Suggestions on data to take for setting up the Muon Detector.

1) The best way to tease out the noise is to increase the width
of the stretched pulse. Then you can try to extrapolate back down to
see what the noise rate is for the smaller widths.

2) Keep in mind, the number of real events per minute is small.
You should take a few long runs to get good statistics.

So, you can take a several minutes long run using the settings of 
the last group. Then try some shorter runs increasing the width e.g.:

Tube V	,Thold	,Width	,#Sngl	,#Coinc	,Time 
1300 (V),42(mV)	,20us	,15000	,300	,50min
1300 (V),42(mV)	,2ms	,3100	,102	,10min

Your data is unlikely to work as well as this example, but lets take a 
look at what we have. At the "standard" setting we expect:

random noise counts of (15000/3000s)*(15000/3000s)*(20e-6)*3000s = 1.5 (tiny!)
            noise rate ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
This implies we see mostly signal at a rate of about 6/min

when we increase the width:
random noise counts of (3100/600s)*(3100/600s)*(2e-3)*600s = 32 

Notice in order to get a signal rate of about 6/min again, we'd
need 112 coincidences. But since this is a random process, there will
be a spread in the returned values. I've greatly increased the time
over which you would normally take data to get the numbers to work 
better as a demonstration.

What does the noise rate imply? Lets assume that we get 6 signal (or real) 
events/min. In a week of data taking this amounts to 60480 real signal events.
If our noise rate is 0.003/min, in a weeks time, we expect 300 events due to
noise. This is a pretty low amount of noise. Typically we get ~1.5% of the
data at random noise, your job is to try and do better than the last group
if possible.

Do try to see if your data taking numbers make sense! As a suggestion, you
may want to take some longer data runs at +/- 100V,+/- 200V  of the standard 
setting and at +/- 30mV (if possible) of the disriminator setting. A possible 
data taking run might look like:

Tube V	,Thold	,Width	,#Sngl	,#Coinc	,Time	,Noise estimate 
1300 (V),42(mV)	,20us	,	,	,5min	,
1100 (V),42(mV)	,20us	,	,	,5min	,
1200 (V),42(mV)	,20us	,	,	,5min	,
1400 (V),42(mV)	,20us	,	,	,5min	,
1500 (V),42(mV)	,20us	,	,	,5min	,
1300 (V),72(mV)	,20us	,	,	,5min	,
1100 (V),72(mV)	,20us	,	,	,5min	,
1200 (V),72(mV)	,20us	,	,	,5min	,
1400 (V),72(mV)	,20us	,	,	,5min	,
1500 (V),72(mV)	,20us	,	,	,5min	,
1300 (V),102(mV),20us	,	,	,5min	,
1100 (V),102(mV),20us	,	,	,5min	,
1200 (V),102(mV),20us	,	,	,5min	,
1400 (V),102(mV),20us	,	,	,5min	,
1500 (V),102(mV),20us	,	,	,5min	,



Then choose "best" setting and increase the width to check the noise:
1200 (V),42(mV)	,200us	,	,	,1min	,
1200 (V),42(mV)	,2ms	,	,	,1min	,
1200 (V),42(mV)	,20ms	,	,	,1min	,
1200 (V),42(mV)	,200ms	,	,	,1min	,

(be sure you can back up your reasoning and choice)

The data taking method and times etc. are really completely up to you.
Keep in mind, dealing with small statistics (less than 40!) can be
a tricky business. For instance, 40 counts is just as likely as 
32 counts if the real number of counts is most likely 36. (Implies
that 6 counts is as likely as 12 counts if the most likely number
of counts is 9) This "simple" reasoning gets more complicated at 
very low counts where Poisson statistics take over.