Best utility for formatting SD Memory Card ?

Ttelmah · Joined: 11 Mar 2010 Posts: 20058

Form the SD spec:

"All hosts shall provide pull-up resistors on all data lines DAT[3:0] as described in section 6 of the SD Physical Specification Version 1.10."

This actually ensures the lines do not float.

The pull up is also required to protect the lines from floating (another thing that can lead to excessive consumption), when the card is removed.

Table 53, specifies a maximum resistance of 100K, and a minimum of 10K on the CMD line and all four data lines.

"RDAT and RCMD are pull-up resistors protecting the CMD and the DAT lines against bus floating when no card is inserted or when all card drivers are in an high-impedance mode.
The host shall pull-up all DAT0-3 lines by RDAT, even if the host uses SD Memory Card as 1 bit mode-only in SD mode. Also, the host shall pull-up all "RSV" lines in SPI mode, even though they are not used."

Note the 'shall'.

I wonder what 'dollar' he is actually using?. The prices mention seem very high. Genuine Kingston 2GB SD cards sell in one off from the manufacturer in the UK, for less than 5$ US, falling if you buy ten off.
It'd seem he could save more money by getting somebody to shop around.

soonc · Joined: 03 Dec 2013 Posts: 215

soonc · Joined: 03 Dec 2013 Posts: 215

Ttelmah · Joined: 11 Mar 2010 Posts: 20058

On the card pricing, I found it cheaper to buy Kingston cards from their own retail shop than from their 'suppliers'...

In V6 specification, you will find it says 'the host shall not leave these unused lines floating". Note the 'not'....
The requirement to keep the PIC DI line from floating is a general one that any undriven line should be biased high or low when it is undriven (when the card is removed).

soonc · Joined: 03 Dec 2013 Posts: 215

In the final analysis the reason for success and failure with several different cards is really due to the mmcsd.c does not comply with the recommend initialization requirements for modern SD cards.

It is MANDATORY to issue at least 74 clocks to switch ANY SD card into SPI Mode.

The mmcsd.c does not do the "at least 73 clocks". During my initial trials I noted that after
I added code to send 80 clocks things started to work.

This begs the question how can other people get mmcsd.c to work at all?

Frankly I don't know: I do know once I sent 80 clocks the card started to respond.

The next big thing wrong with mmcsd.c is the lack of sending CMD8 after successful CMD0.

Version 2.00 or later compliant Cards will most likely fail using the standard mmcsd.c code
because CMD8 is not in the code.
I think the mmcsd.c is really oriented towards the MMC cards.

Some of the cards I have require the CMD8.

Cards that may not be 2.00 or later compliant will succeed without CMD8, and fail with CMD8 which is the signal to proceed with the code as if the card if less than V2 compliant or is a MMC card.

That was the case with the one card I have that always worked, and the others were V2 compliant and failed.

So basically all the suggestions about poor hardware were simply WRONG but may help in
situations where circuit board layout and power supply are lacking proper design.

As I said at the time that was not the issue in my particular case.

If using HARDWARE SPI mode Pull up Resistors are NOT required if the supply and board layout are done properly.

Referring to Elm-Chan's excellent page on how to use the SD card he recommends a single pull-up on
the DO line or some SD card may not initialize. All of my cards initialized but failed
because of the other reasons stated above.

My comment: I suspect (but cannot prove) when SOFTWARE SPI mode is used on a non-spi compliant port pin the resistor compensates for the improper output driver that most SPI compliant ports have.

Link to Elm-Chan: http://elm-chan.org/docs/mmc/mmc_e.html

Link to SPI Flow Chart: http://elm-chan.org/docs/mmc/pic/sdinit.png

My lesson learned is: READ THE SPEC..... !

I was in a hurry and as is the case "the hurrier I go the slower I gets there."

Thanks for all the help and suggestions.

asmallri · Joined: 12 Aug 2004 Posts: 1663 Location: Perth, Australia

Ttelmah · Joined: 11 Mar 2010 Posts: 20058

and on the74 clock, yes it is mandatory to issue this many clocks, but there is a section in the SD paperwork that says it is acceptable to do this by issuing repeated status fetches from the chip. I must admit I prefer to see the code issue 74 simple fetches, and only then ask for he status, but is is acceptable to do it the way CCS does it....