User avatar
kerobaros

Posted Thu May 21, 2020 8:26 pm

It occurs to me. If we have ...is there any reason we couldn't design an FPGA-based, entirely-new Amiga 1200? I mean, sure, it would probably cost a fair bit, but something like the Ultimate 64 in a brand new case with a brand new mechanical keyboard would be incredibly attractive. Am I crazy?

User avatar
nonarkitten

Posted Thu May 21, 2020 8:36 pm

I wouldn't call MiSTer a faithful replica of the AGA chipset. Getting ALL the schematics of the chips would be the first step. I have Alice and I've shared it with some already. Does anyone have Lisa/Denise and Paula?

User avatar
intric8
Seattle, WA, USA
Website

Posted Fri May 22, 2020 10:02 am

Have you seen the Unamiga 1200? It started shipping this month.

User avatar
nonarkitten

Posted Fri May 22, 2020 10:41 am

intric8 wrote:
Fri May 22, 2020 10:02 am
Have you seen the Unamiga 1200? It started shipping this month.
Not dissing the incredible amount of work in recreating the Amiga 1200 motherboard, but not all of us are either skilled enough or willing to solder on a few thousand pins of surface mount chips. The Re-Amiga (and all bare PCB recreations) really only suit a very small percentage of users.

Edit: Unless you're referring to Edu Arana's excellent Unamiga 500 adapter. Sadly it's sold out and not A1200-case compatible. It shouldn't be too hard thought to wiggle some things around for it to be so.

To my knowledge there are only two sets of cores that are based on actual per-chip reverse engineering. The cores that Jens has and sells as part of his Indivision ECS and AGA products (and the defunct Clone-A project). This is only Denise/Lisa and it's not known if they ever completely reverse engineered Agnus and Paula.

The second is FrenchShark's efforts for the MCC-216 project. I've uploaded the source he released onto my github. You can tell they're MUCH closer than MiniMig because they're a fraction of the size. The original Amiga engineers were very miserly with gates and the logic on these chips is really mind numbing. But even FrenchShark's designs were missing a FEW features and had to break perfect compatibility because the original chips are a) NMOS based and b) used a lot of multi-phased clocks -- neither of which translate well to FPGA.

MiniMig and the Vampire are both reimplementations based on "best effort" forward-engineering from the schematics, hardware reference manuals and testing with software. In some cases, they've leveraged work from Toni for verification. In both cases, they lack a LOT of critical features of the chipset. At no point could either of these serve as drop-in replacements to the real chips.

Of the top of my head, the big issues on MiniMig are the lack of productivity modes and superhires. The Vampire also has this limitation along with broken interlacing that can LITERALLY DESTROY YOUR LCD.

I believe Jeri has the complete schematics of OCS Denise and I have some incomplete schematics from Agnus/Alice. Not sure if there are copies of Paula around, but out of the three chips, Paula was the best documented and it's relatively well understood.

MikeJ also has the complete NETLIST from the decapsulated AGA chipset. However, it's still going to be a ton of work to generate gates and logic from that.

User avatar
nonarkitten

Posted Fri May 22, 2020 11:09 am

The second problem with a new Amiga 1200 is sourcing 5V parts. Ideally you'd want to redo the Amiga 1200 using 3.3V logic that is still readily available (especially the FPGAs). There are 5V FPGAs from Microchip, the AT40K, but the largest is the AT40K40 which only has 50,000 gates (40K optimally). For comparison, the MiniMig OCS used about 82% of a 400K gate Xilinx for the complete chipset. Now, FrenchShark's implementation should fit, but the other problem is that these old 5V FPGA's are **EXPENSIVE**.

If we go the other way and make it 3.3V then we're limited on our CPU choices. Only the 68SEC000 and the 68060 can run at 3.3V unless you're willing to port AmigaOS to the ColdFire CPU. That being said, 68060's (especially the LC and EC variants) are still widely available, as are the 68SEC000's which can also overclock comfortably to around 50MHz.

I suppose you could also just make it MOSTLY 3.3V and use a FET bus switch to the CPU. The 68EC020 is super available, there are tens of thousands in stock at Rochester Electronics at what I'd consider a reasonably fair price. Not sure how overclockable these are, but 28MHz seems pretty common. Keep in mind that with a 24-bit FET bus, you'd need four chips for all the IO. Which I guess isn't too bad.

User avatar
kerobaros

Posted Fri May 22, 2020 7:41 pm

I admit a complete lack of experience with these things, but why couldn't you put the CPU in the FPGA as well? My first thought, before I was told earlier in this thread of Mister's relative incompatibility, was just a board with an FPGA in the middle, comprising the CPU and the Amiga chipset, with some external SDRAM and whatever else necessary to interface with the outside world. A CF or SD slot, a keyboard interface, an HDMI output (and an RGB output too, if we're dreaming) and maybe, maybe a floppy interface, if you've got enough pins left over on the FPGA to drive it. Probably not though.

User avatar
nonarkitten

Posted Mon May 25, 2020 7:10 am

kerobaros wrote:
Fri May 22, 2020 7:41 pm
I admit a complete lack of experience with these things, but why couldn't you put the CPU in the FPGA as well? My first thought, before I was told earlier in this thread of Mister's relative incompatibility, was just a board with an FPGA in the middle, comprising the CPU and the Amiga chipset, with some external SDRAM and whatever else necessary to interface with the outside world. A CF or SD slot, a keyboard interface, an HDMI output (and an RGB output too, if we're dreaming) and maybe, maybe a floppy interface, if you've got enough pins left over on the FPGA to drive it. Probably not though.
What you're describing is MiSTer. Or, if you really want more classic IO, FPGA Arcade. And the appetite for such an "Amiga" has been traditionally quite small. Using FPGA as an accelerator has been a little more accepted, but not wide spread. Getting AGA or the 68K exact is the problem and to do so, we need the original schematics. You'll never get those for the 68K since they're still manufactured by NXP -- so your best option is still real silicon.

User avatar
nonarkitten

Posted Mon May 25, 2020 7:13 am

There's also the "UnAmiga" based on a considerably cheaper FPGA you can get from AliExpress. Edu Arana has also made an IO expander for it that give you an Amiga 500 compatible back side, but still no floppy header. I don't see why adding floppy would be especially hard. I know FrenchShark wrote a quick-and-dirty floppy implementation for Paula in the MCC-216, but I don't know if this was ever totally finished.





Return to “Hardware”