fmadec I looked at the data that had the same value for attenuator (at 220) (expID 15764 for no stop and 15812 for f/ratio of 2.8). I used the 11748 as example of the ``old'' data with the larger f/ratio, which had attenuator value of 165. I converted that flux to what I would expect at 220 by using the formula determined in PIPE2D-323.

Then I measured flux at a single wavelength (positioned at e.g., x=2099, y=1657 in the new data), in fibers that are the same as in the ``old'' data. I find that

1. Flux in exposures at f/2.8 is 1.1x higher than flux in the data taken with larger f-ratio. I think that is not unreasonable. There might be also some mistakes due my attenuator converting and similar. I would be tempted to not change exposure times much, or decrease them by 10% in future experiment. (e.g., instead of standard 15 seconds exposures, due 14 seconds).

1. Flux in exposures without the stop is 3.4x higher than flux in the data taken with larger f-ratio and 2.1x times higher than in the data with f/2.8?!? I am a bit confused as how is this possible? I had a short conversation with Jim how also did not know, but wanted to see the defocused images in order to better understand this (in addition to other reasons why we would like defocused images). Is this large difference consistent with what you are finding fmadec?

I am adding a single figure showing the same spot in different exposures, showing this large difference in flux.

Answer by Graham about this issue

Is this a uniform effect seen across all fibres, or is there a significant fibre to fibre variation?
Do you see a similar scale of effect over the full wavelength range, or is there a variation with wavelength? The collimator alignments are established with a HeNe laser and the Hastings triplet lenses used will exhibit a focal length shift at shorter and longer wavelengths, but I don't see how this could be contributing to the effect you are seeing.

Some further thoughts: Without the stop, the limiting aperture is the NA of the F = 40.3 mm triplet lens itself (NA = 0.29). Using that, I calculate the difference in pupil area between the collimators with and without an F/2,8 stop is x 2.64, so simplistically (assuming reasonably uniform flux across the pupil) there is 2.64 x more light arriving at the connector port coupling to DCB. Can this be measured, perhaps with the small stand-alone collimator pair I additionally shipped? I understand the short fibres feeding from the integrating sphere are overfilled at the sphere output port (in area and angle) and cladding modes may be present also, so that may be complicating things. I wonder though if the significant difference in flux you are seeing is a consequence of inherent FRD / geometric FRD downstream of the collimators. Without any stop, light from angles faster than F/2.8 will enter the fibre. Light in the range F/2.8 - F/2.5 will increase flux by a maximum x 1.26 (with some loss due to FRD). But FRD will steer a portion of propagating rays beyond F/2.5 back within the spectrograph clear aperture, also contributing to an increase in total flux. The fibre inherent NA is 0.22, but that is not a hard cutoff and skew rays will propagate at faster angles.

I hope we can get more data to investigate this effect before the instruments gets shipped to Hawaii, but it is looking increasingly unlikely.