In WO4681-001-DJD001-B, page 16, section 5.8 analyzes the “the global tilt (including x and y axis) of the output image”.  To be clear, our specification asks for a limit on “the common angular error among all five output chief rays”, not the output image.  Do you mean to say chief rays?

a.  Also, we wish to clarify the requirement.  Currently it says “the common angular error among all five output chief rays shall be less than 0.25 degrees around the … z axes.”  But it is better to say, the rotation of the output image around the z-axis shall be less than 0.25 degrees.  To be clear, KPF plans to correct common errors below these limits with shims at the mounting interface, and the intention of this Req.8 is to specify a limit on how many shims we will need.  In this section, will you respond to this requirement, now that we say it more clearly?

Thank you, that is clear now.

Please also answer our other comment about this requirement:  2a.  Also, we wish to clarify the requirement.  Currently it says “the common angular error among all five output chief rays shall be less than 0.25 degrees around the … z axes.”  But it is better to say, the rotation of the output image around the z-axis shall be less than 0.25 degrees.  To be clear, KPF plans to correct common errors below these limits with shims at the mounting interface, and the intention of this Req.8 is to specify a limit on how many shims we will need.  In this section, will you respond to this requirement, now that we say it more clearly?

Please show us the plot and the calculation.  We are confused by the calculation, and it would be helpful for us to read it.

For example, in the last version of WO4681-001-DJD001-B that we have seen, the data is fit with a line, 𝛿𝜃 = 47165𝛿𝑦 ∓ 6.0345.  We would expect 𝛿𝜃 =0 at y=0, but that is not what the linear fit shows.  In fact the error is 6.0345 microrad, which is roughly the value that we are seeking!  So it seems that the fit needs to be better.  Perhaps the data is better fit with a polynomial?  Please review the calculation and consider how to fit the data better. 

Yes, we do want to add this accommodation. 

We will discuss the specific masks we wish to use, and provide you with more information soon.

(For the pinhole, we originally showed this product in our slides: https://www.thorlabs.com/thorproduct.cfm?partnumber=LMR05AP.  However, we now understand that this part is 25 mm diam, so it is too large - it would block the beam between the two relay mirrors.  We will consider an alternate.)

We have considered the specific masks that we want, and after more thought, we believe that we can make custom masks and support them without needing an interface from Winlight.  For example, we can design a foil aperture that rests on the Winlight mask at the output image (#5 below).  For our alignment, our mask does not even have to be in perfect focus at the output image, rather it could be defocused on the order of a millimeter.  The lateral tolerances are also loose.  In summary, we do not need Winlight to include accommodation for pinholes.

However, we do need the volume around the output image to be accessible, so that we may put in our own apertures.  When Winlight shows us their design for the output image mask, then we can consider how the volume is accessible for our needs.  And of course, we will need to remove the cover temporarily.

This also means that the Winlight mask at the oversized output image (see #5 below) does not need to be removeable.  Winlight can permanently install the mask during their integration.

The analysis of Req.15 (page 26) says that a pupil mask before the second relay mirror is possible, but it is not included in the design.  It seems that this is useful to control stray light.  Why is this pupil mask not included? 

We suggest that the mask attached to the slicer mirrors for the first slit image might be extended to allow for masking very near the relay pupil and also the final image – see the attached drawing.   Another possibility is that the pupil mask could be added to the final mask at the output image.

-

Bandpass

OK: Latest WL Zemax file (RD1) has wavelengths 445-870m.

Some throughput analysis curves end at 860 nm, but no large deviation or change from trend expected between 860 nm and 870 nm.

The sizes of the three input fibers and their positions are defined in the accompanying documents.

The science fiber, sky fiber, and cal fiber are all octagonal fibers.  (The accompanying documents may show the cal fiber as being circular, which is incorrect.)

Science fiber fields:  OK
Sky fiber fields:  OK

Cal fiber fields:  OK  (undersized to account for spot blur at slicer mirror)

Photometry fields:  OK (updated since CDR)

Cal mask:  Rev D report still talks about black chromium to mask sides of slice.  Non-sequential Zemax has the bevel.

Bevel is preferred (but have question about sharpness of bevel edge).

The input fibers will emit light rays at f/3.3. The output f/# of each light path shall be f/8.0 ± 0.1, including design and fabrication errors.

Rev. D of optical report (pg 12) now shows X and Y f/#s, and both are within the requirement.

The output images shall be arranged at the exit image as shown in the accompanying documents.

Cal fiber image at virtual slit was at 1.518 mm from axis, requested it to be at Zemax 34 value of 1.534 mm.  (CENX, CENY measurements of central field). 

Rev. E Zemax has this at 1.532, so OK.  (2 um offset at f/8 is 0.7 um at CCD)

The reformatter shall be aligned so that each of the five output images shall be located at their nominal positions to within 5 microns in both x and y directions, with respect to each other.

The five chief rays at the reformatter output shall be optimized to be as parallel as possible.  The goal is for all five beams to 1) match the exit pupil size, 2) be aligned to the exit pupil, and 3) be in focus at the exit image.

The footprint of all the beams in enclosed by a circle of 213 mm (pupil is 204 mm).  This gives f/7.51.

This is the best alignment WL has ever achieved, and there does not seem to be a way to make this alignment perfect (or better that this value).

The position of the output image is specified on the drawing with respect to the mounting interface, in all six degrees of freedom.

The alignment of the output image slices and chief rays may be offset slightly with respect to their nominal positions.  KPF plans to correct small errors in the output image alignment of the group of all five slices, by using shims in the reformatter mounting pocket in the KPF bench, given knowledge of the common alignment errors.

The common positional tolerance among all five exit images is shown on the drawing: for reference, the tolerance shall be less than 100 microns in x, y, and z (TBC).  The common angular error among all five exit output chief rays shall be less than 0.25 degrees around the x, y, and z axes (TBC).  KPF plans to correct common errors below these limits with shims at the mounting interface.

Knowledge of the common alignment error for the group of five output beams shall be provided to < 20 microns in x, y, and z (TBC), and <0.05 degrees around the x, y, and z axes (TBC).  Knowledge will be reported with respect to the alignment cube.  KPF plans to align the reformatter to the spectrometer using this knowledge of all six degrees of freedom.

CDR report states: "The position of the five output minislits will be measured with a sight attached to a CMM as it was foreseen on WFIRST IFU instrument. The absolute position accuracy of this setup is below 15 μm in x and y."

In order to fully understand the metrology system limitation, the resolution of the sight (and not the accuracy of the CMM) needs to be known:

For clocking, the sight can be aligned to an edge of the alignment cube. If we assume that the alignment cube is 1/2" in size and accurate to 5 arcsec, the total clocking deviation is 0.3 um (i.e., negligible).

Once the sight is aligned to the cube edge in clocking, a slope can be fit to the slit output. As stated in report 0.05 deg at the slit output represents 3.4 um of max slope value.

Both the cube and slit measurements will be presumably limited by the resolution of the sight. So the clocking knowledge should be RSS of the sight resolution.

Steve's line tilt analysis says we are ok to open up the clocking tolerance to +/- 0.25 deg (+/- sub half-pixel on spectrometer CCD). Update requirement and flow back to Winlight for concurrence.

The position of each of the five exit images, including the Sky fiber (TBC), shall be stable over one hour, moving less than 4 nm in the y-direction, 100 nm in the x-direction, and 1000 nm in the defocus direction (where the exit image slices are arranged along the lateral x-direction, and the lateral y-direction is orthogonal.)

Re-verified our requirements with latest Zemax model:  OK.

I agree with their analyses that show this requirement is met (and with significant margin).

The stability for rotation of the entire exit slit image (the ensemble of all five output beams including the Sky fiber (TBC)) about its center shall be stable over one hour, changing at the exit image by less than 6 microradians in both pitch and yaw, and 2.5 microradians in clocking at the exit pupil.

Re-verified our requirements with latest Zemax model:  OK.

Is there a typo in the units on page 23 of Rev D report?

The photometry paths will be focused by a lens element into a final image that is sized to 95%EE at 250 um in diameter.

The positional error of the photometry output image shall be less than 25 microns in x and y, and 50 microns in defocus (TBC).

The wavefront error of the reformatter design shall be less than 1.5 waves PV at the exit image for each light path, including all design and fabrication errors, and after correcting for common defocus (see #8).

Stray light shall be controlled in the reformatter using appropriate baffles and masks although no image-limiting mask shall be used at the exit image location. Methods of stray light control shall be described including baffle and mask positions and attachment to the reformatter substrate.  

PDR report described a general approach with no details - fine at the time.

The CDR report lists the 4x baffles they have added, and points to their locations: a pupil mask right after the triplet, an image mask at the intermediate image past the slicer, a pupil mask in the relay, and a mask at the reformatter output.  These are appropriate, given our spec.

The CDR report also includes a NSQ stray light model; an accompanying images show that a ghost is blocked by one mask - fine.

The Zemax model is missing the relay pupil mask, and the final image mask, so can't tell how they are mounted.  The other two masks are shown as solids with mounts, as we requested.

We did specify the final mask shouldn't limit the image, we should confirm this is the case.  The mask dimensions are not given anywhere (and the mask is missing from the model).

The model is described as "simple" and "representative", so that the final design may differ.  This should be fine.  The report also claims the model doesn't take into account diffusion of the surface (probably they mean glass surfaces?)  That's worth looking at more, although that is beyond the spec we gave them which only requested baffles - maybe we do a calc based on their expected surface roughness, or we ask them to include in model.

The PDR report raised two poss stray light concerns: the sides of the slices, and the lateral sides of the cal fiber.  But the CDR report doesn't include those concerns, and they aren't otherwise addressed.  Doesn't seem to be in the Zemax model.  The cal fiber is noted elsewhere in the report as simply being masked by chromium outboard of the slicer submirror.  We should ask Winlight about how light is controlled around the areas of these slices.

CDR report gives two options for coating invar foil: black chromium or PU1 paint.  A graph shows the total reflectance for black chromium, but otherwise no info nor comparison is given.  Either one should be acceptable though.

The throughput of the reformatter shall be > 85%.
Calculation of throughput will include the effects of the exit pupil of each sliced path not overlapping exactly with the nominal exit pupil.

An estimated throughput budget that itemizes the contributions to loss shall be provided.

CDR report shows throughput similar to that in PDR report, and still well above spec.

The report shows triplet AR coatings based on real data, so that's reasonable.

The report shows a mirror coating curve, but isn't clear whether it's a theoretical curve or it's a curve based on real achieved coatings.  The coating is used 4x times, so any difference could add up.  This is a question for WL.  Later: WL clarified that this curve is based on a real achieved coating.

Mirror coatings shall be high stability, durable coatings (for example densified by ion-assisted deposition (IAD) or equivalent), to ensure performance over the range of vacuum and humidity specifications.

The three input fibers will be mounted in a fiber block, provided by KPF, which will be mounted to the reformatter on a base block.  The mounting location is defined on the drawing KPF-RFM-ICD-001. The base block shall be contact bonded to the substrate. The mounting interface on the base block should allow for the removal and replacement of the fiber block for installation or maintenance, such as by using spring force to retain the fiber block.  For reference, KPF plans that the input fiber faces will be coplanar to the fiber block exit face to within 5 microns (TBC) and laterally spaced to within 3 micron (TBC) tolerance. The fiber input beam chief rays will be co-aligned to within 6 arcmin (TBC). Other relative alignments of the fiber beam chief rays may be requested.

The location and orientation of the output image is defined on the drawing KPF-RFM-ICD-001, with respect to the mechanical mounting interface, in all six degrees of freedom.

We anticipate that Winlight might prefer to verify this specification by measuring the output image and the mounting interface in relation to the alignment cube.

The photometry slices’ output will exit the reformatter at a location defined on the drawing KPF-RFM-ICD-001 and the Zemax file. The photometry light will be collected by a 300-micron fiber mounted on a fiber block provided by KPF, which will be mounted to the reformatter. The mounting interface to the fiber block is provided by the reformatter and is notionally a base block that is contact bonded to the substrate. The mounting interface should allow for the removal and replacement of the fiber block for installation or maintenance.  The photometry base block at this interface may be the same as the base block used for the input fibers.

The reformatter shall fit within a volume defined on the drawing KPF- RFM-ICD-001. Note that this constraint is to avoid interference with the light rays and intermediate fold mirror within the separate KPF spectrometer.

The reformatter shall mount on the KPF optical bench using a spring and shim mounting scheme as defined on the drawing KPF-RFM-ICD-001.

The reformatter shall meet all of its performance specifications in the operational environment.

At high vacuum: < 1E-3 mbar.
Thermal environment: 20 ± 0.1 °C Vibration environment:

0 to 100 Hz: 5e-7 G2/Hz

The reformatter shall be capable of being aligned and functionally verified in this environment.

In air.
Humidity: < 70% RH
At altitude of 0 meters (1000 mbar) or 4200 meters (600 mbar).

Thermal environment: 20 ± 5 °C
Vibration environment:

0 to 100 Hz: 5e-7 G2/Hz
100 to 2000 Hz: 5e-8 G2/Hz

The reformatter shall survive but is not needed to perform in this environment.

Thermal Environment: -10 to 30 °C
Vibration environment: ± 0.7 g

The reformatter shall survive while its shipping container is subject to this environment.

Thermal Environment: -33 to 71 °C
Vibration environment profile:

1 Hz: 1e-5 G2/Hz
4 to 100 Hz:  0.01 G2/Hz
100 to 300 Hz: 0.01 G2/Hz
300 to 4 kHz: 1e-5 G2/Hz

Bandpass

OK: Latest WL Zemax file (RD1) has wavelengths 445-870m.

Some throughput analysis curves end at 860 nm, but no large deviation or change from trend expected between 860 nm and 870 nm.

The sizes of the three input fibers and their positions are defined in the accompanying documents.

The science fiber, sky fiber, and cal fiber are all octagonal fibers.  (The accompanying documents may show the cal fiber as being circular, which is incorrect.)

Science fiber fields:  OK
Sky fiber fields:  OK

NOTE:  cal fiber fields have fiber at 64 um wide.  OK (and necessary) for SEQ Zemax but we should mention to WL we've chosen an oversized 120um fiber here.

Photometry fields:  Fields slightly wrong: +/- 0.0431 → +/- 0.0631 and +/- 0.1165 → +/- 0.10425

Change since V11:  WL agreed to KPF's request to make cal slicer mirror a square pyramid with a 64 um square top (measured at fiber).  This is reflected in draft optical report.

Cal mask:  CDR report has moved away from pyramid and now using black chromium to mask sides of slice.

Also, CDR text has it at 512 um x 512 um at slider mirrors (OK).  However RD1 Zemax has it at 570 um x 570 um.

Sky mask:  Page 11 in optical report says sky fiber is 250 um.  (Note Zemax RD1 has it correct at 225um).  At the slicer mirror, this would then be 512 um x 1800 um (not 512 um x 2000 um as in optical report).  The smaller size is preferred, as it will help constrain the position of the sky slice (and help with compliance of Req 6 below).  Note RD1 Zemax has this mirror sized at 512 um x 3000 um.

The input fibers will emit light rays at f/3.3. The output f/# of each light path shall be f/8.0 ± 0.1, including design and fabrication errors.

Table pg 12 in document, each fiber/field cone has f/8.0, but overall aiming of all cones is f/8+/-0.16 worst case (f/8.0+/-0.12 errors in quadrature)

The output images shall be arranged at the exit image as shown in the accompanying documents.

Comparing Zemax RD1 slice positions at f/8 virtual slit to Ed's AD2 doc, slide 6:

Science slices (x3) all compliant within 2 um (based on CENX, CENY measurements of central field)

Cal fiber measures at 1.518 mm from axis.  Note in Zemax 34 it was 1.534 mm.    (CENX, CENY measurements of central field).  Difference is ~ 1/2 pixel at CCD. 

I think we want it back at 1.536, as Zemax 34 was used for orderlet spacing analysis.

The reformatter shall be aligned so that each of the five output images shall be located at their nominal positions to within 5 microns in both x and y directions, with respect to each other.

I think reducing size of sky slicer mirror would at least constrain the top and bottom edges of the sky slice.  On one side the fiber edge might move inside the mirror edges (would reduce flux), but the fiber image couldn't move outward (and affect orderlet spacing). Constrains stack height since outboard slices then both set by Zerodur.

They created a substantial tolerance table that shows this requirement is met.  I have an open question to them on why decenter of the mirrors wasn't included.  Analysis they give is the worst case, so likely OK.

Requirement Check:  5 um here equates to ~ 2 um at CCD (well under a pixel).  Given this is an alignment req that's fine.

The five chief rays at the reformatter output shall be optimized to be as parallel as possible.  The goal is for all five beams to 1) match the exit pupil size, 2) be aligned to the exit pupil, and 3) be in focus at the exit image.

The position of the output image is specified on the drawing with respect to the mounting interface, in all six degrees of freedom.

The alignment of the output image slices and chief rays may be offset slightly with respect to their nominal positions.  KPF plans to correct small errors in the output image alignment of the group of all five slices, by using shims in the reformatter mounting pocket in the KPF bench, given knowledge of the common alignment errors.

The common positional tolerance among all five exit images is shown on the drawing: for reference, the tolerance shall be less than 100 microns in x, y, and z (TBC).  The common angular error among all five exit output chief rays shall be less than 0.25 degrees around the x, y, and z axes (TBC).  KPF plans to correct common errors below these limits with shims at the mounting interface.

Knowledge of the common alignment error for the group of five output beams shall be provided to < 20 microns in x, y, and z (TBC), and <0.05 degrees around the x, y, and z axes (TBC).  Knowledge will be reported with respect to the alignment cube.  KPF plans to align the reformatter to the spectrometer using this knowledge of all six degrees of freedom.

CDR report states: "The position of the five output minislits will be measured with a sight attached to a CMM as it was foreseen on WFIRST IFU instrument. The absolute position accuracy of this setup is below 15 μm in x and y."

In order to fully understand the metrology system limitation, the resolution of the sight (and not the accuracy of the CMM) needs to be known:

For clocking, the sight can be aligned to an edge of the alignment cube. If we assume that the alignment cube is 1/2" in size and accurate to 5 arcsec, the total clocking deviation is 0.3 um (i.e., negligible).

Once the sight is aligned to the cube edge in clocking, a slope can be fit to the slit output. As stated in report 0.05 deg at the slit output represents 3.4 um of max slope value.

Both the cube and slit measurements will be presumably limited by the resolution of the sight. So the clocking knowledge should be RSS of the sight resolution.

The position of each of the five exit images, including the Sky fiber (TBC), shall be stable over one hour, moving less than 4 nm in the y-direction, 100 nm in the x-direction, and 1000 nm in the defocus direction (where the exit image slices are arranged along the lateral x-direction, and the lateral y-direction is orthogonal.)

Re-verified our requirements with latest Zemax model:  OK.

I agree with their analyses that show this requirement is met (and with significant margin).

The stability for rotation of the entire exit slit image (the ensemble of all five output beams including the Sky fiber (TBC)) about its center shall be stable over one hour, changing at the exit image by less than 6 microradians in both pitch and yaw, and 2.5 microradians in clocking at the exit pupil.

The photometry paths will be focused by a lens element into a final image that is sized to 95%EE at 250 um in diameter.

The positional error of the photometry output image shall be less than 25 microns in x and y, and 50 microns in defocus (TBC).

The wavefront error of the reformatter design shall be less than 1.5 waves PV at the exit image for each light path, including all design and fabrication errors, and after correcting for common defocus (see #8).

Stray light shall be controlled in the reformatter using appropriate baffles and masks although no image-limiting mask shall be used at the exit image location. Methods of stray light control shall be described including baffle and mask positions and attachment to the reformatter substrate.  

PDR report described a general approach with no details - fine at the time.

The CDR report lists the 4x baffles they have added, and points to their locations: a pupil mask right after the triplet, an image mask at the intermediate image past the slicer, a pupil mask in the relay, and a mask at the reformatter output.  These are appropriate, given our spec.

The CDR report also includes a NSQ stray light model; an accompanying images show that a ghost is blocked by one mask - fine.

The Zemax model is missing the relay pupil mask, and the final image mask, so can't tell how they are mounted.  The other two masks are shown as solids with mounts, as we requested.

We did specify the final mask shouldn't limit the image, we should confirm this is the case.  The mask dimensions are not given anywhere (and the mask is missing from the model).

The model is described as "simple" and "representative", so that the final design may differ.  This should be fine.  The report also claims the model doesn't take into account diffusion of the surface (probably they mean glass surfaces?)  That's worth looking at more, although that is beyond the spec we gave them which only requested baffles - maybe we do a calc based on their expected surface roughness, or we ask them to include in model.

The PDR report raised two poss stray light concerns: the sides of the slices, and the lateral sides of the cal fiber.  But the CDR report doesn't include those concerns, and they aren't otherwise addressed.  Doesn't seem to be in the Zemax model.  The cal fiber is noted elsewhere in the report as simply being masked by chromium outboard of the slicer submirror.  We should ask Winlight about how light is controlled around the areas of these slices.

CDR report gives two options for coating invar foil: black chromium or PU1 paint.  A graph shows the total reflectance for black chromium, but otherwise no info nor comparison is given.  Either one should be acceptable though.

The throughput of the reformatter shall be > 85%.
Calculation of throughput will include the effects of the exit pupil of each sliced path not overlapping exactly with the nominal exit pupil.

An estimated throughput budget that itemizes the contributions to loss shall be provided.

CDR report shows throughput similar to that in PDR report, and still well above spec.

The report shows triplet AR coatings based on real data, so that's reasonable.

The report shows a mirror coating curve, but isn't clear whether it's a theoretical curve or it's a curve based on real achieved coatings.  The coating is used 4x times, so any difference could add up.  This is a question for WL.

Mirror coatings shall be high stability, durable coatings (for example densified by ion-assisted deposition (IAD) or equivalent), to ensure performance over the range of vacuum and humidity specifications.

The three input fibers will be mounted in a fiber block, provided by KPF, which will be mounted to the reformatter on a base block.  The mounting location is defined on the drawing KPF-RFM-ICD-001. The base block shall be contact bonded to the substrate. The mounting interface on the base block should allow for the removal and replacement of the fiber block for installation or maintenance, such as by using spring force to retain the fiber block.  For reference, KPF plans that the input fiber faces will be coplanar to the fiber block exit face to within 5 microns (TBC) and laterally spaced to within 3 micron (TBC) tolerance. The fiber input beam chief rays will be co-aligned to within 6 arcmin (TBC). Other relative alignments of the fiber beam chief rays may be requested.

The location and orientation of the output image is defined on the drawing KPF-RFM-ICD-001, with respect to the mechanical mounting interface, in all six degrees of freedom.

We anticipate that Winlight might prefer to verify this specification by measuring the output image and the mounting interface in relation to the alignment cube.

The photometry slices’ output will exit the reformatter at a location defined on the drawing KPF-RFM-ICD-001 and the Zemax file. The photometry light will be collected by a 300-micron fiber mounted on a fiber block provided by KPF, which will be mounted to the reformatter. The mounting interface to the fiber block is provided by the reformatter and is notionally a base block that is contact bonded to the substrate. The mounting interface should allow for the removal and replacement of the fiber block for installation or maintenance.  The photometry base block at this interface may be the same as the base block used for the input fibers.

The reformatter shall fit within a volume defined on the drawing KPF- RFM-ICD-001. Note that this constraint is to avoid interference with the light rays and intermediate fold mirror within the separate KPF spectrometer.

The reformatter shall mount on the KPF optical bench using a spring and shim mounting scheme as defined on the drawing KPF-RFM-ICD-001.

WL had invar buttons in design at PDR.

CDR Mech Report still talks about invar buttons.

The reformatter shall meet all of its performance specifications in the operational environment.

At high vacuum: < 1E-3 mbar.
Thermal environment: 20 ± 0.1 °C Vibration environment:

0 to 100 Hz: 5e-7 G2/Hz

The reformatter shall be capable of being aligned and functionally verified in this environment.

In air.
Humidity: < 70% RH
At altitude of 0 meters (1000 mbar) or 4200 meters (600 mbar).

Thermal environment: 20 ± 5 °C
Vibration environment:

0 to 100 Hz: 5e-7 G2/Hz
100 to 2000 Hz: 5e-8 G2/Hz

The reformatter shall survive but is not needed to perform in this environment.

Thermal Environment: -10 to 30 °C
Vibration environment: ± 0.7 g

The reformatter shall survive while its shipping container is subject to this environment.

Thermal Environment: -33 to 71 °C
Vibration environment profile:

1 Hz: 1e-5 G2/Hz
4 to 100 Hz:  0.01 G2/Hz
100 to 300 Hz: 0.01 G2/Hz
300 to 4 kHz: 1e-5 G2/Hz