What are the current records for LHC fills?

The place to discuss the LHC. Commissioning, operation, issues, events ....
mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Mon Jun 27, 2016 6:48 pm

Depends on the bunch structure. The PS can deliver up to 72 bunches without gap. Nominally the SPS accumulates four of those trains with a short gap of 8*25ns (not sure about the 8) in between (necessary for the SPS injection magnets), then injects them into the LHC, where a longer gap (28*25 ns) is necessary to ramp the LHC injection magnets.

The long SPS gaps need space in the ring, so you want to minimize the number of SPS injections and the number of bunches per injection. Long bunch trains without gaps are the worst case for heat load, however.
To make things worse, you have two opposing beams and 4 collision points in a 45°-pattern - assembling a scheme that leads to many collisions for all four experiments is challenging.

If the SPS can handle 144 bunches, which is unclear, I could imagine some 48+48+48 pattern first, for ~2300 bunches. Needs more space than 72+72 (which can probably fit 2450 bunches in), but leads to a lower heat load. But that is just a guess.


Edit: They really seem to try to push this run to its limits, although dump and refill should give more collisions. 670/pb for ATLAS, 644/pb for CMS, 36/pb for LHCb, luminosity is down to 35% the design luminosity after 31.5 hours in stable beams. 45% of the initial protons got destroyed in collisions. CMS and ATLAS now show nearly identical lumi values, maybe some nonlinearity in the calibration?

sciing
Posts: 27
Joined: Mon Dec 07, 2009 9:18 am

Re: What are the current records for LHC fills?

Post by sciing » Tue Jun 28, 2016 7:43 am

All records in one fill, longest fill after 37h operator dump. Finally 737/pb for ATLAS and 710/pb for CMS. Ini Lumi see above.

User avatar
Tau
LHCPortal Guru
Posts: 153
Joined: Tue Jan 05, 2010 10:16 pm
Location: Heemskerk, Netherlands

Re: What are the current records for LHC fills?

Post by Tau » Tue Jun 28, 2016 9:13 am

sciing wrote:All records in one fill, longest fill after 37h operator dump. Finally 737/pb for ATLAS and 710/pb for CMS.
And then, in four hours and three minutes, we are back in business! That's not the record, but in combination with the fill length we have over 90% stable beams.
Not bad at all...
- Tau

mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Tue Jun 28, 2016 2:35 pm

~35 hours luminosity lifetime (approximated from last fill) and 4 hours turnaround time would suggest an ideal fill length of just ~16 hours, for 0.65 times the maximal luminosity as average over the whole cycle. With 7 hours of turnaround time ~20 hours of fill time are optimal (57%), and we need 10 hours turnaround time to make 24 hours optimal (52%). 24 hours of turnaround time make 37 hours of stable beams optimal (38%).

0.65 times the design luminosity as average would be an incredible 3.9/fb per week. A more moderate estimate is 90% the design lumi as maximum, and 10 hours of average turnaround time (including early dumps), for 2.8/fb per week - if everything runs smoothly and the fill lengths are optimized.

jmc2000
Posts: 146
Joined: Sun May 08, 2011 2:51 pm

Re: What are the current records for LHC fills?

Post by jmc2000 » Tue Jun 28, 2016 3:48 pm

mfb wrote:~35 hours luminosity lifetime (approximated from last fill) and 4 hours turnaround time would suggest an ideal fill length of just ~16 hours, for 0.65 times the maximal luminosity as average over the whole cycle. With 7 hours of turnaround time ~20 hours of fill time are optimal (57%), and we need 10 hours turnaround time to make 24 hours optimal (52%). 24 hours of turnaround time make 37 hours of stable beams optimal (38%).
I make the luminosity half life around 18 hours giving ~400\fb. I don't understand the reasons behind the very long fills > 24 hours; unless the strategy is to use faults to dump the beam, making the whole process far more efficient. Otherwise an operator dump would be followed a few hours later by an unplanned dump.

Jmc

mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Sat Jul 02, 2016 2:42 pm

You cannot really "use" faults. They can happen, and ruin your fill. Maks keeping the beam a bit better, of course, if they occur frequently.

They managed to get another luminosity record this morning, ~110% design lumi for ATLAS and more than 100% for CMS. The following run was worse again - but they are still working on getting more collisions even with the SPS constraints.

RocketManKSC
Posts: 36
Joined: Mon Mar 29, 2010 10:09 pm
Location: Titusville Florida

Re: What are the current records for LHC fills?

Post by RocketManKSC » Sat Jul 02, 2016 9:07 pm

As I pointed out in another post the optimal time formula only works if you can alwasy run that long. You must take in account short runs.

1 Used the actual data for runs over 2040 bunches
from https://acc-stats.web.cern.ch/acc-stats ... uper-table fills 4976 to 5068

2 Build a curve fit for atlas and CMS delivered Data vs Time
Atlas y = 5.067x + 28313
CMS y = 5.2158x + 35638
Chart 1

3 Build a Table of delivered for Time to Operator Dump (T2OD)
FOR 2 to 40 hours
if time in SB < T2OD then = sum of Atlas and CMS Delivered
else = sum of curve fit

4 Calculated the total delivery for each T2OD

5 Adjusted the time Between Fills (TBF)
if TBF > the 12 Hours then TBF= 12 Hours
if SB > T2OD then TBF=TBF-(SB-T2OD)

6 Calculated total time for each T2OD

7 Delivery per second = Total Delivered / Total Time
Chart 2
Cern.png
Cern.png (26.31 KiB) Viewed 1465 times

mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Sun Jul 03, 2016 1:52 pm

2 Build a curve fit for atlas and CMS delivered Data vs Time
Atlas y = 5.067x + 28313
CMS y = 5.2158x + 35638
Chart 1
That would suggest a constant luminosity plus some magic lumi offset - clearly not a realistic model.
Something your method is missing is the correlation between initial luminosity and operator decisions - the worse the luminosity the more likely is an earlier operator dump.

If you assume a constant luminosity, then clearly longer runs are better, but that is not what actually happens. You get a better model if you take the luminosity evolution of the longest runs.

And where do you take unplanned dumps into account?

mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Tue Jul 05, 2016 8:15 pm

They now plan to dump the current fill in about an hour, at ~40% design luminosity after 24 hours in stable beams. The second planned dump in the last days, at a similar time. Looks like they go for 24 hours now.

CMS has some weird feature in OP VISTARS where they split up the luminosity data in multiple blocks.

More than 10/fb for ATLAS and CMS now.

tomey36
Posts: 101
Joined: Wed Apr 14, 2010 11:39 pm

Re: What are the current records for LHC fills?

Post by tomey36 » Wed Jul 06, 2016 7:02 pm

Any news on the SPS beam dump problem? Are they going to try and fix it this year? I haven't seen them mention it at all, all they have said is that it is stopping them from injecting more bunches.

Follow up question why is it stopping them from injecting more than 2076 bunches? If there is a problem with the dump system wouldn't it be unsafe to inject any # of bunches?

mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Wed Jul 06, 2016 9:17 pm

Didn't hear anything new about the SPS beam dump for a while. I guess a replacement is in production, but it is unclear if it will be installed this year.

Currently they can't get more bunches in simply because the LHC ring is full. See the end of the previous page and the start of this post for an analysis. Longer SPS trains would allow to inject the bunches with fewer large gaps, allowing more bunches in the ring.

mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Mon Jul 11, 2016 8:25 pm

3.1/fb in a week (Monday->Monday) is a new record. 80% time in stable beams.
12 weeks of physics left this year. I don't expect all of them to be that good, but we are well on track for more than 30/fb.

Looks like the SPS beam dump won't get fixed this year, but going to 144 bunches per injection (if possible) and the BCMS scheme discussed here might give a slightly higher luminosity.

tomey36
Posts: 101
Joined: Wed Apr 14, 2010 11:39 pm

Re: What are the current records for LHC fills?

Post by tomey36 » Mon Jul 11, 2016 9:16 pm

Ya, that sucks i was really hoping to see the LHC finally run with the maximum number of bunches. Still, they have had great performance running with ~2000 bunches so in the end it might be a good thing they are not able to inject more than that. We don't know if they would have had as high as of availability as they do now. Plus it forced them to get creative to maximize the luminosity with a fixed # of bunches. I feel that will pay off later if/when they reach ~2800 bunches.

Also 3.1fb in a week is awesome!

mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Sat Jul 16, 2016 11:31 am

The new BCMS ("Batch Compression, Merging and Splitting") works, the first run using it for a lower emittance directly reached 110% design luminosity, a new record but probably not the last one.

More details on how it works (2012). The basic idea: reduce bunch splitting in the PS to keep a lower emittance.

Edit: ~120% design luminosity (CMS) Tuesday afternoon. ATLAS was oscillating between 0 and 150% design luminosity, both values are wrong... later CMS went to zero while ATLAS showed 120%. A bit messy, but the actual luminosity seems to be 120% the design value.

mfb
Posts: 168
Joined: Sat Oct 03, 2015 7:58 pm

Re: What are the current records for LHC fills?

Post by mfb » Sun Aug 06, 2017 2:42 am

Collecting the records in a single post:

Peak luminosity: A bit above 1.6E34/(cm^2*s), several runs in July/August.
Highest beam energy: At 3E14 p/beam and 6.5 TeV, we had 310 MJ per beam in various runs.

Luminosity in 24 hours: 0.7/fb in early 2017 according to this article, not sure if that is still the record.
--Edit: ~0.8/fb from 5. August 14:00 to 6. August 14:00.
Luminosity in 7 days: ~3.5-4/fb last year?
Luminosity in a run: ~720/pb, 28th June 2016

Longest fill: 37 hours, 28th June 2016
Time in SB in a week: ~135 hours (80%!), early July 2016

Post Reply