Dear Taka, Alessandro, Firstly, thanks to Taka for going through the differences between the models and clarifying several things. I think we are converging on how to implement the changes to the material. I have added a few more comments below (prefixed CPW). On Thu, 21 Sep 2006, Tricoli, A (Alessandro) wrote: > Dear Taka, > > Thank you for going through the differences between your and my model in such a detail. I know that understanding someone else work with no much documentation is very difficult and remembering the work done a year before is equally difficult. I appreciate your effort, which is very useful for all of us. > > A few general comments. I think that the differences between my new estimate and the current simulation for the on-barrel services are well under control. > In this email I try to clear the few points left open for discussion. This and wext weeks I will implement the final minor changes on my model, afterwards, after the final agreement by all of us, we can probably start implementing the changes in GeoModel. > > My new comments in response to Taka's email are lables "-A.T.(2)". > > > > Dogleg > ====== > Current Alessandro > ------- ---------- > Thickness/mm (r) 1.0 0.34 > Length/mm (phi) 36.0 76.8 > Width/mm (z) 30.0 26.0 > Volume/cc 1.08 0.78 > Mass/g 6.9552 5.71 > Density/(g/cc) 6.44 8.41 > > * Propose to keep current dimensions and position but update > mass and composition. Length cannot be extended without > overlapping cooling pipe. > > -A.T.: the thickness value I chose (0.34mm) includes only > the power tape layers, does not take into account the opto- > package, fibers etc. for which I could not find dimension > information. So I am happy to have it extended to 1 mm in r. > The length in phi in the current simulation seems rather > short to me but this value has probably been chosen as the > maximum allowed to avoid clashes. So that's fine. > > -T.K.: The dogleg length in phi of the current simulation > was determined by watching page 1/20 of > http://atlas.kek.jp/si-soft/geometry/SCTBarrelMount.pdf, > in which you see most of dogleg parts are within 30mm in > phi. (BTW, the current length (phi) is 30.0mm and width(z) > is 36mm.) We can make phi length longer without clashes, > though. > > > -A.T.(2): from Taka's drawing I understand that the dogleg's length in phi is 30mm approximately between the power tapes to the bracket end, however the dogleg extends farther to the module connector and this length is approx. 30mm or so, see also http://hepwww.rl.ac.uk/Atlas-SCT/engineering/material_budget/measurements/Barrel_SelectedPictures/B3/Harness/IMG_1218.jpg and http://hepwww.rl.ac.uk/Atlas-SCT/engineering/material_budget/measurements/Barrel_SelectedPictures/B3/OxfordAssembly/DSC_0770.jpg . I would personally be more inclined to have the dogleg length in phi longer than 30mm, if there are no other restrictions. CPW: As I understand the code, the length and width in the current simulation are 36mm in r-phi and 30mm in z, as I said originally, rather than the other way round as Taka believes - maybe they are swapped from what was intended? It is possible to make the dogleg longer in the direction towards the centre of the module, i.e. away from the cooling pipe, I'll investigate by how much. > > > * Why is new mass lower? This is 'dogleg+top bracket' option. > > -A.T.: the lower mass in my computation is due to the > placement of the "module clips" and "dogleg clips" into > the "cooling block" model, the placement of the "lmt i/f > board" and "redundancy arms" in the "straight harness" model > and the addition of the opto-fibres in the "dogleg", not > present before. > > -T.K.: I agree with AT. It is better to move both clips to > the cooling block. Thank you for adding opto-fibers which > was missing from my work. > > -A.T.(2): ok. > > > Cooling Block > ============= > Current Alessandro > ------- ---------- > Thickness/mm (r) 1.5 5.0 > Width/mm (phi) 8.0 8.0 > Length/mm (z) 62.0 62.0 > Volume/cc 0.74 2.48 > Mass/g 2.418 3.62 > Density/(g/cc) 3.25 1.46 > > * Suspect mass difference because Alessandro has included > solder, clips, etc. and current value only include block > (current material = Al). Is this so? > > -A.T.: Yes, I think so too. I cannot find any Taka's spread > sheet with the list of items that enter the cooling block, > but from the composition (only Al) and the weight it seems > so. In fact in my calculation, the cooling block itself > weighs 2.24g the rest is due to solder (most of the weight), > shunt shield, screws and clips. > > -T.K.: Again, thanks to Alessandro, my incomplete work was > completed by adding solder, screws etc. > > * Propose to keep current dimensions and position but update > mass and composition. There is no room to increase thickness. > > -A.T.: the thickness value I quote is what I found in > drawings, but I guess in the simulation has been reduced to > avoid clashes. So that's fine. > > -T.K.: Yes, in the simulation I had to separate the cooling > block (attached to each module) from the cooling pipe which > is one in row, the reason why the cooling block is made to > be thinner in simulation. > > -A.T.(2): Ok, so let's keep the cooling block thickness as it is in the current simulation. > > Cooling Pipe > ============ > Current Alessandro > Drawing > ------- ---------- ------- > Radius/mm 2.0 2.1 > Length/mm 1571.76 1471.6 > Radial position/mm (3) 304.8 306.9 307.1/309.7 > (4) 375.7 377.9 378.0/380.6 > (5) 447.4 449.9 449.7/452.4 > (6) 518.1 520.9 520.4/523.1 > > * Current positions (estimated from entry points of > particles) are 2-3mm inside values suggested by Alessandro > and drawings. Propose to leave them there - moving likely to > cause huge problems. > > * Length will be updated to 'whatever is convenient'. I > think the radius can be increased, but am not sure it is > worth it. Mass and composition will be updated. > > - T.K.: If the radius is increased to 2.1mm, the pipe edge > becomes closer to the cooling block for the lower case. Let > not change it. As I said, in order to separate the cooling > pipe from the cooling blocks safely, I had to move the > cooling pipe about 1.5 mm farther. I got 305.8mm instead of > 304.8mm (B3 case) from my drawing of > http://atlas.kek.jp/si-soft/geometry/rphiView-B3lower.pdf . > Can Pat check it again though it is a tiny difference? > > -A.T.(2): Ok, let's keep the pipe radius as it is in the simulation. As far as the pipe lentgh is concerned I leave the decision up to Pat as changing it might cause clashes with other volumes. CPW: OK, we should keep a radius of 2mm. I estimated the position by printing out the coordinates of particles entering the volume (which I thought more reliable that trying to calculate from the code). I may have made a mistake transcribing the numbers, but I've lost the output file of the job so I'll have to run it again to check. But in any case, the difference seems small enough that I don't think we need worry about it. > > > Bottom Bracket > ============== > Current Alessandro > ------- ---------- > Thickness/mm (r) 4.0 3.2 > Width/mm (phi) 8.0 7.5 > Length/mm (z) 51.0 54.0 > Volume/cc 1.63 1.30 > Mass/g 4.94496 2.68 > Density/(g/cc) 3.03 2.07 > > * Why the mass difference? Think Taka includes an 'interface > PCB' and Alessandro does not (I think this is in > Alessandro's harness). This needs checking. > > -A.T.: it is true that the "interface PCB" is placed in the > harness in my model, but this accounts for only 1.43g extra, > which added to 2.68 g gives 4.11 g. That means that 0.84g > are missing. I notice that in Taka's excel sheet in > http://atlas.kek.jp/si-soft/geometry.html > the "bottom bracket" is actually 4.36g. The difference wrt > my estimate is only 0.25 g in this case. Going through > Taka's calculations I notice that he used both THIRD > MOUNTING POINT UPPER and LOWER, whereas I only used an > average between the two, since (if I understand correctly) > either the lower or upper third mounting point is on each > bracket assembly, depending whether the module is in the > upper or lower position. Furthermore, I have one more screw > (M1x4=0.027g) than Taka. In conclusion the differences > between my description and taka's description are minor, > whereas I do not know where the current value in the > simulation (4.945g) comes from. Any idea Pat? > > -T.K.: > (1) The interface PCB is actually on the cylinder attached > to the power tape. Thus Alessandro's model of including it > to the Harness is better. I accept this change. > (2) It is my mistake. I double counted THIRD MOUNTING POINT. > Thank you for pointing it. About the screws, I do not find > Alessandro's additional screw of > Third Point Peek Finger Screw (M1x4) (<- material is > Stainless Steel??) > Please take a look of page 18/20 of > http://atlas.kek.jp/si-soft/geometry/SCTBarrelMount.pdf > in which I find one SS screw of M1x4 and one Al screw of > M1.6x6. Can Alessandro check the related drawings? > (3) There was a mismatch between my drawing and my > spreadsheet of > http://atlas.kek.jp/si-soft/geometry/topView-B3.pdf ->51x8x4 > http://atlas.kek.jp/si- > soft/geometry/BracketDoglegPowerTape.xls > --> 60x15x1.6 > As you pointed, they are inconsistent. Let us stick to > 51x8x4 and change density. > > > -A.T.(2): regarding Taka's points: > > (1) ok > > (2) I think it is just a mislabeling error on my side. Taka and I name this screw differently, but it is included in both models. In fact from the attached file (just B3) by Eric Perrin in cell A33 there are 3 Stainless Steel screws M1x4 per bracket (0.027g each). I put 1 of these 3 screws in the "top-bracket" and 2 in the "bottom-bracket", whereas Taka does the other way round. I agree with Taka's choice, so I will change my spread sheet accordingly. I think the problem is solved. > > (3) I agree. > > > * Propose to keep current dimensions and position, update > mass and composition. > > -A.T.: that's fine with me. > -T.K.: But after we identify Third Point Peek Finger Screw > (M1x4). > > -A.T.(2): I think the discrepancy is understood (see above). > > -T.K.: Now this may be new. As I remember correctly, two > sets of screws and nuts for each bracket were added in the > middle of cylinder assembly around Spring of 2004. These > sets of screws and nuts (imbeded into the Cylinder) were > added because some of the brackets were not fully glued on > the cylinder. These screws and nuts must not be written in > the original drawings. I would be happy if Alessandro could > ask this matter. If I can find these sets at KEK, I make a > weight measurement and let you inform the results. > > -A.T.: I think that these objects are what I call in my spread sheet "Pad Reinforcements" (4 S.S screws M1x3, 4 S.S. washers and 4 Peek bushes) and they are also included in Taka's model. Both of us use Eric's excel sheet as primary source of information. So this should be fine in both models. > > > Harness (Power Tapes) > ===================== > Contained within a volume: > Current Alessandro > ------- ---------- > Thickness/mm (r) 1.221 1.97 > Width/mm (phi) 21.0 24.66 > Length/mm (z) 1571.76 1423.1 > Mass/g 33.5025 93.4 > > * Within this volume are 2*6 tapes (each thickness > 1.221/6mm) stacked in r, running from the interlink to each > dogleg. Propose to keep this structure, but increase tape > thickness (to ?) and reduce length to make end at end of > cylinder. > > -A.T.: I agree. The thickness is difficult to model since on > top of the tapes there are the fibres and the clamps. All > these items have different volumes along z. Any value of > thickness between 1.22 and 2 mm would be fine as long as > they don't cause any clash with other volumes. > > -T.K.: Alessandro's NarnessSplit model with N1 to N6 > requires quite a change in programme. For minimum changes, > it may be better to keep the current dimensions (except > thickness which you can increase without clashes). But as > Alesssandro points out below, the biggest contribution is > from mass is from the "permanent clamps", and in that case > we may better switch from my power-tape model to N1-N6 model > for more realistic mass distribution. Let me draw a picture > for myself and think it for decision. > > -A.T.(2): After talking to Pat, we agree that it is probably more difficut to apply my "HarnessSplit" model in the simulation than for me smear the permanent clamps weights along the 6 power tape volumes. I will try to do this change keeping a realistic description of the mass distribution along z. > > * Why now so much more mass? Seems to be because the > current 'power tapes' include only power tapes, whereas > Alessandro's harness includes fibres, interface PCB, > connectors as well. The power tape component of this is > 33.34g which is similar to current power tapes - is the rest > really so much more? I think some of the stuff Alessandro > has included in the harness has probably been included by > Taka in the dogleg or brackets - they should check, if they > haven't already done so. In which case, do we want to move > it into the harness, smeared out along z, or keep it in the > doglegs and bracket, concentrated in z? > > -A.T.: Yes, most of the difference between the current value > and my value for the harness is due to the absence in the > current simulation of the "permanent clamps", 2.1g each > (25.82 g per row) the interconnects and the fibres. > Furthermore, as already mentioned, some of the connectors > and the i/f PCB are placed in the harness in my model, > whereas are in the dogleg in the current simulation. So the > mass difference is explained. I have no strong arguments to > keep the i/f PCB and the connectors on the harness, but I > don't see any big impact on the simulation either if we move > them into the dogleg: in fact in the simulation the straight > part of the harness is split into 12 subsections along z (as > you can see in the "HarnesSplit" sheet) so that the masses > of these components are smeared only over a small area. > However I am happy to follow Taka's approach if you think it > is more appropriate. > > -T.K.: Thank you again to Alessandro to list up all missing > items. Just like Pat, I am a bit surprised to hear 25.82g > per row by permanent clamps. I am happier if Alessandro can > double check it. > > > -A.T.(2): This information comes from Lewis Bachelor's spread sheet which I attach. The masses of the permanent clamps depend on z, left-hand- side and right-hand side position on the barrel and are in the range 1.84g-2.65. > > > ------------------------------------------------------------ > In addition, for material composition, Alessandro is > proposing to use matter-level breakdown like Al, Cu-Ni, > CFRP, Kapton ...... instead of my atomic-level breakdown to > H,C,O,Al...... I agree with the poposed matter-level > breakdown, because it is more transparent and easy to > understand. I would behappy if Pat can think about this > change, hopefully easy. > > > -A.T.(2): Probably Pat can also comment on this, but I think this change should not be difficult to implement in GeoModel. CPW: Technically, either an atomic-level breakdown or a component material level breakdown is possible, and both involve a similar amount of work for me. But I think the component material breakdown is more transparent: so I think we are all happy with this. > > > I would be happy if I receive further comments from all of > you. BTW, I am coming to CERN for nextATLAS Overview week at > CERN in the week of October 2. > > -A.T.(2): Unforttunately I will not be able to make it to the ATLAS overview week, but I think we can find an agreement by email. Pat, Stephen and I have been all at Durham for the ATLAS-UK physics meeting so we have discussed some of the issues you brought up in your email. CPW: Unfortunately I will not be in CERN for the Overview week either, I'll be there next week for ID week. But I think we are agreed on most thing now, and will soon sort out the rest. Regards, Pat