August 11, 2004 Beams-doc-1301-v1 Rob Kutschke Earlier today Bob Webber sent around the figure bpms2.jpg which is stored in this document. He pointed out that there are a lot of noise points which appear to be associated with HPPA34 and VPPA35, but not with HPPA15 or VPPA14. In short the new BPMs produce outliers with a frequency of maybe 100 points per hour. The other two files in this document, noise.ps and noise.pdf show my investigation into this. The two files are the same, just in different formats. The top left plot shows the proton position at HA34 as a function of time for about 20 minutes this morning. I computed the position offline from the IQ data in the data logger. The top right plot also shows the proton position at HA34 as a function of time. But, in this case, the position was read from T:HPPA34. This is the same data plotted by Bob and I will call this the online data. The bottom plots show the projections of the top plots onto the position axis, with the maximum scale set to 10 counts to cut off the main signal and highlight the noise. Comparing the online and offline data we see: 1) The trivial sign difference 2) The online data has about 50 or so bad outliers for this 20 minute internal.but the offline data has none. 3) For 20 minutes at 15 Hz we expect 18000 data points. We see 17999 online data points but 18079 offline data points! I can easily understand missing a point or two. And I can easily understand goalpost problems giving one or two extra points. But I don't understand how we get 79 extra points! I poked a little by hand and I cannot find a way to identify which points are the extra points - so I can't argue that the extra points are the bad ones. My working hypotheses are: 1) The data is OK - somehow the online calculation is getting screwed up but the IQ data are passed on OK. There are other things about the online calculation which we do not yet understand. or 2) Maybe the big clue is the extra measurements - I have no idea how to track this down.