Conducted Emission Fail at Low Frequency

I have the same problem in our lab. I have what I think is a symetrical two wire AC input. However, powered through the LISN, the reading the L and N measurements, one is much bigger than the other. If I switch the input lines, then the L and N measurements switch. I don't see how the LISN can see different things when I swicth the wires. To me the view into either of the two lines is identical. RC

Does your customer actually required that a "Peak" reading be under the "Average" limit. In the graphs the Red line is the Quasi-Peak limit (QP), while the fuchsia line is the Average limit (AV).

If it is still a problem you should verify the filter component being used is properly grounded (if there is a chassis type ground). Also check if the problem frequency matches a specific area of the system (for example a switcher supply near 205khz). If the noise can be attributed to a specific area the inputs to that section should have added filtering, (for example proper filter capacitance on an input line).

With the noise unequal between the battery terminals it may be that the internal positive line already has some extra filtering. The customer might consider adding a similar filter network onto the negative line. Even a simple capacitor across the battery lines (internal) might help to average out the noise peaks and bring the negative line under the "so called" limit.