Summary notes of the forty-second meeting of
the LHC Commissioning Working Group
Tuesday March 11th, 14:00
CCC conference room 874/1-011
Minutes of the Previous Meeting and Matters
Arising
There were no minutes of the 41st
meeting.
Roger introduced the schedule for
the 42nd meeting, which focused on three topics: the maximum
strength of the LHCb spectrometer at injection
(Werner and Malika), the bunch number distribution
(Philippe and Mike), and the commissioning of the beam instrumentation
(Jean-Jacques).
Responding to a remark by Roger,
Massimiliano commented on the intended switching frequency of LHCb and ALICE: In the long term LHCb
would want to switch the polarity once per week, ALICE once per month (both
solenoid and dipoles together). However, in 2008 the switching could be more
flexible. For the early running, both experiments wish to have one polarity
flip, simply to get some data with both polarities. Massimiliano also mentioned
that both ALICE and LHCb plan to run with full
spectrometer field even if the machine collides at lower energy, e.g. 5 TeV.
How Do We Have to Ramp the LHC Spectrometer Magnet? (Werner)
Werner prepared this presentation
together with Malika Meddahi
and Yannis Papaphilippou. He noticed that the
question about ramping the LHCb
had come up in regular intervals, with slightly different details since the
boundary conditions had been changing. Instead of the originally suggested
title of his presentation “can we excite the LHC spectrometer with ½ strength at injection”, Werner addressed the more
general question “how do we ramp the LHC spectrometer magnet?”
He recalled his pertinent
presentation on “The Effects of Solenoids and Dipole Magnets of LHC Experiments” at Chamonix 2006. In IP8 the crossing angle is in the
horizontal plane, and bunches always have to cross from the inside to the
outside. Namely, the sign of the effective crossing angle is fixed to avoid
additional crossings. The effect of the LHCb
spectrometer is compensated with three close-by dipole magnets. A polarity
switch of the spectrometer changes the sign of the internal crossing angle,
which must be overcompensated by an external crossing angle. The effective crossing angle always is the
sum of the two.
Massimiliano asked what exactly was
meant by “negative” polarity. Werner replied that for negative polarity the
spectrometer bends the beams to the left when looking from above, or, put
differently, in the "problematic" polarity (or "bad"
polarity) the LHCb spectrometer B-field is pointing
up.
For the bad polarity the external
angle must be much larger than the angle introduced by the spectrometer. The
effective crossing angle is different for the two polarities (smaller for the
bad one) due to aperture limitations. The full spectrometer field at injection
would result in an internal crossing angle of +/-2.1 mrad.
Werner posed the question whether for positive polarity the overcompensation
was possible at all.
To answer this question he looked at
different scenarios corresponding to the various commissioning phases. For 43
or 156 bunches no external angle would be needed. However, more bunches with
bunch spacings of 75, 50 or 25 ns require the
external angle. Werner assumed no beta* values smaller than 10 m for all cases
with crossing angle.
For operation with 43 and 156
bunches, the spectrometer at full strength generates +/- 10 mm orbit bumps at
injection energy. The aperture nevertheless was OK. Werner concluded that in
cases without a crossing angle we are safe with respect to aperture, and we
could accept the spectrometer at full strength in this case. The reason is that
the longitudinal extent of the internal crossing angle bump is rather small.
Next, Werner showed that in cases where we need a crossing angle, the case of
negative polarity would also be OK at full spectrometer strength with +/- 170 microrad external crossing angle. Mike asked if Werner
really advocated this scenario. Werner repeated that from the viewpoint of
aperture there would be no objection.
Then Werner demonstrated that for
the bad sign of the spectrometer polarity (“plus”) and with the spectrometer at
full strength we would need a very large external angle for providing
sufficient separation of the parasitic encounters, so large indeed that the
beam would be sent through the triplet magnet.
Werner demonstrated that with the
bad polarity we cannot even accept 20% of the spectrometer field at injection,
which would still yield too low an aperture of n1=5.5. By scanning the field,
computing n1, and trying to optimize the crossing angle bumps for maximum n1,
Werner obtained a case with 8.2% of the maximum field and n1=6.1, which might
be almost OK. He drew the conclusion that realistically the maximum field we
can hope for is 7.5%. This would make no big difference to ramping
the spectrometer field linearly with the beam energy, which yields an injection
field of 6.4%. Werner stressed once more that the aperture problem arises only
for one polarity.
Roger commented that the experiments
would need to accept this conclusion. Massimiliano remarked that in the
beginning LHCb will request to have full field at 5 TeV (whether ramped with the beam or not). Full field means
the 7-TeV nominal LHCb field. LHCb
would prefer to keep the magnet fixed at full field, even at injection if
possible (but this depends on the LHC commissioning group: one has to balance
ease of operation, risks with beam, etc.). Andreas Schopper
asked if LHCb can hope for lower beta*values than 10 m. Werner
responded that without crossing angle the beta* value is uncritical, since the
spectrometer bump is completely independent of the optics, and that there was
no beta* limit from the aperture in LHCb. Roger
commented that operationally other aspects could be important. Werner clarified
that the internal crossing angle bump at injection with full spectrometer field
was not “huge”, only 11 mm. Helmut suspected that the aperture alone might not
be the full story, but that detector background and other constraints could
enter.
Werner turned to collisions at lower
energy, asking at which energy the full spectrometer field would be possible.
From the previous results it was clear that without a crossing angle the full
field was always possible. The answer for cases with crossing angle depends on
the beta* and again on the polarity. As an example, Werner looked at 4 TeV and beta*=10 m. He found that under these conditions
the full spectrometer field would be possible for both polarities.
Werner commented that for
beta*<10 m, one would need to increase the crossing angle to provide an
efficient separation. The resulting limits on beta* at 4 TeV
would need to be assessed. Massimiliano suggested to check beta*=2 m at 5 TeV beam energy. Werner expressed doubts that this beta*
value would be possible for the bad spectrometer polarity, but he was ready to
explore it.
After the meeting Massimiliano
summarized his understanding of the bump aperture problem for LHCb as follows:
1. Is it possible to keep a fixed full field at any energy, i.e. without bump field ramp ? (we only talk here about the aperture problems)
A.
zero external angle filling schemes: no problem for both polarities,
at any energy, at any foreseen beta*.
B. external angle filling
schemes:
a. B-field down:
no problem, at any energy, for beta* of 10m.
b. B-field up: problem. This may
depend on beta* and energy.
Option to cure this with mixed vertical/horiz
crossing was mentioned and yet to be investigated (but not most urgent).
2. At what energy can the full field be used ? Is 5 TeV OK ? (with ramping the
bump field if needed)
Given
the answers to 1.A, we only need to talk about a possible problem
with
external angle filling schemes (75-50-25ns).
The answer depends on
beta*. For 10m: OK at 4 TeV, hence 5 TeV as well.
For 2 m: this may be more
difficult.
3. A (not so urgent) question: what is the minimum possible beta* at 5TeV with external crossing angle and full field for Alice and LHCb? (with ramp allowed). For Alice this is only relevant in the “new” 50ns scheme with few colliding bunches; for 75ns and 25ns they do not want any squeezing.
Roger concluded that there was no
problem for the 2008 physics runs, namely we could always leave the magnet
powered at full strength, from the aperture point of view. Werner added that
for ALICE the situation was even easier, thanks to a smaller crossing angle and
the vertical crossing plane. John cautioned that in ALICE some new constraints
related to tertiary collimators had recently emerged for heavy ion runs. During
a subsequent discussion on the spectrometer ramp, Mike recalled that the maximum
ramp rate of the spectrometer was 4 A/s, which could limit the overall speed of
the LHC ramps. Jean-Jacques observed that the undulator
ramp would take some time too, and one such ramp would need to be done prior to
injection. The spectrometer ramp could conceivably be performed in the shadow
of the undulator ramp.
Bunch Numbering (Philippe Baudrenghien)
Philippe described the proposed
numerology of LHC bunches, in terms of several conventions. Convention 1: The 400 MHz RF defines 35640 buckets,
spaced by one RF period, and numbered from 1 to 35640. Convention 2: Bucket 1 is the first bucket after the 3
microsecond long abort gap (defined from bucket 34442 to 35640). Convention
3: bunches in bucket 1 of the two rings
collide in IP1 (or any other IP to be agreed upon).
Massimiliano and Roger commented
that Philippe’s proposal was accepted. John cautioned that for ions one might
prefer to have the bunches no. 1 collide in IP2.
Philippe presented several examples,
e.g. for 25ns operation the bunches will occupy
buckets 1, 11, 21, etc., with gaps at every PS or SPS kicker gap. The Revolution Frequency is a train of
pulses with one 5-ns long pulse per turn. At a given place in the machine, and
at a given beam energy (that is at a fixed RF frequency) the delay between the
pulse and the passage of a bunch in bucket 1 will be fixed from run to run.
Drifts during the ramp are insignificant. The Bunch Clock is a square wave at 40 MHz obtained by dividing the RF
by 10. The divider is synchronized with the Revolution Frequency. At a given
place in the machine, and at a given beam energy (RF frequency) the delay
between the edge of the Bunch Clock and the passage of a bunch will be fixed
from run to run.
Alick asked who was fixing this
delay. Philippe replied this would be addressed later during his presentation.
Now turning to the clock generation,
Philippe described how all signals are generated in SR4, and then are directly
transmitted to CMS (point 5) and BT (beam dumps). Philippe pointed out that the
requested delay for injection pulses to the kickers is very different from the
delay requested for BI, a difference which prevents using the same modules for
BT and BI, and which should be re-discussed.
A reference clock is generated for the physicists. The experiments
only care about this signal during physics. RF will provide a stable 400 MHz
reference. On the flat top the beam would be rephased
to that reference. The presence of the signal and its validity will be
guaranteed only from shortly before the filling until the beam dump.
The RF synchronization equipment is
being installed and commissioned in SR4. Philippe expected that all this equipment
would be running by mid-April. He reviewed the RF synchronization for the
SPS-to-LHC bunch-to-bucket transfer and its associated functionalities. Two
different rf frequencies for
the two rings can be used to rotate the two LHC rings against each other prior
to an injection.
Philippe outlined the commissioning
procedure of the synchronization and bunch clock systems. The first bunch
injected will be bunch no. 1 by definition. Then one will adjust the delay in
the reference frequency sent to the beam dump, in order for the dump kicker
rise to occur just before bunch no 1. Philippe assumed that BT must have the
tools to make this adjustment. Jean-Jacques commented that for the abort gap
monitor a similar adjustment is done in BI. He added that it was important to
keep the relation between bucket 1 and the RF frequency signal fixed from run
to run. Massimiliano remarked that all experiments will adjust this phase, and
that this delay would indeed never change. Replying to a comment by Alick, Andy
clarified that the bucket selector setting changes at every injection to inject
bunches into different buckets.
Philippe stressed that in order for
bunches in bucket 1 to collide in IP1, we must rotate ring 2 with respect to
ring 1, and for this we need information from the experiments or from BI on the
time difference between passages of the two pilots in IP1 or IP5, e.g., we will
need one pick up which sees both beams.
Helmut
commented that the installation of a special pick up and electronics near the
IP had been agreed upon with BI earlier. Helmut added that an extra card had
been added for the purpose of detecting the time difference of the two beams.
Jean-Jacques cautioned that nevertheless assumptions on the cable lengths were
needed, and that one would need some calibration measurement. After the meeting
Gianluigi and Rhodri explained that BI plans to measure the time delay between
beam 1 & beam 2 at the first BPM on both sides of the IP (in fact, BPMSW 1,2,5,8 L & R,
BPMR-Q6L2 BPMR-Q6R8, + 2 BPMs in Pt4). If
the BPMs, cables and geometries are symmetric then the time difference between
both should be the same when the beams collide in the centre of the IP.
Frank asked whether the final
resolution of the collision adjustment at the IP was not better than 1.25 ns.
Philippe replied yes, it was, and that this better centering was achieved by
fine tuning with phase shifters. Alick asked for the target value of such fine
tuning, and Philippe replied better than 1 degree, adding that we would use
information from the experiments to make this fine adjustment. Massimiliano
asked when this would be happening. The tolerance of experiments was said to be
of the order of cm.
During and after the meeting John
commented that when the idea of adjusting the longitudinal position of the IP
with BPMs had come up in Jan 2007, the accuracy hoped for was of order of a few
mm. However, Rhodri had possibly
reported in the extended LTC held in the previous week that the extra
electronics needed would not be installed at startup (to be confirmed with
him). John added that this discussion had started because he had heard that
ATLAS (and presumably other experiments too) would be able to measure the
longitudinal position of the IP with their zero-degree calorimeters. This should perhaps be kept in mind as
possible verification for the accelerator tuning.
Philippe addressed the question if
we could change the convention that bunches no. 1 collide in IP1. His answer
was no, but he suggested that changes also would not be necessary, since one
could always use the bucket selector
to shift the collision point, e.g. by one quarter or octant. Massimiliano
remarked that in this approach, however, the abort gap would always collide in
IP1-IP5, while for ion collisions it would be better to have abort gaps collide
in IP2 to maximize the luminosity. On the other hand ALICE might also plausibly
prefer the opposite configuration for beam-gas background studies. John
stressed again later that one might sometimes want to collide in IP2 instead of
IP1.
Finally, Philippe answered a couple
of other FAQs: Counting in 400 MHz or 40 MHz is equivalent. We should not
number the bunches, but stick with bucket numbers or bunch clocks.
Massimilano raised a question about satellite bunches. According to Jean-Jacques
in this case the bucket number would be used. Gianluigi pointed out that most
of the instrumentation would not see any signal of the satellites. Massimiliano
recommended that we agree on the terminology between the experiments and
accelerator. Alick recalled that the orbit marker was fixed to bucket no. 1.
Massimiliano asked how the filling pattern was distributed to the experiments.
This was answered in the following presentation by Mike.
Distribution of Bunch Configuration to Experiments (Mike)
Mike described how the LHC bunch
configuration will be transmitted. The communication of the machine with the
experiments proceeds via three channels: (1) the General Machine Timing (GMT)
network [see e.g. J. Serrano et al, “Nanosecond level UTC timing generation and stamping in CERN's LHC,” ICALEPCS-03 Korea], (2) the Beam Synchronous Timing (BST)
network [see D. Dominguez et al, “An FPGA based multiprocessing CPU for Beam Synchronous Timing in CERN's
SPS and LHC,” ICALEPCS-03 Korea], and (3) the CERN
Data Interchange Protocol (DIP) [https://edms.cern.ch/file/457113/2/DIPDescription.doc ].
Mike highlighted that 21st
century tools will be used for the data transfer: ORACLE, JAVA API for
Parameter Control (JAPC), and Controls Middleware (CMW) with get/set methods
and publication&subscription facilities [see e.g.
K. Kostro et al, “Controls Middleware - The New Generation,” EPAC02 Paris]. All these aspects are well covered. Basic timing concepts include the following: Events arrive asynchronously and can be
subscribed to. Each event is a 32-bit quantity. The payload is 16 bit. Telegrams are sent at fixed frequency. Safe Machine Parameters (SMP) are
distributed in the form of events and telegrams. Mike presented examples of
telegram groups descriptions (slide 4) and distributed events (slides
5-6). He also recalled Philippe’s three
conventions.
The injection request involves several events: BTNI (next injection
beam type), BKNI (next injection rf bucket), RNGI
(next injection ring), and BCNT (number of CPS batches). 2x4 telegrams are used for defining the beam type (spacing, intensity. particle
type etc). Telegrams exist both for injected and circulating beams.
Roger asked if the beam type would
change from cycle to cycle. Mike replied that, yes, the beam type could change,
for example, if we go from the pilot bunch to an intermediate beam. Gianluigi
inquired whether this information would be sent out at every injection. He
suggested that one must also specify the number of batches as part of the beam
type. John asked if the particle type could have other values than 0 (protons)
or 1 (ions) to encode ions other than lead. Mike answered that the data
structure allowed 16 bits to encode the particle type, so that “2” might
eventually represent argon, etc.
Mike presented a proposal for
defining the bunch configuration (slide 10), while also referring to an alternative
proposal by Andy. Mike’s proposed scheme covers all cases. It is set on the
database, which stores the LHC run configuration and which is published to DIP
(response time ~second) and via JAVA Messaging Service (JMS) to any high level
application. He recommended to set the RF FESA property rather than to
distribute it over the timing system.
Philippe asked if the distribution
is not done via the timing, how one could be sure to get this information on
time. Andy pointed out that the time was not critical, since a time interval
equal to a full SPS cycle is available to do it. Massimiliano posed the
question how much could be distributed over the timing. Andy answered 4
integers – sufficient to describe the incoming beam. Mike announced that a dedicated
discussion on this topic would happen immediately after the 42nd
LHCCWG meeting.
The next item addressed was the bunch intensity from the
FBCT. Mike stated that he would push the average bunch intensities over the
DIP.
Massimiliano resumed his earlier
question on satellite bunches and he asked whether those would show up.
Jean-Jacques replied that the satellites will not be seen here (due to the 40
MHz data acquisition rate); the only monitor that would notice them is the
longitudinal density monitor - which would not be available for phase 1.
However, Jean-Jacques offered as a temporary replacement the dedicated pick-up
(APWT in point 4) that BI will connect to an oscilloscope. This should
hopefully be able to detect satellite bunches of a few % of nominal intensity.
He added that if the satellite bunches were big enough they could lead to a
strange behavior of some applications, but there was no easy way to find them.
Roger asked if all questions from
the experiments had been answered. Massimiliano replied that all these points
and conventions should be defined in an up-to-date document e.g. in EDMS, since
it was difficult to refer to presentations. Roger asked Philippe whether he
would like to document the bunch numerology in this way.
Jean-Jacques pointed out that BI
counts in 40 MHz slots, not in buckets. The BI instruments would therefore not
resolve the individual RF buckets. Gianluigi suggested that one might possibly
infer the presence of satellites from the DCCT reading. However, Jean-Jacques qualified this option
by saying that the DCCT and fast BCT will both integrate satellite bunches, so
the difference will not say anything. The difference could give theoretically
an idea of the ‘de-bunched’ intensity (that we may lose during the ramp) but
the final resolution of the DCCT (a few 10^9 p) and the uncertainty on
cross-calibration will tell what we can do. Jean-Jacques also reminded that the
BCTDC data transmitted via the timing system for the validation of a safe beam
will only have 1 bit data corresponding to 1e10 protons, and that e.g. the
intensity shown for the pilot beam on the general machine timing network will
be zero.
Recognizing that this theme concerns
different groups, Massimiliano stressed that at such interfaces weaknesses usually
occur. He reiterated that, therefore, a write up would indeed be marvelous. He
recommended a continually updated official documentation. Philippe agreed to
write a document for definitions and conventions.
Philippe inquired if all the information shown by Mike will
be available from the start up. Mike replied in the affirmative, saying that it
was already available in large parts.
Commissioning of
the Beam Instrumentation (Jean-Jacques)
Jean-Jacques explained how the
machine check out for the beam instrumentation will be organized. It will be
scheduled after the BI HWC. The goals of the instrument machine check out are
several: to verify the integration in the control system, to check the
functionality, to debug the devices as much as possible before beam, and , last
but not least, to ‘have people involved’ (‘commissioners’ from OP/ABP – already
identified) which will then know and ‘own’ the systems. This activity will be
based on the AB-BI LHC Technical Board Wiki : Dash Board . Jean-Jacques provided important background information
for the various columns of the dash board (slide 4).
The BI SW responsible of the
instrument knows best what can be tested and when. He/she will organize the
continuous assessment of his/her instruments, which implies to find/organize
time slots outside official dry runs where conditions allowing testing of the
dashboard are met, to invite the instrument team members to the tests, and to
maintain & update the dash board information. BI would participate in the
LSA dry runs, and the dash board should show to the LSA dry-run organizers the
instruments available. In addition, to work efficiently, BI was asking for
certain timing events which LSA should play regularly outside the dedicated dry
runs. Some example events were mentioned – all of these were normal events.
Roger asked whether there could be
any interference with the hardware commissioning Jean-Jacques gave the
reassuring answer that in principle no, none had been identified yet. The only
tricky event is the post-mortem event and it’s not part of the list for the
moment.
The requested timing table contains
events related to injection beam#1, injection beam#2, energy ramp, beam dumped,
etc., which are interleaved with minutes-long sleeping periods. The underlying
idea is to simulate a realistic sequence and play it for testing all
functionalities of the instruments. A “start simul
table” event (TBD) is sent prior to the simulated beam injection to pre-check
the instrumentation. The ramp events would be used e.g. to test BLM threshold
usage or to protect wire scanners. Jean-Jacques pointed to the need of sending
the next injected beam telegram group RNGI in LSA and of agreeing on a
temporary central timing (CTIM) for the “start simul
table”.
Assuming that his proposal was accepted,
Jean-Jacques presented a corresponding planning for the BI access. He pointed
out that due to RF tests there could be no access in point 4 in the 2nd
half of May and June. Most BI instruments should be ready for checkout much
earlier, during April. He recommended that the LHCCWG and BI work together on
this. As immediate next steps, LSA has to create, test and run the agreed-upon
timing table, and BI will organize dedicated meetings for each instrument
before the end of March.
Jean-Jacques also suggested that a
few events for BT and/or RF could be interlaced in the proposed sequence,
allowing e.g. BI tests between the dry runs. Reyes confirmed that a similar
tool had been prepared for BT, which was now in place for the EMC test, and the
next reliability run. The same could be done for BI. Jean-Jacques fully agreed
and he advocated to integrate Etienne’s needs with those of BI.
Gianluigi asked whether an energy
ramp was possible. Jean-Jacques responded that no real ramp was needed, but
that the ramp could be entirely simulated. Mike commented that Julian could
provide this simulation. Jean-Jacques said that he did not know of any side
effects of the “beam dumped” events and the energy ramping. Roger asked if the
potentially interfering events could not be replaced by fake events.
Jean-Jacques replied that this was not easily possible, e.g. for the injection
of beam#1 Lars would not like to change his 64 crates to a dummy trigger. The
easiest solutions would be if all the teams concerned would disable their
response to certain events, so that one e.g. would not perform a wire scan at
every event. Lars mentioned that one would also need to test post-mortem crates
with possible side effects on the HWC of power converters and QPS, and this
latter test must be executed in any case.
In conclusion, and RF, BI, BT will
agree on a common list of events to be played regularly by LSA. Gianluigi
highlighted the organization of the mini-teams.
AOB
Roger announced that the next LHCCWG
meeting would take place on Tuesday April 8 to avoid a possible conflict with
Easter and to be able to take into account the feedback and conclusions from
the extended LTC.
Next Meeting
Tuesday April 8th, 14:00
CCC conference room 874/1-011
Provisional agenda
Minutes of previous meeting
Matters arising
Other agenda items yet to be
determined
AOB
Reported by Frank