybody knows how to build a console around them now. Anybody with access to a hardware factory can slap together a NoaC console for relatively little money up-front. That's the same reason a lot of plug-and-play consoles supposedly based on systems like the Atari 2600 and the Intellivision are actually NoaC's in disguise. It actually cost less to hire some programmers to write Nintendo ports of those games than it would have to design new hardware compatible with the original code.

 

Since the original NoaCs first appeared, there has been plenty of time for someone to design a new NoaC with better compatibility. You might think, even with the costs involved, somebody would have given it a shot by now. But really, why re-invent the wheel when the original wheel is "good enough" 99% of the time? And even if someone did come up a better NoaC, all the other knock-off companies would still be flooding the market with stuff based on the old design.

 

Gradually the better retro-catering companies are listening to their customers, but at the same time, bits like general purpose ARM processors have become dirt cheap, so again there is no need to come up with a better NoaC. Just put together an ARM-based console and run the game code through an emulator, and you're good to go, especially if it's a decent emulator. The RetroN 5 reportedly has such a design, though of course Hyperkin isn't saying what exactly is under the hood.

[JamesD]

So why can't anyone make a NES ASIC that is 100% accurate? Are there unknowns pertaining to the original hardware, or do the NOAC designers simply intentionally take shortcuts?

From a technical standpoint, they can. But only Nintendo has the original design for the chips. That means chip timing and any audio filters not obvious from the docs, etc... have to be reverse engineered.

A lot of that has been done by individuals that have created emulators or FPGA based systems but I haven't seen an exact version of the audio... which seems to be the only issue with my NOAC console... so far.

 

From a commercial standpoint, creating a custom ASIC is expensive. The upfront cost is more than what you'd spend for a board that uses off the shelf parts. If you sell millions then things turn out ok. If you sell hundreds or thousands then you have to raise the price or go broke.

In addition, your product will probably get ripped off by the Chinese manufacturer. This is basically what happened with the C64 in a joystick.

You almost have to include some sort of programmable part on the board that can't be ripped off to avoid this and then you are using a CPLD in addition to the ASIC and you might as well go with a lower cost FPGA.

FWIW, the smaller FPGAs have come down enough in price and have enough capacity that using them in a product like this will soon be feasible. It won't be as cheap as an ASIC but the price will be acceptable.

 

 

As for the existing NOAC systems... if I had to guess, a university student published source to an FPGA NES he created and that was ripped off to create the first NOACs. The people that ripped it off apparently aren't smart enough to fix it though or we would have close to perfect units out there by now.

 

 

 

[CatPix]

An ASIC is real hardware. An FPGA is real hardware. If by "real" you mean separate parts rather than logic integrated into a single part then yes and they are only as accurate as the clone chips which is the same issue as the FPGA or ASIC.

I meant "real" as is "produced the same". ASIC and FPGA only "emulate" physically, if you get what I mean.

 

The thing is that the original Nintendo hardware had specific specs, and also issues, that need to be reproduced to emulate the NES perfectly.

Even Nintendo themselves would probably not push that deep if they made a NOAC themselves.

An interesting example is Amstrad and their CPC computers. When first released in 1984, they used a custom "Gate array". To cut costs when the CPc line became obsolete, they replaced the Gate array by an ASIC chip.

Well, this ASIC behave a bit differently fro mthe original IC. Not much, but it does behave differently. And we're talking about an official ASIC, not some aftermarket home made replacement part.

 

Another interesting example is that good old Stella emulator.

If you read through the history of the emulator (it's somewhere on Internet) you can read that at first, the emulation was focusing on getting everything right. Well, now, they focus on also replicating all the things that goes wrong. Because you always get bugs into processors, video chips, etc... And to be accurate, you also got to reproduce those bugs. And some of them are so deeply rooted that it would take only retrogamers to hunt and find them.

 

Now, surely clones chips can have the same issues than the real hardware... Especially for custom parts like the PPU. But the Ricoh processor itself can easily be copied accurately. As well as the RAM chips, all all kind of microchip that can all behave a bit differently than what the specs says.

 

Now, either way, making a NES clone without having access to the original chips specs is an hard way. And probably yeah, ASIC/FPGA is the easiest way to go as you can change parameters easily.

[gamecat80]

Just put together an ARM-based console and run the game code through an emulator, and you're good to go, especially if it's a decent emulator. The RetroN 5 reportedly has such a design, though of course Hyperkin isn't saying what exactly is under the hood.

 

Does anyone know when the RetroN 5 will be released? I have looked around and don't see it for sale anywhere. IMO, the games look great on it!

http://www.youtube.com/watch?v=-fLLIH2lVNg

 

http://www.youtube.com/watch?v=DH_Oa3hZkhw

 

Picture is crystal clear and I like the 'video smoothing' effect; looks better than the original, imo. The NES already has high-res graphics for an 8-bit console. But the RetroN 5's 'video smoothing' makes the graphics/sprites look completely rounded; no 'staircasing' of the sprites' pixels, even up close. Some may prefer the original, but some may like this too - I know I do!

 

Sound seems good and mostly faithful to the original. But I do hear a slight vibrato at the end of the music when you beat a boss on Rescue Rangers on the RetroN 5 (going by the video). There is no vibrato on those sections of Rescue Rangers on the original hardware. I don't mind though, still sounds and looks good.

[godslabrat]

 

Does anyone know when the RetroN 5 will be released? I have looked around and don't see it for sale anywhere. IMO, the games look great on it!

 

It was supposed to be this month, but it's been pushed off until next month. Bad cart connectors, IIRC. A fixable problem, but it'll take a little extra time.