stephenbrooks.orgForumMuon1GeneralStephen - Request for a progress report & understanding the graphs
Username: Password:
Search site:
Subscribe to thread via RSS
[TA]Assimilator1
2008-04-26 12:24:10
Looking around the website & forum I've found nothing written that states how far we've progressed nor what we have progressed from.
I'm guessing the graphs on the front page have some info related to that but I don't understand them & their is no explanation of them anywhere AFAIK.  What's the muon % about?  % of what?  What's 'Linac88MHz900MeV6'?

If you go to the Folding@home front page, http://folding.stanford.edu/ you'll see near the bottom a paragraph titled 'What have we done so far?' with links to various reports so you get a good idea of what progress has been made. 
I've seen your link to 'Tech reports' but a quick look at the bottom 4 reports didn't seem to relate to DPAD, or if it did it didn't say so, or maybe it's just too complicated for me to see??. I think we need something bridging those reports to how we've advanced the accelerator design to get some idea of progress made.
Surely I'm not the only one thinking this??

Also is there a labelled diagram anywhere showing the build/setup of the particle accelerator we're currently simulating?

I figured after running this project for over 2 yrs now I ought to at least partly understand some of it's workings.

Thanks
Mark
Stephen Brooks
2008-04-28 17:48:40
A month or two ago I was thinking of doing some diagrams of the bits of the accelerator each lattice involves, because it does get a bit confusing.  But I didn't get around to it because I was supposed to be writing my thesis right now, though actually a later section might require such a thing anyway.

My last couple of documents have been about further investigations of the rod target right at the beginning of the simulation.

The Muon % figure is the percentage of muons that get to the end of the accelerator being tested (out of all muons produced in the target); in recent lattices, they must also be within an energy band that the following part of the accelerator can accept.  So the purple line on the graphs that keeps increasing is the efficiency of the current best desigh Muon1 has found.

Linac88MHz900MeV6 is a lattice that produced a range of accelerators aimed at accelerating the muons into an energy band around 900MeV (several times what they typically start off with).  This is a useful energy because it matches with the first dogbone RLA in the scoping study (2006) design.
[TA]Assimilator1
2008-04-28 21:18:05
Thanks for your replie Stephen.

About a diagram for a particle accelerator, maybe a generic diagram would do the job?  (if their is one?), in which case you could just copy someone elses diagram or just simply link to it in the FAQ.  Is there anything here of any use?  http://hepunx.rl.ac.uk/uknf/

What's a lattice?
Is Linac short for linear accelerator?
What's the difference between Linac88MHz900MeV6 & V1?
What's the best efficiency you've seen so far?
Is the 'Old optimisations' list at the bottom of the front page all the different design of PAs that DPAD has tried so far?
Stephen Brooks
2008-04-29 11:29:53
Yes I was thinking of making a fairly generic/block diagram just for showing the lattices if I do that.  There are at least one or two less-specific diagrams of the neutrino factory as a whole in the UKNF site in various places, as well as in the reports on this site, for instance this poster.

A lattice in general is just the name for the arrangement of magnets and particle beam optics in an accelerator.  In Muon1, lattice files are the .txt files in the lattices directory that specify an optimisation _range_ rather than just one design.  The program simulates designs from within that space.

Is Linac short for linear accelerator?  Yes

What's the difference between Linac88MHz900MeV6 & V1?  The '6' lattice has more parameters, specifically it breaks the linac into 6 sections, each of which can have different characteristics.

What's the best efficiency you've seen so far?  It's shown for each optimisation on the graphs on the main page!

Is the 'Old optimisations' list at the bottom of the front page all the different design of PAs that DPAD has tried so far?  They are all the different previous lattices, yes.  Some of them are different designs, others are more slight variants like breaking pieces up into more sections.
[TA]Assimilator1
2008-04-29 19:15:16
Thanks, your site went down briefly at the time I was originally about to post the above replie so I couldn't check out the different lattices effiency before hand.
However I had meant what was the best effiency out of any of the lattices?, I noticed that some of the solenoids were upto about 16% but I'm guessing that's not the 'whole story'.
By knowing the efficiency of the 1st lattice at the very beggining to where we are currently will give us some idea of progress, though I assume that's not the only factor?, I noticed in an article that some designs consume much more energy than others & are more costly than others to build too. 
RGtx
2008-04-29 20:08:24
"But I didn't get around to it because I was supposed to be writing my thesis right now...." After your D Phil, are there any plans to continue the DPAD project?
[TA]Assimilator1
2008-04-30 19:48:11
D Phil??
[TA]Assimilator1
2008-05-17 10:01:48
I'll ask again (you really need forum email subscriptions here).

What's the best efficiency you've seen so far out of all the lattices?
[TA]Assimilator1
2008-05-17 11:42:47
I thought I'll see if I can answer that by listing the efficiencies from old optimisations.

Lattice__________ max efficiency _ Approx peak date

ChicaneLinacA ____ ~1%* _______ Mid 04
ChicaneLinacB ____ 1.557285 % __ 1Q 06
ChicaneLinacB90 __ 1.477859 % __ Mid 05
DecayRot ________ 0.920487 % __ Apr 06
DecayRotA _______ 0.334014 % __ Apr 06
DecayRotB _______ 1.214223 % __ Jan 08
PhaseRotA _______ ~0.8 %* _____ Mid 04
PhaseRotB _______ 4.254614 % __ Jan/Feb 06
PhaseRotC _______ 4.094615 % __ June 06
PhaseRotC_bigS1 _. 1.788399 % __ Jan/Feb 06
PhaseRotD _______ 0.336316 % __ May/June 07
PhaseRotDD ______ 0.448531 % __ Oct 07
PhaseRotEb1 _____ 3.355024 % __ Late Jan 08
PhaseRotEb1a ____ 3.358101 % __ Late Jan 08
PhaseRotEb6 _____ 2.076617 % __ Dec 07
PhaseRotEb60 ____ 3.372407 % __ Feb 08?  (can't see the full curve of the graph on this one)
PhaseRotEb6a ____ 3.480755 % __ Feb 08
PhaseRotEby5 ____ 2.158242 % __ Dec 07
PhaseRotEby5a ___ 2.677150 % __ Feb 08
PhaseRotLinac1 ___ 6.977837 % __ Late April 08
PhaseRotLinac6 ___ 6.866721 % __ Mid April 08

*(not sure how accurate these ones are as the graph looks messed up & thier are no numbers in the list)
I'm assuming atm that the solenoids are a simulation of just part of the accelerator & therefore not comparable, can someone confirm this?

Looks like the best muon transmission efficiencies were achieved by PhaseRotLinacs last month at ~6.9% & they were quite a jump up from the previous best PhaseRotB at 4.25% over 2yrs earlier .
Currently Linac88MHz900MeV6 is only at ~2.27% & Linac88MHz900MeV1 is at ~2.4%, but they both only started recently in January & their curves are still climbing ATM.

Stephen
What did DPAD start from?
Are their any other factors I've missed out that would be included under desugn progress?  power useage?  build cost?

On a side note, I hope the formatting above holds or the table is going to look a real mess! 
[TA]Assimilator1
2008-05-17 11:43:11
Yes it did hold! 
[TA]Assimilator1
2008-05-17 11:45:28
Opps typo above, I meant design progress
[TA]Assimilator1
2008-05-17 11:55:53
(God I really wish you had an edit function here )

I saw the following from this thread by Stephen, http://stephenbrooks.org/forum/?thread=1340&bork=exipbtetru

>>>>>I'm now focussing the whole project on the Linac88MHz900MeV6 lattice, see if that can 'get around' the impasse.  The other PhaseRotLinacs were interesting but didn't accelerate the beam enough for those high 7% yields to really translate into something the accelerator can use.<<<<<

So those designs whilst retaining a higher % of the muons weren't much good because the beam was too slow?  how do you know what the beam speed is?  (looks like that's 1 factor I've missed out ).
Stephen Brooks
2008-05-19 15:47:07
In the above list (plus the current lattices) there are four different groups and yields can only be compared inside a group.  All the ChicaneLinac* can be compared.  DecayRot* and PhaseRot* (excluding PhaseRotLinac) can all be compared to each other.  PhaseRotLinac* are on their own, as are Linac88MHz*, though the latter are "at least as useful" as the PhaseRot* or PhaseRotLinac* groups.

So within each group, the best are:
ChicaneLinacB ____ 1.557285 % __ 1Q 06
PhaseRotB _______ 4.254614 % __ Jan/Feb 06

Oh wait, the PhaseRot and DecayRot group splits into two by initial proton energy (2.2 or 10GeV).  The above is for 2.2. Our best for the 10GeV pion file (PhaseRotC and later) is

PhaseRotC _______ 4.094615 % __ June 06

...but that was with Pre-4.43 versions of the Muon1 code.  At version 4.43 I improved the particle decay distribution a lot, which changed the results.  Thus, the real winner in the PhaseRot group has to start from PhaseRotD/DecayRotB, which makes it

PhaseRotEb6a ____ 3.480755 % __ Feb 08

With that sorted, the final (recent) two groups have easy winners:

PhaseRotLinac1 ___ 6.977837 % __ Late April 08
Linac88MHz900MeV1 (current!) 2.430991%


Now some criteria for counting the muons are "easier" than others.  At the moment it depends only on the energy band (assuming you're using version 4.43+ and 10GeV protons).  The bands for the various groups are:

ChicaneLinac* 400±100MeV
PhaseRot*,DecayRot* 180±23MeV
PhaseRotLinac* 400±100MeV
Linac88MHz900MeV* 830.41MeV to 969.64MeV
Stephen Brooks
2008-05-19 15:51:14
Generally, the narrower the band is compared to its energy, the harder it is to get muons into it.  Thus high scores on ChicaneLinac* or PhaseRotLinac* are easy to get.  ChicaneLinac lattices didn't even score highly with this "easy" figure of merit, so that's why we don't use a chicane any more.  In fact you can compare PhaseRotLinac* directly across to ChicaneLinac* and see that it's getting over 4x the yield!

However, what is more interesting are the "harder" energy bands; the 900MeV one being the hardest.  If you like to simplify it-

Best score on "easy"/400±100MeV: PhaseRotLinac1 ___ 6.977837 % __ Late April 08
Best score on "medium"/180±23MeV: PhaseRotEb6a ____ 3.480755 % __ Feb 08
Best score on "hard"/830.41MeV to 969.64MeV: Linac88MHz900MeV1 (current!) 2.430991%
[TA]Assimilator1
2008-05-20 19:37:28
Interesting , how do the beam speeds compare between those last 3?

What sort of efficienies did DPAD start at?
Stephen Brooks
2008-05-26 18:22:00
Zero, technically... all the lattices in the last few years have evolved out of the "primordial soup" with no seed design.

If you want to compare with a previous lattice, there aren't many direct comparisons to make keeping everything the same.  On the "easy" figure you could compare the current PhaseRotLinacs at nearly 7% with the ChicaneLinacs that didn't reach 1.6%. The "medium" doesn't have many comparisons before the 4.43 code change apart from PhaseRotD, which is nearly 10x worse, though it was really a side-branch.  The "hard" figure I've only just started using and is the furthest down the accelerator of them so far.
[TA]Assimilator1
2008-05-27 18:29:10
Ok so that's quite a big improvement then , is their a known physical limit as to what efficienies could be reached?  Could DPAD ever hit say 50% in the years ahead?

Btw are the narrower 'harder' beams also the faster ones?
Stephen Brooks
2008-05-28 14:25:31
Faster = more MeV, but "more difficult" also means a narrower band.  If we had a beam with no energy spread whatsoever, it would be fairly easy to accelerate to any higher energy.

Generally you could catch larger and larger amounts of beam but they would not be "useful", they would not be tightly focussed.  Catching a larger beam would require a lot of the components in the rest of the accelerator to be enlarged, adding much to the cost.  There's also a tradeoff between channel width and the amount of acceleration you can get, and the efficiency of that acceleration.  Because muons decay, if you accelerate too slowly, you lose them before you can turn them into neutrino beams in the way that you want.  So 100% is certainly impossible.  You can imagine very high percentages but they either require extremely expensive or just technically infeasible components.
[TA]Assimilator1
2010-01-23 10:23:20
Ttt, useful info for anyone who missed it.

Linac88MHz900MeV1 ended at 2.605366% ~Sept 08
Linac88MHz900MeV6 ended at 2.481781% ~Sept 08


[Edited by [TA]Assimilator1 at 2010-01-23 10:36:03]
[TA]Assimilator1
2010-01-23 12:16:23
Looking at newer designs :-
(Date refers to last increase in %, but also corrolates closely to when the lattice simulations were stopped).

Linac900NoGaps1 ______ 2.280941% ~Sept 08
Linac900Removable1 ___ 2.928274% ~Oct 08
Linac900Removable3 ___ 2.984858% ~Jan 09
Linac900Removable6 ___ 3.684663% ~Dec 08
Linac900Removable12 __ 3.670822% ~Mar 09
Linac900Removable6c2 _ 3.891455% ~June 09 (Has note of 'The computed design is 6.01% better than the baseline one')
Linac900Ext6d2 ________ 1.432695% ~Oct 09
Linac900Ext7d2 ________ 0.781853% ~Dec 09
Linac900Ext8d2 ________ 2.673065% ~Nov 09
Linac900Ext9d2 ________ 0.789847% ~Aug 09

Currently running designs :-

Linac900Removable8c2 _ 2.280928% (that has just flatlined over the past few days, maybe it's peaked?  Unless it gets another boost?)
Linac900Ext10d2 _______ 1.522498% (increases in % have slowed right down over the past few days, maybe also near it's end??)

Looks like Linac900Removable6c2 is the best so far amongst the '900s', with it being nearly 50% better than the 1st '900' series Linac88MHz900MeV1 (err assuming I've understood correctly!).

Btw these 2 current '900s' are the only ones to suddenly have a large boost in % after going flatline for a while, all the other 900s more or less go up quickly at 1st, then slow level off before flatlining.  Wierd!, maybe as Yoyo said it's to do with the different simulations he chooses???  Any further ideas on that guys?

Under those graphs for older lattices it's showing that their are still some active users, am I reading that right that some people are still crunching long since retired lattices??  Or is it that was just what they were last crunching when they stopped?

(Thank god for editing!  )

[Edited by [TA]Assimilator1 at 2010-01-23 12:28:53]
[TA]Assimilator1
2010-01-28 18:45:05
Well I was wrong about the current lattices, their still going up, currently at 2.421% & 1.642%
Stephen Brooks
2010-03-16 17:58:10
In the comments I had on my thesis first draft, my supervisor asked for a diagram of the lattices I'd discussed in the final chapter.  So here is a partial "roadmap" of what Muon1 has been doing over the last few years.

[TA]Assimilator1
2010-03-18 07:32:38
Cool, thanks
: contact : - - -
E-mail: sbstrudel characterstephenbrooks.orgTwitter: stephenjbrooksMastodon: strudel charactersjbstrudel charactermstdn.io RSS feed

Site has had 25163623 accesses.