FPGAARCADE minimig compatible board, comments?

[mikej]

The problem is the DVI driver chips I can use will not embed the audio, you can't buy them without a HDMI license. Nothing to stop you wiring the DVI up to a HDMI and using either a separate analog or the digital connection for the sound.

In the future I will look at direct driving the DVI from a Spartan6 with SerDes.

The standard stereo jack works fine, I have it wired up to my amplified PC speakers. Sounds rather nice

As soon as I have finished testing the ARM, memory controller and basic video I'll get some more boards made.

/Mike

[BlackMonk]

I believe DisplayPort offers much of the same benefit as HDMI and it is specifically royalty-free.  So that might be an option, though it sounds like you're already set with the DVI and separate audio out anyway.  You can probably get DP to HDMI or DP to DVI adapters, too.

http://en.wikipedia.org/wiki/DisplayPort
http://www.displayport.org/

[mikej]

Display port would have been nice, although it didn't exist when I started this
/Mike

[jkonstan]

To modify the example WM8729 Low pass filter to run from a single ended positive power supply:

1. Use single ended opp amp such as MC33204 where ground is now used for the -VEE supply for opp amp.
2 Make VCC/2 supply via VCC to 1K/1K divider to Ground & use @ 1uF cap from VCC/2 supply to ground.
3. Ground side of 47K resistor should now connect to VCC/2 supply.
4. Ground side of 4.7K resistor should now connect to VCC/2 supply. Note that this part of filter forms a noninverting amp with a DC gain K=2.
5. Ground side of 680pf connect to VCC/2 supply or keep to ground.

[mikej]

jkonstan, great thanks for the advice.

One small problem I have is the 47KR resistor is on the bottom board and the feed to the expansion board is taken from there. Perhaps if I do a board respin I can take the audio connection directly from the DAC before the cap, but then it is biased around the DAC vmid which I don't have on the top board. mmm

Having said that, I could add another 1K to VCC and 1K to GND at the left hand side of the 1.8KR, and ignore the 47K to GND. I would also need an output capacitor as well after the 51R, or would you not bother?

The DAC maximum output is 0.726V RMS so 1.1V peak. With a gain of 2 we are getting close the the op amp supplies.

I think I need to simulate this.

/Mike

[jkonstan]

This is a Sallen & Key 2nd order VCVS low pass filter.
I have used this for LPF for audio in the past and here is a link for a description of it.

http://www.ecircuitcenter.com/Circuits/opsalkey1/opsalkey1.htm

The transfer function for this filter is interesting.
http://en.wikipedia.org/wiki/Sallen_Key_filter

You would need another series cap after the 51R.

I would simulate it in PSpice or TinaPRO.

[freqmax]

The "filter" we are talking here is it to do the same task as the A500 filter that was directly connected to the power LED on the A500 ?

[mikej]

Not really in this case but same sort of idea. The main reason here is to better drive the cable and whatever is connected to the board, as well as further attenuate any nasty aliasing from the DAC.

The filters on the A500 had a much lower cut-off freq and this could be approximated with a digital filter in the FPGA.

/Mike

[xyzzy]

Will the add-on board be able to handle a 68060 processor and whatever RAM it needs, possibly 1 stick of SDRAM and maybe a fast (UDMA) IDE interface? It would probably need an extra FPGA to coordinate everything though.

I was thinking that this would make for a killer A1200 replacement once an AGA core was available for the main board :-D

[mikej]

Possible, we have gone as far as a 68030.
To be honest, the soft cores are likely to be able to significantly out perform even the 68060 given some time.
Best,
Mike

[freqmax]

matthey mentioned "Superscaler would be super difficult to implement". Softcores is the thing, but they might be a pain to program

However I suspect 68020 can do everything 68030-68060 can, albeit slower.. So it's a good starting point, speed can initially be achieved the Intel way.. more MHz & cache :-D

[mikej]

Exactly. The aim is to get cycle accurate 68000 & 68020 cores. Not sure what the limits are yet but 30-40MHz should be possible.

I think it is unlikely we can clone the 68060 exactly, and I also don't think it is the right way forward. It is much easier to design a software compatible processor with the implementation targeted to the FPGAs architecture.

I am all for adding new video modes etc which can be used easily. Adding new processor instructions is less useful, who is going to code for it now? Anybody fancy doing the compiler backend work?

/Mike

[freqmax]

I agree. New processor instructions would be highly unsupported. But my thought were more like "if that feature had be implemented then we could.."

New video modes would benefit workbench. But no demo/games I suspect.

Btw.. AMD Am7900 Ethernet interface seems to be quite widespread and QEMU has source code to clone it.

[ChaosLord]

@Mike

Here is the deal:
I already have a 32MB A1200 retro machine.  I really need more ram.

If the FPGAarcade machine came with 512 MB ram then I would have to buy it.  Or if I could replace the 64MB ram with my own 512MB.

But as it is... 64MB is not enough of an upgrade for me to justify the cost.

May your bugs be few and your sales be many.:-)

[NovaCoder]

mikej wrote:
Exactly. The aim is to get cycle accurate 68000 & 68020 cores. Not sure what the limits are yet but 30-40MHz should be possible.

I think it is unlikely we can clone the 68060 exactly, and I also don't think it is the right way forward. It is much easier to design a software compatible processor with the implementation targeted to the FPGAs architecture.

I am all for adding new video modes etc which can be used easily. Adding new processor instructions is less useful, who is going to code for it now? Anybody fancy doing the compiler backend work?

/Mike

I would like to see it with just a 'real' 030 + FPU on a daughter board as planned and the extra effort spent on new video modes instead.  We all know that the AGA implementation was a bit of a bodge and was also hampered by slow and limited chipram + a slow blitter and copper.  What I would like to see is something closer to what the AAA chipset would have been (new screen modes, better performance, more chipram....and some chunky VGA displays).

I've already got a classic 1200 running an 030 so to replace it I want something that's going to not only be 100% compatible but also something that's a bit quicker (eg can actually run in 256 colours @ 1024x768 without dying in the ass).

[mikej]

Hi,

The board can be fitted with 128MB of ram. I really thought that would be enough for anybody!

256MByte is possible with a flip mounted chip underneath the top DRAM. This would not normally be mounted so it would not increase cost. I have some spare wires for chip selects. I'll consider this if I have to do a full board respin.

The other possibility as we discussed is memory on the daughter card. The problem here is the IO is 3.3V. You would need a slave memory controller FPGA on the daughter board. I could imagine a 68060, small FPGA and DIMM slot.

Then the main board would be used as the "chipset" and the lower memory as video display buffer.

It's going to be an interesting ride

Best,
Mike.