# Count vs Count-per-channel

Hi there Wise people of the internet,

I am trying to do analysis some data gathered from a gamma scintillator setup, its stored in root.

So i have to do some coincidence measurements, and i found that in Krane, you normally use a TAC (Time to Amplitude Converter) to check for it. However, this gaussian peak expected there has count per channel in y axis. Should i just divide the counts that i recieve for each channel by channel number? I dont see how that will be useful.(this will disproportionately eat away any meaningful data stored in higher channels, which corresponds to higher energy,after calibration, breaking expected stable background of chance coincidence)

I am kinda new to detectors. Any help is much appreciated and also any resources/references will also be.

Regards

Counts/channel is perhaps a bit misleading from a dimensional analysis point of view. Take a channel. Counts/channel in that channel is actually just the count of events that showed up in that channel during the acquisition. It's dimensionless.

Don't divide by the channel number. If you want a rate take something like (counts/channel)/(channel width)/(acquisition live time).

its stored in root.

My condolences.

However, this gaussian peak expected there has count per channel in y axis. Should i just divide the counts that i recieve for each channel by channel number?

You should not.

"Counts per channel" tells you that this is a differential spectrum. If you wanted the total number of counts, you would integrate over all of the channels — which is the same as summing, in this case, because each channel is one channel wide. If you discovered through some valuation that each channel was 100 keV wide, then you could think of the distribution as counts per keV (with a conversion factor), but you would still integrate to the same total number of counts.

This becomes more important if your channels are unevenly spaced. For instance, I have done cold neutron experiments where you get equally spaced wavelength bins in a natural way, but you want to convert to get neutron counts per unit energy, with $$\lambda\propto E^{-1/2}$$. You have probably encountered this sort of thing in the Planck blackbody spectrum, which has a $$\lambda ^5$$ in the intensity per unit wavelength, but a $$\nu^3$$ in the intensity per unit frequency (or something like that), and therefore has a different "most intense color" depending on whether you are separating by wavelength or by frequency.