RF feedback for non-uniform fill patterns

Moderators: cyao, michael_borland

Post Reply
Teresia
Posts: 30
Joined: 04 Oct 2018, 08:42

RF feedback for non-uniform fill patterns

Post by Teresia » 01 Sep 2021, 05:48

Hi,

I am trying to simulate transient beam loading with main cavity + harmonic cavity, but I have some trouble with finding good settings for the update_interval of the RF feedback when running with non-uniform fill patterns. According to the manual the update_interval should preferably be set as mNb where Nb is the number of buckets for equally spaced bunches. I have attempted to follow this principle as close as possible, but sometimes it seems like the feedback actually performs better by not doing that. The manual mentions that the reason for this is to read the state at a fixed timing relative to the bunches so I have tried to figure out if there might be a connection to how the phase transient looks and which update_interval is preferable (and if it could help to change the read_offset), but I haven't seen a clear pattern. It only is more difficult for cases with long gaps where the phase transient is larger.

Unfortunately our harmonic number is 934 so only dividable with 2, which makes it a bit tricky to make fill patterns with gaps that are completely symmetric in case that affects this? I have some figures below that shows what happens to the main cavity generator voltage for different update_intervals for one of the fill patterns I simulate. This fill pattern is made up of the following pattern:

- 211 bunches separated by 1 bucket with intensity 1
- 50 empty buckets
- 1 bunch with intensity 3.44597 (corresponding to 3 nC charge)
- 49 empty buckets
- 212 bunches separated by 1 bucket with intensity 1
- 50 empty buckets
- 1 bunch with intensity 3.44597 (corresponding to 3 nC charge)
- 49 empty buckets
- 211 bunches separated by 1 bucket with intensity 1
- 50 empty buckets
- 1 bunch with intensity 3.44597 (corresponding to 3 nC charge)
- 49 empty buckets

Generator amplitude and phase:
MC_generator_amplitude.png
MC_generator_phase.png
Depending on which update_interval I choose the generator voltage starts to oscillate. For this fill pattern that has three gaps over one turn I would expect that an update_interval close to 934/3 = 311.3333... would work best, but that doesn't seem to be the case. It seems like rather 467 is preferable which isn't at all symmetric with the fill pattern. I then thought that perhaps 467 is preferable because it's a fraction of the harmonic number, but when I tested it for fill patterns where I have one or two gaps over one turn I also get the oscillating voltage sometimes with that update_interval so it doesn't always work.

So I'm wondering what "read the state at a fixed timing relative to the bunches" means? Could it be of importance for the feedback how much the bunches I read the status before moves in phase so I should try to use the read_offset to read before bunches that move as little as possible? I made some attempts with that, but it also didn't seem to provide a clear pattern for when it works on not.

Best regards,

Teresia

michael_borland
Posts: 1927
Joined: 19 May 2008, 09:33
Location: Argonne National Laboratory
Contact:

Re: RF feedback for non-uniform fill patterns

Post by michael_borland » 07 Sep 2021, 02:09

Teresia,

You don't want the feedback sample time to walk relative to the bunch pattern if you have significant transient beam loading, which I'm guessing you do. I would try update intervals of 1 or 934. With an interval of 1, your should be able to compensate for average beam loading effects. With an interval of 934 and relatively low gain, you should do the same. Raising the gain may allow compensating for transients.

I wouldn't bother with the read offset parameter. I don't think that will help.

--Michael

Teresia
Posts: 30
Joined: 04 Oct 2018, 08:42

Re: RF feedback for non-uniform fill patterns

Post by Teresia » 15 Sep 2021, 10:32

Thank you, Michael. Using an update interval of 1 so far works much better for the cases where I have a large transient beam loading. I also tried 934, but that seems to update too slow for the beam to not drift out of the binned area.

For the cases where I have a small transient beam loading or even an uniform fill I however found that at an update interval of 1 doesn't work well and that I need to increase it to get stable results. The energy of the beam is starting to oscillate with an exponentially increasing amplitude. Do you have a guess for why this happens? I was wondering if might be that the feedback in that case mostly is acting on noise between the bunches which might drive a resonant behaviour if it is updating too frequently?

Best regards,

Teresia

Post Reply