MicroSD question (SOLVED!)

[dbotkin]

Been trying to get this working for a while now... and it's close, but so far no joy.

CPU: PIC18F46J53
Compiler: PCWH 4.128
#include "mmcsd.c" - with changes for MBR suggested by andrewg.
#include <fat.c>
Hardware: MicroSD card socket, wired like this:

Pin 1: No connection
Pin 2: To PIC pin 2, RD4 (#define MMCSD_PIN_SELECT PIN_D4)
Pin 3: To PIC pin 40, RD2 (#define MMCSD_PIN_SDO PIN_D2)
Pin 4: 3.3V
Pin 5: To PIC pin 3, RD5 (#define MMCSD_PIN_SCL PIN_D5)
Pin 6: Ground
Pin 7: To PIC pin 41, RD3 (#define MMCSD_PIN_SDI PIN_D3)
Pin 8: No connection

I also have 10K pullup resistors to 3.3V on all four lines.

OK... so apparently the PIC can see the memory card. I tossed in calls to mmcsd_print_cid() and mmcsd_print_csd() and get this:

[ckielstra]

What voltage is your PIC running? If it is 5V than you'll need some 'glue' logic to convert from / to the 3.3V for the SD card.

SD cards start with their data bus in Open Collector mode until the initialization to SPI mode databus has completed. Do you have a pull up resistor at the DO from the Mini SD card (pin 7, 10k will do).

What decoupling capacitors do you have at the SD card?

When writing data the power consumed by the SD card is a lot higher than when reading. Perhaps the 3.3V power supply can't cope with this power surge?

[Ttelmah]

The 46J53, is a 3.3v PIC, so he should be OK on this one, unless he has copied circuits with a resistive divider to the card.

Other comment he has set the file system type in fat.c to FAT32?.

He says he has pullups on all four lines (essential on CS as well, since otherwise it can 'glitch select' when the unit is waking up).

Obvious comment card must be MicroSD, not MicroSDHC.

He seems very close, getting the ID etc..

Best Wishes

[dbotkin]

[Ttelmah]

The weak pull-ups are not enough. The maximum specified is for SPI is 100K, and they don't switch on till the code says to do so. This can still give the CS glitch on boot. The external resistors are the way to go.
I'd triple check the mods you have done for MBR.
Also try FAT16, and see if anything changes.

I don't remember any specific problems around 4.128, though there did seem to be some oddities with some chips around then. Not a version I've 'kept' so can't try to see if it works.

Best Wishes

[dbotkin]

Progress...

Between andrewg's fixes to fat.c (and using my LOCAL copy with those changes applied, instead of the original from CCS), and finding a format that it likes, I'm able to do most of the things I need to. fat_init() works, and with the routines from ex_fat.c I can see the directory, tail a file, etc. From my program I can use fatread() to fill a buffer. There are some unexplained delays I need to chase down, but I'm almost there.

[dbotkin]

Now all I need to do is figure out how to make this all work with #pin_select, SPI2 and FORCE_HW. I can read a file... but slowly, around 3K byte/sec. Ugh. I need about 4x that speed at minumum, preferably 10x.

Back to the grind... sigh.

[asmallri]

[dbotkin]

And how would you suggest I change that? Looking at #use_spi...

[dbotkin]

OK, so a little more digging into mmcsd.c... not the one I had, but a new one Darren emailed me. Thanks, Darren!

mmcsd.c sets the baud to 400000 during mmcsd_init(). Darren's sets it to a more realistic speed, IF you're using the MSSP hardware. He also thoughtfully included some new code to make it easy to specify the MSSP hardware SPI. So... great.

Unfortunately, it still fails to detect the card any time I specify the hardware SPI. Software works great... hardware no. MOSI looks normal, but not CS or CLK. And of course without that, there's no MISO... and no love from the SD card.

I tried setting BAUD to several values far higher than the 400K to which it's initially set with the software SPI. No dice. The improvement in transfer rate is very slight. While waiting on more from CCS support, I'm continuing to look for any mistakes in my own code.

[ckielstra]

I don't have your v4.128, but early versions of the mmcsd.c contained a bug where the SPI was initialized in a wrong SPI mode (mode 2). SD cards only work in mode 0 or 3.
In V4.138 the setup is correct:

[Ttelmah]

It's still there in the spec for SPI mode.

SPI mode itself is now 'optional' for MicroSD, _but_ since the cards all generally use variants of the same chipsets used for SD ((where it is still required), I've yet to see a card that doesn't have it.

It also raises a question, since how could a MicroSD card then work in an adapter to SD, where this mode is required....

Best Wishes

[dbotkin]

[dbotkin]

OK, so we've pared this down to the bare minimum. Maybe someone here can see something that explains what we're not getting.

When this code is run, I get positive-going pulses on the chip select pin (PIN_D4) -- and nothing else. No SPI clock or data. It's as if the SPI2 module isn't even there. I've been through the .LST file and everything looks OK to me -- PPS seems to be getting set up correctly, SPI2 initialized, data getting stuffed into the SPI2 buffer. But I get nada on the actual device pins. It's enough to make me think about getting out the heat gun and replacing the chip.

[Ttelmah]

You are using SPI streams, and then _not_ using these for the transfer.
You need to use:

SPI_XFER(SPI2,0x55);

Then, get rid of the setup_spi2.
There are two distinct ways of using SPI with CCS. #use spi with spi_xfer, and streams, or setup_spix with spi_write/spi_read (no streams). You need to use one or the other, not both.

Your #use spi, needs to select SPI2 and the correct format.

Best Wishes