A study of worse energy resolution observed in E949
 

1. We observed a worse energy resolution in E949 data than that in E787 data, the difference is about 0.7 MeV. Also note that the total measured energy downshifts by about 1.5 MeV in E949 data ( see the comparison in the kp21 monitor data ).
2. Monte Carlo study shows there is no difference in the expected energy resolution between  E949 and E787 ( see the comparison in the kp21 trigger ) except for a much better expected energy resolution in comparing to those observed in data , which is still a mystery. Therefore, we exclude the possibility of any existing bug embedded in our analysis codes.
3. From a study of (sqrt(pdc^2+m_pi^2)-m_pi-Ers), where pdc is the pion momentum before entering the Range Stack, m_pi is the charged pion invariant mass and Ers is the pion kinetic energy measured in the Range Stack ( see the plots ), the Range Stack energy measurement looks unnormal. We separated the contribution from the Range Stack and the others (target and I-counter) and studied the correlation between the expected and the measured,  the Range Stack contribution in E949 data  is not so stable as that in E787 data . While  the other contributions in E949  looks quite stable as observed in those in E787 data . The 0.2 MeV worse energy resolution in E949 data can be the part of contributions for the worse total energy resolution (about 0.7 MeV worse than that in E787 data), but it apparently cannot be the main contribution. The Range Stack energy resolution in E949 data doesn't show too much difference from those in E787 data.
4. A further study on the ratio of the measured energy over the expected energy indicates that the energy deficit in E949 data is very likely due to the improper energy calibration  in the middle layers of the Range Stack, more specifically, from Layer# 6 to Layer# 11 ( see the comparison , where the black dots and the open triangles are for 1998 and 2002 data separately). We also observed a better energy resolution from Layer #1 to Layer#5 in E949 data, which is what we expected after replacing the inner 5 Range Stack layers. For some unknown reason, the energy resolutions beyond Layer 10 look a bit worse, but I am doubt it is the major contribution. Even checking the absolute energy measurement in each layer after requiring an roughly the same  incident momentum (|pdc-190|<5 MeV and |cos3d|<0.1), we cannot see too much difference between E949 and E787 ( see the comparison ).
5. There was a discussion before on the influence of higher Kaon flux. We investigated the possible correlation between the energy resolution and either C_pi, C_K, C_K/C_pi and C_pi+C_K ( see these plots ), there is no evidence indicating such correlation.
6. Well, then what causes the problem? Let's check the energy peak and resolution as a function of run for both  E787 1998 kp21 data  and  E949 2002 kp21 data . The E787 data shows a fairly stable energy peak position around 105 MeV except for some offset below run 38000, which we have already understood to be from the calibration. The energy resolution stays at 3.8 MeV if my eye inspection doesn't foolish me. While looking at the E949 data, apparently the energy peak position dances with the run number, and this is definitely due to a problem of energy calibration. Interestingly, the energy resolutions are not so active and  line up at 4.0 MeV (even a bit smaller). The difference of energy resolution (say 0.2 MeV at maximum) cannot be the reason for a 0.7 MeV difference  observed in the comparison of energy resolution between E949 and E787 Kp21 monitor data . Therefore, I would like to jump to my conclusion, the worse energy resolution observed in E949 data mainly come from the calibration. Lining up the energy peak position can simply fix most of the problem. If we can go further to improve the energy resolution of the target and I-counter,  then the resolution of energy measurement in E949 can be achieved as good as that obtained in E787.
7. Using a latest RS energy calibration, which takes care of the variation within a few runs, the energy resolution now looks in good shape despite of the annoying long tail effect ( see the plot ).