|September 10, 1997||All That Fits is News to Print||Vol.10,No.3|
Page contact and owner at end of this issue.
July 17, 1997
|Author: Grossberg,Hendrickson||Subsystem: Fast Feedback||User Impact: Large|
|Panel Changes: Few||Documentation: Yes||Help File: Yes|
The luminosity optimization at the SLC has been limited by the precision with which one can measure the micron size beams at the Interaction Point. Ten independent tuning parameters must be adjusted. An automated application has been used to scan each parameter over a significant range and set the minimum beam size as measured with a beam-beam deflection scan. Measurement errors limited the accuracy of this procedure and degraded the resulting luminosity.
A new luminosity optimization feedback system has been developed using novel dithering techniques to maximize the luminosity with respect to the 10 parameters, which are adjusted one at a time. Control devices are perturbed around nominal setpoints, while the averaged readout of a digitized luminosity monitor measurement is accumulated for each setting.
The loops typically collect data at 120 Hz, and take about two minutes to average data for each calculation. A parabolic fit determines the optimal setting. Loops are available to control SQ3, X and Y waist, and X and Y ETA for both FF01 and FF11. Currently the CCS feedback for a given micro may not be in feedback while an optimization loop is running; this is expected to change soon so that the CCS loop can be on while optimization is running.
Each loop uses one multiknob. The knob is moved through a sequence of 3 settings: nominal, above and below nominal (also known as center, up and down steps). The knob incorporates the same quadrupole and/or LGPS devices which are used in the pre-existing optimization scans.
Each of the luminosity loops has a single state element. This is the offset of the parabola, which is in the same units at the associated knob. So for example, for a waist knob, if the optimal point is estimated to be at 2 cm, then the state would equal 2.0. When the offset is near zero, we know the feedback has converged.
Each loop includes two kinds of measurements: raw and derived. Raw measurements are data collected at 120 Hz, and include the luminosity monitor ARRY readouts in addition to the intensities for the electron and positron beams. Derived measurements are calculated typically using averaged values from 10,000 pulses. For each of the 3 dither settings (center, up and down), the luminosity monitor readout is averaged, after being normalized to both of the beam intensities for each pulse. The last derived measurement is the estimated offset of a parabola, which is calculated by solving the parabolic equation.
While a loop is in dither mode, it has control over all of the associated actuator devices (such as correctors, quads, etc). So it would not be reasonable to perform a scan-style optimization using the same devices while a loop is in dither mode. In dither mode, the parabola offset is estimated, but no changes are made to the control knob. In feedback mode, the control knob is adjusted to correct the offset. A gain factor is applied to the calculated correction, but the applied correction is limited to the size of the dither step. Note that no exponential averaging is included in the feedback matrices for these loops, unlike standard feedback loops. The straight average for a user-defined number of pulses is all that is used.
Some special handling has been added to the fast feedback code, activated when optimization loops are turned from Feedback or Dither to OFF. For the fast correctors of all optimization loops, BACT is not written to BDES, as is usually done when feedback loops are turned off. Instead, BDES is retained at the value it held before the loop was turned on, and the magnets are trimmed to that value. For the two QUADs of the ETA loops whose values must always be equal and opposite, the value of the second QUAD is set equal to the value of the first QUAD, with opposite sign.
A slow feedback scheduler is available to turn on and off the optimization loops one at a time. While the slow feedback is off, the SCP software provides some support to avoid conflicts when optimization loops are turned on. No more than one CCS or luminosity optimization loop is allowed to run in a given micro at a given time. If one of these loops is already in feedback or dither on the micro of the requested loop, then it is turned off before the requested loop is turned on. More specifically, the optimization loops are turned to OFF and the CCS loops are turned to compute.
While the slow scheduler is not running, the following procedure is recommended. Before turning on an optimization loop, first check that no optimization loop is running in the other micro. It is not recommended to have optimization loops on in FF01 and FF11 at the same time, and it is up to the user to avoid this. Turn the selected optimization loop to dither or feedback. The associated CCS loop is automatically turned to compute and a message is issued. When finished, you may either turn on another optimization loop in the same micro or turn the CCS loop to feedback. This will automatically turn off the optimization loop if it is running.
For example, if FF01_CCS and FF11_CCS are running in feedback on FF01 and FF11 respectively, and an operator requests that FF01YETA be turned to dither, then FF01_CCS will be turned to compute before the request is filled. If, subsequently, the operator requests that FF01_CCS be turned to feedback, then FF01YETA will be turned from dither to OFF before the new request is filled. These changes will not affect what is going on in FF11, and FF11_CCS will continue to run in that micro until a request is made to turn on one of its optimization loops.
Note that the software will not automatically turn loops back on, so it is the user's responsibility to make sure the CCS feedback is turned back on as necessary. Note also that we do not have enough CPU to run all of the optimization loops in compute at once, and the software does NOT protect against this. So the user is requested to turn optimization loops to OFF when they are not in use.
For each feedback loop, parameters associated with the dithering pattern and averaging time may be entered. The luminosity parameter panels for the North and South are accessible from the fast feedback magnet index. The user can control the size of the dither, averaging pulses and settling times. Furthermore, help on these touch panels gives details of the feedback implementation.
August 7, 1997
|Author: Mike Zelazny||Subsystem: PEP2||User Impact: Large|
|Panel Changes: Many||Documentation: Yes||Help File: Yes|
The BPM software in the control system has been upgraded to accommodate the PEP2 storage rings.
There are three new concepts in display group management.
When selecting data collection and display ranges the SCP will allow you to select a starting and ending position on any micro boundary or at the point of beam injection and will allow up to one full ring turn. In the high energy ring, for example, the beam is injected into PR10 and travels clockwise through PR12, PR02, PR04, PR06, PR08, then back through PR10, and so on. Selecting a range of PR02 to PR02 will display data for that micro only. Selecting PR02 to PR12 will display a full ring turn. PR10 is divided into two parts: the area just after injection and the area just before injection. The SCP prompt will allow you to start or end your display at either of these locations in order to get one full ring turn starting at the point of injection. In the low energy ring the beam is injected into PR08 and travels anti-clockwise, but the same concept of the division of PR08 applies.
At LEP a very useful and accurate method to measure the betatron phase advance has been developed using the 1000 turn BPM system. At SLAC we have a similar system for PEP2. The basic idea for PEP2 is that the tune measuring system is used to excite the beam at the tune frequency. This is done in either the horizontal or vertical direction. Synchronized data for 1024 turns are collected. An oscillation with a frequency given by the fractional part of the tune should be evident. That is it should have the form x = A (2* i*q + phi) where i is the turn number and q is the fractional tune. The BPM processors can fit this oscillation and return an amplitude, A, and phase, phi to the SCP. This phase is the desired phase advance except for an arbitrary phase offset. The phase advance data can be used to calculate the beta functions using some information from the model. This only requires that the transfer functions near the BPM are accurate.
From the PEP2 High Level Applications SCP panel, a link exists to the PEP2 BPM Phase Advance and Beta Function panel. From here you can collect and display the BPM sine fit data, and save and load configuration files containing this data. Several plots are available:
phi(measured) - phi(model)
phi(measured) - phi(config1)
phi(config1) - phi(config2)
beta(measured) and beta(model) on the same plot
July 31, 1997
|Author: Greg White||Subsystem: Correlation Plots||User Impact: Small|
|Panel Changes: Few||Documentation: Yes||Help File: No|
Control of the PEPII Master Oscillator has been added to Correlation Plots. Two variables are available for both stepping and sampling, named ``PHYS HER RF_DFREQ" and ``PHYS HER RF_DENGY" for frequency and energy respectively. When the LER comes online they will work copacetically.
Their values are ``delta'' to exactly 476,000,000 Hz, and given in kHz. Now, at the time of writing the Master Oscillator actually reads nominally in the range approximately 476,000,320 to 476,000,350 Hz, so for instance RF_DFREQ will be nominally valued about 0.32 to 0.35. Correspondingly, when stepping RF_DFREQ, the steps will be in Khz from 476,000,000 Hz exactly.
When stepping, Correlation Plots does take care of unlocking the ring from the linac, stepping the acquisition, and then re-locking the ring.
Just to state the obvious, stepping both HER and LER isn't allowed.
August 27, 1997
|Author: Ron Chestnut||Subsystem: History Correlation||User Impact: Small|
|Panel Changes: Few||Documentation: No||Help File: None|
Epics channels are now supported by the History Buffer Correlation facility. These channels can be attached or entered just like SLC database items. Epics channels are now supported in most SLC subsystems:
History buffer correlations
Summary Information Plots (SIP)
For reference, please see the November, 1955 special INDEX PANEL issue on Epics Integration.
July 30, 1997
|Author: Ron Chestnut||Subsystem: Wire||User Impact: Small|
|Panel Changes: None||Documentation: No||Help File: None|
A new Hardware Descriptor (HDSC) bit has been added for the WIRE primary. This bit indicates that the wire sizes (secondary SIZE) are given in nanometers instead of the usual microns.
This bit is currently set for FB69 WIRE units 535 and 536 (the laser wires). Be advised that the bit is valid for all wires in a wire unit, so if micron and nanometer wires are mixed, a 40 micron wire would have the value 40000 as its size.
Back to top of this issue
|July 30, 1997||Index Panel||indx103|
Translated from original PlainTeX by index2html.pl.
*Links followed by an asterisk are limited to SLAC clients only.