Super Amiga Hobby Project (Just a fantasy ATM)

[MskoDestny]

Someone recently sent me a link to a low-cost Spartan 3 (an FPGA) dev board and it revived an idea I had for a hobby project.  No idea whether or not I'll attempt it (or if I could complete it if I tried), but I think it's a fun idea to dream about.

Basically it would be a Coldfire V4e based computer with a mostly AGA compatible chipset and modern connectivity (PCI, USB, Ethernet).  The AGA-like stuff would be implemented on FPGA and would support the classic modes and sound formats, but would also have updated modes for things like 24-bit color and 16-bit 5.1(6) channel sound.  Nothing too fancy like 3D acceleration or a sound DSP, though I suppose with a PCI card and appropriate drivers you could get those as well.

Ideally it would be able to run Amiga OS 3.X (ROMs and install disks would probably have to be provided by the user) with the help of that 68K emulation and a new 680XX.library and AROS (obviously the Amiga port would have to be revived).

Personally, I think this would be really cool.  It would retain the classic hardware that helped make the Amiga so cool, but would also bring the machine forward enough to be useful for lightweight computer tasks. Obviously it would be a rather foolish commercial venture, but I see this as more of a hobby project than anything.

So I guess the question is, in the unlikely event that I actually do this, would anyone buy one and how cheap would it have to be.

[MskoDestny]

In theory patents could be an issue, but I seem to remember someone saying that the patents on the chipset have since expired.  It's also possible the design would end up being different enough that it wouldn't be considered infringement.  My friend's brother is a patent lawyer so in the unlikely event this goes anywhere I could ask him for advice.

As for Amiga OS copyrights, if I bundled it at all I would buy existing legitimate ROMs and disks or would see if Amiga Inc. would be willing to license it (I kind of doubt it, but if they're desparate for cash...).

I doubt I could do it for 150 euros.  In low quantities (40+) the 266MHz V4e chips cost around $40 US.  An appropriate FPGA would probably be in the neighborhood of $50 (not really sure on this one, I haven't done any FPGA work, just CPLDs so I'm not sure how big of an FPGA this will need).  Add in board manufacturing and I might be able to sell one for $300, but that's just a guess.

[MskoDestny]

Depending on how many gates/cells it would require a Soft-core 68K might not be a bad idea for a compatibility mode, but I don't think it makes much sense for a main CPU.  At the speeds I'd likely to be able to get I would probably just be better off with a 40MHz EC030 (~$60 from Digikey).  Doesn't look like they have a 68000 at opencores anyway, just a 6800.

For me, the XScale + off-the-shelf solution would be cheating and not really a "real" Amiga anymore.  Still, not a bad idea to keep in mind.

I think if I do attempt this I'll start with a Spartan 2 in a PCI form factor (fpga4fun sells a $99 board in this config) and work on AGA compatible video hardware.  I could then modify UAE to send video writes to this card for testing.  That way I can break down the system into pieces for testing.

[MskoDestny]

Ah, okay. I didn't see it because I didn't turn off the filter that prevents beta cores from being shown.

It looks like it's relatively small.  If I use a large enough Spartan 3 (maybe the 1000 part) and I'm thrifty with the AGA stuff I could probably fit it in.  Definately a possibility for a compatability boot mode.  If it doesn't fit, I suppose I could make a second core with just a standard AGA or even ECS to save space and just add an option at boot to choose a core.

[MskoDestny]

Well I don't think testing the bits and pieces in an Amiga would be the most practical route for me.  It gets me involved in more of a mess timing-wise and requires a logic analyzer much earlier in the process.  If I test bits of pieces by attaching the FPGA to a PC and use a modified version of UAE, I can debug problems using simple testing software and an FPGA simulator.

I'm definately going to break it down into small pieces though.  That's a lesson I've learned from application programming.

I'm willing to invest in a logic analyzer and scope when it becomes necessary (fortunately these are becoming available on ebay for reasonable prices), but I'm going to try and avoid that at least until I get the first prototype with the Coldfire and FPGA working together.

It is probably too much work for one person, which is why I think it unlikely I'll ever complete this project.  I might try to rope someone else in at some point, but I think it will be difficult unless I can find someone locally (since there would need to be a lot of duplicated equipment otherwise) and currently I don't have any local friends who are interested in stuff this low level.

We'll see how it goes.  I have to wait until I get a new job to start on the hardware (interview this Friday, here's hoping), but I might take a look at the AROS source again tonight and see just how much of a pain it will be to get a basic Amiga port going again.

EDIT:

Since you're at it why not finish the AAA chipset instead and make it AGA compatible. Then you could implement it in FPGA in a PCI card so that any future Amiga could have classic compatibility. I think like that would make it much more comercially viable, just look at the Classic Amiga sales on eBay.

Is there any real solid information on the operation of the AAA chipset other than basic specs?  Even so, the AAA chipset was designed to work with classic 68Ks and all of those are rather slow compared to the Coldfire and the faster ones (040 or 060) are relatively expensive in comparison.

[MskoDestny]

Finances will not be a problem if this interview goes well.

Again, this is really just a hobby project.  If I do it, I'll be doing it to have fun and that will be my measure of success.  Most of the equipment I'd need for testing I was planning on buying eventually anyway.  Only reason I plan on selling them because I'd like to share my work with the Amiga community and I can't do that for free.

[MskoDestny]

Hyperspeed wrote:
What if someone wants a repair? You could get into a DCE situation where they struggled to find parts for the Cyberstorm PPC cards and you could have a linch mob on your hands!

V4e coldfires and Spartan 3 FPGAs should be available for a long time.  Besides unless it looks like these things are going to be fairly popular (at least >100 units) I doubt I'll sell these with a warranty (or at least not a very long one).

Don't you think Eyetech have the motherboard market wrapped up now though with the distinctive AmigaOne brand and also there are Pegasos computers too!

There seem to be a fair number of people who don't see these as "real" Amigas since the hardware isn't anything special.  I would imagine the primary market for this device would be among those people, but once again this isn't designed to be a financial success, just a fun project.

Not many people understand what Aros is, myself included. Sounds a bit like DOpus Magellan to me - a sort of AmigaOS compatible replacement.

AROS is to Amiga OS as Linux is to Unix as FreeDOS is to MS DOS as ReactOS is to Windows etc.  It's a clone.  It's written from scratch to mimic Amiga OS.  Also, this hardware should run Amiga OS 3.X with the appropriate ROMs and install disks.  No AROS required.

01) GFX card for desktop A1200

I would think that a Mediator+PCI card would do better than I could.

02) AGA compatible laptop

If the desktop version sells well I'll consider converting it into a laptop form factor.  It will be a relatively low-power machine so it won't take much jiggering to make a laptop with the same basic design.

03) Dual monitor capable PCI/Zorro GFX card

Such PCI cards exist, it's just a matter of hacking OS 3.X to support multiple monitors and writing a driver.

05) Video digitiser with hi-speed Teletext saving

Don't think I can help you there as I live in the US and the Television standards are different and I don't have any PAL equipment to test on.

06) 24-Bit capable scandoubler/flicker-fixer

Aren't there already a couple (admittedly hard to find) cards that can do this?

08) Coldfire accelerator with integrated 3D GFX accelerator

Supposedly the folks over at Elbox are working on this one...

09) A tablet Amiga

See 2.

10) An Amiga TV-Games style joystick with 2-player support

If the softcore fits along with the chipset I could theoretically make one of these too, but getting all those old games to work would probably be a pain because of timing issues.

[MskoDestny]

Tahoe wrote:
Depending on the degree of compatibility with the original platform I would be VERY interested in one, and maybe some Amiga stores would be too.
Here in Amsterdam I met a few people that just recently bought a brand new A1200, which should say somthing about the demand at the moment...

My goal is for most software that would run on an 040 A4000 would run on this machine.  There will be some code that will break on the ColdFire because not everything can be emulated via trapping unimplemented instructions,(flags not being set on certain multiply instructions for instance) but it doesn't look like that should kill a huge number of applications.  I hope to increase compatability with either (or maybe both) a softcore 68K in the FPGA or a modified version of UAE which can pass hardware access to the chipset in the machine and thunk OS calls to the OS running outside the emulation environment.

Any external scandoubler should be 24-bit capable (being an analog device), but I'm not 100% sure about that, so please correct me if I'm wrong.

I was under the impression that most external scandoublers worked by digitizing the analog signal into a framebuffer and then spitting the digitized image back out through a DAC at VGA timings.  At least that's how I seem to remember some little homebrew one working.  If that's the case, it really depends on the ADCs and DACs used. Of course, I could be wrong too.

[MskoDestny]

MuLTiViEW wrote:
 I would take an A1000 motherboard,  measure it,  and re-design an AGA (or even a simple ECS version) and try and somehow dual processor a pair of 060's right on the motherboard.   (I mean actually create a brand new mobo replacement that fits exactly in the A1000)

 Maybe even add PPC somehow.   This coldfire CPU is also interesting.  If you really want to do a hobby puter.  Add a 3rd different type of CPU to the mix like that and come up with some new "tWisTeD AmIGA"  in a 1000.

Let's not get carried away now...
A single 266MHz V4e Coldfire is probably faster than 2 060s anyway at least when running code compiled for ColdFire (not sure how much overhead implementing the missing instructions adds or how frequently those instructions are used).