I just realized that the negative bin widht is not explicitly mentioned in the quoted paper. So let me explain it here:
The negative value of cell 498 is correct and "real" in the sense that the signal is first captured in cell 498 and later in cell 497. This is due to the exact layout of the cells on the chip and the input signal. Cell 498 is simply much closer to the input, so sees the signal earlier than cell 497, even if it's triggerd after cell 497. So nothing to worry about.
If I remember right the negative sampling width happens only for 498 and at high sampling speeds. It is described in a paper from Stefan:
“Novel Calibration Method for Switched Capacitor Arrays Enables Time Measurements With Sub-Picosecond Resolution”( IEEE Transactions on Nuclear Science 61 (2014),Nr. 6, 3607–3617)
I am using four DRS4 v5 eval boards to digitize 16 channels of data. I have recently changed from saving the timing information of the waveform using GetTime() to GetTimeCalibration(). When changing over, I noticed that some values for fCellDT for cell 498 are negative. Over the 16 channels used, 4 of them have negative time bin widths for cell 498 while the other 12 channels are very close to 0 (in the ~10 ps range). One of the eval boards has no negative fCellDT whereas the other three boards have one or two channels with negative values.
Upon further inspection, I checked the time between samples of GetTime() and found the same results in cell 498. After finding this, I did a timing calibration again with CalibrateTiming() even though in a different post on the discussion forum you said it was valid for a wide range of temperatures and a long time (years). This still allowed the negative fCellDT values to persist.
Is this a common occurance? If so, is there a method to fix this issue? Is there a reason for cell 498 to have a small value for fCellDT? I searched the discussion forum and did not find anything relating to this issue.
Attached are a couple waveform traces using GetTime() zoomed in on cell 498.