minimig 4000

[A6000]

I would like to see some of the talented designers on this forum design a replacement motherboard for the amiga 4000D.
Many systems are being eaten by battery and capacitor ooze, and will not last another 15 years.

The board should have a minimig FPGA to replace the AGA chipset, it would start with ECS but later could be reprogrammed for AGA, AAA or HDTV feature set, with akiko chunky to planar conversion.
There would be a socket to take either an 040 or 060 with a jumper to select 3.3v or 5v cpu voltage.The clock speed could start at 25mhz and be raised through the bios, there could be sixteen speeds ranging from 25 to 120mhz for easy overclocking.
The board could use DDR2 ram - 2GB - all chip ram.
And I suppose we might as well put a cheap PPC on it.
The Walker used a Multi I/O chip which could simplify the circuit, but the unreliable CIA's have been incorporated in all amigas since the A1000 are they indispensible because of code in Kickstart?
The hard drive interface should be either ultra ATA/133 or SATA 2/300 either would be faster than the old scsi drives we may be using now, and an SCSI controller could be plugged into the zorro bus if necessary.
Eventually there could be replacement mainboards for all big box amigas. If we were to produce an ATX board our systems would look the same as any "dead common" pc, and I don't think any of us wants that! :-)

[Akiko]

Will you be financing this ambitious creation?

[A6000]

This would be a volountary project, a "hobby" if you like.

[A6000]

I am sure that anyone who thinks they can make a profit  supplying bare or ready bult boards will put in whatever money is needed.

[Flashlab]

Sounds nice! But I'm afraid that it's a bit unrealistic. Let's just start with AGA and a faster processor?

[AJCopland]

I don't necessariyl disagree with your ambition but I think that you're going to need to put some effort in yourself if you want to get this started.

Firstly however a couple of things. SDRAM PC100 or PC133 is seemingly much easier to implement and is on my own list of things todo right after "build my first MiniMig v1.1". DDR and DDR2 has been mooted as difficult for hobbyists due to timing constraints. So for now I'd stick with SDRAM, its cheap and plenty fast enough for MiniMig.

Forget sockets and just go straight for an '060 if you're wanting to upgrade the CPU. It makes the design easier as its running at the same 3.3v as the FPGA so theres no level shifting.

CIA's are, as far as I know, replicated within the FPGA using Verilog. No need for anything external.

To support AGA and '060 you're gonna need a 32bit data bus, might as well support a 32bit address bus at the same time. That means quite a few more pins available on the FGPA so you'll be going to BGA based FPGA.

ATA/100/133 IDE is nothing I've heard talked about so I don't know how much harder it'd be than any other IDE mode.

Zorro support. I have not a clue :-D

Basically I think that if you want to see this become a reality then you should break it down into the tasks required for each stage. Once you've done that pick the first one that you feel is crucial and look into the parts and things that you'd need to change to make it happen! You don't need any experience to that bit but it'll mean that when you come back and need experienced help you've done most of the legwork!

Good luck.

Andy

[A6000]

AJCopland wrote:
Firstly however a couple of things. SDRAM PC100 or PC133 is seemingly much easier to implement and is on my own list of things todo right after "build my first MiniMig v1.1". DDR and DDR2 has been mooted as difficult for hobbyists due to timing constraints. So for now I'd stick with SDRAM, its cheap and plenty fast enough for MiniMig.

Forget sockets and just go straight for an '060 if you're wanting to upgrade the CPU. It makes the design easier as its running at the same 3.3v as the FPGA so theres no level shifting.

CIA's are, as far as I know, replicated within the FPGA using Verilog. No need for anything external.

To support AGA and '060 you're gonna need a 32bit data bus, might as well support a 32bit address bus at the same time. That means quite a few more pins available on the FGPA so you'll be going to BGA based FPGA.

ATA/100/133 IDE is nothing I've heard talked about so I don't know how much harder it'd be than any other IDE mode.

Zorro support. I have not a clue :-D

Basically I think that if you want to see this become a reality then you should break it down into the tasks required for each stage. Once you've done that pick the first one that you feel is crucial and look into the parts and things that you'd need to change to make it happen! You don't need any experience to that bit but it'll mean that when you come back and need experienced help you've done most of the legwork!

Good luck.

Andy

Well I will admit that this task is beyond me, but well within the capabilities of others on this forum. but I would start with the circuit and pcb design of the 4000D and:-
1. replace fast slot with cpu socket, this can accept 040 or 060 since there will be more 040's available than 060's. the programmable clock speed allows any speed version to be used.
2. replace AGA chipset with minimig FPGA.
3. implement memory, SDR is old and expensive, DDR2 may be more economical.
4. implement hard drive interface.

Job done?

P.S The PPC can go on a board plugged into the cpu socket.

[alexh]

I am sure that anyone who thinks they can make a profit supplying bare or ready bult boards will put in whatever money is needed.

There is no large scale profit, which is why no-one is mass producing MiniMigs.

I'll do it for £15k upfront, which will be paid back on first sign of a profit.

They still make both 040 and 060. I've bought E41J 060's before at €75 each. No-one will care about 040.

I know the 060 is 3.3v core voltage but I think it still has 5v I/O

Freescale  Data

SDR is still cheap and sold in very high quantities as it's in everything from DVD players to washing machines.

Personally I would prefer MegaMig. A MiniMig form factor with an 060.

[HenryCase]

Akiko wrote:
Will you be financing this ambitious creation?

This project isn't ambitious enough for the amount of work necessary. I do agree that if A6000 wants to get this started he should put up some of the funds though.

A6000, the project you want isn't best served with a Minimig-based solution, you'd be much better off with Clone-A.

http://www.amigahistory.co.uk/clone-a.html

Only mentions OCS in that link, but AFAIK work on AGA is well under way.

Way I see it is Clone-A and Minimig complement each other. Clone-A is about perfect reproduction of the past, whilst Minimig (+AROS) gives an opportunity to take complete control of the future.

[FrenchShark]

Let's call it Maximig :-)

Seriously, my remarks:

- 2GB of Chip RAM will break compatibility, the Amiga architecture is limited to 8 MB of Chip RAM due to the memory map.
- To limit the the number of I/Os on the FPGA you can use the 040/060 in multiplexed bus mode (32 wires for address and data instead of 64).
- A coldifre is less expensive than a 040/060 and you can achieve a 100% accurate emulation (not the illegal instruction hack) with an average of 20 CPU cycles per instructions.
- A drop-in replacement of an A4000 board will be very expensive because of the board size, most of the PCB will be empty since the FPGA can integrate most of the A4000 components.
- Using a off-the-shelf evaluation board is less expensive: Altera is selling its NiosII evaluation kit WITHOUT the software for $395. The Altera FAE told me that the new Cyclone III NiosII evaluation kit should be available for December. It should integrate a VGA output. The price is still $395.

[downix]

Ok, fine, my 2 bits:

You must crawl before you can walk.  I am working on an A500-like expantion bus for my MiniMig, you'll need that before even trying Zorro III.  You'll need ECS before you get AGA, so someone needs to make ECS.  Come on guy, be realistic.  You have to take these steps first.

[A6000]

downix wrote:

Ok, fine, my 2 bits:

You must crawl before you can walk.  I am working on an A500-like expantion bus for my MiniMig, you'll need that before even trying Zorro III.  You'll need ECS before you get AGA, so someone needs to make ECS.  Come on guy, be realistic.  You have to take these steps first.

ECS is in the minimig FPGA which is used in place of the AGA chipset, it may be possible to reprogram this later to be AGA as I said in my earlier posts.

[downix]

A6000 wrote:

downix wrote:

Ok, fine, my 2 bits:

You must crawl before you can walk.  I am working on an A500-like expantion bus for my MiniMig, you'll need that before even trying Zorro III.  You'll need ECS before you get AGA, so someone needs to make ECS.  Come on guy, be realistic.  You have to take these steps first.

ECS is in the minimig FPGA which is used in place of the AGA chipset, it may be possible to reprogram this later to be AGA as I said in my earlier posts.

Only if you supply an FPGA large enough.  The current FPGA is not. How about we get an A2000 replacement going first, then let's work on the next step?

[alexh]

FrenchShark wrote:
- A coldifre is less expensive than a 040/060 and you can achieve a 100% accurate emulation (not the illegal instruction hack) with an average of 20 CPU cycles per instructions.

Yeah right... NOT!

Emulation is good in a closed environment, as soon as you introduce external inputs, interrupts, exceptions etc. the house of cards comes tumbling down.

If it was possible and the speed was anywhere near a 50MHz 060 we'd have Dragon(?) boards by now.

[alexh]

A6000 wrote:
ECS is in the minimig FPGA which is used in place of the AGA chipset, it may be possible to reprogram this later to be AGA as I said in my earlier posts.

MiniMig is not ECS, it is currently a slightly buggy OCS. (Still great work though, a major achievement!)