fullsnes: Difference between revisions

Revision as of 22:04, 2 July 2023

Fullsnes is a large hardware reference for the SNES created by nocash with several ascii art illustrations.

Errata

In https://problemkaputt.de/fullsnes.htm#snesapumaincpucommunicationport, fullsnes only mentions storing the entrypoint to 2142h and does not mention whether the entrypoint is a word or byte (the other places in this section where the bracket notation omits "Byte" or "Word" it is a single byte being stored). The Official Super Nintendo development manual is clear that the program start address for the uploaded SPC700 driver is stored to both PORT 2 and 3 (as part of step 11 on page D-2).
In https://problemkaputt.de/fullsnes.htm#snescartridgecicnotes, the formula for the Polynomial Counter is missing an inversion in the second term.
In https://problemkaputt.de/fullsnes.htm#snesapuspc700ioports, it could be clearer that bits 0-3 of TnOUT are reset to 0 automatically by hardware after reading - it is not the programmer's responsibility to zero them after reading. [1]
In https://problemkaputt.de/fullsnes.htm#snesapudspvolumeregisters, MVOL should stand for "main volume," not "master volume" because MVOL does not control the echo volume, and "Main Volume" is how the official docs expand "MVOL." [2]
In https://problemkaputt.de/fullsnes.htm#snesapublockdiagram, the word "Reverb" is used (the only place this word is used in fullsnes) but should probably say "Echo" instead like the rest of the times this feature is mentioned, as they have slightly different musical meanings.
In https://problemkaputt.de/fullsnes.htm#snesapudspechoregisters, "bit" should be plural in several places when describing the ESA buffer entry bytes
In https://problemkaputt.de/fullsnes.htm#snescartridgeromimageinterleave, the worst case scenario can much worse than triple-interleaved or double-mis-de-interleaved - in fact ten interleavings of a normal 320Kbyte file is the identity operation, so in this particular case the worst is quintuple-interleaved.
"Game Boy" is spelled several times as one word
In the https://problemkaputt.de/fullsnes.htm#snesapumaincpucommunicationport notes there is a typo: "injumping"
In the IPL Boot ROM Disassembly, both the zerofill and transfer loops have a typo where "loop" is spelled with one "o"
The four communication ports between the APU and the 5A22 are prefixed CPUIO (cpu input/output) on the 5A22 side but APUI0 (apu eye zero) on the APU side

External Links

https://problemkaputt.de/fullsnes.htm

References

[1] Page 3-5-2 (NCL PG 11) of Book I of the official Nintendo documentation: "When CN is read, the 4-bit up counter alone is cleared through IC internal timing"
[2] https://archive.org/details/SNESDevManual/book1/page/n175

@@ Line 2: / Line 2: @@
 === Errata ===
+* In https://problemkaputt.de/fullsnes.htm#snesapumaincpucommunicationport, fullsnes only mentions storing the entrypoint to 2142h and does not mention whether the entrypoint is a word or byte (the other places in this section where the bracket notation omits "Byte" or "Word" it is a single byte being stored).  The Official Super Nintendo development manual is clear that the program start address for the uploaded SPC700 driver is stored to both PORT 2 and 3 (as part of step 11 on page D-2).
 * In https://problemkaputt.de/fullsnes.htm#snescartridgecicnotes, the formula for the [[Polynomial Counter]] is missing an inversion in the second term.
 * In https://problemkaputt.de/fullsnes.htm#snesapuspc700ioports, it could be clearer that bits 0-3 of TnOUT are reset to 0 automatically by hardware after reading - it is not the programmer's responsibility to zero them after reading. [1]