Lyot filter tuning shows strange results when tuned to the RED_TUNING continuum

[bberkeyU]

Looking at the L0 FLAT values across our wide (21 pts?) scans, we see some odd results when the continuum was set to RED_TUNING.
One example is a set of flats taken in BOTH_TUNING, BLUE_TUNING, and RED_TUNING on 2021-07-19 in the 691 prefilter.

In the figure all of the LYOT tunings had TCAM onband. And the 6 curves are the on/off band tunings across all 3 continuum options. Data were minimally processed; the FLAT images were dark, and NUMSUM was corrected and averaged across the 4 polarization states. We then report the median value of the averaged flat within an annulus that approximates the UCOMP science field. (The exact masking is not important. This behavior appears with no mask or different choices in mask radii.). 20210719 691 RED was chosen because it was one of the last wide scans we did with a RED_TUNING, so it would hopefully represent all the optical and software fixes we made since deployment. However, these results are not unique to this day or the 691 bandpass; we see broadly similar results in most of the prefilters (It is hard to tell with 530 due to SNR).

The plot's most interesting feature is that the BLUE, GREEN, and BROWN curves all have the same shape, but the BROWN curve has an X offset of ~0.75nm. Further, the BLUE and GREEN curves represent ONBAND tunings (of the BLUE_TUNING and BOTH_TUNING continuum), while the BROWN curve represents OFFBAND tunings (or the RED_TUNING continuum).

This behavior would be expected if the stage0 was phase-shifted by Pi/2.

Below is an idealized 1074.7 LYOT filter response plotted for ONBAND and OFFBAND tunings. The lots are then repeated with +pi/8 and +pi5/8 phase shifts on stage0. When pi/8 represents our planned RED_TUNING phase shift, and pi5/8 represents the RED_TUNING shift with an extra pi/2 phase shift. The plots were computed for 1074.7+/- one free spectral range; this shows the complete 1074.7 tuning and 1/2 of the adjacent orders. In the actual instrument, we knock down the adjacent orders with the prefilter, but the lyot filter repose rings to +/- infinity.

The +pi/8 (RED_TUNING) shows that the transmission peaks are in the same location as the ideal lyot response. But instead of 100% and 50% transmission at the onband and off-band peaks, we see 85% and 15% transmission.

In the +pi*5/8 (RED_TUNING + PI/2), we again see the same transmission peak locations but not the central 1074.7 tuning, which is one of the 15% transmission peaks instead of the expected 85%.

This

[bberkeyU]

Spot checking the 20210719.213635.04.ucomp.1074.l0.fts headers, the V_LCVR1 voltage changes between the red, blue and both continuum for 1074.7 show values consistent with the correct (+/-)pi/8 phase shifts. So, at least for the wavelengths checked, the real-time code appears to be applying the tuning tables in the expected manner.