FPGA Replay Board

[billt]

They are shipping now, just slowly as I work through the pre-orders.
It takes a lot of time to hand assemble and test each board.
Now I have FCC and CE approval I'm going ahead with mass production.
/Mike

How does one go about getting a product like this approved by them? And does that get you worldwide certification, or are there other things to be done for some parts of the world?

[billt]

I received yesterday a box of brand-new genuine MC68060RC50s ... but they were the 1st mask set so I have rejected them.

What happens if you reject them? Do they take them back and refund you, or do you have some unrefunded trash as part of your total costs?

Once we get the bigger softcore 680x0 going, I can imagine this problem going away, puttin gan FPGA on a small PCB with 680x0 PGA pins on the bottom to plug into any board with such a socket. Not sure if it would be relevant for FPGA Arcade at that point, put the softcore inside the mainboard FPGA, but interesting for others perhaps.

[billt]

I wish the softcore could be used for an super fast 020 compatible FPGA on an accelerator card for classic Amigas. No need for PPC

Understand that softcore 68k in FPGA will never compete with high-end PPC or ARM or x86. It should be somewhat better than Motorola 040 or 060 in modern FPGAs, but that's the best we'll hope for. Not a 2GHz 680x0.

[billt]

There are plenty of FPGA development boards that let you load the core over USB from a pc. It's a matter of using a controller cpu with a USB port.

Some FPGAs allow you to do partial reconfiguration of then on the fly. Use the PCAP or ICAP interfaces (I think Spartan3 is external so needs kindof connected to itself, Spartan6 has this internal) Make a controller capable of sending a selected reconfig bitstream over that interface and you should be good, and that controller should be able to live inside the FPGA. It's been a while since I've looked into that, but it made sense a few years ago.

I'd imagined a PCI (or Zorro) card that could be various things at different times, pick a partial-bitstream to send to the card and have it reconfigure itself to that. Think hardware codecs for mp3, avi, mpeg2, mpeg4, fivx, etc. that would change out based on what file you were playing. I'm finally very recently learning how to do such things, but not sure what I'll actually end up tinkering with.

http://www.cmpe.boun.edu.tr/caslab/publications/selfReconf_final.pdf
http://www.mn.uio.no/ifi/english/research/projects/cosrecos/publications/paper/fpt10koch_demo.pdf

http://electronics.stackexchange.com/questions/34524/loading-an-fpga-image-with-selectmap

[billt]

WHOA !! New drop-in replacement 68k softcore

You'd have to test to compare to see which is "better". but I'd be surprised if you could afford a commercial core, let alone figure out how to be allowed to put it into a GPL project.

http://www.digitalcoredesign.com/ipcore/101/d68000/#info

There's other cores out there to play with as well. I suggest trying the one in the Suska atari project. It's free and GPL like Minimig is.

[billt]

They've said they've tested the core in an A500 ... so would it be possible for them or somebody else to take that core and put it on a suitable FPGA that is mounted onto a PCB that itself can be mounted into the 68000 socket? (A600 is a bit more complex)

Bam! Super-fast 68000 ... still, it's not as interesting as a 68020, but I guess they're working on that as well.

If they tested it in an A500 then either someone there is an old Amiga fan or they found one cheap while looking for some old thing to plug a 68000 into for testing. And either they did make a special board for that purpose with an FPGA on it that plugs into socket, or they ran a cable from socket to some standard FPGA board similar to this, and Shadowfire there is essentially doing what you speak of. Neat.

I'm sure you could buy their IP and make an FPGA board to put it into an A500 socket. Majsta is trying to do essentially the same thing for his A600 using the TG68. I'd have used the Suska's 68000 core if I was doing such a thing as a starting point, as it's already much closer to the 68000 socket. I understand that is being improved in TG68 and TG68 is also getting 020 and I think 32bit bus. More neat.

EAB's Robinsonb5 is also essentially doing the same thing.

[billt]

i remember majsta on natami forum saying hes gonna look at alternative cores suspecting they might be a better hoice. what cores, he didnt mention.

I haven't felt convinced that he ever did look elsewhere though. I really wanted him to look into some simulations of the TG68 he was looking at and compare that with simulations of the Suska's 68k block to help him understand what was so weird about the TG68 and help him figure out what to wrap around TG68 to make it work in a normal 68000 socket. I don't think I succeeded in getting him to run simulations either, or in convincing him of their utility or importance. (I'm a verification engineer of ARM SoC chips, and I think that part is tremendously important and useful for what he's trying to do so blindly) And he seemed to me that he absolutely must do it with TG68 and not anything else for some reason. I tried suggesting that he work with the already much closer to what he needed Suska 68k, see that work better much sooner to help him debug his board design, then go back to TG68 if he must, but he seemed adamant that he must complete his design using TG68 and must avoid others for some reason.

I very much like his idea, and look forward to his results, as well as whatever comes from the other guys.

[billt]

http://www.natami.net/knowledge.php?b=6¬e=32232&x=2:

sounds somewhat weird, and he doesnt seem very experienced, but looks like he actually wanted to look for alternatives. i think him being not a professional he might try to figure his way around that doesnt look reasonable to you, thats frequent with autodidacts..

The ao68000 he mentions is not on a 68000 bus, it's a direct to wishbone bus processor, so that would have required a bridge and thus a lot more work. (I think that porting its companion aoocs project to FPGA Arcade would be fun too) Yea, he sounds like a hobbyist, no problem with a non-engineer trying to learn to do engineering work. We all had to start somewhere...

[billt]

The FPGA on the fpgaarcade is a Xilinx Spartan 3. You can get Xilinx's design software free from them. Either download it... but have lots of time, it's over 4GB, or get a free DVD, which also takes time. ;-)
XESS - makes some Spartan-based development boards. They have some free tutorials on their web site.

The WebPack version of ISE is free cost, but you do need to sign up for a license to be able to use it. Just so people aren't confused at the end of install and first run, the free WebPack license is needed at that point.

[billt]

@all,

can we please take the emulation/not emulation argument elsewhere? Perhaps the FPGA for Dummies thread where it has already overtaken more useful aspects of the topic, so we can keep this thread about the Replay product?

http://www.amiga.org/forums/showthread.php?t=59952&highlight=fpga+dummies

[billt]

Why are you so persistent to ruin it for us. We want implementation not emulation. I will not pay money for emulator in a hardware, please don't ruin FPGA for me! I want a reimplemented upgraded new technology to the old technology of the Amiga classic. FPGA Replay is doing that for me so don't enforce "emulation" into this. Let me enjoy it. Please.

One thing to consider, software "emulation" requires parsing, understanding, translating, and execution of the results of that translation. This FPGA stuff is a circuit. The Verilog or VHDL is a circuit design, and then the FPGA IS that circuit. There is no translation  involved of the Amiga application or operating system you are running.

Some FPGA experts wish the FPGA inventors had used the term "configurable" rather than "programmable" for an FCGA instead of an FPGA, but that's all completed history.

Now, I don't understand why it's so important to decide to buy somethign or not based on a single word applies or not. Silly. So is the endurance of argumentation over such a trivial and unimportant word. The people who are adamant will not have their minds changed, let them believe they are right when you know better. (That goes for all of you, on both sides of the fence) Let's get back to productive, interesting discussions that have a point to them.

[billt]

FPGA emulation - using a multitude of look-up tables and other basic logic elements embedded in a programmable integrated circuit to implement high-level logic algorithms.

 * parallel by nature, real-time processing of input & output
 * compact, most bus connections are internal
 * recompilation needed for use on a different host platform
 * host platform can be expensive

Just a couple clarifications there:

it's not real-time processing of inputs & outputs, it's real-time propogation. There's no "processing" going on.

You don't recompile, you resynthesize to another FPGA architecture, or to another "size" of the "same architecture".

I'm not sure about compactness. The way an FPGA is made to be generic, and (re)configurable, makes it kindof big compared to the ASIC equivalent die space for the given user logic circuit. I'm not sure what you're comparing with.