Joe Seele found some evidence that high p_T tracks were being lost due to a single line of code changed between SL10f and SL10g in Sti/StiNodePars.h. So I tried processing ~1000 events from a Run 9 st_W file (st_W_10103041_raw_4180001), yielding almost 1M global tracks with flag>0, in both the standard SL10g, and SL10g_mod (where I changed the single line of code back to that from SL10f, and also re-compiled the dependent libraries of Sti, StiMaker, and StDetectorDbMaker). I then used code I had to match tracks between the two productions, requiring the following for a match:

• Same event
• Same charge sign
• |Δη| < 0.1
• |Δ(Z_proj)| < 12 cm
• |Δ(Ψ)| < 10 °

where Z_proj = Z_first - R_first * (Z_last-Z_first)/(R_last-R_first) [roughly a projection to the z-axis]. Nearly all tracks got matched (910 are unmatched from SL10g_mod, and 892 from SL10g, with 973214 matched tracks), but it's interesting to learn from the small differences.

First, the effects on matched tracks are essentially negligible. For example, the h^-/h⁺ issue is on the scale of d(1/p_T)/q ~ 0.002, but the differences here are on the level of 10^-(5±1) or so.

Turning to the unmatched tracks, these are the log₁₀(p_T) and N_hit distributions unmatched in SL10g (back set) and SL10g_mod (front set):

There's definitely an alteration of high p_T tracks due to the code change. There seem to be some effects in the N_hit distribution too.

Here are the distributions of unmatched/matched tracks for these same two observables, with red for SL10g_mod, and blue for SL10g:

The ratio for SL10g_mod exceeds 1 at very high p_T (so more tracks are unmatched than matched)! The N_hit distributions are similar. The same data is shown here in log scale:

Plotting vs. both quantities (for SL10g_mod on the left and SL10g on the right, set to the same vertical scale):

So the unmatched tracks are worst from SL10g at high p_T and perhaps even increasing in percentage with increasing N_hit, until the statistics expire at high p_T and high N_hit. Here are the stats from the unmatched tracks.

-Gene

[note: this page was updated from an initial version after slightly loosening cuts to ensure unmatched tracks were really un-matchable, and after a bug fix which prevented some unmatched tracks from being recorded]