1. DCOPS Monitoring of Iron Bending
DCOPS Readout during MTCC
James Bellinger
13-September-2006
V1.0

2. Outline Layout Original Plan Hardware and Software Complications
DCOPS readout
Results
Further Efforts

3. Review of Layout
The DCOPS laser system as implemented in the MTCC consisted of
1) 6 half SLM lines at the ME+1 position
2) 3 full SLM lines at each of the ME+2, ME+3, and ME+4 positions. Each has two lasers, one at each end.
2 partial transfer lines, consisting of 1 laser and DCOPS at the ME+4/3/2/1 and YE+2/+0/-0/-2 positions (4 others had the ME+ positions, and one of those a YE+2 as well).

4. DCOPS
Each DCOPS has channel CCD's arranged in a square, with light filters in each face to try to equalize the light intensity.
The HSLM's have 1 reference DCOPS and 3 mounted on chambers.
The ME+2/3 SLM's have 2 reference DCOPS and 8 mounted on chambers
The ME+4 SLM's have 2 reference DCOPS and 4 mounted on chambers.
The Transfer Lines have 12 DCOPS, one on each reference plate.

5. Usage The transfer plates represent knowable points in the system.
We ideally can measure the Z distance from one to the next, measure the angle the steel on which they are mounted bends using inclinometers, and use the laser transfer lines to establish the radial positions of each plate.

6. Usage
When the disk bends, DCOPS near the center where the steel bulges forward will record different positions for the laser beam. If the laser were always normal to the Z axis you could easily measure the displacements.
The edge of the disk bends backwards, though, and so each laser points in a new direction.
We should be able to determine the new direction, and from this the new DCOPS positions, using the laser on the other side as a reference, using inclinometers, or using assumptions about how uniformly the disk bends.

7. Original Readout Plan
Communication with the DCOPS is by a serial line: RS422 to an adapter and RS232 into a terminal server.
A java-based DIM program (begun by Evgeniy Sharapov) would alternately turn lasers off and on and read out the on-board averages for the laser beam positions, collecting these and handing them off through PVSS to a database.
Periodically it would read background light levels which can be automatically subtracted on the DCOPS adapter board.

8. The Devil in the Details
Testing was delayed through lack of a timely permit to run the lasers.
We discovered after mounting and carefully aligning the lasers so that they would pass through the DCOPS along the SLM lines, that the green framework was not where it was thought to have been, and almost all the lasers had to be dismounted for closing and then remounted again after the disks closed.
There was not time enough to realign all of the lasers once remounted, and the lack of access and visibility made the job even slower.

9. Consequences
The result of the alignment problem was that for the first part of the magnet test we had only two functioning SLM lines, and even then only one of the two lasers in each was aligned to go all the way through. (The HSLM lines worked fine, but were short and did not see all the bending.)
Very hard work during the times when the magnet was off won us more of the lasers aligned, although not perfectly.

10. Consequences II
It turned out that the on-board calculation of an average to determine laser beam position was often inadequate.

11. First a reminder
Before I go on, I need to emphasize that a great many of our beam profiles look like this. These look great, and the onboard calculation of the average is probably just fine for them

12. Why simple averaging doesn't work

13. Even worse

14. Why?
There appears to be glare from the laser, producing a background under the signal that shifts the average substantially. This mostly effects the DCOPS nearest the laser.
In these plots the upper and lower plots represent parallel CCD's in a DCOPS, and measure roughly the same thing. I surmise there was some kind of local reflection.

15. Combination Deal
This has 4 pathologies: blockage or bad filter (??), nearly off the edge, partial saturation, and a reflection. (Full saturation suppresses the middle of the peak, as Jeff Klukas showed.)

16. We can handle these, but
Some need a little help. A different on-board algorithm could probably deal with these. Until we test and install one, we have to read back the entire spectrum and fit it.
I chose to fit it online, rather than store all the date. That may have been a mistake.

17. Hardware status
Some of the exceptionally low signals may be due to shadowing, or possibly damaged filters, and at least in principle are fixable: in the first case by adjusting upstream apertures and in the second by repair.
Some lasers were broken, and at least four DCOPS was broken during disk assembly. We hope that better care will prevent this from happening again.

18. DCOPS Readout
The Evgeniy-inspired java-based readout system turned out to be over three orders of magnitude slower than expected, thanks to a quirk in the java implementation on Linux and the fact that we couldn't use the on-board averaging reliably.
The existing system uses two java-based utilities: one for controlling the lasers and one for communicating with DIM to pause or stop the program.

19. DCOPS Readout 2 The existing system uses two program streams.
There is a DIM server that creates sentinel files to control the operation of the main readout system. It could in principle be used to monitor the status of the main readout system.
The main readout system is a bash script which reads a configuration file and from it generates and executes auxiliary scripts to control the lasers, read the DCOPS, and write the information to the database.

20. grandloop.sh
Reads (and rereads) control.list, which contains the database information to define which SLM/Transfer lines to read and what they contain and what their database indices are
Periodically turns off all the lasers and reads the background levels for each DCOPS, which information is stored in the DCOPS for background subtraction later.

21. grandloop.sh 2
Turns off all lasers and processes the information for a single Line (SLM or Transfer)
Loop over the lasers for a Line, turning on each in turn
Loop over the DCOPS in a Line, telling the terminal server to tell the DCOPS to read (4 sec) and return the background-subtracted results in hex format (5 sec); at which point the output is pre-processed and handed to a C++ routine running under root

22. grandloop.sh 3
The output of the C++ routine consists of fit values and and error word: This is formatted into an SQL command
When done with the DCOPS loop, the command file containing the SQL commands is executed to store the info in the development database and the laser is turned off

23. Known problems
The development database is used instead of the online database.
If you tinker with the sentinel files by hand, the java DIM server gets confused and the program is no longer controllable remotely.
We don't have a DCOPS account we can all reference

24. Unknowns How often do we have to redo the background reading?
Heavy network traffic in the online subnet can DOS the readout by making the readout time exceed the timeout limit. How often will this happen?
Why does the fit sometimes become unstable?

25. Results
The DCOPS readout runs continuously in a loop taking a little over an hour, so if the magnetic field changes during that time the results are likely to be inconsistent.
We have data from several magnet ramp-ups, and are able to look at displacement vs time and displacement vs current.
We are able to estimate disk bending angles using Transfer Line 5

26. Laser Beam Slope vs Magnet Current

27. Further Results
We find that the fits are not always stable. This is not understood yet.
As expected, the steel does not return to its original position.
It is doubtful that a Transfer line laser will be able to reach the other side of the detector: the steel bends the beam too much.

28. DCOPS ME+3/1-14 (SLM2) Laser PT5

29. Interpretation
As you can see from the previous slide, the center of the laser peak does not return to quite the same place once the magnetic field turns off.
We also see some long-term drifts, which may be relaxation of the iron.

30. Transfer lines

31. In the following slides, the laser (at ME+4) is nearest the upper left
In the following slides, the laser (at ME+4) is nearest the upper left. As you go from left to right or down a row you get farther from the laser.
Plots are omitted if there was no data in that plot. These plots are for stations ME+4/ME+3,ME+2, then ME+1/YE+1/YE+0, then YE-0 and YE-1
The TP2 stations are less constrained than the TP5 stations
The horizontal axis is magnet current in amps
Positions for CCD2 and CCD4 are plotted

34. Behavior of Transfer Line 5
The DCOPS next to the laser shows hardly any variation, of course, but the next two stations show a good deal more. They are both fixed to the same disk, one cantilevered forward and the other backward, which provides a lever arm to enhance the displacement of the first and reduce that of the second.
It is plain from the bottom two MAB DCOPS that the angle through which the beam bends as field is applied will result in a shift that exceeds the aperture of the DCOPS at the far end of a Transfer line.

35. Transfer line
Using the data for ME+1 on Transfer Line 5, I estimate that the laser's mount was bent backward by about 1.1 milli-radian
Using the data for ME+3 ME+1 I estimate a change in the laser direction of .9 milli-radian.
Since the distance between the two ends is about mm, the change in beam position at the far end would be of order 21 mm, although the deflection at YE-0 and YE-1 (at less than full field) suggests that the deflection might be substantially greater.

36. ME+2/SLM2 : PT5 Laser

37. ME+2/SLM2 : PT5Laser
There are 10 DCOPS in this laser line. The one nearest the laser is the lower left, and they lie farther from the laser the higher the row is and the farther to the left.
The laser goes offscale on the upper left DCOPS before the magnet reaches full current.
The near side of the disk does not bend much relative to the laser line: there's a jump as the laser reaches the other side.
The deflection at the center must be

38. HSLM1

39. HSLM1
The laser is nearest the upper left DCOPS, and the DCOPS are farther from the laser as you go right and down.
The deviation is only .5 mm in the farthest DCOPS at 3.7 Tesla. The disk must be bending roughly in the shape of a cone.

40. ME+3/SLM2 : PT5 Laser

41. ME+3/SLM2 : PT5 Laser
In this the laser is nearest the bottom left, at TP5
I do not understand exactly what is going on with the “near side” DCOPS. Station 5 is more constrained than the rest, so perhaps the laser moves less at first, but unless there is some threshold for “give” in the disk mount at high enough force I don't understand the reversal

42. More plots
See for references and plots.

43. Near Term Plans (week)
Try to understand the stability issue. I miss quite a bit of the high field data because of this. It would be very useful to have a full test SLM working, though if this problem actually is related to the magnetic field I won't learn much.
Fold the shimming into a fit, using the complete calibration.

44. Medium Term Plans (1 -2 months)
Incorporate suggestions from Vladimir to speed up the readout.
Tinker with the fitting algorithm and fold several of the processing programs into a single executable.