Other Parts Discussed in Thread: LAUNCHXL-F28379D
Tool/software:
In the recent past, we purchased 3 units of eZdsp F28335, socketed version. At the time, we were already using the launchpad 28379d in the laboratory, but we prioritized the use of eZdsp due to the external memory, as we considered storing several voltage and current curves of the plant.
During the first tests, we noticed errors in the stored curves (spikes) that did not make sense compared to the oscilloscope signals. We then verified that some data lines presented errors in the reading of specific bits.
The fault was present in all 3 units, but two returned to working correctly after cleaning the socket contacts. However, the third unit continues to have a problem in two data paths (lower byte).
To get around this problem, we write each float data twice, and the second time we shift the lower byte to the upper position. We transfer the curves to the computer using Code Composer and then use a program that recovers the curves by shifting bits. In addition to the delay in saving the curves, this strategy consumes more memory.
Has anyone experienced a similar problem? Is there any way to fix this? It's hard to consider scrapping a costly unit because of a contact failure.
