|
|
| (3 intermediate revisions by 2 users not shown) |
| Line 1: |
Line 1: |
| [[Category:iNES Mappers|069]][[Category:in NesCartDB|069]]
| | #REDIRECT [[Sunsoft FME-7]] |
| Here are Disch's original notes:
| | {{DEFAULTSORT:069}}[[Category:iNES Mappers]][[Category:in NesCartDB]][[Category:Mappers with cycle IRQs]] |
| ========================
| |
| = Mapper 069 =
| |
| ========================
| |
|
| |
|
| |
| aka
| |
| --------------------------
| |
| [[Sunsoft FME-7]]
| |
| [[Sunsoft 5B audio|Sunsoft 5B]]
| |
|
| |
|
| |
|
| |
| Example Games:
| |
| --------------------------
| |
| Gimmick!
| |
| Batman: Return of the Joker
| |
| Hebereke
| |
| Gremlins 2 (J)
| |
|
| |
|
| |
| Notes:
| |
| --------------------------
| |
| This mapper is FME-7 and compatible. Sunsoft 5B operates the same as FME-7, only it has [[Sunsoft 5B audio|Sadditional sound
| |
| hardware]]. For a long time, it was thought Gimmick! uses FME-7, so the expansion sound is labeled as FME-7
| |
| in various places -- however -- technically FME-7 has no extra sound.
| |
|
| |
| Gimmick! is the only known game to use the extra sound found on Sunsoft 5B
| |
|
| |
|
| |
| Registers:
| |
| --------------------------
| |
|
| |
| Range,Mask: $8000-FFFF, $E000
| |
|
| |
| $8000: [.... AAAA] Address for use with $A000
| |
|
| |
| $A000: [DDDD DDDD] Data port
| |
| R:0-7 -> CHR Regs
| |
| R:8-B -> PRG Regs
| |
| R:C -> Mirroring
| |
| R=D-F -> IRQ Control
| |
|
| |
| $C000: [.... AAAA] Address for use with $E000 (sound)
| |
|
| |
| $E000: [DDDD DDDD] Data port (sound -- see sound section)
| |
|
| |
|
| |
| PRG Setup:
| |
| ---------------------------
| |
|
| |
| R:8 controls $6000-7FFF. It can map in PRG-RAM, PRG-ROM, or leave it unmapped (open bus), depending on the
| |
| mode it sets:
| |
|
| |
| R:8: [ERPP PPPP]
| |
| E = Enable RAM (0=disabled, 1=enabled)
| |
| R = RAM/ROM select (0=ROM, 1=RAM)
| |
| P = PRG page
| |
|
| |
| if E=0 and R=1, RAM is selected, but it's disabled, resulting in open bus. In case it's still unclear:
| |
|
| |
| R=0: ROM @ $6000-7FFF
| |
| R=1, E=0: Open Bus @ $6000-7FFF
| |
| R=1, E=1: RAM @ $6000-7FFF
| |
|
| |
|
| |
| R:9 - R:B appear to be a full 8 bits: [PPPP PPPP], and select only ROM.
| |
|
| |
| $6000 $8000 $A000 $C000 $E000
| |
| +-------+-------+-------+-------+-------+
| |
| | R:8 | R:9 | R:A | R:B | { -1} |
| |
| +-------+-------+-------+-------+-------+
| |
|
| |
|
| |
| No games seem to use more than 8k PRG-RAM, so I'm unsure whether or not it's swappable when selected. I
| |
| don't see why it wouldn't be.
| |
|
| |
|
| |
|
| |
| CHR Setup:
| |
| ---------------------------
| |
|
| |
| $0000 $0400 $0800 $0C00 $1000 $1400 $1800 $1C00
| |
| +-------+-------+-------+-------+-------+-------+-------+-------+
| |
| | R:0 | R:1 | R:2 | R:3 | R:4 | R:5 | R:6 | R:7 |
| |
| +-------+-------+-------+-------+-------+-------+-------+-------+
| |
|
| |
|
| |
| Mirroring:
| |
| ---------------------------
| |
|
| |
| R:C: [.... ..MM]
| |
| %00 = Vert
| |
| %01 = Horz
| |
| %10 = 1ScA
| |
| %11 = 1ScB
| |
|
| |
|
| |
| IRQs:
| |
| ---------------------------
| |
|
| |
| This mapper has a 16-bit IRQ counter which decrements every CPU cycle. When it wraps from $0000->FFFF, an
| |
| IRQ is tripped.
| |
|
| |
| reg R:E sets the low 8 bits of the counter
| |
| reg R:F sets the high 8 bits
| |
|
| |
| Note the regs change the actual counter -- not a reload value.
| |
|
| |
| reg R:D is the IRQ control:
| |
| [C... ...T]
| |
| C = Enable countdown (0=disabled, 1=enabled)
| |
| T = Enable IRQ triggering (0=disabled, 1=enabled)
| |
|
| |
| In order for IRQs to work as expected, both bits must be set. If either bit is cleared, an IRQ won't occur:
| |
|
| |
| C=0, T=1: IRQs are enabled, but the counter will never decrement
| |
| C=1, T=0: Counter decrements, but IRQs are disabled
| |
|
| |
| Acknowledging IRQs can only be done by disabling them (T=0).
| |
|
| |
|
| |
|
| |
| Sound:
| |
| ---------------------------
| |
|
| |
| Sunsoft 5B appears to be identical to the AY 3-8910 (or a similar chip -- possibly a different AY 3-891x or a
| |
| YM2149). The only game to use the sound, Gimmick!, does not use the envelope or noise functionality that
| |
| exists on the AY 3-891x, however, through testing it has been shown that such functionality does in fact
| |
| exist.
| |
|
| |
| The sound info below is a simplified version of the behavior. Envelope and Noise are not covered (aside from
| |
| the noise shift formula), and registers relating to those areas are not mentioned. However the information
| |
| below is enough to satisfy Gimmick! If you want further information and full register descriptions, consult
| |
| an AY 3-8910 datasheet or doc.
| |
|
| |
| Sunsoft 5B has 3 Square channels (no configurable duty cycle -- always play at 50% duty). Each operate
| |
| similarly to the native NES sound channels. They output sound at 1 octave lower than what may be expected,
| |
| though (see below).
| |
|
| |
|
| |
| Sound Regs:
| |
| ---------------------------
| |
|
| |
| $C000: [.... AAAA] Address for use with $E000
| |
|
| |
| $E000: [DDDD DDDD] Data port:
| |
|
| |
| R:0 -> [FFFF FFFF] Chan 0, Low 8 bits of Freq
| |
| R:1 -> [.... FFFF] Chan 0, High 4 bits of Freq
| |
| R:2 -> [FFFF FFFF] Chan 1, Low 8 bits of Freq
| |
| R:3 -> [.... FFFF] Chan 1, High 4 bits of Freq
| |
| R:4 -> [FFFF FFFF] Chan 2, Low 8 bits of Freq
| |
| R:5 -> [.... FFFF] Chan 2, High 4 bits of Freq
| |
|
| |
| R:7 -> [.... .CBA] Channel disable flags (0=enabled, 1=disabled)
| |
| C = Disable Chan 2
| |
| B = Disable Chan 1
| |
| A = Disable Chan 0
| |
|
| |
| R:8 -> [.... VVVV] Chan 0, Volume
| |
| R:9 -> [.... VVVV] Chan 1, Volume
| |
| R:A -> [.... VVVV] Chan 2, Volume
| |
|
| |
|
| |
| Operation:
| |
| ---------------------------
| |
|
| |
| For tone generation, a counter is counted up each CPU cycle. When it reaches the given 'F' value, it resets
| |
| to zero, and another step through the duty cycle is taken. These squares' duty cycles are fixed at 50%
| |
| (AY 3-8910 docs say 8/16, but see below).
| |
|
| |
| Emulating in this fashion, with a 16-step duty, these channels play 1 octave higher than they should!
| |
| Therefore, either channels are only clocked every other CPU cycle... or (what I find to be easiest to
| |
| emulate) the duty is actually 16/32 instead of 8/16, or something else is going on. I do not know which is
| |
| actually happening.
| |
|
| |
| The generated tone in Hz can be calculated with the following:
| |
|
| |
| CPU_CLOCK
| |
| Hz = -------------
| |
| (F+1) * 32
| |
|
| |
|
| |
| When the duty cycle outputs high, 'V' is output, otherwise 0 is output. When the channel is disabled (see
| |
| R:7), 0 is forced as output for the channel.
| |
|
| |
|
| |
| Non-linear volume:
| |
| ---------------------------
| |
|
| |
| Output volume is non-linear... increasing in steps of 3 dB.
| |
|
| |
| Output can be calculated with the following pseudo-code:
| |
|
| |
| vol = 1.0;
| |
| for(i = 0; i < 0x10; ++i)
| |
| {
| |
| sunsoft_out[i] = vol * base;
| |
| vol *= step;
| |
| }
| |
|
| |
| Where 'base' can be adjusted to match your native NES sound channel levels, and 'step' is "10^(dB/20)".
| |
|
| |
| For 3 dB, 'step' would be ~1.4125
| |
|
| |
|
| |
| Noise Formula:
| |
| ---------------------------
| |
|
| |
| >> >>
| |
| +-->[nnnn nnnn nnnn nnnn]->output
| |
| | | |
| |
| | | ++
| |
| | | |
| |
| | v v
| |
| +-------------------XOR
| |
|
| |
|
| |
| - 16-bit right-shift reg
| |
| - bits 0,3 (before shift) XOR to create new input bit
| |
| - bit 0 is shifted to output
| |
| - initial feed is 1
| |
