I am working on a custom design based on the i.MX 8M Plus (Quad) processor. We are using the MIMX8ML8CVNKZAB processor for design. I ran into failures using the mscale_ddr_tool _v3.30 when trying to test my board from -25C to +85C, the rated temperature of the LPDDR4, MT53E1536M32D4DT-046 WT:A. I was seeing failures at both temperature extremes and have solutions for each, but I am wondering if the software can be updated to help with them.
-25C : I found that the default boot software is running the ARM core at 1800Mhz, which is higher than the I-Temp version of the processor is rated for (1600MHz). I changed the "Clock" setting in the DDR Tool GUI to 1600Mhz, but the output was still showing 1800MHz for the ARM clock. A setting of 1200MHz for the "Clock" did work in the output, and my LPDDR4 Stress Test passes at -25C with this setting. Is it possible to update the software to support the I-Temp version of the processor and the max speed of 1600MHz on the ARM core? Since the RAM is still operating at 2000MHz I am pretty confident in our system design, but it would be nice to run everything at full speed.
+85C : I was seeing failures on the RAM when the ambient temperature reached about +75-80C. I was able to get the boards to pass by manually turning off the DERATE_ENABLE and setting the T_RFC_NOM_X32 to half the value the RAM chip says for room temperature operation. A similar scheme was utilized for the Mini/Nano version of the processor to work around a silicon bug (ERR050805). However, it is not clear to me why I need to make this adjustment for the Plus processor; I was expecting that the DERATE_ENABLE should be working for the Plus. Can you tell me what the RAM controller/code does if the RAM chip MR4[2:0] (Refresh Rate) reports back 3b111: SDRAM High temperature operating limit exceeded? I would think it keeps operating at 0.25x refresh, with de-rating, but I wanted to verify. Related, I noticed the RPA spreadsheet sets DERATE_MR4_TUF_DIS to 1 to disable the Temperature update flag from the LPDDR4 chip; I assume that means the software/controller checks the Refresh Rate on a regular basis instead of using an interrupt?
Thanks
