NIKA2+ hardware adaptation to focuses differences

Contents

NIKA2+ hardware adaptation to focuses differences

Historical context

In the first runs of NIKA2 started in October 2015 we noticed that the 2mm beam was severely degraded (see offline processing wiki pages of runs 1, 2, 3, 4) and concluded that the dichroic #1 was severely distorted (suspicion firstly mentioned early 2016 in run 3 daily report, and regular NIKA2 meetings end January, begin February). We decided to change it (in march 2016), which we did (in September 2016, as described in the run 5 daily report). After years using the dichroic #2, which was flat but had a problem of misplaced transition in the 1 mm band (explanation discovered end 2017), we decided to go back to a hot-press dichroic, like the dichroic #1. Unfortunately, similarly to dichroic #1, this dichroic #3, installed in August 2018, turned out to have also a distortion problem, described in run 20 daily reports, though it was mounted with PEEK rings supposed to avoid this. So we had to put back dichroic #2 in September 2018, and had to live with it the time to find a better solution. After a few years of developing and doing lab tests on dichroic #3 and #1 to find a way to mount them without distortion (regularly mentioned in the regular NIKA2 meetings, in particular the following key ones in 2019, 2021, 2022, 2023), we finally found a solution to mount the dichroic #1 in an acceptable manner and started to define a plan to upgrade NIKA2, not only with this dichroic, but also with other elements, in particular new 1mm arrays. This plan was discussed between NIKA2 instrumentalists and concerned IRAM personnel in special meetings, in June 2023, November 2024, and finally January 2025. The reason for the 2 years delay between the solution and the implementation is due to the fact that we had to wait for the completion of the 30m telescope refurbishment (new motors and servo control, and new painting of the primary mirror). The development steps of the plan was presented in the regular NIKA2 meetings, the most important being in June 2024 when Samuel presented a summary table of the upgrade plan, with estimations of expected performances improvements. In this plan we decided to go back to dichroic #1 for the NIKA2+ upgrade, mounted with the best method found in the lab tests that has no impact on the overall transmission, but presented a risk possible residual distortion, that is possible to mitigate as shown with Zemax simulations presented in a January 2025 regular NIKA2 meeting.
The NIKA2 upgrade occurred in March 2025, as reported in the dedicated daily reports, and mentioned in a March 2025 regular NIKA2 meeting.

==> So, as described in the various documents referenced previously, this time we mounted the dichroic #1 differently than in the early runs 10 years before and than the dichroic #3 in 2017: we mounted its membrane outer part in sandwich between 2 aluminum rings maintained in the copper mount with a copper flange screwed on the mount with a torque of 0.5 N*m. From the the last labs tests done in April 2023 we were expecting a residual distortion in NIKA2 at cold temperature between 0 and 1 mm amplitude. The first order effect of the distortion is a displacement of the image surface in the cryostat, that can be compensated thanks to a displacement of the KID array. Beyond 1 mm distortion some second order aberration effect start to be more and more significant and degrade the beam even at the best focus.
A collection of copper rings between 1 and 10 mm thickness was fabricated, to be used as possible shims to be placed between the KID array holder and its mount on the cryostat, in order to shift the array position along the optical axis. But for the March 2025 intervention we decided to not change the position of the arrays since we didn't knew exactly what would be residual distortion in the NIKA2 cryostat environment.

During one year of using NIKA2 on sky, between run -ref- and run -ref-, we monitored the focus differences between the arrays. All these values are gathered in a document showing the evolution of the focus values and differences in a few plots -ref-. We noticed a constant focus offset between Array 1 and Array 3 (possibly due to a small residual distortion of the polarizer or a small difference in optical path length), and changing focus offset between Array 2 and the other arrays, passing from a small value with consequence on the beam to an unacceptable value where the 2 mm beam quality is very severely affected when the focus is set for the 1 mm Arrays. Somme Zemax simulations allows visualizing this effect with plots quantifying this effect -ref-. After each full warmup of the cryostat, the focus difference between Ar2 and the others at the following cool down was a bit more degraded, up to reaching an apparent asymptote at the last warmup in January 2026.

The residual distortion of the dichroic, its evolution with warmups and the possible measures to mitigate is were discussed at several NIKA2 meetings, in particular the following regular meetings -ref-, and the following dedicated one -ref-. In these meetings were discussed also another non optimal element in NIKA2+, the new Array 3, which is behaving better than the old one but performing significantly less good as the new Array 1, so that its contribution to the overall 1mm channel is weak. The discussions in the dedicated meeting leaded to create the following table of possible actions -ref-. From all these discussions we ended up with the following plan for this intervention in March 2026:

Change the current Array 3 installed since March 2025, with a another one, which is a twin of the current Array 1, fabricated at the same time (2018) with the same parameters, but for which 1 of the 8 multiplexed lines of KID has stopped working during a lab test a few years ago. Connect this faulty line to box-S, which has always shown noisier signal than the others, probably due to a LNA not performing as well as the others.
Insert 10 mm thick shim in the Array 2 mount to compensate for the 1.2 mm focus difference with Array 3 caused by the distorted dichroic, assuming the distortion is downward, like in all the lab experiments.
Insert 2 or 3 mm thick shim in the Array 1 mount to compensate for the 0.25 mm focus difference with Array 3.

Actual plan for the intervention starting on March 10th, 2026

One week before the intervention, comparing focus values over the last year with typical EMIR focus values, Samuel came to the conclusion that the dichroic distortion must be upward and not downward as we thought. He came to this conclusion because in the optical design of NIKA2 the optimal focus for NIKA2 is obtained by increasing the distance between M1 and M2 by 1 mm compared to the optimal focus for EMIR. The value to enter in Zemax for NIKA2 to displace M2 in that direction is -1 mm, while in PaKo this is +1 mm. Several colleagues from IRAM Granada looked for information attesting with 100% certainty that a positive value entered in PaKo correspond to an increase of the distance M1-M2. They found some clues, like this drawing found by Miguel -ref-, but no document clearing totally the uncertainty.
A distortion upward of the dichroic shortens the optical path on the 2mm band, so that we should not introduce 10 mm thick shims to compensate to focus shift, but reduce the mount length by 10 mm, which is not possible. An alternative is to make the focus on Array 2 with M2, and insert shims on Ar1 and Ar3 to compensate for the corresponding focus shift. But as Zemax simulations show -ref-, this has consequences on the 1 mm beam quality. Hopefully, this affect the beam only by a few %, so that the loss is small enough to be acceptable. As a consequence the plan changed to the following actions:

Measure with a laser the displacement of M2 when a command is entered to change the focus Z.
Change current Ar3 with the twin of the current Ar1, connecting its missing line to box-S.
Insert 10 mm thick shim in the Ar3 mount.
Insert 2 or 3 mm thick shim in the Ar1 mount.

Preliminary work before March 10th, 2026, and staff for the intervention

NIKA2 warmed-up.
Array 3 spare prepared.
Spare leak detector shipped from IRAM Grenoble to IRAM Granada.
Shims at the telescope.
Tool box for opening NIKA2 cryostat.
Torque wrench at the telescope.

Intervention team: Alessandro Monfardini, Martino Calvo, Emile Prele (PhD student at Neel), David John, Samuel Leclercq

AM, MC, EP, DJ at the telescope on Monday, March 9th to prepare everything for the intervention on Tuesday morning.

March 10th, Tuesday

Measuring displacement of M2 with a laser, when changing focus Z value with OHB control system, and with PaKo:

Laser in M1 vertex distance to M2 before any focus command: 13.061 m. - few mm in OHB control soft => 13.058 m ; 0 mm => 13.060 m ; +10 mm => 10.071 m ; -8 mm in PaKo => 13.052 m.
==> As expected, for both the OHB control software and in PaKo a positive value increases the M1-M2 distance. This means the dichroic distortion is upward, as feared.

Opening NIKA2:

First SURPRISE: the dichroic is much more distorted than we expected, the distortion is visible with the naked eye !! As expected it is upward. We measure it:

This is logic that it's bigger at warm temperature where the membrane is at its maximal thermal expansion, than the 1.17 mm at cold temperature deduced from focus differences. But this is so spectacular that everyone here wonder whether we should really stick to the plan or rather reconsider it and try to do something about this distortion, namely dismounting the dichroic and reset its mechanical constraints to let it retrieve a flat shape. We decide to push a bit the Zemax simulations to weight various scenarios, and inform the persons involved in emails discussions since the discovery that the distortion should be upward and not downward the week preceding the intervention (NP, MS, SB, CK, AS, AB, MCF, IR, JT, JMP).

Other important observation looking at the Ar2 mount: it's possible to reduce its thickness by 2 mm, if we dismount it and mill it in the telescope machine shop. In terms of beam quality, it's preferable to do that rather than introducing thicker shims on the 1 mm arrays. We decide to go for it, Victor takes care of the milling.

After lot of discussion here we come with 2 options:

1) we insert shims 8 mm on Ar1, 10 mm on Ar3. That's it.
2) we flatten the dichroic, reduce Ar2 holder thickness by 2 mm (max we can do), introduce 3 mm shim on Ar3, don't touch Ar1.

Option 2 is more uncertain result in terms of final Ar2 focus difference with the others and on its stability, but the potential gain in beam quality is bigger, which let more margins if we are a bit off. Option 1 has better certainty but not 100% (look at the fluctuations of recent Ar2 focuses), if it is successful the beam quality is lower than a successful option 2, and it gives no margin for being a bit off.

Several replying our email prefer option 1, we still hesitate at the telescope but have a small preference for option 2. Karl is put in the loop, calls us, and after explanations express also a preference for option 2. 3 arguments in additions to the one stated previously make the balance weight in favor of option 2, that we eventually decide to apply: [1] if we are a bit off, this is not so dramatic as only very few proposals require having both the 2mm beams and 1 mm beams optimal, for the other the focus can always be done on the most important band (we know this represents additional work for data processing, but the advantage is that the beam quality on the most important band would be better than with option 1)*, [2] if the dichroic continues to distorts more, we are not sure its physical integrity is insured, [3] we can learn much more for future optimization if needed with option 2) than with option 1).

* Update 1 week later after, done during the March 18th regular NIKA2 meeting: Stefano warns that such a solution is not possible, as it requires having 2 calibrations, with all it implies in terms of housekeeping and overheads, plus need to keep track of the focus configuration chosen for each scan. So if we are in a situation that is not optimal for all arrays, will have to decide on the best compromise focus for all arrays.

Dismounting dichroic to rest its distortion to 0. Second SURPRISE: after releasing the rings, the dichroic membrane does not recover a flat shape as it did in the lab !!

==> Distortion height = distance from a flat reference to the dichroic edge perpendicular to its surface - distance from a flat reference to the dichroic center perpendicular to its surface = 16-13 = 3 mm! The membrane is not flat, but at least the distortion reduced by a factor 2.
After discussions on the various options offered to us we decide to fix again the membrane in its mount, applying a torque of 0.15 N*m torque on the screws of the copper flange instead of 0.5 N*m previously (the hope is that this gives more room to the dichroic to expand during warmup before being blocked again by the pressure of the rings, and thus reduce or even cancel the hysteresis effect that we suspect at each thermal cycling of the cryostat).

We decide to introduce a 3mm thick shims on Ar3, and not touch Ar1. So the total displacements between Ar3 and Ar2 is +3-(-2) = 5 mm. This is half the thickness of the thickest shim from the original plan. So if we managed to reduce the dichroic distortion by a factor 2, we should be right on spot in terms of focus equality between all arrays.

We mount the new array 3 with its 3mm shim, re-mount array 2 with its 2 mm thinned mount, and measure again the dichroic distortion after mounting it back in NIKA2.

End of Day 1.

March 11th, Wednesday

Closing NIKA2:

Pumping, leak check, putting IKA2 back on its pinned position:

Samuel refined Zemax simulations. The distortion of optical elements, like the dichroic has the first order consequence that leaded to the present intervention, of shifting the image surface along the optical axis. But it has several second order consequences which are harmless as long as the distortion is equivalent to a M2 focus shift lower than 1 mm, but that can be harmful to the data quality when the focus shift is higher than 1 mm. The most obvious, already presented in a document shown in the last regular NIKA2 meetings -ref- is an aberration of the beam such that even at the best focus the beam size and amplitude are not as sharp as it can be at best focus without the distortion. Another consequence not shown till now is that the distortion changes the size of the field of view (FOV) on the cold pupil and on the image surface. When the distortion is upward the FOV is smaller on the cold pupil and on the image, which means a larger illumination of the primary mirror (M1) at the risk of picking more spill over, and a reduction of the FOV on the KID array without change of the angular resolution (so the beams are bigger on the array). When the distortion is downward this is the opposite: smaller illumination of M1, less spill over, larger FOV on the image with smaller beams. Few screen shots to illustrate:

Upward distortion of the dichroic equivalent to 1.1 mm M2 focus shift, compensated by Ar2 displacement by -10 mm. Ray tracing layout on image surface and on cold pupil (contour of M1 for each field):

Downward distortion of the dichroic equivalent to 1.2 mm M2 focus shift, compensated by Ar2 displacement by +10 mm. Ray tracing layout on image surface and on cold pupil (contour of M1 for each field):

This shows that a downward distortion is more favorable. To keep in mind in case one day we want to consider trying to flip the dichroic now that it seems plastically distorted along one of its faces.

Now, if we are right on spot with the actions done during this intervention, that is to say distortion reduced by a factor 2, Ar2 position -2mm, Ar1 position unchanged, Ar3 position +3mm, all beams should be fine for the same M2 focus position; ray tracing layout on KID array, on cold pupil and spot diagram (with lowest Strehl ratio across FOV written on it) for Ar2, Ar1 and Ar3 respectively with M2 focus Z = 0.7 mm (e.i. -0.3 mm compared to the average best focus for Ar3 until now):

End of Day 2. No leak was detected, we started cooling down NIKA2.

March 12th, Thursday

Cooling down is going well. Pressures seem a bit high after 12h compared to the usual after a full warmup, but this is probably due to out-gazing of the various internal components which were exposed to air during the opening of the cryostat. ALESSANDRO: nothing strange I think. I am not sure that comparing "vacuum pressures" while cooling down the cryostat is relevant.

Alessandro, Martino, Emile, Samuel leave the 30m. We'll know in a few days when NIKA2 is cold if all KID lines are OK (update March 16th: yes, all KID lines seem OK, box-S is out as expected, Box-P to watched). Then a new sweep will be necessary, at least on the new Array 3. Re-commissioning planned for the March 24-31 week -ref-.

New NIKA2+ configuration after intervention

Configuration after the intervention is as discussed in previous teleconfs and approved by IRAM. The actions were: a) replace array 3; b) move the focus of arrays 2 and 3; c) "reset" the dichroic shape (as best as possible). Here is the arrays configuration we've left inside the cryostat:

Array 1 (pol H, refl): NIKA2_1MM_MS_v3_10 - 0mm wrt nominal focus position - LO=1.755GHz - lines 1-8 on boxes 5-12 (E,F,G,H,I,J,K,L)
Array 2: NIKA2_2MM_MS_v0_02 - -2mm wrt nominal position - LO=0.94GHz - lines 1-4 on electronics boxes 1-4 (A,B,C,D)
Array 3 (pol V, transm): NIKA2_1MM_MS_v3_09 - +3mm wrt nominal position - LO=1.890GHz - lines 1-8 on boxes 13-20 (M,N,O,P,R,S,T,U; knowing that feed 7, which is NOT WORKING, is plugged on box S)

The status of the cryostat on March 13 is "cooling down", expected to be cold early next week (16/03/26). The priority will be to see if the resonances are OK on all the 19/20 lines where they are expected to be. In array 3 we expect if everything goes well, and considering that one line is missing, around 60-70% (700-800/1140) of the resonances.

March 19th, Thursday

Cryostat is cold (T_MC = 0.165 K; T_Still = 1.1 K; T_4K = 8.37 K).
Lines transmission check and a new sweep (only array 3) were performed by Alessandro, Martino and Dave. Line S (ar3) is dead. Current bad KIDs counts (ar1, ar2, ar3; window closed/Al tape sealed): 18/1000, 7/574, 11/749. Means respectively 88%, 93%, 66% KID wrt. the theoretical number on the masks.
Screenshots of DAQ monitor and noise traces sampled afterwards are given below:

NIKA2+FocDiffAdaptWork