Afterglow Studies

Afterglow StudiesEric Torrence

University of Oregon

183nd LMTF Meeting10 October 2013

2Eric Torrence September 2013

Overview

•Test afterglow model by single-bunch addition

-Originally studied by Mika

-Will need to be used for 25 ns operations

•Try to test assumptions in method

-Afterglow is universal with time and μ

-Afterglow is additive to the prompt luminosity signal(not true if there are migration effects)

•Look at expected afterglow levels in 2015(using 2012 templates)

•Look at 2012 25 ns fills


Fills

•Single-bunch templates

-r200804 - 2 bunches, μ~20, first fill of 2012

-r206717 - 1 bunch, μ~50, high-mu test

•Test runs

-r212529 - 6 bunches in mini-train

-r214651 - 50 ns physics fill - during BCM noise period

-r215541 - 50 ns physics fill - after BCM noise period

-r216399 - 25 ns fill - 97 bunches

-r216432 - 25 ns fill - 373 bunches

Large range in mu, 3 months apart


Details

•Looked only at OR algorithms (simpler), mainly:

-BcmH_EventOR

-BcmV_EventOR

-Lucid_HitOR

•No absolute calibrations applied, everything scaled to some relative luminosity

•Simple log formulas, no complicated Lucid mu dependence L = -ln(1-Rate)


BCMV single-bunch response

r206717 - BcmVOR

Single-bunch, high μPeak = 1, used to normalize

relative responseAfterglow falls below noise

level after ~500 BCIDs

Colliding bunch

Afterglow

Noise


BCMV single-bunch response

r206717 - BcmVOR

Single-bunch, high μ

Averaged over many LBs

AfterglowReflections

More plots in appendix...


Stability over short time

QuickTime™ and aVideo decompressor

are needed to see this picture.

BCMV: http://physics.uoregon.edu/~torrence/BcmVOR.movLucid: http://physics.uoregon.edu/~torrence/r200804_LucORA.mov


Stability over long times

r206717

r200804

BcmVOR

~ identical over x2 in muand three months!


Lucid HitOR

r200804

No reflectionsShort-term falloff

Similar mid-term slopeLonger tail

(or just lower noise?)

10

Eric Torrence September 2013

Single-beam Templates

r200804BCM - 500 BCIDsto reach ~10-7 level

LucidHit - 1500 BCIDsto reach ~10-7 level

(ran into next bunch)

Constant backgroundsubtracted from -100 BCIDs

11


Maximum BCM afterglow

Just add 500 copies of this template (without peak), each shifted by 1 BCID

Asymptotic value 0.8% reached in ~200 BCIDs

12


Maximum Lucid Hit afterglow

Even with longer tail, less ‘integral’ afterglow (no reflections)

Asymptotic value 0.3% reached in ~300 BCIDs

13


50 ns limits

BCID-1 worksfor BCM in 2012!

(coincidental,only for BcmV)

Afterglow undercollisions ~ 10-3

Difference w/ BCID-1~ 1 x 10-3

(worse in 2011)

0.4% is half of 25 ns limit (expected)

colliding

bcid ± 1

14


Realistic 25 ns fill pattern

14

25ns_2604b_2592_2288_2396_288bpi12inj.sch

2604 bunches collidingin P1/5

Mostly saturated(except at

start-of-train)

15


Data subtraction procedure (per LB)•Start with raw luminosity by BCID for each lumi block

•Identify colliding bunches, and divide out average colliding luminosity from full distribution

-not strictly necessary, but useful for averaging over LBs

-also avoids need for calibration, everything relative...

•Build model of afterglow by adding up templates, one template per collision BCID, weighted by relative lumi

•Subtract this afterglow from raw luminosity in all BCIDs

•Iterate if desired (practically makes little effect)

•Measure residual background in abort gap (last 50 BCIDs)

•Subtract background as well to produce corrected lumi

16


Example

6 colliding bunches (normalized response)

500 BCIDtemplate length

constant term‘fit’ here

Raw Luminosity

After.+Bgd. Prediction

17


Example Zoomed

6 colliding bunches

rather excellent agreementbetween predicted and observed afterglow

undercollisions

18


Lucid artifacts

Finite Lucid template leads to (small)artifacts with few bunches

1500 BCIDs

constant term‘fit’ here

19


Lucid HitOR

Prediction close to luminous bunches looks right on(high μ template, 28 tubes)

20


Lucid - 50 ns fill

Works fine withfull fill pattern

Remember: simple logformula applied,

no mu dependence

21


BcmV noise comparison

During BCM noise period

After BCM noise period

<< 10-3 discrepancies

22


25 ns runs

Can’t prove afterglow under collisions is correct,but procedure seems to work fine

23


25 ns runs

Can’t prove afterglow under collisions is correct,but procedure seems to work fine

Same train length and gap expected in 2015

24


Lucid reduced HV

Lucid ran with reduced HV for most runs from r215433,includes all 25 ns runs

Template clearly not accurate

25


Lucid reduced HV 2

Scale up fast component (BCID+1, 2) by ~30%

Seems to work fine!

26


25 ns Lucid Hit OR

Use scaled template to look at 25 ns Lucid dataMuch larger afterglow (~2%) due to HV settings

normalized

background error

slight mismatch

27


25 ns comparison

Compare afterglow-subtractedrelative luminosity

collisions only

Shape looks familiar, butmagnitude is larger...

28


50 ns comparison

28

Similar to Benedetto’s plots?

Remember, no complicated Lucid corrections, just log formula

29


Trigger Counters•Special-purpose counters to look at 6 L1 items

before and after veto to study deadtime by BCID

•Can try to use before veto as a proxy for luminosity

•Triggers available- Counter 0 is trigger 93 L1_MU11

- Counter 1 is trigger 85 L1_EM30

- Counter 2 is trigger 102 L1_J50

- Counter 3 is trigger 128 L1_FJ75

- Counter 4 is trigger 118 L1_XE50

- Counter 5 is trigger 97 L1_J10

•Must be skeptical, many trigger-related issues with bunch train position...

•Trigger rates ~1% error per BCID (over many LBs)

Most linear with lumi

30


50 ns run

L1_EM30

L1_MU11

L1_MU11 low at start of train (retriggering?)L1_EM30 rises in early train (calo noise?)

31


50 ns run

L1_EM30

Now referenced to Lucid, L1_MU11 looks pretty OK...

32


25 ns run

Back end of bunch train seems to be more consistent with BcmV

33


25 ns run

Really had to draw any conclusions from this

34


Conclusions•Single-bunch template method seems to work for

2012

-Templates quite universal over all 2012

•Afterglow error using BCID-1 appears smaller than 2011for BcmVOR

•No evidence of anything weird in BCM noise period

•25 ns data shows larger (as expected) but manageable afterglow levels, Lucid larger due to HV settings

•First look at Lucid/BCM ratios is rather alarming, but probably lots to understand here

•Trigger rates don’t seem to help

35


Appendix A

Single-bunch plotsRun 200804

2 bunches, μ ~ 20

36


Lucid Hit OR

37


Lucid OR

38


Lucid OR A

39


Lucid OR C

40


Lucid AND

Raw rate only!

41


BcmH OR

42


BcmV OR

43


BcmH OR A

From Rates: A + C = OR + AND

44


BcmV OR A


45


BcmH OR C

46


BcmV OR C

47


Appendix A

Single-bunch plotsRun 206717

1 bunch, μ ~ 40

48


Lucid Hit OR

49


Lucid AND

Raw rate only!

50


BcmH OR

51


BcmV OR

52


BcmH OR A


53


BcmV OR A


54


BcmH OR C

55


BcmV OR C

Afterglow Studies

Documents

Transcript of Afterglow Studies