ex_fat.c example - mmcsd_init()

[miro]

Hi, is there any hands-on experience with the ex_fat.c example?
An sdcard connected to dspic33 (all 3v3, pullups on cs, sdi, sdo), all signals are there.

The function r1=mmcsd_go_idle_state() in mmcsd_init() always returns 0x80, so I get "ERROR INITIALIZING FAT".. I've decreased the SPI speed at the beginning of mmcsd_init() with

setup_spi(SPI_MASTER | SPI_MODE_0 | SPI_CLK_DIV_512);

and tried mode 0,1,2,3.. The fram device on the same SPI1 (but different /CS) works fine. M.

[temtronic]

The driver isn't at fault as CCS doesn't publish anything they haven't checked out several different ways.

You are aware that not ALL SD cards are the same ??? The 'standard' is not always followed !!

Simple test101...

does the card work in a PC ?

can the PIC read data previously stored from a PC ?

who made the SD card(mfr/make/model) . What are the specs (timing/pullup values,Vcc tol/etc.)

You've tried the 4 SPI modes but WHICH one is correct for your card ???

[miro]

First issue I've found is the bug in setup_spi():

.................... setup_spi(SPI_MASTER | SPI_MODE_0 | SPI_CLK_DIV_512);
04994: BCLR.B 240.6 <<<< this shall be 241.7
04996: MOV #1A0,W4
04998: MOV W4,242
0499A: BSET.B 241.7
So it does not set the proper spi clock rate (measured with LA).

Done manualy:
.................... #WORD SPI1CON1 =0x242
.................... #BIT SPI1EN =0x240.15
.................... SPI1EN=0;
0498C: BCLR.B 241.7
.................... SPI1CON1=0x120;
0498E: MOV #120,W4
04990: MOV W4,242
.................... SPI1EN=1;
04992: BSET.B 241.7

That sets the spi clock rate properly..

[miro]

@temtronic: did you run that successfully?? Or you do think it works?

[FvM]

With "bug in setup_spi()" you mean, that the module isn't disabled before reinitializing it,
e.g. to switch the SPI speed to high after card initialization?

This can be easily achieved with setup_spi(0)

[miro]

That bug means (at least with dspi33) the spi frequency does not change if you change the prescaler value in setup_spi().

[Ttelmah]

Miro's bug is not with the MMC code, but with the settings for his chip.
CCS (unfortunately rather habitually), adds chips to the database, but often gets the control bits, port addresses etc., wrong. In most cases these are fixed after a few more releases, but in some cases they stay for a long time.
You need to contact CCS, with the chip number and compiler version number, pointing out that the code is accessing the wrong bit in the setup_spi function and hopefully they will fix this in the next release.
This is not a fault with the ex_fat example, but is specific to your chip.
If you have the IDE, then you can change the bits used yourself via the device editor.

Best Wishes

[temtronic]

yes, code is good....

what PIC are you using ?

[miro]

However, not to waste too much time - is there somebody who is running the example with an sdcard (does not matter which chip, 4GB SDHC cards)? I've seen some patches related to fat32 but frankly I am in doubt whether the code actually works (ie the timeouts are mostly 0xFF which is a too small value for todays cards, etc). M.

[ckielstra]

Most libraries I've seen are lacking the initialization code for the SDHC cards, so only cards up to 4Gb are supported.

[asmallri]

[FvM]

[miro]

@FvM: the change of 241.7 did the trick - SP1STAT, bit 15, SPIEN.

This works as well:
SPIEN=0;
setup_spi(...);

[FvM]

As said, the problem occurs when the SPI interface had been enabled before. setup_spi(0) also helps in this case.

The problem is mentioned in the PIC24F family reference manual, but possibly missing in the dsPIC30/33 documents.

[ckielstra]

Has someone reported this to CCS?