PCB C300565 questions

[khryssun]

Hi all,

 

I have two questions about 7800 PAL carts using the PCB C300565 model :

 

 

1) Does all the 7800 PAL games using this PCB have the 16 K Super Chip RAM

included on the PCB ?

 

2) If not, does someone is able to tell me what are the carts built with this PCB

and including the 16K SC RAM.

 

Thanks a lot.

[Bruce Tomlin]

Not necessarily.

 

I had to go through all this last week and I'm all fresh about '565, so you're in luck.

 

First of all, there were only five (yes, five!) games which used RAM. Impossible Mission, Jinks, and Tower Toppler were 8K, Summer Games, and Winter Games were 16K. I think you will find '565 boards in all of these. PAL vs NTSC shouldn't matter, because that wouldn't make a game have or not have RAM. I'm presuming you want to know this so you can do a conversion?

 

Basically, when it came to RAM support, the C100339 was just plain broken. It gated the chip select with the phase-2 clock, when it should have gated the write-enable line with the phase-2 clock. I found this out the hard way by actually trying to put RAM in one. So they had to make a new board for the games that used RAM. Which is where '565 came in. But I don't see any reason (other than having boxes full of unused '339 boards) why they would only use it for RAM games.

 

In fact, I have found a Crack'ed with a '565 board.

 

FWIW, I still haven't looked at a "low mode" (bank N-1 at $4000-$7FFF) board yet. Looks like the least rare games of that category are Ace of Aces and Touchdown Football. Which is interesting, because there are 16 or 17 games in that category. (with Crack'ed, Dark Chambers, Planet Smashers, and maybe Barnyard Blaster, NOT being low-mode games)

 

Here are a couple of schematics that I've been working on:

 

http://atari7800.net/files/C100339-schematic.gif

http://atari7800.net/files/C300565-schematic.gif

[Harry_Dodgson]

I'm confused...

 

1) I've been dis-assembling Planet Smashers and it writes to 16K of RAM starting at $4000 from a routine at $F0EE.

 

2) All the 7800/2600 Monitor cartridges are based off the Hat Trick board - which I guess is C100339 and the 6264 RAM chip works okay on them.

 

3) I think your schematic is incorrect regarding U2

 

Harry

Edited July 8, 2005 by Harry_Dodgson

[Bruce Tomlin]

My reference to memory sizes came from the CC2 rom list. Planet Smashers is listed as a 78SC_LOW, which means the CC2 puts ROM bank 6 at $4000-$7FFF, but I got it to work before I had implemented low-mode. Looking back at an old cartridge dump I made of it years ago, it definitely did not have RAM, and bank 6 appeared at $4000-$7FFF, even if it wasn't used there.

 

As to memory working or not working on the Hat Trick board, apparentlly it works okay for the CPU, but not for Maria DMA. I know that in my initial tests, I was able to use my own monitor to write and read back RAM locations manually. But until I made the change, Summer Games and Winter Games showed a bunch of semi-identifiable junk on the screen, and so did the other games that used RAM. Impossible Mission even crashed soon after entering the first room. So it does work, but not really.

 

Which board's U2? Which gate? I can always go back and double check. I know I somehow ended up with different pin numbers from your version on the '339 board, but that was a long time ago that I traced it out, and I simply never bothered to double-check. With the '565 I was working from a desoldered board this week, and maybe my eyes slipped where there were three or four traces together. Anyhow, the '565 board is where I got my updated equations for RAMCS/RAMWE in my Ultracart project.

 

In fact, what really tripped me up last week was that I had long assumed the Hat Trick board supported RAM just fine, and "everybody" knew it did. I guess nobody had tried to copy-cart Summer Games or Winter Games because there was no point to it.

[Harry_Dodgson]

Looking at the schematics - I see why Maria can't use the RAM on that board. The phase 2 signal is only generated by the 6502 and is much too slow for Maria RAM accesses. For on-board RAM, the Maria generates the chip selects directly.

 

As far as U2, that's the NAND gate on the Hat Trick board. If you compare our schematics, you can see the differences easily. It could be that mine is wrong too.

 

So "Planet Smashers" copies the data, then never uses it? That's not the only example of bad coding in that game. Tempest (not the game) just piqued my interest in finding a cheat code for the game so I started looking around.

 

Harry

[Bruce Tomlin]

I'll put a double-check of that board on my list of things to do. I have a desoldered board handy, so it's just a matter of checking continuity in a lot of places.

 

And with your explanation of the Maria issues, it's all starting to make sense. The RAM on the Hat Trick board really did work properly for the CPU, but the Maria DMA accesses work in a different enough way to not work with the older board.

 

So once programmers started using the extra RAM, of course they wanted to use it with the Maria, and it was quickly found out that this basically didn't work at all. I'm pretty sure in my testing that all five games had problems, meaning that all of them were pointing the Maria to cartridge RAM.

 

And if, as you say, Planet Smashers writes to the $4000-$7FFF area, that could be fun for the output drivers. Neither of the two supercart boards I've looked at so far gate a ROM at $4000 for reads only. Now I really want to crack open an Ace of Aces or a Touchdown Football.

 

EDIT: I finished rechecking, and all the U2 and U3 pinouts on my C100339 schematic are correct. Maybe you traced a different board with the same logic? I do still need to update the text to indicate that the RAM problem is only with Maria DMA. I've also got the C300595 board from Touchdown Football mapped out and ready to start making a schematic from.

Edited July 9, 2005 by Bruce Tomlin

[khryssun]

Thanks for the infos !

 

This was very helpful for me

Edited July 11, 2005 by khryssun

[R.Cade]

I noticed that the documents to make RAM carts don't use the C100339 design, except for the version with the built-in BIOS. Does anyone have instructions for making a RAM cart from the '339 without the BIOS? I already have the DEVOS EPROM in my 7800...

 

Pete

[PacMan]

Something to think about for the MARIA and her accessing memory for DMA.

 

The MARIA has two different modes of DMA, fast DMA for Zone-list, display-list and character maps. These DMA cycles needs to have access to a memory with an access time of maximum 165nS (According to the datasheet). Note that this applies to a memory connected directly to the MARIA to one of the chip select signals RAM0CS- or RAM1CS-.

 

There is also a slow DMA mode for graphic objects which is much more relaxed and would normally never cause any problems.

 

If we analyze the fast DMA mode more closely and start tracing the signals involved in the DMA cycle out from the MARIA we can deduce the following:

- There is a 15-18 nS penalty for the 74LS08 gate onboard the A7800 motherboard.

- There is 15-18 nS penalty per gate (two of them) in the C300565 cartridge.

 

The 15-18 nS delays were taken from old TTL data sheets found in the old section of my data book library. They are maximum times at a load of 50pF which is very realistic, all lines going out to the cartridge can be expected to see that kind of load.

 

Totaly there is a line delay of maximum 54 nS out to the Cartridge RAM. Calculating backwards we now see that the cartridge need to be equiped with an SRAM no slower than 111nS accesstime. This means that using a standard 120nS or 150nS SRAM in the cartridge will make the MARIA fail in its fast DMA transfers. Slow DMA's and CPU accesses will be OK though.

 

I have been experimenting a lot lately with this while trying to design a debug cartridge for the A7800 and found that even a 100nS RAM will fail occationally with my board layout. 85nS seem to do the trick though.

 

The XBoard that I designed previously also works OK with the fast DMA so I think I am in the correct direction.

 

These are just my findings when working on the hardware and anyone with better knowledge is encouraged to correct my statements.

 

/PacMan

[gdement]

As to memory working or not working on the Hat Trick board, apparentlly it works okay for the CPU, but not for Maria DMA.  I know that in my initial tests, I was able to use my own monitor to write and read back RAM locations manually.  But until I made the change, Summer Games and Winter Games showed a bunch of semi-identifiable junk on the screen, and so did the other games that used RAM.  Impossible Mission even crashed soon after entering the first room.  So it does work, but not really.

...

In fact, what really tripped me up last week was that I had long assumed the Hat Trick board supported RAM just fine, and "everybody" knew it did.  I guess nobody had tried to copy-cart Summer Games or Winter Games because there was no point to it.

888381[/snapback]

 

I must have missed this thread before. Assuming you still stand by these findings, then I should mention that I have a Winter Games cart which was produced on a C100339 board. It does not have any graphical glitches, but I have noticed that it has a lot of sounds missing which are present in the ROM file I downloaded from the internet. I mentioned that once in another thread and at least one other person also said they had the same "missing sounds" issue on this game. So while I can say my C100339 Winter Games does seem different from the popular ROM, it's not showing any apparent Maria related issues.

[gdement]

After looking more at my Winter Games cart, I've realized it differs slightly from the Dan Boris schematic.

 

U4-pin27 (RAM WE) has no continuity to U3-12 or U5-11 as shown in the schematic. However it does have continuity directly to the edge connector R/W.

 

In addition, there is some rework at U3 (74LS10 NAND gate). The connection to CLK2 has been severed and pins 10-11 have been shorted, so both are connected to +5V. The date code on this board is 41 87.

 

 

Unfortunately, I think the CLK2-Maria issue has undermined my attempt to make a Flash cartridge. Mine works for about 10 minutes at a time, then it has to cool before it works again.

[edit: It seems to work fine with 70ns flash chips, but slower ones are glitchy.]

 

FWIW, if anyone wants to add RAM to a C100339, you could just replicate the changes above. I flashed the popular version of Winter Games (with full sound) and it works 100% for me on this board.

Edited July 1, 2006 by gdement