We are having a problem with the waveforms of EMI of i.MX28.
There are overshoots/undershoots which are more than 300mV on Address/Command signals.
The image is A6 probed at a via near DDR2.
Are there any way to reduce the shoots from Address/Command without modifying hardware?
The conditions of our board are as follows:
-The board is 6-layer.
-The pattern length of each EMI line is 30.2-34.0mm.
-There are no dumping resistors on the Address/Command line.
-The only register settings which is modified from the reset value is bit 1 of EMRS(1), or HW_DRAM_CTL183. (Reducing the drive strength of DDR2.)
-Adding dumping registers to the Address/Command lines are too risky.
The board is our original design and the pattern of EMI is not copied from EVK since our board is too small to do so.
Instead we simulated the pattern before manufacturing the board and the result was that the overshoots and undershoots are less than 300mV.
Since the vias we used to probe were a little apart from the balls of DDR2, the shoots are not as bad as it seems at the balls.
But as there is no way to observe the exact waveform at the ball, we'd like to improve the waveform we can see.