SD card initialization fails on FPGA

[jbush001]

Times out initializing card in SPI mode. It seems to never read anything but 0xff from the data out (DO) pin after sending CMD0.

1. Connected pins to GPIO block. I verified I could read the input pins (printed to the serial log) by shorting the solder pads to ground, and drive output pins by controlling with software and measuring with a multimeter.
2. I tried putting the chip into 4 bit mode. I could see correct data on DAT1-3, but DAT0 was stuck at VCC. I wrote various patterns on the SD card on my laptop, then re-read on the board to confirm.

[jbush001]

SD Pin	SD Mode	SPI Mode	FPGA PIN
1	CD/DAT3	CS	PIN_AC14
2	CMD	DI	PIN_AD14
3	VSS1	VSS1
4	VDD	VDD
5	CLK	SCLK	PIN_AE13
6	VSS2	VSS2
7	DAT0	DO	PIN_AE14
8	DAT1		PIN_AF13
9	DAT2		PIN_AB14

Schematic for DE2-115 board:

Pin information from DE2-115 data sheet:

SPI/SD pin mapping:

https://en.wikipedia.org/wiki/Secure_Digital#Transfer_modes

FPGA pin configuration:

https://github.com/jbush001/NyuziProcessor/blob/48f71f80f91e50ba7d54f324b76a6903f78c5538/hardware/fpga/de2-115/de2_115.qsf#L186-L189

I/O configuration:

https://github.com/jbush001/NyuziProcessor/blob/48f71f80f91e50ba7d54f324b76a6903f78c5538/hardware/fpga/de2-115/de2_115_top.sv#L62-L65

Mapping to SPI controller:

https://github.com/jbush001/NyuziProcessor/blob/48f71f80f91e50ba7d54f324b76a6903f78c5538/hardware/fpga/de2-115/de2_115_top.sv#L196-L202

[jbush001]

Simulation waveforms. The times at the top aren't accurate, is they assume a clock speed of 1 GHz rather than the 50 MHz that board runs at. Here is the initialization clock sequence (80 clocks with CS high).

Here is CMD0:

01000000 00000000 00000000 00000000 00000000 10010101
40       00       00       00       00       95

Closeup showing bit phase:

output in falling edge, sample on rising edge (mode 0)

[jbush001]

Verified 250 system clocks per spi clock: 200kHz SPI clock at 50Mhz system clock. I tried slowing this down by increasing the divider on the dev board to ensure there wasn't an issue with clock speed, but there was no effect.

[jbush001]

While waiting for response. CS is still low, DI is high.

[jbush001]

Tried with 3 cards, all failed in the same way (no response to initial CMD0)

• 8 GB Samsung microSD (in adapter)
• 4 GB Monster SDHC
• 8 MB Canon (old)

[jbush001]

I hooked up a logic analyzer and it responded to the first command.

Unplugged the logic analyzer and it started failing again... Seems like this might be a connector issue.

(I should be sending CMD8 instead of CMD1 after CMD0, but that is a separate issue)

[jbush001]

I also had to turn the board upside down to attach the probes. In the normal position, gravity pulls the card down away from the pins. Upside down, it pushes them against them. I might try sticking a piece of paper in with the card to apply more pressure.

[jbush001]

I tried sticking paper in, turning the board over with no success. However, it works consistently if the probes are attached...

[travisg]

Hmm, what board is that BTW?

[jbush001]

DE2-115

[travisg]

Ah didn't know if you were still using that. I actually bought one of those for myself when this project got started up years ago. I should brush it off.