Brad Sawatzky's comments on scalers¶

In general the scaler trees for each arm expose the scaler data for
modules associated with that arm (likely stating the obvious). However,
a number of signals are fanned out to two scaler modules, one in each
arm. If you're running in 'COIN mode' (single DAQ for both arms), then
the counts for those duplicated signals should be (basically) identical.

Strictly speaking the content of the two scaler trees is entirely
controlled by the analyzer software configuration (THcScaler* configs).

Known reasons for discrepancies between the two scalers viewing the same
signal could include:
- There appears to be a problem with some of the "G0 scalers".
Apparently they have a bit-flip issue. Steve documented this here:
https://hallcweb.jlab.org/DocDB/0010/001058/001/Scaler_Problem.pdf
- If the scaler channels you're seeing issues with are in the same
physical modules, then that is almost certainly your problem.
- Mitigations are discussed in that file as well. (Carlos should
be familiar with this.)

- Electronic artifacts and low-level timing differences in the gating
between the two scalers
- Should only be a difference of a few counts at most on high-rate
channels

You do need to be mindful that the HMS 'trigger scalers' may be
different than the 'SHMS trigger scalers'. Each arm has a scaler that
counts the 'T1, T2,...' trigger signals plugged into each arm's Trigger
Master module. Those are not necessarily the same triggers, but the
names can be confusing because each arm can have its own set of T1, T2,
.... The trigger assignments are documented in HCLog entries, and
should also be reflected in the hcana scaler configuration git commits.

If you're looking at Coincidence-mode runs, and you think you're looking
at the same hardware signals in two different modules, then try plotting
the /difference/ between the two duplicate scaler channels for each
scaler read (note scalers are read out periodically, not on every
trigger). It would be interesting to see one duplicate is always
reading X% more than the other, or if they are generally identical
except form some occasional outliers.

the scaler dbase/param files should shed a lot of insight into
the hardware configuration (without having to get eyes on the physical
setup). For direct duplications, we typically jumper an entire 16
channel ribbon cable across 2 scaler modules, so you'll see two 16
channel groups in the dbase files that have identical names in the same
order.

If you see a situation where an HMS scaler channel and an SHMS scaler
channel have the same 'name' but they are not embedded in a block of 16
other identical channels, then you'll want to look a little more closely
at the details to see if they are really individually fanned out copies
of the same channel, or if it is a 'similar name' issue. For example,
one channel is the SHMS arm 'T3', and the other is the HMS arm 'T3'.
(In that case it is possible they are looking at different input
signals.)