1284P addon

[Harty123]

... about the purchase oft ATMEGA1284(P): I will ask my AVR distributer from the company were I work for samples and prices... and a SMD assembly of such DIP40 board is also not really a problem!

-Harty

[uze6666]

P.S. I'm not to familar with the 1284. Do we need a recompilation of all existing games for this Controller or do we have only to set a special fuse (like in the ATMEGA8515 for the former AT90S8515) ?

From what I know, it's backward binary compatible with the 644 at 100%. I programmed HEX compiled for the 644 and they worked fine on the 1284.

[martinsustek]

Atmel has single datasheet for 164, 324, 644 and 1284, so only difference is memory sizes, I think, and codes are compatible (except for memory size, of course).

Step-up converter is a good idea, if it will be cheap enough, had enough power and stability, and will not make a lot of interference to audio/video signal. I don't have experience with them, but one my switching power adapter makes enough interference to disturb ADSL modem in another room.

Diode hack is a good idea, but is Euzebox-specific and influences more parts. Personally I think, that bringing one wire from before stabilizator to on-adapter stabilizator is better idea.

What about controllers? Have anybody tested them with bigger voltage? Or maybe we can use 5v zener diode on chip output to joystick (and other parts where we can't compute proper series resistor).

[Harty123]

Atmel has single datasheet for 164, 324, 644 and 1284, so only difference is memory sizes, I think, and codes are compatible (except for memory size, of course).

Step-up converter is a good idea, if it will be cheap enough, had enough power and stability, and will not make a lot of interference to audio/video signal. I don't have experience with them, but one my switching power adapter makes enough interference to disturb ADSL modem in another room.

Diode hack is a good idea, but is Euzebox-specific and influences more parts. Personally I think, that bringing one wire from before stabilizator to on-adapter stabilizator is better idea.

What about controllers? Have anybody tested them with bigger voltage? Or maybe we can use 5v zener diode on chip output to joystick (and other parts where we can't compute proper series resistor).

In the EUzebox we have - except of the resistor network for the RGB signals and for the SD-Card - no other influenced parts. For the 6V we have also a very simple solution:

http://de.farnell.com/jsp/search/produc ... KU=1087087

In the original controllers Nintendo used a CD4021. This IC have 15V Vmax. So I can't see a problem for the controllers.

Another point is the ISP connector. I'm not sure that AVR ISP MKII can handel 6V?!

-Harty

[martinsustek]

Yes, in case of EUzebox it shouldn't be problem.
I have original controllers, if someone build custom controllers, probably used similar ICs, to 15V.
It's normal stabilizator, not step-up, isn't it? Maybe we should use some trimmer-configurable stabilizator, to be able to set voltage high enough for overclocked chip to work, but not too much higher.
I don't know how about ISP, because I don't use it. I've programmed my ATMega in external programmer (VUSBtiny I guess), I've built (using 5V from USB).

[Harty123]

P.S. I'm not to familar with the 1284. Do we need a recompilation of all existing games for this Controller or do we have only to set a special fuse (like in the ATMEGA8515 for the former AT90S8515) ?

From what I know, it's backward binary compatible with the 644 at 100%. I programmed HEX compiled for the 644 and they worked fine on the 1284.

Fine! In this case we have only to recompile (& change) the bootloader.

It's normal stabilizator, not step-up, isn't it? Maybe we should use some trimmer-configurable stabilizator, to be able to set voltage high enough for overclocked chip to work, but not too much higher.
I don't know how about ISP, because I don't use it. I've programmed my ATMega in external programmer (VUSBtiny I guess), I've built (using 5V from USB).

Yes this is a very important point - normally such 7806 regulators have only 5% acuracy.

I will test the 1284 in my "Meine WII" DIY Uzebox with a LM317 voltage regulator. I will report my results after this test...

-Harty

[mapes]

Out of curiousity, with the additional PROGMEM and large bump in RAM, what will change?

For instance, will I still be stuck with only 5 sprites per scanline, and 20 sprites on the screen at one time for mode 3? Or will I get double the sprites? The clock isn't changing, and since the architecture is the same, there will be no speed increase that I can see.

What will I be able to really do that I couldn't with the 644? I'd like a good comparison of maximum useage between the two.

For example: I think the Mega Bomber probably takes the most RAM and PROGMEM for the 644 (of all the games/demos I've seen).
But for the 1284, what would be a game that would stretch its capabilities if it isn't in the graphics department? All I can think of is a new Dragon Warrior (IE DW 1 to DW 2 to DW 3, etc...) RPG game (which currently we have none in the community... probably because they take a LOT of work to design and a compelling story to make it fun from start to finish).

Any other comparisons?

-Mapes

[DaveyPocket]

P.S. I'm not to familar with the 1284. Do we need a recompilation of all existing games for this Controller or do we have only to set a special fuse (like in the ATMEGA8515 for the former AT90S8515) ?

From what I know, it's backward binary compatible with the 644 at 100%. I programmed HEX compiled for the 644 and they worked fine on the 1284.

Fine! In this case we have only to recompile (& change) the bootloader.

It's normal stabilizator, not step-up, isn't it? Maybe we should use some trimmer-configurable stabilizator, to be able to set voltage high enough for overclocked chip to work, but not too much higher.
I don't know how about ISP, because I don't use it. I've programmed my ATMega in external programmer (VUSBtiny I guess), I've built (using 5V from USB).

Yes this is a very important point - normally such 7806 regulators have only 5% acuracy.

I will test the 1284 in my "Meine WII" DIY Uzebox with a LM317 voltage regulator. I will report my results after this test...

-Harty

I actually did use the LM317 to create an adjustable regulator to step down form 12 volts

Mentioning ISP programming. The AVR ISP MKII will program the AVR, but it will continuously blink indicating voltage problems, and the voltage level won't show up in the programming window either.

[martinsustek]

Out of curiousity, with the additional PROGMEM and large bump in RAM, what will change?

Count of sprites in mode 3 is limited by RAM size (number of ramtiles), so double RAM means double sprites per screen (in fact more than double, because sprites are not whole RAM). Number of sprites per line is not limited in mode 3 as I know.

Bitmap modes like 6 or 8 will be able to have bigger resolution or more colors.

Modes like 1 (and possibly others) can use more tiles (more can be stored in FLASH and addressed by 2 bytes in RAM). Mode 9 can use more tiles (they are stored in FLASH).

Game logic computations can be better (more data in RAM, bigger algorithms) so more complex games can be developed (like that with many dynamic objects - bigger worlds).

You'll be able to have more "savegames" (stored states in EEPROM).

I can't say surely, if better games will arise, but it's opportunity for it.

And it's also solution to long-standing problem of having the best (what you can get of this platform) .

[D3thAdd3r]

I agree with 95% your analysis

Count of sprites in mode 3 is limited by RAM size (number of ramtiles), so double RAM means double sprites per screen (in fact more than double, because sprites are not whole RAM).

The count has always been just about equally limited by time also, now actually it's more time limited. With the newest kernel upgrade, I think most games can have more ram tiles than they can even use. Without any code running for a game engine, just the time it takes to ProcessSprites() and blit ram tiles, would limit us to much less than double that of say newest MegaBomber version(37 ram tiles with 30x24 vram, quite short screen..). We can probably max out time for 28mhz on the 644 right now, before we run out of actual ram(at least for a simple test!) Now I'm not against the 1284 idea, flash would improve games greatly, but I think a lot of people would be disappointed that our sprite capabilities would not increase much(it wouldn't feel like that much of an upgrade for the hardware change). I will do a test, without sound, default vram height and see what the max possible is for our clock speed and one for 30x22 vram(about the minimum useful height for say a big sprite fighter game). To go further(reach maximum blits every frame) Kernel would need redesigns and existing games reprogrammed to run on opposite ticks from blits, game logic one tick, sprite blit next tick(max possible vblank time for blits). I'm experimenting with just that and it does make a big difference!