AGS F18_house (Mei's) Tune Meter - as it exists today
12 January 2006 L. Ahrens
note: some details of this setup will evolve over the next weeks.
A little system description:
It might be clearer to call this system a coherence monitor/tune meter rather than just a tune meter, but we do not for simplicity. Any intentional transverse exciting of the beam (pulsing the A10 kicker for example) is a separate part of the drill not discussed here.
Signals from horizontal and vertical pick-up electrodes in the F20 straight section are analyzed on every turn during the acceleration cycle. The analog difference and sum signals from the bunch on a given turn (there is only one bunch in the ring for '06 polarized proton operation) are integrated within an rf-locked gate that is somewhat wider than the bunch. If cable delays are properly chosen the gate will remain very well locked around the bunch. The integrated difference signals and the integrated sum signal (only one for the system) are sampled in an analogue sense once per turn. (The electronics are a "track and hold" evolution generated by Siegfried Naase). The resulting levels are held until just before the next turn sample occurs. The resulting difference "steps" are then ac coupled (with a very long time constant ~ milliseconds) and sent into the F18house scope. The ac coupling allows the scope channel gains to not be limited (saturation) by equilibrium orbit offsets. The three channels are digitally sampled once per turn with the sampling time (now the scope clock) again tracking with the rf revolution frequency, though the exact timing of the sample is greatly relaxed relative to the integration gate timing. Three numbers per turn (per ~ 3 usec) are recorded and transmitted from the scope to the application for further analysis. Numbers proportional to difference positions are calculated in the code by dividing the difference voltages by the sum. (This step is not essential for getting coherence and tune information but it allows the simplification of talking in terms of mm's. Whether the step can stay will be a question of signal to noise.) The application then performs Fourier analyses on subsets of the data and makes the resulting spectra, plotted as betatron tune, visible. The peak and area for the two individual spectra are displayed as numbers.
The application is started from StartUp / AGS_applications / AgsTuneMeter
Invoking this application brings up a graphics page that looks very like the old Coherence Monitor page. Data acquisition should automatically begin (a little patience, we are pushing things) - fuzz appears on the two upper (position vs turns) plots. Perhaps 500000 turns (the maximum and also the starting default) are displayed. At 3 usec/turn this covers about 1.5 sec, a nice length for the AGS. Reducing the number of turns requested (an expert task - see below) reduces the load on the data acquisition system, but of course also provides a reduced view of the acceleration cycle.
If in fact nothing happens when tha application is started, or if an error message occurs, then probably there is a problem with the scope or the FEC. This system has not a good track record but a lot of effort has been put into making it more robust. At this point players are Jon Laster (application) and Joe Piacentino (scope interface), Mei wrote some underlying pieces of the analysis. Leif is coordinator at this point.
Via the "fit" program the system can be shocked - possibly into life - in two ways: 1) The scope can be sent an ac reset (Startup/ General programs/ fit/Others/RFD-SCOPE-F18), and then AC reset. 2) The FEC can be reset (fit/FECs/cfe-911b-f18scope1), and then reset. The sequence for recovery is less than clear to me, and perhaps under steady running conditions the system will behave. Joe suggests if the system is hung then if working through FIT, do a scope ac reset and immediately do an FEC reset.
In the menu bar (across the top) click Setup, and then in the popup click Acquisition Mode and then following the arrow Full Ramp… This results in the new interface that looks a lot like the original one only now with additional display boxes at the bottom.
First we investigate options in the menu bar:
On this page again go into the menu bar (across the top). Click Setup. Find some choices (in the following exp = expert = touching may get you into soup):
Modify Ags Variables…(exp)
This allows entering parameters required to connect turns to time. Usually not relevant.
Timing for the A10 kicker machinery, and not relevant here.
This menu button brings up the cursor control used to select the beam turns (when and how wide) to be included in the Frequency spectrum - Tune display discussed below.
FFT X Autosc ale
Provides a toggle to determine whether the Frequency Spectrum display is or is not autoscaled.
provides low level control of the V102 timing delay for the trigger to the F18 scope that is the tune meter data acquisition system. This delay usually listens for AGS event link event #37 ags_tm_tr_trig and usually has a delay of 1 tick = 1 usec after that event occurs before triggering the scope.
TuneMeter PetPage (exp?)
Access to tune meter kicker controls in the A10 house: timing, power supply state.
Scope PetPage (exp?)
This is the scope setup. The memory size and the acquisition sensitivity for the three channels are set here. If channel saturation is a problem this is where the sensitivity can be reduced. Also a shorter record can be requested. To keep the scope in "external clock" mode, the sample clock has to be set (click and enter) to TTL when some other settings are changed. Hitting the explicit Send button on the Time/Div line moves the scope from listening to an external clock (synched with the rev tick) to listening to a real-time clock. Doing this ends tune measurements of course.
Also the start of the scope data readback arrays are shown. These readbacks are a useful diagnostic if things are not happy. Otherwise having them appearing here just loads down the system.
This scope interface is not complete and is rather primitive which is both efficient and confusing.
Back to the Menu Bar: click File. Here are data saving and recovery options. Also a connection to the Main Magnet program for connecting turns with time from T0 (more below).
Now to the stuff on this page:
The top boxes (left, right = horizontal, vertical - there are labels but small) display "Coherence Amplitude", the turn-by-turn beam position. The display is an ac coupled version of the beam position of one bunch on every revolution starting from the Trigger time until the turns requested (500000 max) are used up. Because the normalization (i.e. dividing difference by sum) is carried out before display, saturation of one of the input scope channels can happen without being very noticeable. (We are working on building some intelligence upstream to warn of this).
The two vertical cursor lines on each turn-by-turn display enclose the subset of this data that is used in doing the FFT displayed in the next horizontal band. The cursors are controlled (in an intuitive way … ) via the "Show Cursor " request in the menu bar under Setup.
The really new stuff lies below. Buttons on the next line are here indicated in bold font. Hit the "Add Data" button and the position data is analyzed across the full scan in "bins" containing however many "turns" are specified (on the same line) and with a "shift size (turns)" between each bin's worth of FFT analysis as set in the next box. The same bin size and shift size are applied to both the horizontal and vertical data. The "Trigger" is just the event generating the scope trigger mentioned above, the start of the data accumulation (usec from AGS To).
This analysis only occurs when the Add Data button is clicked and then uses data already accumulated. The time stamp on the resulting display identifies what data set was used for the ramp analysis.
The display is similar to the RHIC tune meter ARTUS display. The frequency spectrum for each bin is plotted as a horizontal line with the color of the line as the frequency or tune slides along being proportional to the height of the spectrum at that point. So each line has the information of a frequency spectrum encoded into the color. The location of the line in the vertical corresponds to the turn number (maybe the starting turn?) for the bin, translated into the time when that turn occurs after AGS T0. The translation uses knowledge of the magnetic cycle that is active. To update this magnet program information: find in the menu bar (for this page) File/Reread B(t). The actual appearance of this tune plot depends strongly on the way the color scaling is set. The program tries to autoscale this, but the result is not necessarily satisfactory. The "Color Map" buttons are a work in progress attempting to make color range adjustments more friendly.
The display can be adjusted to give most of the monitor screen real estate to the upper or lower plotting and the "usual" zooming is possible.
In the menu bar under File there are Save and Save As… options which allow processed data to be saved and then recovered into this program. I have not used these enough to go farther.
Below are attached a few pictures from the application.
Figure 1 shows a typical display. The turn-by-turn pickup data fed to horizontal and vertical is identical (and faked, but at a low level). The horizontal (left) spectrum spans 1000 turns, the vertical spectrum 100 turns giving different spectrum widths. The horizontal color plot is "zoomed". The fake data has a fixed amplitude oscillation and uses the actual ramping rf for the revolution frequency - hence the jog in reported tunes.
Figure 1 A possible display from the tune meter of up-the-ramp data.
Next we use the zoom on the turn-by-turn plot just to show the difference motion - here effective motion of about 50 microns (if the scaling is right). This is taken with most power supplies off in the AGS - a very clean environment.
Figure 2 Here the horizontal turn-by-turn data has been expanded
And then we superimpose one of the pop-ups - this of the cursor control for the single frequency spectrum setups - responsible for the different spectrum widths mentioned above. The calculations done for the spectrums (to be evolved) are the tune for the peak, and the sum of all the points across the full (8.5 - 9) frequency range.