Electronic – the noise level as read on the oscilloscope

Looking at the zoomed versions, and assuming the X axis is time in seconds, you can see a 10ms periodicity in the strongest component, and many of the other components show the same periodicity.

That points at full wave rectified 50Hz mains as being the underlying cause. It provides a rich source of switching spikes and current transitions at 100Hz.

If I read the first zoomed one correctly, there is something oscillating at 27MHz, frequency modulated by 100Hz spikes. If you can identify the source of that frequency, you may be a lot closer to your answer.

Where you go from here is not so clear :

1. there could be EM coupling from high power supplies into sensitive circuitry : screening would reduce the electric field coupling, but magnetic screening (iron or mu-metal) may be required if the coupling is magnetic. (If your sensing circuit is a high impedance node, suspect electric coupling; if low impedance, suspect magnetic). Walk round the room turning everything else off while someone watches the scope and shouts when you make a difference. This includes heating, air conditioning, computers, and especially lighting since you mention fibre optics.

2. there could be noise on your power supplies. If you have sensitive amplifiers powered from a switching PSU, run them from a linear regulator instead. Improve decoupling. Measure the noise on your power and ground lines. Isolate any high current circuitry from this system. Adopt star earthing and eliminate any ground loops.

3. There could be noise on the scope's own power supply. Amplify the signal (without introducing any of the problems above) to make the scope's own noise less significant. Or buy a better scope...

Think about the problem. You want to see a small AC signal superimposed on a much larger DC voltage.

The obvious answer is AC coupling. That will block the average DC voltage, showing you short term variations from that voltage. With AC coupling and 10 mV / division, the noise should be about 3 divisions tall, although there could be bigger peaks.

One problem with this method is that you are basically high-pass filtering the signal, but the high pass rolloff frequency is chosen for you by the scope. Sometimes that's not optimal. For example, if you are looking for 100 kHz switching noise on the 12 V signal, but it also has some 120 Hz power line ripple, then that ripple may overwhelm the noise you are looking for. In such cases, make your own high pass filter. Put a cap in series with the signal, then a resistor to ground that is significantly lower than the scope probe resistance. That way your capacitor and resistor dictate the high pass rolloff frequency, so you can place it where you want by using different capacitors or resistors. In this example, 2.2 nF and 10 kΩ would work well. That puts the high pass rolloff frequency at about 7 kHz, which will attenuate 60 Hz by over 100 while leaving 100 kHz essentially untouched.