Picozed Rev.E won't boot from sd card with Out-of-Box designs or 2017.4 BSPs

Question

Hello,

I am facing an issue with the Picozed Rev.E we bought at our organization which has taken me a lot of hours to debug. My tools version (Vivado and Petalinux) is 2017.4 on Linux Mint 19.3 Xfce, 64 bit. I am connecting to the board using the Xilinx Platform Cable USB II for JTAG.

I designed a custom carrier card for the Picozed based on the Carrier Design Guide (http://zedboard.org/sites/default/files/documentations/5283-CDG-PicoZed-V2.4.pdf ). I downloaded the Petalinux 2017.4 BSP packages (both standard and enhanced) and created a petalinux project using the bsp. I am able to boot using JTAG and receive UART communication with no problem.

I cannot seem to manage though to boot from the SD card at all. I copied the files that are created under /tftboot, BOOT.bin and image.ub. If this is of importance, the BOOT.bin file I package it myself since it's not created automatically by running the command 'petalinux-package --boot --fsbl /tftpboot/zynq_fsbl.elf --fpga /tftpboot/system_wrapper.bit --u-boot'.

I have also tried to use the out-of-box designs with no luck at all. In both cases (use of BSPs and out-of-box design), the results I am observing are the same. Of course, I have changed the DIP switch to SD Card boot mode configuration.

I am not interested in the onboard eMMC, so the jumper resistors have been soldered so the MUX connects to a Murata WiFi Module, but at the moment, due to a pcb footprint error, the module and related smd components are not soldered.

I have tried different SD cards, all with the same result. The current one is a SanDisk Ultra 16 GB, class 10.

Some things I observe which I don't fully understand and would lilke some explanation:

During JTAG boot, I can see the SD card being detected, but I also see the message 'sdhci@e0100000: 0 (SD)Card did not respond to voltage select!'
After JTAG booting finished, running 'fdisk -l' shows no results, so it seems the SD card is not visible. My question here would be if it's possible to mount the SD card from JTAG boot, to prove that we can write and read to the card.
Is the JTAG somehow overriding all the other boot modes? No matter in what boot mode the DIP switch is, the JTAG can always boot if I run the command 'petalinux-boot --jtag --prebuilt 2 --hw_server-url tcp:127.0.0.1:3121'.
During JTAG boot, there is no clock towards the SD card, that is in both cases of DIP switch mode (JTAG & SD card). If I unplug the JTAG cable and reset the board, the clock signal is present. Why would the clock be affected?

I have tested the DIP switch to check if it does change the boot mode in case of a faulty DIP switch, but the connections are solid. At this moment, I am out of ideas on why I can't boot from SD card so any insight would be appreciated.

Thank you .

bhfletcher · Accepted Answer

You do need to have the data/cmd traces be +50 to +200ps relative to CLK. The skew is important. Of course, with the PicoZed system, the trace delays consist of multiple layers:

Xilinx package delay
PicoZed SOM trace lengths
SOM/Carrier Connector delay
Carrier trace lengths (on both sides of level translator)
Level translator delay

Are you using a level translator on your carrier? The FMC-V2 uses the TXS02612RTWR.

I haven't done the analysis myself, but I assume since the Avnet FMC-V2 Carrier works that the lengths were skewed correctly after adding up all the other delays. An easy thing to do is try to match the trace lengths on the Avnet Carrier. You can find the list of trace lengths here:

PicoZed FMC Carrier Card V2 Kit

PicoZed FMC Carrier V2 REV1 NetLengths_160318.zip

I did a quick calculation to add the traces on both sides of the level translator, then compared to the clock, then converted to picoseconds using 165 ps/inch.

You'd have to add in the Xilinx package delay, SOM trace lengths, Mating connectors, and level translator to get the complete picture. But, we know this works, so if you do something close to this, your's should work as well.

Bryan