Hi,
As we know, Common Mode noise in SMPS is always the hardest one to minimise
Is it? A common mode choke should work. Or a ferrite ring around the supply wires...
So we are thinking, can we take the earth out of this situation
You talk about common mode noise
current.
Indeed it can be minimzed by disconnecting EARTH.
By why and how is EARTH involved now in noise current?
* either just by stray capacitance
* or by installed noise filter capacitors?
If you now installed filter capacitors (usually "Y" type safety capacitors), then they can not work anymore, thus instead of noise current you get noise voltage....which again may cause radiated noise.
Klaus
Added:
You say it's a power supply installed at a light pole.
Is the light pole a metal pipe?
If so, the pipe should be earthed in either case and it will act as a shield and will reduce radiated noise.
Additionally one can expect that no critical device will be operated in close proximity to the supply.
Thus I recommend to check if you are using correct EMI limits. Maybe a more "realistic" EMI measurement may be used (higher distance of "receiver antenna", maybe DUT installed in a metal pipe during test.
If you do this, then you might add these "operating conditions" in the manual of the supply/light.