Log entry time 16:41:10 on September10,2009

Entry number 288994

This entry is a followup to: 288949

To eliminate the "glitches" in the double differences which
Ahmed observed (run 12208 in halog 288949) a cut on the so-called
"DAC noise glitch" of the 18-bit ADCs is necessary and sufficient.
This is proven in fig 1.  The outliers disappear when cutting
within 20 events of the glitch, for example.

More on these glitches:  They happen once per 256K events, so
they happen once per 2.37 hours at 30 Hz.  They happen on all
channels of the ADC simultaneously and on all the ADC18 in a 
crate simultaneously.  And we don't know why.  They are a 
universal feature/bug of the ADC18.

See fig 2 for the raw ADC18 data for this run 12208.
One simply needs to find the discontinuity in the base, 
which is a variable like adcx0_0_b0 for adcx #0 channel #0.