stephenbrooks.org : Quarantined Results and Retesting

Stephen Brooks

Fri, 27 Apr 2007 15:21:22 +0000

Ah, I think I see why - in the PhaseRotB lattice


seed 0
seedpitch 1

...so each recheck chain uses the same random seeds (0,1,2,3,4) for decays and some randomisation of the initial particle production. That means #3 just uses an "unluckier" set of decays (probably early on).

Stephen Brooks

Fri, 27 Apr 2007 12:39:09 +0000

That is weird! Can you send across (e-mail) the raw data from those 25 rechecks (x4 runs, I guess). Do you count Mpts or muon % or both?

[DPC]DanC

Wed, 25 Apr 2007 00:44:27 +0000

Oops... Let's try that again. My changes are WAAAY down the pipe from the first five solenoids though. My current run just finished #3 predictably: 1. 617.805 2. 621.357 3. 552.617 The run before this went 614, 618, 554, 587 This may just be some anomaly that I'm seeing, and it has no bearing on anything... I just thought it somewhat odd. You'd think occasionally a #3 run would be higher than #1 or #2. Out of the last 25 rechecks - #3 is always significantly lower than 1 or 2. 4 usually balances 3 out a little, but is also lower than #1 or #2.

Rgtx

Tue, 24 Apr 2007 20:33:05 +0000

From my latest queue.txt file "0.595414,0.568889,0.580954,0.587214 (409.0 Mpts) [v4.43d] <DecayRotB>". This is from modifying the first five solenoids, where it can be seen, that of the first four trials, the second run gives a lower muon%. I would guess, that the spread is due to the designs themselves.

[DPC]DanC

Tue, 24 Apr 2007 15:45:51 +0000

Since I no longer have many boxen to produce volume, I've been concentrating on quality - and trying my own design mutations. One thing that I find curious about the rechecks is that (seemingly) EVERY time - the third run on a quarantined result is (seemingly) always VASTLY lower than the first. This pattern seems to hold regardless of design. It would seem that at least occasionally - the client would produce similar results when retesting an identical design. So - on to the question. Is there really that much standard deviation from result to result?... or is the observed behavior more a part of the design, and I'm just missing something? How is the seed used, and does the seed account for the deviation? Not really a gripe, just trying to understand, and improve my designs. Cheers! DanC