This issue was already resolved by the time I'm posting this, but I want to make sure it is documented.

The problem was initially seen as a hit in the z distribution of TPC hits (actually, clusters).

But this is more likely to be a peak in the time distribution. Here is z vs. timebucket for TPC clusters (shown in log z, which suppresses the bump at z = +/- 100cm, but highlights the noise):

There appears to be some noise at timebucket of ~500, but it is minor. The timebuckets should stop just short of 400, where we record some "post-membrane" hits which are useful in understanding pileup (the hits are not from the triggered event).

Anyhow, I can plot the cluster position vs. cluster size (in time buckets, log z):

Here we see that the feature in question is a peak of maximum cluster width (31 buckets) generally centered at time bucket 185 (approximately the center of the time bucket range). If I cut tightly on this peak cluster position and width, I can see where these occur in the TPC. Here is the sector vs. padrow, and then sector vs. event and padrow vs. event for the first 1000 events of a file:

What is clear is that these clusters come from a few specific places in the TPC, and they appear and disappear with time.

G. Van Buren
gene@bnl.gov

Tonko looked at this and replied in the hypernews on March 6, 2009:

Gene showed me the timebin plots of the TPC
and there is a very pronounced peak at his
timebin 185 (which corresponds to TPX timebin 207)
in the cluster timebin distribution.

This effect comes from whole pads flaring up
causing a "cluster" to be found in this mid-range
timebin (415/2).

I saw some effects like this at some point but
I thought I understood them and got rid of them.
Obviously not.

While I look into this, I disabled cluster
finding for pads where the whole pad (all 415 timebins)
is lit. I simply skip those pads. The data is worthless
under this case anyway.

There are about 200-250 of such pads in every event
and they seem to move around event to event. I have some suspicion
this is due to an ALTRO bug and that they depend on
the past history, the specific data of the preceding
event. No such effect is observed in pulsers or empty
events but I'm going to re-verify this as well.

This is the reason, I believe, that the effect is 0 for
pulsers/empty and gets more pronounced as more
charge hits the TPC at the last timebin(s).

So from now on one should not see these peaks
in the timebin distribution and if you _do_ see them
please let me know ASAP!

       -- Tonko