PEPII Interaction Point Beam Trajectory Control

Preface: Control of the beam trajectories at the PEPII Interaction Point is a somewhat complicated topic which has evolved as the PEPII machine has matured. This page is intended to serve as guide to control room staff who wish to become familiar with the methods (and their evolution) of PEPII IP trajectory control. What Devices are involved? All of the trajectory control is implemented with steering magnets (XCORS and YCORS). Monitoring devices include BPMs and the PEPII luminosity monitor. There exists a long list of interleaved feedback loops which measure the monitors and control the steering magnets. Each of these will be discussed further below.

Individual steering magnets:

Sixteen steering magnets are involved in the IP steering control of the two beams. The magnets are:

Multidevice Knobs:

Multidevice knob files have been constructed to steer each beam's ordinates at the interaction point. There exists a knob file for each of the following:

The idea behind these knobs is that each adjusts the IP trajectory ordinate intended (i.e. HER IP X POSITION) without effecting any of the other IP ordinates (i.e. HER IP X ANGLE or HER IP Y POSITION or ANGLE) or the orbit outside the knob region. Due to design interaction region coupling in both HER and LER, each of these knobs adjusts four XCORS and four YCORS, so a complement of sixteen steering magnets (four HER XCOR, four HER YCORS, four LER XCORS, and four LER YCORS) comprise the set of devices used to control the two beam's IP steering.

A table of coefficients for each of these steering multiknobs is available here.

Pictures of what these bumps do to the HER/LER orbits can be seen here.

Beam Position Monitors

Most of the PEP IP feedbacks involve states which are determined by readout of beam position monitors.

Support tube BPMs:

The two BPMs closest to the interaction point are termed the "support tube BPMs". There is one on each side of the IP. Buried inside the BaBar detector within the permanent magnet support tube, these BPMs are each 0.8 meters outboard of the IP (between QD1 and B1 permanent magnets).

Being inboard of IR "Q2" magnets, the support tube BPMs see both beams.

• In a by-2 collision pattern (bunch separation 1.3 meters) the first parasitic crossing happens at 0.63 meters from the IP.
• With the support tube BPMs at 0.8 meters from the IP, bunches from the two beams (in a by-2 pattern) are separated by only 1.1nS.
• Even in single bunch collision the time separation of the two beam's signals at the support tube BPMs is only 5.3 nS.
• This small timing separation of the two beam's signals precludes readout of one beam's position in the presence of the other.

So the support tube BPMS are only useful if there is substantial (10's of RF buckets or 20's of nS) timing separation between the two beams. During collisions, the support tube BPM readout is not meaningful.

The primary roles of the support tube BPMs include:

• Steering the IP orbits of the two beams close enough to find overlap using other means (beam-beam deflection, beam-beam tune shift, luminosity signal).
• Calibration/tuneup of the IP steering multiknobs¹
• Providing a "pick-up" signal for IP beam timing diagnostic (oscilloscope).
• Providing history monitor of IP positions/angles sampled during "single beam" conditions as a general diagnostic tool (see PR02HIPS and PR02LIPS feedbacks below).

¹ The technique for measuring the two beam projected beam size at the PEPII interaction point (in X,Y,U and V directions) involves scanning the HER beam across the LER beam using the IP steering multiknobs and measuring luminosity signal vs. beam separation. The reported overlap width turns out to be in knob units. So the beam size calibration comes from the knob calibration, which in turn comes from the support tube BPM readout.

PEPII Luminosity Monitor

The PEPII luminosity monitor is used by one of the IP trajectory control feedbacks ("IPXY"). This device measures the flux of radiated photons from the IP e+/e- interaction to determine a luminosity measurement with bandwidth ~300ms. See SLAC Pub 8688 for more information on the PEPII luminosity monitor.

Feedback Loops:

During most of the life of PEPII, the complement of feedback loops associated with IP steering and monitoring has been evolutionary. The following table summarizes the names and status (as of January 2005) of each of these loops:

PR02IPXY

PR02IPXY is the primary feedback loop use for keeping the PEPII beams in collision. It is a "dither style" feedback. Feedbacks of "dither style" move (dither) some control device seeking to optimize or maintain some measured parameter. In the case of PR02IPXY, the dithered device alternates between multiknobs HER_IPX, HER_IPY, and HERIP_YP (although it is possible to exclude HERIP_YP by changing a parameter on the PR02 parameters touchpanel ¹). It was chosen to move the HER beam (as opposed to the LER beam) since the HER IP steering multiknobs do not span any sextupoles, while the LER IP steering knobs do span the SCX1 and SCX2 sextupoles.

The dither process entails moving the dithered device between three settings; nominal, nominal + δ, and nominal - δ. At each of these settings the luminosity signal is measured (averaging available see SCP help for details). Once the dither sequence is complete, a parabolic fit is made to the measurements of luminosity vs. dithered parameter and an optimal luminosity value is determined from the parabolic fit. The feedback then moves the dithered device to the optimum value (moving at most the dither step size δ). It then repeats this process for the next dithered parameter, cycling between HER_IPX, HER_IPY, and HER_IPYP.

Experience with the PEPII machine showed that the dither amount δ was important to the performance of the feedback loop. Large δs are desirable when the beams are far from being centered (i.e. while filling) since the maximum correction per iteration (which effects the correction rate) is δ. During stable beam conditions when the beams are nominally centered, small δs are preferable for reasons which include:

• Only small corrections and correction rates are required.
• Large δs lower the average luminosity.
• Large δs result in a fitted correction with larger absolute uncertainty.

To address this matter of maintaining a reasonable dither amount δ, a process called "dynamic dithering" was developed. Dynamic dithering adjusts the dither amount δ according to a formulation based upon the stored beam currents and number of buckets². The formulation is described in detail in the SCP help but many find the formulation confusing. An overly simple explanation would be that the dither size falls between the range of a (SCP specifiable) minimum and maximum; minimum dither size being used for large beam currents, maximum dither being used for small beam currents.

This dynamic dithering calculation and update of the dither sizes is performed by a slow feedback process called "DITHWTCH".

¹As a footnote it is worth recognizing that in the past IPXY dithered only HER_IPX and HER_IPY. A separate (no longer existent) "HER_YANG" loop would dither the HERIP_YP (or at one point the combined HER/LERIP_YP) multiknob, but could not be done simultaneously with IPXY. A complicated slow feedback scheduler process would alternately schedule dithering control between the two dither loops, with HER_YANG scheduled far less often than IPXY. Presently SCP help still refers to this scheduler process.

²Since inclusion of the HERIP_YP dither into IPXY came after the development of dynamic dithering, dynamic dithering of the HERIP_YP parameter is not available. The static value for HERIP_YP dither amount is specifiable on the SCP PR02 parameters panel.

PR02LERO

PR02LERO is BPM based steering loop which corrects anomalous steering effects to the global LER orbit caused by kicks generated in the IP region.

PR02LERO (also called "PR02 LERO ORBIT2") is the second incarnation of this feedback loop (read on further to learn about this second incarnation).

Early in the commissioning phase of the PEPII rings, it was observed that there was (significant) global orbit variation with stored beam current. Orbit steering efforts quickly revealed that these current-dependant orbit variations were due to kicks originating in the IP region. It was quickly reasoned that the origin of these kicks was likely thermal motion of IR region focusing elements (quadrupole raft motion) due to synchrotron light heating. These kicks were seen in both rings, with each ring having components related to both ring's currents. The kick amplitudes have slow time-constant response to an impulse change in current consistent with magnet displacement due to thermal heating. A correction algorithm was quickly developed.

The correction algorithm employs BPM readouts far outboard from the IP; out in the adjacent arcs. The scheme to determine kick amplitude entails fitting the incoming and outgoing arcs' trajectory measurements into four states (per plane); incoming position and angle, kick angle at the IP, and kick angle at the chosen correction magnet; the last of these becoming the error signal for the feedback loop.

The first incarnation of PR02LERO used as correction PR02 XCOR 3027 (HK1R) and PR02 YCOR 3024 (VCOR1R).

Note that under this technique, a global betatron oscillation originating outside the arcs adjacent to the IP result in no error signal in this feedback's (controlled) states.

Judicial choice of the arc BPMs used to determine these kicks is important. The reason being that in order to properly fit for these IP region kicks, each arc's ensemble of BPM units must adequetly span both phases of betatron motion for each plane. Additional complication arises with the (later) advent of the PR02SEXT feedback loops which steer local bumps in the IR adjacent arcs; LERO's arc BPM complement needs to be chosen such that changes associated with the PR02SEXT loops are transparent to LERO.

BPM readout jumps plague LERO. A random non-physical readout change in one of LERO's BPMs causes a random change in the resultant IP kick determination...causing the feedback to steer the beam...sometimes out of collision enough to cause a beam abort. Similarly, a single BPM with sufficiently large influence on the IP kick determination whose readout drops (say NoQ status) will also cause a change in the kick determination. Towards desensitizing feedback to these pathologies, a process of "chi-square filtering" was developed.

"Chi-square filtering" refers to a technique in which measurements can be rejected. Since the IP kick determination is made from a chi-square minimization fit, there is available fit error (difference between measurment and fit) for each BPM readout. This difference is called the "chi-square residual". If a given BPM's chi-square residual is too large (allowed range specified for each measurement on the SCP feedback measurements panel) that BPM's measurement is flagged as "bad" and will not be included in the overall fit. In this condition the feedback loop searches for the offending BPM measurement (not necessarilly the unit with the out-of-limit residual... see SCP help for details) and flags it as "suspect".

Some operational principles which have been suggested to help minimize suseptablility of LERO to excursions due to BPM pathologies include:

• Avoid allowing the "number of bad measurements allowed" parameter to become excessive. If a measurement with large influence on the IP kick determination drops out and is allowed to be excluded, the resulting kick determination will change unphysically causing the feedback to steer undesirably.
• Re-gold the feedback loop when the chi-square residuals become large. All of the fitting is done based upon measurements with the gold orbit subtracted. If the actual orbit and the gold orbit diverge over time in a manner inconsistent with the parameters being fit, the chi-square residuals will increase. In this condition, a dropout in a single BPM readout has greater influence on the resultant fitted kick, and suseptablilty to these BPM pathologies therefore increases. (For an demonstration of this arguement look here.)
• Be thoughtful about taking BPM measurements offline. Remember that both sides of the IP need enough measurements to adequately span both phases of betatron motion in both planes in order to extract two phases (per plane) of IP region kick.

The Second¹ Incarnation: "LERO2"

Somewhere along the PEPII operational time line it was observed that during the sequence of stable beam running→abort→filling→ stable beam running, corrections made by IPXY were largely due to steering effects attributable to the LERO feedback. In this condition it was difficult for IPXY to maintain the beams in collision since it was "chasing" the steering caused by LERO.

Plausable cause for this condition was that the LERO technique was correcting for orbit kicks generated elsewhere than the location of the LERO correctors; leaving a significant IP orbit excursion in the LER for which IPXY needed to use the HER ordinates to correct.

LERO's first incarnation used as correctors HK1R and VCOR1R. IR2 raft motion studies later indicated that motion was seen on the opposite side of the IP. It was further recognized that the kick location(s) and the corrector locations probably did not well coincide. To address these issues two changes were made for the second incarnation of LERO:

• Correctors were moved from the right (or forward) side of the IP to the left (or backward) side of the IP (HK1L instead of HK1R, VCOR1L instead of VCOR1R).
• Instead of using a single corrector per plane which probably did not have the same betatron phase as the thermal kick(s) being compensated, a second pair of correctors (XCOR 2081 and YCOR 2186 with different phase than the original pair) were added to better match the correction to the source(s) of the thermal kicks.

¹Note that while LERO has been updated with these enhancements, to date (January 2005) no such update has been made to HERO.

PR02LER1

PR02LER1 (aka PR02 LERO OLD) is the leftover database setup for the first incarnation of LERO (see link to PR02LERO above for futher details).

PR02HERO

PR02HERO is BPM based steering loop which corrects anomalous steering effects to the global HER orbit caused by kicks generated in the IP region.

Early in the commissioning phase of the PEPII rings, it was observed that there was (significant) global orbit variation with stored beam current. Orbit steering efforts quickly revealed that these current-dependant orbit variations were due to kicks originating in the IP region. It was quickly reasoned that the origin of these kicks was likely thermal motion of IR region focusing elements (quadrupole magnet motion) due to synchrotron light heating. These kicks were seen in both rings, with each ring having components related to both ring's currents. The kick amplitudes have slow time-constant response to an impulse change in current consistent with magnet displacement due to thermal heating. A correction algorithm was quickly developed.

The correction algorithm employs BPM readouts far outboard from the IP; out in the adjacent arcs. The scheme to determine kick amplitude entails fitting the incoming and outgoing arcs' trajectory measurements into four states (per plane); incoming position and angle, kick angle at the IP, and kick angle at the chosen correction magnet; the last of these becoming the error signal for the feedback loop.

HERO uses as correction PR02 XCOR 8020 (HCOR1R) and PR02 YCOR 7052 (VCOR1L).

Note that under this technique, a global betatron oscillation originating outside the arcs adjacent to the IP result in no error signal in this feedback's (controlled) states.

Judicial choice of the arc BPMs used to determine these kicks is important. The reason being that in order to properly fit for these IP region kicks, each arc's ensemble of BPM units must adequetly span both phases of betatron motion for each plane. Additional complication arises with the (later) advent of the PR02SEXT feedback loops which steer local bumps in the IR adjacent arcs; HERO's arc BPM complement needs to be chosen such that changes associated with the PR02SEXT loops are transparent to HERO.

BPM readout jumps plague HERO. A random non-physical readout change in one of HERO's BPMs causes a random change in the resultant IP kick determination...causing the feedback to steer the beam...sometimes out of collision enough to cause a beam abort. Similarly, a single BPM with sufficiently large influence on the IP kick determination whose readout drops (say NoQ status) will also cause a change in the kick determination. Towards desensitizing feedback to these pathologies, a process of "chi-square filtering" was developed.

"Chi-square filtering" refers to a technique in which measurements can be rejected. Since the IP kick determination is made from a chi-square minimization fit, there is available fit error (difference between measurment and fit) for each BPM readout. This difference is called the "chi-square residual". If a given BPM's chi-square residual is too large (allowed range specified for each measurement on the SCP feedback measurements panel) that BPM's measurement is flagged as "bad" and will not be included in the overall fit. In this condition the feedback loop searches for the offending BPM measurement (not necessarilly the unit with the out-of-limit residual... see SCP help for details) and flags it as "suspect".

Some operational principles which have been suggested to help minimize suseptablility of HERO to excursions due to BPM pathologies include:

• Avoid allowing the "number of bad measurements allowed" parameter to become excessive. If a measurement with large influence on the IP kick determination drops out and is allowed to be excluded, the resulting kick determination will change unphysically causing the feedback to steer undesirably.
• Re-gold the feedback loop when the chi-square residuals become large. All of the fitting is done based upon measurements with the gold orbit subtracted. If the actual orbit and the gold orbit diverge over time in a manner inconsistent with the parameters being fit, the chi-square residuals will increase. In this condition, a dropout in a single BPM readout has greater influence on the resultant fitted kick, and suseptablilty to these BPM pathologies therefore increases. (For an demonstration of this arguement look here.)
• Be thoughtful about taking BPM measurements offline. Remember that both sides of the IP need enough measurements to adequately span both phases of betatron motion in both planes in order to extract two phases (per plane) of IP region kick.

HLERYANG

HLERYANG feedback is under development as of January 2005.

In a previous implementation, this feedback loop was of "dither style" (similar to IPXY) dithering a multiknob-type combination of the HERIP_YP and LERIP_YP knobs to optimize luminosity. In this previous implementation, a slow feedback scheduler alternated between IPXY and HLERYANG. This implementation of HLERYANG, together with the slow feedback scheduler, was abandoned with the incorportation of the HERIP_YP dither state into the IPXY feedback loop.

The present development is intended to have this loop actively control a combined multiknob of the HERIP_YP and LERIP_YP knobs (a "common HER/LER" Y angle). The measurement to be regulated will be the HER beam's IP Y angle as measured by HER BPMs (in the same manner as monitor only loop PR02HIPP).

An observant reader will immediately identify an apparant conflict between this development for HLERYANG and IPXY feedback. The apparant conflict being that both loops proport to control the HER Y angle; IPXY based upon dithering for luminosity optimization, and HLERYANG for HER BPM readout.

To resolve this apparent conflict the following lessons from PEPII Operation should be considered:

• Angular overlap with the LER beam: Both beams have a finite length in z, a small IP βy^*, and a large IP θy^*. The "overlap area" (in Y-Z space) and therefore luminosity decreases as the relative vertical crossing angle between the two beams increases. As the bunchlengths and train phase transient increase (both with increasing current), this luminosity sensitivity to vertical angular overlap increases.

  This effect is much smaller for the horizontal plane due to the larger horizontal beam size.

• Pointing of the beams towards the luminosity monitor: The vertical angular acceptance of the luminosity monitor is at most several sigma of beam divergence (guess ~700uR FHWM). If the IP angle does not point to the center of this acceptance, some luminosity signal will escape the luminosity monitor.

• IP vertical dispersion: HER angular IP trajectory transposes to a position offset in the final douplet quads Q4/Q5. Offset trajectory in quadrupoles introduces dispersion which dilutes the IP beam size correlated with energy spread.

Now recognize that IPXY feedback dithers the HER Y angle to optimize luminosity based upon whatever LER IP trajectory might exist. If the LER happens to be 500uR from the optimum, then IPXY will find a HER Y angle luminosity maximum (maximum angular overlap) at 500uR away from the optimum. So with IPXY only, the HER IP Y angle tends to optimize angular overlap with the LER regardless of where such overlap leaves the (absolute) HER angle.

So the intent of this new development for HLERYANG (relatively slow) feedback is to force the IPXY (relatively fast) dither feedback solution for optimal HER Y angle to find it's optimum at a prescribed HER Y angle (based upon BPM readout); adjusting the LER (and HER) Y angle as necessary.

As of January 2005 this technique is awaiting commissioning.

PR02LIPA

PR02LIPA (aka LERIP X&Y ANG) measures LER BPMs in the same manner as monitor-only loop PR02LIPP. This loop has provision to control the LER IP X and Y angle using the LERIP_XP and LERIP_YP multiknobs, similar to now extinct "LER IP CONTROL" which controlled the LER IP X angle.

PR02HEXP

PR02HEXP is a novel feedback which takes advantage of complication which arises from interleaving multiple feedback loops.

The complication is that many of these IP feedbacks share correctors. To allow this, a "dual-feedback" scheme was developed. With this scheme a corrector's output is controlled by the sum of the requests made by the feedback loops which use said corrector.

PR02HEXP takes advantage of this dual-feedback capability. This loop has as actuators (presently) the HER X angle multiknob and the combined HER-LER Y angle multiknob.

PR02HEXP makes no (useful) measurements. Yet changes to it's states (ordinates mentioned above) are made in addition to whatever other feedback's requests happen to be. This allows Operator control of these ordinates without the need to negotiate the status of all of the other feedback loops which control the steering magnets desired.

So in the classical sense, PR02HEXP is not a feedback at all. Rather, it is merely an open-looped trick which allows the Operator to change the desired ordinate while the other feedback loops remain operational.

PR02FFWD

PR02FFWD is under development.

PR02FFWD is intended to be used in a manner similar to PR02HEXP. The devices to be controlled will be some ensemble of the IP steering magnets (as of January 2005 the HER IP X Position knob and the HER IP Y position knob) to be determined by further study.

PR02FFWD is not a feedback loop, since it does not attempt to correct a measured state. It is rather a "feed-forward" in which the control devices are actuated directly (or "blindly") based upon the measurement of some external set of devices.

The external measurements are the HER stored beam current and the LER stored beam current.

One must have the proper perspective about the behavior of the PEPII IP steering during filling to appreciate the motivation for PR02FFWD. See PR02HERO and/or PR02LERO for perspective about anomalous IP steering due to beam current dependant heating.

The idea behind PR02FFWD is that the behavior of the IP position ordinates as the beam current increases can be predicted. With an accurate prediction, a correction can be applied preemptively. Without such preemptive correction, this correction needs to be made by the PR02IPXY dithering feedback loop which has limited correction rate.

So the plan with PR02FFWD is correct as much of the beam current based IP steering as can be predicted, leaving IPXY with less correction to make and thereby leaving IPXY better opportunity to maintain collisions while filling.

PR02HIPP

PR02HIPP is a "monitor only" feedback. BPMs monitored are PR02 7052 and PR02 8012 both X and Y. These two BPMs are the closest HER only BPMs to the IP, situated just outboard of Q4 between Q4 and VCOR1. There are no routine steering elements between them, so they provide a stable representation of the HER IP position and angle.

PR02HIPP's primary role is to provide a mechanism for reporting the HER IP ordinates to the control system database for stripchart monitoring and for history buffer.

PR02LIPP

PR02LIPP is a "monitor only" feedback. BPMs monitored are PR02 2203 and PR02 3024 both X and Y. These two BPMs are the closest LER only BPMs to the IP, situated just outboard of Q2 between Q2 and VCOR1. There are no routine steering elements between them, so they provide a stable representation of the LER IP position and angle.

PR02LIPP's primary role is to provide a mechanism for reporting the LER IP ordinates to the control system database for stripchart monitoring and for history buffer.

PR02HIPS

PR02HIPS is a "monitor only" feedback. BPMs monitored are the "support tube" BPMs PR02 7083 and 7087 X and Y. The only optical elements between them are the IP B1 permanent magnets, so these BPMs are the best tool for monitoring the HER IP ordinates.

However since both beams traverse these BPMs, their readout is not meaningful unless there is only one beam present.

Special software is used in PR02HIPS to monitor the e+ beam intensity, and only update the e- states when the e+ beam is absent.

PR02HIPS's primary role is to provide a mechanism for reporting the HER IP ordinates to the control system database for history buffer monitoring.

PR02LIPS

PR02LIPS is a "monitor only" feedback. BPMs monitored are the "support tube" BPMs PR02 3013 and 3017 X and Y. The only optical elements between them are the IP B1 permanent magnets, so these BPMs are the best tool for monitoring the LER IP ordinates.

However since both beams traverse these BPMs, their readout is not meaningful unless there is only one beam present.

Special software is used in PR02LIPS to monitor the e- beam intensity, and only update the e+ states when the e- beam is absent.

PR02LIPS's primary role is to provide a mechanism for reporting the LER IP ordinates to the control system database for history buffer monitoring.

Another Very Useful Tool:

There exists a slow feedback process which samples the BACT values of all of these IP steering magnets, and at each update deconvolves from the various magnets' readout and the known multiknob coefficients a value for each multiknob. SCP help on slow feedback group select panel "MKB Hist. Monit." explains the process in detail.

This multiknob reconstruction monitoring (available in history buffer) can prove to be very useful in sorting out issues with the multiple interleaved IP orbit control feedbacks.