Version:           1.6.0 (using KDE 3.5.9)
Installed from:    Compiled From Sources
OS:                Linux

The following was requested by the LFI team:

LFI reference loads are at a different temperature then the sky signal, therefore in order to reduce the 1/f noise it is necessary to include in the difference a scale factor r (sky - r * ref). 

The default LFI differencing procedure is to take the mean of 1 hour of sky and ref, compute the r factor as mean_sky/mean_ref, and then produce the differenced timestream by computing sky - r * ref. 

differencing plugin implementation
Inputs: 
  sampling_freq (the same for sky and ref) 
  sky datavector 
  ref datavector (of the same lenght and sampling rate) 
  timespan (hours) where to compute the r factor, default is 1, if empty the complete vectors are taken into account 

Pseudocode: 
  totaltime = (number of samples) / sampling_freq 
  for i  from 0 to totaltime/timespan:
    data_range = [ i * timespan * sampling_freq : (i+1) * timespan * sampling_freq ]
    mean_sky = mean(sky [data_range])
    mean_ref = mean(ref [data_range])
    r[i] =  mean_sky/mean_ref
    diff[data_range] = sky[data_range] - r[i] * ref[data_range] 

output: 
  r (vector containing computed r for each timespan, dimension is the number of timespans in the sky and ref datavectors) 
  diff (datavector of the same dimensions of sky and ref)