View source for Comments from M.M. Ito on FADC FCAL NIM 2010

Nice techniques, never seen stuff like this done so systematically.

My comments:

* p. 2: <s>"250MHz" should be "250 MHz"</s>
* p. 2: <s>"[1]" should be "Ref. [1]" or "Ref. 1"</s>
* p. 3: <s>"digitizing the same pulse twice" How about "effectively digitizing the same pulse twice". When I first read that it made me crazy thinking about how you do that. You don't actually do it, you just come very, very close. If you qualify it a bit, the reader will know to expect an explanation of what you really mean.</s>
* p. 5: "Any one channel may be compared to four different control channels" vs. "The uncertainties on the numbers are derived from the standard deviation of the resolutions for all eight channels." The former makes it seem like you are really going to compare to only one control channel at a time, the latter like you will use all four at once. I think the former is referring to the cross check only. If so that should be made clear, perhaps by putting the words "cross check" earlier in the paragraph.
** I don't know if I understand this.  In order to study a channel, one needs a control channel to do so.  The answer one gets is independent of the choice of control channel (the cross check stated in the paragraph).  One can study up to 8 different channels (each channel under study requires appropriate choice of control) and results can be summarized for these 8 channels.
* p. 5: "The average value for $a$ over all eight channels..." since $a$ is the slope of the pulse-height variation, how is it obtained for channels that have no pulse-height variation, i. e., the four control channels.
** Now I think I understand your confusion -- we can do everything with one choice of control channels and then repeat the process with a different choice allowing us to study all 8 channels.  I tried to make the text more clear.
* p. 6: Caption to figure 5, "for the other six channels". Again I thought that only four of the eight channels had pulse-height variation.
** see comment above
* p. 6: After equation (13) "the residual systematic contribution" you mean the scatter left after time-walk correction, no? That should be made explicit.
** it is made explicit in the next sentence
* Question: It seems kind of contradictory that the algorithmic contribution to resolution is bigger for the linear method than it is for the Gaussian (as the delay time is varied), but when used to anaylze the statistical contribution, the linear does better than the Gaussian. I would think that in measuring the statistical contribution, the algorithmic contribution is folded into the measurement. Is there an explanation?
** It is possible that the Gaussian algorithm is more sensitive to variations in shape than the linear method.  Once we allow channel to channel fluctuations in shape the resolution for the Gaussian algorithm is degraded significantly.  These statistical shape variations are not present in the signal channel studies.