INES Mapper 096: Difference between revisions
m (refile, add cat:nescartdb) |
(describe hardware) |
||
Line 1: | Line 1: | ||
[[Category:iNES Mappers|096]][[Category:Discrete logic mappers|096]][[Category:in NesCartDB|096]] | [[Category:iNES Mappers|096]][[Category:Discrete logic mappers|096]][[Category:in NesCartDB|096]] | ||
==Hardware== | |||
The game uses a [[74161]], [[7402]], and [[7474]]. The 74161 is wired exactly the same as in all of Nintendo's discrete logic mappers, complete with bus conflicts. | |||
The 7402 generates the clock signal for the 7474 by calculating <tt>(PPU /A13) NOR (PPU A12)</tt>, which is equivalent to <tt>(PPU A13) AND (PPU /A12)</tt>. On a rising edge of this signal, the 7474 latches the value of <tt>PPU A9</tt> and <tt>PPU A8</tt>, i.e. when the PPU fetches the nametable after having fetched from the pattern tables. Because <tt>PPU /RD</tt> isn't involved and attribute fetches are immediately after nametable fetches, attribute fetches are ignored. Another two NOR gates take these two latched values and produce 4+4F CHR banking. | |||
To rephrase: it is the act of changing the PPU address from $0xxx, $1xxx, or $3xxx ''to'' $2xxx that latches A9 and A8. There is no special logic that guarantees attribute fetches don't cause bankswitching, and Y scroll values greater than 240 will choose the same bank as reads from any of the rest of $23xx. It is only because of the PPU's fetch pattern that attribute fetches have no impact. | |||
==Disch's Notes== | |||
Here are Disch's original notes: | Here are Disch's original notes: | ||
======================== | ======================== |
Revision as of 07:30, 24 April 2013
Hardware
The game uses a 74161, 7402, and 7474. The 74161 is wired exactly the same as in all of Nintendo's discrete logic mappers, complete with bus conflicts.
The 7402 generates the clock signal for the 7474 by calculating (PPU /A13) NOR (PPU A12), which is equivalent to (PPU A13) AND (PPU /A12). On a rising edge of this signal, the 7474 latches the value of PPU A9 and PPU A8, i.e. when the PPU fetches the nametable after having fetched from the pattern tables. Because PPU /RD isn't involved and attribute fetches are immediately after nametable fetches, attribute fetches are ignored. Another two NOR gates take these two latched values and produce 4+4F CHR banking.
To rephrase: it is the act of changing the PPU address from $0xxx, $1xxx, or $3xxx to $2xxx that latches A9 and A8. There is no special logic that guarantees attribute fetches don't cause bankswitching, and Y scroll values greater than 240 will choose the same bank as reads from any of the rest of $23xx. It is only because of the PPU's fetch pattern that attribute fetches have no impact.
Disch's Notes
Here are Disch's original notes: ======================== = Mapper 096 = ======================== Example Games: -------------------------- Oeka Kids - Anpanman no Hiragana Daisuki Oeka Kids - Anpanman to Oekaki Shiyou!! Notes: --------------------------- These games use the Oeka Kids tablet -- so you'll need to add support for that if you really want to test these. These games use 32k of CHR-RAM, which is swappable in a very unique fashion. Be sure to read the CHR Setup section in detail. Registers: --------------------------- I'm unsure whether or not this mapper suffers from bus conflicts. Use caution! $8000-FFFF: [.... .CPP] C = CHR Block select (see CHR Setup) P = PRG Page select (32k @ $8000) CHR Setup: --------------------------- This mapper is tricky!!! Firstly, this mapper divides the 32k CHR-RAM into two 16k blocks (above 'C' bit selects which block is used). The selected pages (including the fixed page) are taken from only the currently selected 16k block. $0000 $0400 $0800 $0C00 $1000 $1400 $1800 $1C00 +-------------------------------+-------------------------------+ | **See below** | { 3 } | +-------------------------------+-------------------------------+ But that's the easy part. This mapper does a very, very cool trick which watches the PPU address lines to effectively "split" the nametable into 4 smaller sections -- thereby assigning a different CHR-RAM page to each section. This allows **every single tile in the NT** to have a unique tile graphic! Long story short: A nametable spans from $2000-$23BF ($23C0-$23FF are the attribute table). The mapper breaks the NT up like so: $2000-20FF = use CHR page 0 $2100-21FF = use CHR page 1 $2200-22FF = use CHR page 2 $2300-23BF = use CHR page 3 the other nametables at $2400, $2800, $2C00 are broken up in the same fashion. Long story long: PPU Address lines are modified as the PPU fetches tiles, and also when the game manually changes the PPU address (via the second write to $2006 --- or by the increment after read/writing $2007). The mapper monitors every change to the PPU Address lines, and when it lies within a certain range, it swaps the appropriate CHR page in. It will only swap CHR when the address falls between $2000-2FFF (or mirrored regions like $6000-6FFF, $A000-AFFF, $E000-EFFF). $3xxx will not trigger a swap. When in that range, it checks to make sure the address is not attribute tables ((Addr AND $03FF) < $03C0). Note I'm not 100% sure if the mapper really does this or not. It's very possible that attribute fetches will also swap CHR... this would not really disrupt anything other than making the game be more careful about its PPU writes. When all that checks out, bits 8 and 9 (Addr AND $0300) select the 4k CHR page to swap in to $0000. Note that the mapper does not distinguish between PPU driven line changes and game driven line changes. This means that games can manually swap the CHR page by doing specific writes to $2006: LDA #$20 STA $2006 STA $2006 ; Addr set to $20xx -- CHR page 0 selected LDA #$21 STA $2006 STA $2006 ; Addr set to $21xx -- CHR page 1 selected And in fact, games would HAVE to do that to select CHR, since that's the only way to fill CHR RAM with the desired data. So make sure your emu supports this.