Reverse-engineering the audio chip in the Nintendo Game Boy Color | Hacker News Reader

Reverse-engineering the audio chip in the Nintendo Game Boy Color

righto.com

80 pointsingve12 comments

A few months ago the whole GameBoy SoC (DMG-CPU) was reversed engineered in the same way [1], except for the CPU part - this means the schematics include pixel processor unit, timer, serial port, IRQ controller, etc. However, it seems that the CPU itself is pretty straight-forward to emulate anyway, since Gekkio, author of the mooneye-gb [2], figured it's a variant of a well-documented SM83.

Anyway, reverse-engineering efforts like this are pretty exciting, as they should allow to create a 100% accurate emulators - and we know they are accurate not only because all the games are running as expected, but because they are based on the actual hardware design (which is even better than the producer's specification).

GB is not a complex piece of hardware, in comparison to eg. SNES, but there are some emulation issues really hard to tackle, especially around the pixel processing unit IRQ timings - it's like solving a complex set of linear equations, but we don't even know what the variables are (eg. what can affect a 4-processor-tick shift in the HBlank IRQ timing this way or another).

Looking forward to a complete, transistor-level GB simulator, something like Visual6502.

[1] https://github.com/furrtek/DMG-CPU-Inside [2] https://github.com/Gekkio/mooneye-gb [3] http://www.visual6502.org/

For any analog engineers in the audience: how would this chip differ if designed and fabricated today (if at all)?

I looked at a random TI headphone amp chip [1] for portable gaming systems. It's tiny (3mm×3mm) and uses CMOS for low power consumption. Internally it has a charge pump to generate the voltages it needs from a wide power supply range. The charge pump also allows the output to be centered around ground so DC blocking capacitors can be omitted. The chip includes click and pop suppression. Inexplicably it is Windows Vista compliant. So audio chips have progressed a lot since this Nintendo chip.

[1] http://www.ti.com/product/TPA6132A2/datasheet

“MicrosoftTM Windows VistaTM Compliant”

Remember when the software had to comply with the hardware instead of the other way round :]

I feel like they've had an incestuous relationship as long as we've had both.

It is a bit goofy to call out a amplifier without even a SPI or I2C bus for config though as compatible. I'm not sure how Vista would know the difference between this and a tube amp.

Microsoft has long had PC design guidelines, starting with PC 97 in the late 90s. "Windows Vista Compliant" would mean it meets the functional and performance requirements found on the hardware design guidelines for Windows Vista.

I understand that. It's just weird that they'd specify anything for components like that.

I think the biggest difference is that you wouldn't use this chip at all, or a chip like it. The entire audio IP stack can be put into a single chip these days, depending on the fidelity and power constraints. I2C on one side of the chip, analog output on the other.

> I2C on one side of the chip, analog output on the other.

Surely you meant I2S [0] (or at least SPI)? I2C doesn't sound very optimal for this kind of application.

[0]: https://en.wikipedia.org/wiki/I%C2%B2S

Am I missing something or isn't this just the amplifier, as opposed to something that generates sound like the SID?

Yes. The Gameboy doesn't have a dedicated sound generator chip, that functionality is integrated into the main CPU.

Yes, it's just the amplifier. I see that the title could be confusing.

Reply on news.ycombinator.com