New Method For DC Serial To SDCARD - InDevelopment

Some free GPIO can be taken directly from SH4.

 

I've had this idea and then went nowhere with it. I would have probably tried to use an AVR instead, but i think i'll try STM32 instead, the STM32F103 barely cost $2 on their own or on a dodgy little board, and genuine STM32F4 Discovery kits are available cheaply locally. Which is not something i can say about PIC18 at all! Don't see a necessity for additional interface IC, 4MBit/s USART should just about work on its own.

But thinking about what SWAT said, it will be curious to see what kind of errors those are exactly! Perhaps that's something we can work around. Maybe we can have high speed hardware DMA, with reed-solomon error correction, to stamp out the transmission errors? I've been wary of the particular kinds of errors that can occur due to clock drift and mismatch at such high clock rates. However, i'm looking now, and jj1odm discovered that pin 2 of connector is in fact connected to SCK2 of SH-4 processor. This means if registers are configured such, external device can be serial clock source for perfect synchronization. Indeed perhaps the CLK output of the USART, the configurable UART/SPI mixed interface.

 

SCIF only USART, SCK2 for external clock
SCI have SPI and DMA, maximal speed 5 Mbps with external clock

 

Yeah i mean USART of STM32, i think it's possible to get its clock out. If not, the clock base which is 8 times higher definitely goes out. The idea is to get STM32 to drive the communication of SH-4 SCIF, without internal hardware modification of Dreamcast. The communication mode needs to be akin to serial UART, async mode, but with a forced common clock between two devices. In normal serial communication, you have two devices which each have a high frequency clock (multiple MHz), which they divide down to arrive at the same low frequency, which is thus hopefully very exact and stable. At MHz baud rates thus low divisor, this assumption breaks down, and the slightest drift and jitter between two clocks will result in communication errors. I'm hoping that having a common clock source may increase communication quality and thus allow higher baud rates while leveraging continuous buffered and DMA communication on the SH-4 side. SCFRDR2 is DMA source.

SCI port probably doesn't go anywhere, does it?

 

max speed SCIF in USART mode with extrnal clock - 2 Mbps
current speed in software SPI mode ~7 Mbps

yes.