Go to Introduction to Proposal | Proposal
During Collider Run I, the low beta squeeze was initiated by TCLCK event $C5. On that event, all the power supplies that participated in the squeeze started playing a time table. All of these power supplies' reference cards were programmed to have synchronous time tables with the break points corresponding to well defined lattices on the way to the minimum beta*. To halt the squeeze process before the final beta* was reached, special "incomplete" ramps had to be loaded. This was done occasionally to tune the intermediate steps and is known as "parsing" the squeeze. To continue the squeeze, all the waveform generators had to be reloaded to begin at the point where they were stopped, and continue to the end (or to the next stopping point). There was never an absolute reference to where we were in the squeeze process. There was only the combined length of time that all of the waveform generators had played a particular time table.
For Run II, we will generate an MDAT frame that broadcasts a squeeze index. All of the power supplies that participate in the squeeze will play a function of that MDAT value. This will provide an absolute reference to the lattice step as well as considerably simplifying the process of "parsing" the squeeze. Later on if luminosity leveling becomes operational, the control of the low beta MDAT frame will make the mechanics of leveling easy.
The MDAT frame will contain an index to the squeeze (sequence number) rather than the actual value of beta* because of operational considerations. There are points in the squeeze where the low beta quad gradients have to remain constant while other devices continue to change. An example of this is when the separators change from the injection helix configuration to the collision helix configuration. There is also the need to hold the low beta quad gradients constant while polarity switches are reversed. The polarity switches and separators play as a function of the squeeze index, so the MDAT frame must continue to change while the actual lattice remains constant.
A consequence of generating the reference for the squeeze in this way is that some of the low beta power supplies will need to make larger transitions than the LSB of their reference card. The power supply references are capable of making a transition as small as .076 amps (the LSB of a 468 card for a 5000 amp supply). The squeeze wave form calls for Q5 to change nearly 600 amps per lattice step in the squeeze (near 35 cm beta*). The supply wave form generator would make 7650 LSB transitions per lattice step. But the low beta MDAT frame will probably have a full range of 0 to 32 (unscaled that would be 0 to 65535). The MDAT frame itself will only make 65535/32 or 2048 transitions per lattice step. For the power supply to track this MDAT frame, the supply will have to make steps of four of its normal LSBs. We are still investigating whether this will be a problem that needs attention.
We are also planning to use this low beta MDAT frame for an unrelated purpose. We plan to have 15 distinct lattice steps that correspond to MDAT indices of 1 to 17 (remember the separators and polarity switches). But, we will set the index to zero when we are in the injection lattice (sequence 1) and in injection conditions. After we've finished injecting both the protons and the pbars and have retracted the injection Lambertsons to their stored beam position, we will increment this index to 1. This will NOT change the lattice, but will remove an orbit bump in the injection area. Details of this scheme are mentioned in the specification for the Tevatron orbit program
Mail comments to annala@fnal.gov
For Collider Run II we are planning to put the low beta squeeze tables into the 46x cards as h(i) tables running off their own MDAT, the LBMDAT. (In Run I, the low beta squeeze was a f(t), a time table.) The main benefit of making the low beta squeeze an h(i) table is to make it easier to do luminosity leveling. This also greatly simplifies parsing the squeeze.
We intend to control the progression through the squeeze from a 467 card which will provide the signal to an MDAT transmitter (a 166 card). This 467 card will have a time table with the dependent values being the scaled sequence number in the low beta squeeze. This time table would cover the entire squeeze. Also, we intend to use sequence number 0 to mean that the injection bump around the F0 Lambertsons is on.
The original plan was to have 4 different time tables for the 467 card controlling the squeeze. These would be for :
Each of these time tables requires its own TCLK event as a trigger. These are shown in parenthesis above. For luminosity leveling, we'll want to stop at intermediate points in the squeeze, wait for a while, and then continue on in the squeeze. To do this, we would issue STOP ($CC) and CONTINUE ($CD) events for the low beta squeeze. A total of 6 TCLK events are required to control LBMDAT, the low beta squeeze MDAT frame.
The four ramps described above are an unusual way to use the time tables. Typically time tables start from a value of 0 at t=0. Here however, the value at t=0 is not zero, but should match the end value of the previous time table. (Will a 467 make this transition smoothly ?)
For luminosity leveling, I am slightly concerned about our ability to precisely time the low beta squeeze STOP and CONTINUE events. I am concerned that we would always end up slightly off an actual table entry and that how far off may vary slightly from store to store. Also, unless we take special measures, the STOP may come out either slightly before or slightly after the desired LBMDAT value. This is probably not a problem in itself as long as it is always on the same side. We may have to take measures to ensure that we always stopped slightly beyond the desired value. (If the MDAT value seen by a card is between two of its table entries when the card gets a "copy f(t) to h(i)" command, both the adjacent slots in the h(i) table will be modified. However, if the MDAT value seen by a card is exactly the MDAT value for a particular slot, the "copy f(t) to h(i)" command will only modify that slot.)
Back to top | Back to Introduction to Proposal
This proposal will reduce to 4 the number of TCLK events, get rid of the timing problems, and is slightly simpler conceptually (though a bit more involved in the details). This is only a suggestion. There may be problems with this scheme that I haven't thought of, but I would like to circulate it and see what others think.
The 467 card controlling the LBMDAT will listen to the $D4 event (the "copy f(t) to g(i)" command) and will have 4 ramps. ( One question: "Will issuing these $D4's affect any other cards ?", that is, "Are any other cards playing time tables at this time ?". This is something I hadn't thought about. Off hand I don't think that any other cards will be playing time tables, but I will have to think about this. ) Each of these ramps has its own time table, but they all use the same g(i) table. The single g(i) table will use MDAT10, the Tevatron Energy, and is very simple, it has only two slots. The MDAT10 value for the first slot is 0 and for the last slot is 32767. The MDAT10 value will always be between these two extremes and so a "copy f(t) to g(i)" command will always modify both slots of the g(i) table. The four time tables will correspond to
These time tables follow our usual convention ; they always have a value of 0 at t=0. The first two ramps simply increment and decrement the LBMDAT by 1. The third ramp is for the low beta squeeze and the fourth ramp is for the unsqueeze.
In proton injection conditions, both slots of the g(i) table would start at 0 (Lambertson Bump ON, Injection Optics). Once we finish injecting both protons and pbars, and the Lambertsons are moved out, we would trigger the first time ramp, wait for it to finish playing ((g(i) = 0) + (f(t) = 1) = 1), then trigger a $D4 (g(i)=1, no time table playing). The acceleration ramp would not affect the output of the 467 because the g(i) table would still be between the MDAT10 values for its two slots and both those slots have identical output values. For the squeeze we would trigger a $C5, wait for it to finish playing ((g(i) = 1) + (f(t) = 18) = 19), then trigger a $D4 (g(i)=19, no time table playing). For the unsqueeze, we would trigger a $C9, wait for it to finish playing ((g(i) = 19) + (f(t) = -18) = 1), then trigger a $D4 (g(i)=1, no time table playing). As with the acceleration ramp, the deceleration ramp would not affect the output. Finally, to put the Lambertson bump back in for pbar extraction, we would trigger the second ramp, wait for it to finish playing ((g(i) = 1) + (f(t) = -1) = 0), then trigger a $D4 (g(i)=0, no time table playing). This returns us to the set-up for the proton injection conditions.
We don't need any additional ramps for luminosity leveling. We would change the squeeze ramp to not increment as far, perhaps only to go from 0 to about 11 and make a corresponding change to the unsqueeze ramp so that it only goes from 0 to -11. Then, as before, for the squeeze we would trigger a $C5, wait for it to finish playing ((g(i) = 1) + (f(t) = 11) = 12), then trigger a $D4 (g(i)=12, no time table playing). Whenever we are ready to advance the squeeze, we could trigger the first ramp, wait for it to finish playing, then trigger a $D4. Each time we did this it would advance the squeeze by one sequence. We would probably set up a repeatable sequencer aggregate that included this. When we were ready to end the store and after the protons were removed, we would trigger the second ramp, wait for it to finish playing, then trigger a $D4. We would repeat this until the ((current LBMDAT value)+(last value in the unsqueeze time table ramp)=1. Then we would do the unsqueeze ramp and continue as before.
We finish with a few additional notes/comments :
Back to Top | Introduction to Proposal | Proposal
Return to Tevatron Home Page
Updated: 08-Sept-1998
minor rev. 31-may-02 E.M.