i.MXRT1021 SDcard Issues

[andrew58210]

We have continued to be plagued with SDcard issues with this processor.  We have adjusted the layout to have a total trace length of < 2" and all the traces are matched to within 0.005".  We are only using 3.3V I/O and have disabled any UHS speed modes that require 1.8V I/O.  We limit the maximum clock rate of the card is 50Mhz through the driver.  We also implemented the Dev Kit SDcard power cycle circuit as it has become clear that is necessary for some odd reason.  What we are finding are the cards are extremely sensitive to CPU pin drive strength.  R0/3 and both R0/2 look really good on a scope but we get errors at cold and hot conditions.  R0/2 as much fewer errors.  R0/1 behaves much better but the Clock and Data lines from the processor look like a triangle waveform.  Certain brands of cards are more sensitive than others (we tested 9 brands)  We see several failure modes.  The card will not mount, it will report nearly 100% full in the FAT file system after mounting, or it will fail at temperature extremes (-40C or 60C).  This does not make sense and my belief now is that there is a setup time violation when the CPU is supplying the clock and the SDcard is supplying the data.  Is there some internal adjustment that the SDK driver is not making as to when the data should be clocked in when it is performing a read?  In all my years engineering and having done numerous design with SDcards, I have never seen a CPU be so stubborn and sensitive for a peripheral that is a universal feature on CPUs

[EdwinHz]

Hi @andrew58210,

Are you certain that the SD card lines comply with the RT1020 Hardware Development Guide specifications from section "7.6 High-speed signal routing recommendations"? You could also refer to the RT1020-EVK Design Files for reference of the layout. It is most likely that the issue you are seeing is HW related, as there isn't any SDK adjustment like what you are refering to.

BR,
Edwin.

[andrew58210]

We are certain they comply.  Total trace length is 1.5".  Everything is length matched to within 0.005".  Layer stack is setup for impedance to target 50ohms.  I find it interesting that the EVK SDcard demo has the drive strength set to R0/1 which at 3.3V gives poor looking signals with a triangular shape.  

On our hardware the only drive strength combinations that "work" over -40 to 60C are the following:

-  R0/1

-  R0/3 (Data & Cmd), R/02 Clock

-  R0/2 (Data and Cmd), R0/1 Clock