# Is there a way to extract the difference between the 2 signals? Fitting?

I'm very new to signal processing and have been struggling this far. I have a device that measures the signal due to the magnetization of iron particles. In order to calculate the response of the particles, I did 2 measurements:

1. Background measurement, so without particles, so the signal should be 0, but as the figure shows, it is not.
2. Measurement with particles.

The problem is that even at the background measurement I get a sort of signal, because of many complex reasons like field through, in the device, there is iron because of the production process and so on... It is true that the magnitude of the signal of the base measurement is 2 orders lower than the one with the particles but the shape of the signal and the peaks are already to be seen at the base measurement. So one should say, that the signal I got while having the particles in the device, does not only come from the particles.

• Question: now my question is how to extract the signal that only comes from the particles? A simple subtraction is not the answer, as the magnitude is not alike. Filtering in the frequency domain did not help as the particles are simulated with a frequency f0 and in the signal, we just collect the harmonics. Can please anyone help?

My suggestion:

• take out the peaks at 44 and 65 because they are certainly not from the particles, as one can definitely see them in the base measurement without the particles.
• Interesting question. If the background reading is 100 times smaller, what SNR do you need? That level seems good --- and ignorable (i.e. Just use the signal with particles as-is). Telling us more about what you want to do with the filtered signal might help. – Peter K. May 14 '17 at 19:05
• Thank you for your response Peter. The SNR is good yes. But I find it questionable that the shape is the same. The filtered signal should give ONLY the reponse of the particles. But if the signal have the same peaks with and without the particles (like at bins 44 and 65) that means that those peaks do not come from the effect of the particles, so they need to be taken out, don't they ? – Tassou May 14 '17 at 19:09
• You keep saying"the shape can already be seen without particles","the peaks are the same",and you infer that this means that what you see is somehow distorted by a seemingly very strongly amplified base signal – but I must admit I don't really see how that's justified.It's just as likely something modulates the entity you're observing or the observer in just the same manner.So,what did a rudimentary calibration by division through the "base signal" yield?Again,we're not experts on the physical aspect of what you want to observe,and I'm not able to "imagine" how the signal should look like! – Marcus Müller May 15 '17 at 7:44
• the y axis is in a.u. derived from a voltage.The signal I want, is the change caused by the particles. The first figure should describe it because the signal is background corrected, but I think that the background correction, which I did not implement my self, is not really working as the second figure exhibits a signal with no particles and the shape is really similar. the particles are a delta-probe, so they should yield one peak at each half period. – Tassou May 15 '17 at 8:10
• "They should yield one peak at each half period"; is this a synchronous system where you have knowledge of the "period" mentioned, perhaps with some interrogation signal? If so, then perhaps multiplying by twice the interrogation signal and low pass filtering will reveal your signal of interest as a DC signal level. – Dan Boschen May 15 '17 at 9:25