Bolko Beutner: Transverse Phase Space Diagnostics for the 250 MeV Injector Room: WBGB 019 Selected questions and discussion HB: (on illustration for MENT algorithm): Why do you have non-equal bins? Is there a smart idea behind this? BB: This is just an illustration for the most general case... BO: You could check the uniqueness of your solution with a sharp pulse image. What if two different images give you the same projections? BB: Identical projections lead to the same reconstruction of course. BS: But this is true for any method! VS: How does the criterion for sharp edges relate to the resolution of the imaging device? BB: The answer is in the diagram on the slide on MENT algorithm! VS: So better resolution is equivalent to more projections? BB: Yes, it is kind of a trade-off. FLP: When do you know that you have enough projections? BB: You look at the diagnostic section, define the requirement on emittance resolution, then determine the number of projections needed. FLP: Will you systematically use MENT? BB: Yes. FLP: How do MENT and the other algorithm compare for many projections - does MENT slowly improve? BB: Yes, MENT slowly goes down. For many projections it can be that the other algorithm is better. RI: It should be pointed out that the phase space rotation between three quadrupoles and screen is restricted to the range between 30 and 150 degrees; in that case MENT is always doing better! Thus for our application (accelerator physics) MENT really is the better choice. BB: The experience from FLASH shows that 3 or 4 projections are good enough already. HB (on SLS Linac results): You are comparing 1-sigma emittance from SLS monitor with 90% emittance from MENT algorithm? BB: Intense pixels at the edge of the phase space disturb the MENT algorithm, thus a median filter and a cut are applied, the profiles are not properly smoothed (at DESY it is done with a filter, to be done in the future for our case). RI: One of the reaons we have to do image analysis is that the intensity has to be positive for the max. entropy analysis. HB: The CCD gives negative values? RI: Yes, it is good practice for an ADC to have a range going negative, to be able to deal with baseline drifts etc.! BS: The noise of the ADC will fluctuate around zero, therefore you wamt to put a positive offset! BB: MENT assumes all profiles come from the same distribution, but noise changes from image to image. We can deal with this via sophisticated smoothing of image profiles. At DESY the wavelength filter works really well. HB: Do I understand correctly that the purpose of such a reconstruction is not to get an RMS emittance figure, for this you would rather use a standard procedure...? BB: You can only do that for a Gaussian phase space ellipse... HB: The definition of RMS emittance does not assume Gaussian phase space. BB: Sometimes you really want to see the tails, or slice emittance, then you need a tomographic method. HB: But you agree that the main goal of this is not a measure of the RMS emittance, the real purpose is to see the real phase space and to shape it with whatever buttons you have? BB: I agree. Phase space tomography can give you much more than just RMS emittance!