Hacking NAND chip

[Gonzo]

I am wondering about the NAND chip THGBM2G3P1FBAI8:

It is stated by 3Dbrew to be 1GB sized, but on the website of a decapping company it is told to be 2GB - refer https://chipworks.secure.force.com/...GBM2G3P1FBAI8&viewState=DetailView&cartID=&g=
So which size is correct? I would assume that the guys with detailed pictures of the internals would knows it better?!

I tried to find some specs about the chip, but I had no luck. Also I didn't find any shops to buy one. The spec can tell us how many pins are required - probably these are only 8 pins CLK, CMD, DAT0-3, VCC and GND. Perhaps only CMD is needed for switching but therefore the data lines needs to be open-drain. Or perhaps there is a separate chip-select...

Background is, that I still have the idea of connecting a second chip in parallel with a logic to switch between them, which would give us the ability to have two separate Firmware versions. But this can only be realized if that chip (or a compatible one) is available and additionally if we can cut the connections of the origin NAND and rewire them to the switching logic without desoldering the complete chip.

 

[charlie22911]

It could be the same chip but at a lower bin with defective NAND fused off, or it could be the generic part number for a line of NAND chips.
Either way I doubt you would be able to switch at a hardware level, I wager it will require a software solution instead.

 

[Cyberdrive]

THGBM2G3P1FBAI8
Not sure if it means what I think it does, though.
As for the hardware switch, I don't see why it can't be done. Here's a Dreamcast example, orginal DC BIOS chip with piggybacked second one on the top image and BIOS switch in the ventilation grill on the bottom one.

Spoiler

Courtesy of http://www.dc-swat.ru/blog/hardware/16.html

 

[Coto]

since much before done the xl mod, this got me thinking so NAND handling would actually be more handy/fruitful:

hardware way: keep all (both) NAND points together and unsolder VCC, use a normally open, normally closed switch, and feed the common with input VCC. Normally open would feed NAND A, Normally closed feeds NAND B.

software way: iirc SDIO api (SD library), derived from SD interface (SPI library) has CS (chip select) for dual-nands. While one's in use, the other is aware and stops sending CLK & DAT0/n. The problem is you need to feed by SPI commands both SD controllers(client 1, client2, and host 3DS controller) so one become active, and the other passive. This would require a mini cpu that handles SPI commands. (think of a cpu that derives from nand A to B by software, even if both are tied to the same circuitry)


as seen here:

https://www.sdcard.org/developers/overview/sdio/sdio_spec/Simplified_SDIO_Card_Spec.pdf

Chapter 2.2

For one, from what i'm reading, you need a programable SPI chip (basically emulates a sd microcontroller, except registers are not fixed, one feed them manually), so you can, then issue a SDIO command, specifically tell the RCA register which SD should be treated now (Relative Card Address).

 

[Gonzo]

In my opinion it is not necessary to have cpu based logic for switching. Man could simply block the signals with an external gate chip plus a flip-flop.

As for the hardware switch, I don't see why it can't be done. Here's a Dreamcast example, orginal DC BIOS chip with piggybacked second one on the top image and BIOS switch in the ventilation grill on the bottom one.

The days of good-old piggybacked chips are gone since smd form is used because now, all pins are located on the (unreachable) bottom side

 

[Coto]

this is SDIO mode: No need to block DAT5-7, these are alternate multi-block transfer DAT lines. MMC could work with just DAT0 connected, and CMD, CLK, VCC as well

edit: http://pinouts.ru/Memory/mmcard_pinout.shtml

The MultiMediaCard card identification and addressing methods are replaced by a hardware Chip Select (CS) signal. For every command, a card (slave) is selected by asserting the CS signal (active low). The CS signal must be continuously active for the duration of the SPI transactions. The MM card bidirectional CMD and DAT lines are replaced by unidirectional DATA in and DATA out signals (single and multiple block read/write commands are supported).

this in SPI mode: you need a micro-cpu that sends SPI signals, to both 3DS microcontroller, e-mmc onboard, and external SD card, all to be set in SPI mode

 

[Gonzo]

Ah, I see.