DMD flatness for spatial light modulator application

Markus,

Welcome to the DLP E2E forum.

I am interested in how you measured the flatness of the DMD mirror array.

The "flat" state of the mirrors in the DMD - that is the "off" state tilt - is not specified. In the "off" state mirror tilt is not controlled, or specified. In the "off" state a mirror is not being actively held in position (+/- 12 degree tilt), and therefore each "off" mirror can exhibit some deviation in tilt from the plane of the DMD array.

What is specified is the "on" state tilt: 12 degrees, +/- 1 degree. There are 2 on states: +/-12 degrees, corresponding to a pixel data value of 1 or 0. The flatness of the DMD window is also specified in (at least) some DMD data sheets.

I'm sorry that there is really no way to select DMDs with maximum flatness. Over time and operation, the mirrors will develop some bias in their off state tilts. This will not affect their on state tilts.

Perhaps there is a way to implement your ideas with the mirrors in their "on" and controlled states.

Best regards,

Pascal

Hi Pascal,

thanks for your email!

What I meant with flatness was wavefront distortion, i.e. when I send in a plane wave, is the reflected wave plane or distorted. I was not talking about tilt angles. 

Flatness (e.g. of a mirror) can be measured with an interferometer, e.g. as can be seen in this article where the flatness of a DMD device is beeing measured:

http://www.astro.wisc.edu/~wilson/Results.html

A normal lasermirror has a flatness of about lambda/10, meaning that the RMS distortion of the surface is less then a tenth of a wavelength, i.e. about 63 nm. The DMD seems to be about 10 - 100 times worse, hence it does not act as a flat laser mirror. Again, I am not talking about flat is in tilted mirrors, rather flat as in the entire chip is not deformed or bent. 

Best,

Markus

Markus,

Thank you for explaining further. Also, thanks for the link - it is very helpful.

As for the comparison of a laser mirror vs a DMD, the linked article does show a contour map of a DMD which is very interesting - taken by a "phase shifting interferometer". It is not clear whether all of the mirrors were activated (tilted) or if the measurement was taken with the mirrors in their "off" or non-tilted state. I can see that the wavefront distortion measurement depends upon only the relative "heights" of the pixels across the DMD surface. I presume that this measurement would show the same results with the mirrors tilted or floating ("off"), although there may be some change in the height of the centroid of an individual mirror as it tilts from "flat" to +/- 12 degrees.

So, the measurement appears to be a valid indicator of the phase height of the DMD array across its surface. This is indeed nowhere near the achievable flatness for a single plate optical mirror. The DMD is a MEMS device, with a 3-dimensional metal structure on top of a silicon chip. The laser mirror is (probably) polished to an optically flat figure.

So - these are the characteristics of a DMD. As the linked article demonstrates, it is possible to use the DMD itself to correct certain optical distortions or shortcomings. But, it can not naively be treated as a simple (!) optical flat.

Best regards,

Pascal

Hi Pascal,

thanks. Due to the symmetry of the mirrors I expect the map not to depend on the mirror state. 

You are right, the dmd chips are amazing enough, and it is not surprising that they are not as flat as a simple mirror. My guess is that the bending only occurs in some late mounting step as the mems chip gets glued to a heatsink - the pattern to me looks like what you would get under such circumstances. Anyways, since it seems that the pattern is benign in the sense that it is smooth we can probably compensate it by measuring it exactly and then adopting the dmd pattern accordingly. Also, I realized that one can use the dmd chip itself to interferometrically measure the higth of each pixel, so this is what we will try.

I was just curious whether you had any info on whether this bending is more severe with some chips then with others, e.g. the .55 inch vs .7 inch vs. 1080p chips. I will probably just measure different chips. Do you know who I could contact to find out whether I could borrow a few samples to perform that measurement? (I don't have to operate the chips - mirror off state would be sufficient).  I would be happy to share the results.

Best,

Markus

Hello Markus

The diffraction efficiency obtained is relatively low, best case would be when you work in on axis symmetric data projection and tilting the DMD at the right angle to reach maximum output optimization same as what is shown in the article you presented. to do that all you need to do is tilt the DMD while projecting a vis laser on the active surface and watching the plane while you tilt it until the astigmatism reaches minimum level. In case you wish to reconstruct the propagating beam without having to deal too much with aberration reading of the DMD itself and then compensate on that in the transmission design, you could use random phase backward propagation signal field technique and this would reduce outcome dependency on DMD level. The initial transmission is obtained from propagating the current signal field with a superimposed random phase backward to the transmission plane. You'd be able to achieve clear data reconstruction at the planned distance (fresnel transform approach- which would be better to work with than Farfield in this case)
It would be better to work with a conjugating beam in which angle of incident has high significance due to the diffraction efficiency level of the transmission in this case and would affect the quality of the reconstructed data you would obtain at the projected plane.

A similar approach would be creating a DOE defuser, this means you'd be minimizing dependency on input whether it comes from the source or the DMD itself.

Sincerely,

Yuval