MiniMig: One Random CRAZY Thought.

[alexh]

Hmmm, an FPGA with an integrated PCIe 1.0 phy eh?... Crap look whats on my desk

[freqmax]

@alex:
Got hit by an Enterpoint product? :-D

[downix]

alexh wrote:
Hmmm, an FPGA with an integrated PCIe 1.0 phy eh?... Crap look whats on my desk

What's your FPGA of choice anyways?  I've had good luck with Altera, but wound up with Xilinx most of the time.

[freqmax]

Just reading a bit on FPGA (Xilinx) and SDRAM & DDR. Which memory type to bet on?, SDRAM memories are available in abundance from old x86-pc ram modules. DDR maybe offers a more future proof path..?

[little]

SDRAM or DDR?

Wouldn't DDR provide more bandwidth just in case someone wants to improve the AGA chipset or implement some kind of RTG graphics to provide 15/16/24 bit modes?

[Belial6]

I would think that whatever is the cheapest and easiest to implement would be the way to go.  It is likely that there will be periodic updates to the MiniMig.  This means that if SDRAM is easier to implement, current MiniMigs can use them.  This will give people plenty of time to work out getting DDR to work.

[freqmax]

Any pipelined dram may actually waste a lot of (dram) clock cycles as the Amiga doesn't really expect such functionality.
So a SDRAM would then receive a read command, wait1, wait2, get data and continue (or something similar).
DDR1 in essence just seems to be about different pipeline depth and voltage.

DDR2 and DDR3 use voltage levels that may be incompatible with the FPGA. At least there's currently no onboard supply to drive them. So EDO, SDRAM, DDR1 seems to be the choice of DRAM in that order.

Sidetrack: Seems SDRAM is tested with 50 pF bus load and each input is max 3,8 pF. So it should be possible to drive an array of 12 SDRAMs with one bus to accomplish a large ramdisc.

@Belial6:
Considering that most pre DDR2 dram is dirt cheap these days. I think the rational choice is to go for the simplest dram to implement.

[CrazyProg]

Good Idea, but as a commercial product very limited. As a hobby type project it could fun.

I have another idea, I was wondering if you got one of the Xilinx chips with a built in PPC, could you make a very small computer capable of running OS4 ?

The minimum of I/O, PS2 keyboard/mouse, VGA video connector, USB, all done by the FPGA.  Use a Compact Flash card as a HDD.  Very small, no noise, and should be fast enough to useful.

There are FPGA packages for the VGA type video, the PS2 keyboard/mouse is simple, I couldn't find a USB implementation for a FPGA (but it should be doable).

Comments!

An open source OS4 compatable computer would be nice.
Would probably need to pay a lincese fee for the OS4 though.
 :-)

[jkonstan]

SDRAM is a move forgiving on the PCB layout than DDR.
DDR with a Xilinx FPGA will work fine; however, PCB layout rules and signal integrity analysis must be followed in order to insure a properly working memory implementation.

 :-)

[little]

Any pipelined dram may actually waste a lot of (dram) clock cycles as the Amiga doesn't really expect such functionality.

Yes and no, isn't the 31 khz modes achieved using the extra bandwidth the installed RAM provides? Back in the day the Picasso IV had to use EDO RAM (45ns) and a 64 bit chipset to have enough bandwidth to display 1600x1280x16; that is the reason I suppose that with faster RAM it would be feasible to implement AGA+ or Picasso96 compatible video modes.

[freqmax]

@jkonstan:
What do I have to do when it comes to pcb layout to make sure SDRAM will work?

Also what DRAM access time does the 7.09379 MHz Minimig demand? (it had four cycles or so?)
The current sdram is 70 ns. But that's maybe faster than needed.

@little:
EDO is the "last" dram that don't use any real pipeline. But it might be harder to get a hold of in the future.

[EDanaII]

To be clear, guys, I'm not talking about a "computer on a card" like the Inside-out. I'm talking about using the lessons learned from Dennis' project to create a graphics card with legacy support.

But, as McVenco points out, there are still some sound issues, so maybe it shouldn't be JUST a graphics card. The point here, however, is to create some kind of card that help supports the legacy issues while supporting all other modern hardware solutions with modern hardware.

So, what I mean is legacy support, but no CPU. The best you can hope for a modern CPU is an m68060 and that running at, maybe, 75mhz. And THAT, even in this day and age, is still not cheap. Let's leave it, as well as any other currently supported hardware options, to modern hardware.

And, of course, who's talking emulation? We are talking about supporting ALL of the currently available options by providing a little legacy support. You need a catweazel to if you want legacy floppy support. Why not a graphics card or audio card for legacy video/audio support?

The modern computer is modular in design. Why shouldn't the current Amiga options also be modular?

And creating a hardware solution that supports ALL of the current options has a better market than creating a minimig which is an island unto itself. (To be sure, I'm not dissing minimig, merely pointing out the advantage of cross solution support. Also, to be clear, I doubt we're talking about a potentially viable market here, but... what the hell. :-) )

Two more cents.

Ed.

[Speelgoedmannetje]

EDanaII wrote:
So far, I've stayed (mostly) quiet about the minimig situation. Don't get me wrong, I think it's fantastic what Dennis has achieved, and I'm curious to see where all this will lead, and I have one thought on this matter that I have no idea has been explored or not, so apologies if it has.

Why a minimig "board?" Or, more specifically, why JUST a minimig running with an m68k supporting only legacy games? Why not a minimig graphics card? In other words, forget keyboards, serial ports and all that jazz and instead create a MODERN graphics card using what minimig has learned to support UAE, AROS, MorphOS and OS4.0? Leave the keyboard, mouse, serial and even USB to UAE and the like, and use minimig to support the legacy software?

Like I said, just a random crazy thought. I'm no expert in this matter, but it seems feasible, and practical, to me to use what we have already today with what minimig can bring tomorrow to make a better Amiga, if only an emulated one.

Two cents.

Ed.

I've been thinking kinda the same thing, but then an USB thingy, with connectors for floppydrive and joystick, and a plugin for winUAE :-)

[AJCopland]

You mean something like an FPGA based graphics card that implemented the necessary parts of the Alice and Lisa (for AGA) or Denise and Agnus (for OCS/ECS) to reproduce the Amigas graphics in hardware.

I think that'd probably be harder than doing the MiniMig as you'd effectively be pulling out and exposing the raw interchip interfaces and then trying to put a clean interface over them so that they could be called from the "other" OS over some kind of bus. PCI/PCIe most likely.

[jkonstan]

freqmax wrote:
@jkonstan:
What do I have to do when it comes to pcb layout to make sure SDRAM will work?

Also what DRAM access time does the 7.09379 MHz Minimig demand? (it had four cycles or so?)
The current sdram is 70 ns. But that's maybe faster than needed.

In order to insure that the SDRAM works, the answer is complex and may be out of scope for a simple post. The proper recipe deals with a static timing analysis & Signal Integrity (PCB stack up, series termination resistors, trace lengths, trace impedances, and a Signal Integrity simulation tool such as HyperLynx from Mentor).

If you really want to know more about this, I can write up more after I finish assembling my MiniMIG1.1.

Also, this book is a good starting point/reference.

http://www.amazon.com/Handbook-Digital-Techniques-High-Speed-Design/dp/013142291X

 :-)

[EDanaII]

@ AJCopland,

You mean something like an FPGA based graphics card that implemented the necessary parts of the Alice and Lisa (for AGA) or Denise and Agnus (for OCS/ECS) to reproduce the Amigas graphics in hardware.

Essentially, yes. :-) But a card designed in FPGA, somebody correct me if I'm wrong, doesn't have to stay in FPGA. Just a thought.

I think that'd probably be harder than doing the MiniMig as you'd effectively be pulling out and exposing the raw interchip interfaces and then trying to put a clean interface over them so that they could be called from the "other" OS over some kind of bus. PCI/PCIe most likely.

Of course. I've never stated otherwise.

Consider, though, that some people on this thread, and elsewhere I'm sure, have been advocating an open-sourced Amiga. What's amusing for me about all this, thanks to Dennis and AROS, is that we already HAVE an open-sourced Amiga. :-) And all I'm doing is advocating one of the directions that it should take.

Ed.