Optomechanics – Using Ivory for Early Structural Concepts


If spherical surfaces didn’t make pretty good images we’d live in a whole different world.  It seems that optics is an art that was just meant to work.  The mechanics?  Well, that’s maybe a whole different story.

I recently helped to demonstrate that a proposed optical system could be made to work.  It’s stability requirements were 2 1/2 times tighter than the earlier system on which the proposal had been based, and the earlier one had been a challenge in its time.  That extrapolation was a risk that the contractor had to eliminate very early. 

So, the CAD engineer and I shared a cubicle.  He collected information on all the stuff that had to go into the system.  I created a structural finite element model to analyze the image stability:  I started with the CodeV prescription, which I read into AEH/Ivory and then imported the Ivory file into Patran; I also imported into Patran the step-files (and ray bundle) for all the optical elements; I attached the Ivory file to the elements (Any time I moved an element I could then read the resulting motion of the image on the detector); finally, I imported into Patran the proposed flat honeycomb plate to which the optical elements were to be mounted.  The boring part was over.

And the real fun began.  I used Patran as a design tool:  I put cells around each of the optical elements; I tied the cells down to the flat plate; I ran a 6 DOF rigid body check and a 3 axis static gravity check in Nastran.  Everything behaved well computationally.  But in random vibration it was out of bed by ~3X.  There was work to be done.

With the AEH/Ivory data imported to Excel I could identify which elements were the big drivers of the image motions.  So, I started beefing up the bracing on those elements.  The CAD engineer was checking my work while he started his own modeling effort.  He guided me in positioning the optical mountings and I guided him in locating the other services (electronic, thermal, servo, mechanical) that had to work in proximity to the optics on the inner gimbal.  The bracing and the services all had to fit.  Ultimately, we (he and I) were able to reduce the image motions by over 3X and show safe margin on the stability requirements with everything on the gimbal.

All of this was done in the opening days of the project.  In fact, if you cannot make the optical system work when the design spaces are malleable, you will be unlikely to make it work later.  It only gets harder (and I’ve been there too). This early structural concept was itself malleable and would change over time as all of the disciplines agreed to the design.  It might be months before all the CAD interfaces would be settled.  Meanwhile, the project had a structural concept that promised to meet the stability requirements and could guide the detail mechanical design.  And an engineering tool for occasional spot-checks and trade-off studies.

Joy and Happiness!

Ahhh… April.

Al H.

Optomechanics – Computational Zero is a Relative Term


You more astute readers have made a good point about my call for completeness and congruence in the modeling data for Nastran optomechanical models:  These properties are a necessary condition but, alas, not sufficient. 

The goal is to achieve computational zeros in the sums of a few dozen terms of a finite element model that may itself contain a few million terms.  But “computational zero” is a relative, not an absolute, term.  The engineer needs to demonstrate that the actual value of the computational zeros are small enough to be ignored.  Rigid body checks alone can’t do that.

And no technology is more sensitive to this issue than optics.  For instance: 

AEH was called to verify the predicted magnitude of dynamic jitter in a spectral imaging system.  AEH assembled an all-up meshed solid finite element Nastran model of the system, put the optical prescription through the AEH/Ivory Optomechanical Modeling Tools to generate the complete and congruent data that control the motions of the image on the detector and ran the rigid body checks.  Everything so far seemed reasonable.

Then AEH ran three-axis static gravity checks.  Considering the frequencies, displacements and accelerations the instrument would see AEH judged that the computational zeros, which dominated the rigid body checks, were two orders of magnitude too large compared to the static gravity checks.  Correcting this required three more significant figures in the optomechanical modeling data (both the geometry and the coefficients) for sufficient precision in the dynamic jitter analysis.

Fortunately, the software also produces both geometric and coefficient data with eighteen significant figures.  AEH was able to edit the additional significant figures into the Nastran database.  Finally, AEH ran the frequency response spectrum.  The image jitter proved to be safely within the system’s specified maximum. 

But, the engineer had to take the critical steps to demonstrate sufficient precision in the optomechanical modeling data.

AEH offers completeness, congruence and precision in those critical optomechanical analyses.

Well, the March Hare is here, “I’m late, I’m late, I’m late!”

And as Lewis Carol further wrote, “The time has come….”

See you all in Baltimore.

Al H.

Optomechanics – Ebony vs CodeV and Zemax


As a coda to last weeks missive I reserved the knottiest part for today, All Hallows Eve.

My optical design friends like to solve for the point spread function in a lens design code like CodeV or Zemax.  To achieve the same accuracy as my Nastran analysis they would require the use of the first 5,280 Zernike polynomials in the series.  With some effort the number could possibly be reduced to just 264 from among those polynomials.  Then they would trace the rays.  Have any of you ever used 264 Zernike polynomials in a lens design code at the same time?  Great Fun!

My Halloween treat for all of you is a direct solution, via AEH/Ebony and Nastran, to structurally deformed point spread functions.  Even More Fun!

Now, Beware the Great Pumpkin!

Happy Halloween.

Al H.

Optomechanics – Optical Analog, OA for Short


It all began with the “Optical Analog,” OA for short. 

OA is what I’ve called my method for modeling the optical point spread function (PSF) in Nastran structural models of optical systems.  I started simple, modeling an axial chief ray and calculating its motion in object and image spaces when I tweek the structure with forces, displacements or thermal gradients. 

After a few successes it became clear that there was a lot more to be learned by modeling multiple optical rays through the system.  Their motions on the focal plane array would not only indicate image motions but also changes to the PSF (and therefore the OTF, a measure of image quality).

A typical application of the OA was to determine the optical effects caused by residual plastic strains in a light weight metallic primary mirror.  The plastic strains were caused by a sudden shock load.  The figure shows two views of the solution. 

The right side shows a 20 degree sector of the primary mirror model with 44 optical rays reflecting from it.  The mirror had 18 such sectors and the problem was axisymmetric.  The left side of the figure shows the results at the center detector of the FPA (blown-up about 5,000 fold).  The black dashed line shows the size of the geometric PSF before the shock load and the red dashed line shows the geometric PSF after the shock load.  The project had a strict requirement for “ensquared energy” on the detectors and I thinned-out the face sheet and webs until the results were just within the specification.

I wrote Ebony, a computer program, to assist in assembling structural models for OA analyses.  It’s one of my optomechanical modeling tools that I use to help guide mechanical designs.  I tend to put them to work in the early days of a project, while the concepts are malleable.  They’re also useful in “Red Team” assignments to find out, after-the-fact, what went wrong and what it takes to fix it.  AEH/Ebony unifies and couples the PSF to all of the structure in the Nastran model.

If you’re waiting for the beginning of Summer, some good news.  You have only eight months to go!

All Hallow Even comes first, of course.

Joy and good health to all.

Al Hatheway