STM32F4 eMMC CMD signal voltage levels

[shannon2]

Posted on July 12, 2018 at 22:23

I'm developing a system that uses an STM32F4 with an eMMC from Micron.  The issue I'm having is related to the voltage levels I'm seeing on the CMD line.  When the STM32 is driving the line I see good 3.3V levels.  However when the eMMC is driving the line the peak voltage drops down to around 2.6V.  This causes erratic data values to be read by the STM32.

I'm using a fixed 4.7k pull-up on the CMD line and the layout is as close to 50 ohm as I can get.  Very short trace as well (<1600 mil).  Signal edges look fine it's just the level that is a problem.

Is there some configuration I'm missing in STM_Cube?

I've seen one datasheet from ISSI that mentions the use of a switched current source or resistor for initialization that gets switched out later but they didn't give specifics.  I feel like I'm missing something simple.

Ideas?  anyone else with this problem?

[shannon2]

Posted on July 12, 2018 at 22:32

Oh I should add that I tried a second assembly with an ISSI part instead of Micron and the response/signal levels were identical.

[Tesla DeLorean]

Posted on July 13, 2018 at 00:02

I wouldn't expect pull-up resistors to have a great deal of impact against a push-pull driver running at a lower voltage. I'd think 2.6V would be clearly in the VIH range. Most of my systems run in the 2.85 to 3V ranges, but I've not reviewed the eMMC bus in that regard. You should perhaps look at the IO voltage applied/selection within the eMMC is calling for.

Had a reasonable about of success with SanDisk/WDC and Samsung eMMC from 4GB to 128GB, on boards including the STM32F412-DISCO and NUCLEO-F412ZG

I'll port a demo/integrity app to a specific platform, I'd need to know HSE clock source type/speed, USART and pins for debug/diagnostic output, and SDIO pins if beyond the stock PC[8..12] and PD2 configuration.

[shannon2]

Posted on July 13, 2018 at 01:06

3 weeks I've spent on this....sigh....  Looking more deeply at the hardware if found the entire power plane for the eMMC is not connected to power.  The device was running parasitically from the data lines of the STM32.  Groan.  All fixed now.

Clive, the next step is to connect this eMMC to the fat file system.  I dread working through this.  I'm a hardware guy not a software guy.  This is a commercial product.  Any chance we can work with you on a consultancy basis?  Is that something you do?  I've read some of your other posts on eMMC and it seems you have had some success in this area.

[Tesla DeLorean]

Posted on July 13, 2018 at 12:39

Been there before, floating islands, where the netnames, or power symbols weren't identical, and now you have two nets where you wanted one.

Yes, replied to you privately.