How to take background counts

[Richard Hull]

Background counts, just like regular counts, tend to rely on the stability of the instrument for repeatability.

I have recommended a pre-fusor run background count followed by a post-fusor run background count. Then find the mean (average) of the two backgrounds as your main background count.

In most real, decently working fusors, a good BF3 counter should be clicking away in steady runs to the tune of over 100cpm with more than 300 cpm being more the norm. In this case, background is of little real concern at even the 0.5 cpm level.

Backgrounds become important for detecting the edge of fusion or when working with known weak sources of neutrons. (cobbled up sources, CF or LENR work if you should be so inclined.)

In an effort to secure the most accurate and repeatable backgound, turn your neutron counter on for at least 20 minutes in a temperature stable environment near where the source will be measured. Note* It is best to place the counter in one spot for all three counts, (PRE, DATA, POST) and move the source into the counter's location. (or, in the case of the fusor turn it on and off as needed). By leaving the counter on for a while in a temperature stable environment and not moving it or disturbing it, you warrant a modicum of stability over a range of possible error inducements.

Take the first background (pre-run) count after twenty minutes of on time and run a good long time. Thirty minutes is good, but sixty minutes is better in ultra low count situations. Take your data count. Finally, your post-run count over the same period as the Pre-run count. Average the two to find the mean.

Let us run a simple example.

pre-run background count average, 0.56cpm taken over 1 hour
data run .96cpm taken over 10 minutes.
post run background count average, 0.69 taken over 1 hour

Total average background would be 0.625 with a range of 0.13

Subtracting mean background, our experiment's data recorded a total of 0.335 cpm

The range error of our two backrounds (0.13) taken 10 minutes apart is rather large compared to the actual data average of .335 CPM This range error occurred over a total of 120 minutes and accumulated at the average rate of .13/120 or ~.001cpm/minute so for our 10 minute data run in the middle we might expect a range error of 0.01cpm over our data run period. This looks a lot better against the 0.335 signal data.

We can assume we have real neutrons, albeit a very weak source.

Classic statistics would demand several runs for data and background. However, we have seen that the background changes very unpredictably and over short periods. The neutron producing experimental parameters can change from run to run as well, especially in amateur work. So a multiple run check of data and background would often be like comparing apples and oranges as they rot or grow. Both the experiment and the background wobble.

We are forced to accept one run statistics when the neutron count is near the noise level, (the worst possible scenario in statistics), and pray that the means show significance against the ranges.

Again, a really nicely working fusor will be so far out of the noise and background that even taking a background might be an effort in overkill. An average of 390cpm data run neutrons with an averaged background of 0.63 cpm background with a range of 0.13 is not worth the time it takes to subtract as you are almost 3 orders of magnitude above background. With poor statistical repeatability, a short run trio count should, hopefully, be 50% above background.

Richard Hull